Upgrading transformers 3 to 4
Breaking3.5.1 → 4.57.6Breaking71Security16Platform2Deprecated5New439Changed58Fixed4895Other355
Fetching 176 releases from GitHub…74 / 176
SecurityMediumHugging Face Transformers vulnerable to Regular Expression Denial of Service (ReDoS) in the AdamWeightDecay optimizerGHSA-4w7r-h757-3r74 · CVE-2025-6921 ↗
BreakingUpgrade verdict
71 breaking changes and 2 platform requirement changes and 16 security fixes.
In plain terms
- Remove invalid @staticmethod from module-level get_device_and_memory_breakdown.
- Remove offload_state_dict from kwargs.
- Remove Group Beam Search decoding strategy.
- Remove Constrained Beam Search decoding strategy.
Must act Top 5 of 94 that need code or config changes
Python requirement changed in 4.47.0: >=3.6.0 → >=3.9.0.
Adds huggingface-hub <1.0,>=0.34.0, pyyaml >=5.1, safetensors >=0.4.3.
Drops protobuf, sacremoses, sentencepiece.
Also installs fsspec, hf-xet, typing-extensions.
No longer installs click, cloudpickle, joblib.
Moves tokenizers 0.9.3 → 0.22.2. Your build may pick other versions.
What you must change
Breaking71
Remove invalid @staticmethod from module-level get_device_and_memory_breakdown.
@staticmethodRemove offload_state_dict from kwargs.
Remove Group Beam Search decoding strategy.
Remove Constrained Beam Search decoding strategy.
[generate] ignore cache_implementation="hybrid" hub defaults.
cache_implementation="hybrid"Switch default compilation to fullgraph=False.
Remove DoLa decoding strategy.
Remove Contrastive Search decoding strategy.
[Trainer] Enable average_tokens_across_devices by default in TrainingArguments.
average_tokens_across_devicesTrainingArgumentsUpdate default dtype for pipelines to auto.
dtypeautoRemove HQQ from caching allocator warmup.
Delete conversion scripts when making release wheels.
to make modeling code simple as single-worker case: all manual TP implementations under if self.config.pretraining_tp > 1 can be removed.
Remove warning v2.
No more default chat templates.
Remove unnecessary guard code related with pytorch versions 1.4.2 ~ 1.7.0.
Remove size check between attn_weights and kv_seq_len for phi3.
TextGenerationPipeline: rely on the tokenizer default kwargs.
Remove ConversationalPipeline and Conversation object.
FLAVA: Remove double softmax.
Remove dataset with restrictive license.
[LlamaTokenizerFast] Refactor default llama.
LlamaTokenizerFastRemove static pretrained maps from the library's internals.
[BC 4.37 -> 4.38] for Llama family, memory and speed (causal mask is no longer a registered buffer).
Removed obsolete attribute setting for AQLM quantization.
Allow check_model_inputs in core VLMs.
check_model_inputsUpdate Glm4V to use config values.
Fix Inconsistant input_feature length and attention_mask length in WhisperFeatureExtractor.
input_featureattention_maskWhisperFeatureExtractorAdd ministral model.
Move variable output controls to _prepare_generation_config.
_prepare_generation_configMake center_crop fast equivalent to slow.
center_cropUse lru_cache for sine pos embeddings MaskFormer.
Always return Cache objects in modelings (to align with generate).
byebye torch 2.1.
[Whisper] 🚨 Fix pipeline word timestamp: timestamp token is end of token time !!!
Don't use cache in non-generative models.
[eomt] make EoMT compatible with pipeline.
Fix and simplify attention implementation dispatch and subconfigs handling.
Fix initialization of Mask2Former.
Early-error🚨 config will error out if output_attentions=True and the attn implementation is wrong.
output_attentions=True[VLM] modeling updates.
rotating_light: :rotating_light: Fix custom code saving.
[core] Completely rewrite the masking logic for all attentions.
[Attention] Refactor Attention Interface for Bart-based Models.
Attention[Attention] Attention refactor for Whisper-based models.
AttentionAdd CB.
rm already deprecated pad_to_max_length arg.
Fix forward of Dinov2ForImageClassification for models with registers.
[VLM] Add base model without head.
Video processors as a separate class.
Allow saving and loading multiple "raw" chat template files.
Update CLIP vision attention to new attention interface.
Setup -> setupclass conversion.
Added segmentation maps support for DPT image processor.
image-classification pipeline single-label and multi-label prob type squashing fns (sigmoid vs softmax) are backwards.
An attempt to fix.
Limit backtracking in Nougat regexp.
Short-form: Returned a ModelOutput or torch.LongTensor, including decoder input IDs and the EOS token ID.
ModelOutputtorch.LongTensorLong-form: Returned a Dict or torch.LongTensor, excluding decoder input IDs and the EOS token ID.
Dicttorch.LongTensor[Whisper] 🚨 Fix whisper decoding 🚨.
Remove conversational pipeline tests.
[Idefics2] Update ignore index.
out_indices always a list.
Deprecate evaluation_strategy to eval_strategy🚨🚨🚨.
evaluation_strategyeval_strategyAdd training compatibility for Musicgen-like models.
Update image_processing_vitmatte.py.
generate fix breaking change for patch.
Fully revert atomic checkpointing 🚨.
Fix EnCodec internals and integration tests.
rotating_light: :rotating_light: Inherited CausalLM Tests.
[Whisper Tok] Update integration test.
Security16
Hugging Face Transformers vulnerable to Regular Expression Denial of Service (ReDoS) in the AdamWeightDecay optimizer
MediumHugging Face Transformers is vulnerable to ReDoS through its MarianTokenizer
MediumHugging Face Transformers Regular Expression Denial of Service (ReDoS) vulnerability
MediumHugging Face Transformers library has Regular Expression Denial of Service
MediumTransformers is vulnerable to ReDoS attack through its DonutProcessor class
MediumTransformers's Improper Input Validation vulnerability can be exploited through username injection
LowTransformers vulnerable to ReDoS attack through its SETTING_RE variable
MediumTransformers vulnerable to ReDoS attack through its get_imports() function
MediumTransformers's ReDoS vulnerability in get_configuration_file can lead to catastrophic backtracking
MediumTransformers Regular Expression Denial of Service (ReDoS) vulnerability
MediumA vulnerability in the
Highpreprocess_string() function of the transformers.testing_utils module in huggingface/transformers version v4.48.3 allows for a Regular Expression Denial of Service (ReDoS) attack. The regular expression used to process code blocks in docstrings contains nested quantifiers, leading to exponential backtracking when processing input with a large number of newline characters. An attacker can exploit this by providing a specially crafted payload, causing high CPU usage and potential application downtime, effectively resulting in a Denial of Service (DoS) scenario.Transformers Regular Expression Denial of Service (ReDoS) vulnerability
MediumHugging Face Transformers Trax Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file.
HighHugging Face Transformers MobileViTV2 Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file.
HighHugging Face Transformers MaskFormer Model Deserialization of Untrusted Data Remote Code Execution Vulnerability. This vulnerability allows remote attackers to execute arbitrary code on affected installations of Hugging Face Transformers. User interaction is required to exploit this vulnerability in that the target must visit a malicious page or open a malicious file.
HighTransformers Deserialization of Untrusted Data vulnerability
LowPlatform2
Merge type hints from microsoft/python-type-stubs (post dropping support for Python 3.8).
microsoft/python-type-stubsDrop support for Python 3.8.
Deprecations to plan for
Deprecated5
[sliding window] revert and deprecate.
[agents] remove agents 🧹.
byebye torch 2.0.
Deprecate and map Causal to Conditional.
Deprecate transformers.agents.
API changes
Generated by DepDelta from published Python source. Not mixed with prose notes.
- removed:
export ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export ALL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export AdamW - removed:
export AdaptiveEmbedding - removed:
export AlbertConfig - removed:
export AlbertForMaskedLM - removed:
export AlbertForMultipleChoice - removed:
export AlbertForPreTraining - removed:
export AlbertForQuestionAnswering - removed:
export AlbertForSequenceClassification - removed:
export AlbertForTokenClassification - removed:
export AlbertModel - removed:
export AlbertPreTrainedModel - removed:
export AlbertTokenizer - removed:
export AlbertTokenizerFast - removed:
export AutoConfig - removed:
export AutoModel - removed:
export AutoModelForCausalLM - removed:
export AutoModelForMaskedLM - removed:
export AutoModelForMultipleChoice - removed:
export AutoModelForNextSentencePrediction - removed:
export AutoModelForPreTraining - removed:
export AutoModelForQuestionAnswering - removed:
export AutoModelForSeq2SeqLM - removed:
export AutoModelForSequenceClassification - removed:
export AutoModelForTokenClassification - removed:
export AutoModelWithLMHead - removed:
export AutoTokenizer - removed:
export BART_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export BERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export BartConfig - removed:
export BartForConditionalGeneration - removed:
export BartForQuestionAnswering - removed:
export BartForSequenceClassification - removed:
export BartModel - removed:
export BartTokenizer - removed:
export BartTokenizerFast - removed:
export BasicTokenizer - removed:
export BeamSearchScorer - removed:
export BertConfig - removed:
export BertForMaskedLM - removed:
export BertForMultipleChoice - removed:
export BertForNextSentencePrediction - removed:
export BertForPreTraining - removed:
export BertForQuestionAnswering - removed:
export BertForSequenceClassification - removed:
export BertForTokenClassification - removed:
export BertGenerationConfig - removed:
export BertGenerationDecoder - removed:
export BertGenerationEncoder - removed:
export BertGenerationTokenizer - removed:
export BertJapaneseTokenizer - removed:
export BertLMHeadModel - removed:
export BertLayer - removed:
export BertModel - removed:
export BertPreTrainedModel - removed:
export BertTokenizer - removed:
export BertTokenizerFast - removed:
export BertweetTokenizer - removed:
export BlenderbotConfig - removed:
export BlenderbotForConditionalGeneration - removed:
export BlenderbotSmallTokenizer - removed:
export BlenderbotTokenizer - removed:
export CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export CONFIG_MAPPING - removed:
export CTRLConfig - removed:
export CTRLLMHeadModel - removed:
export CTRLModel - removed:
export CTRLPreTrainedModel - removed:
export CTRLTokenizer - removed:
export CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export CamembertConfig - removed:
export CamembertForCausalLM - removed:
export CamembertForMaskedLM - removed:
export CamembertForMultipleChoice - removed:
export CamembertForQuestionAnswering - removed:
export CamembertForSequenceClassification - removed:
export CamembertForTokenClassification - removed:
export CamembertModel - removed:
export CamembertTokenizer - removed:
export CamembertTokenizerFast - removed:
export CharacterTokenizer - removed:
export Conversation - removed:
export ConversationalPipeline - removed:
export DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export DPRConfig - removed:
export DPRContextEncoder - removed:
export DPRContextEncoderTokenizer - removed:
export DPRContextEncoderTokenizerFast - removed:
export DPRPretrainedContextEncoder - removed:
export DPRPretrainedQuestionEncoder - removed:
export DPRPretrainedReader - removed:
export DPRQuestionEncoder - removed:
export DPRQuestionEncoderTokenizer - removed:
export DPRQuestionEncoderTokenizerFast - removed:
export DPRReader - removed:
export DPRReaderOutput - removed:
export DPRReaderTokenizer - removed:
export DPRReaderTokenizerFast - removed:
export DPR_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export DebertaConfig - removed:
export DebertaForSequenceClassification - removed:
export DebertaModel - removed:
export DebertaPreTrainedModel - removed:
export DebertaTokenizer - removed:
export DistilBertConfig - removed:
export DistilBertForMaskedLM - removed:
export DistilBertForMultipleChoice - removed:
export DistilBertForQuestionAnswering - removed:
export DistilBertForSequenceClassification - removed:
export DistilBertForTokenClassification - removed:
export DistilBertModel - removed:
export DistilBertPreTrainedModel - removed:
export DistilBertTokenizer - removed:
export DistilBertTokenizerFast - removed:
export ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export ElectraConfig - removed:
export ElectraForMaskedLM - removed:
export ElectraForMultipleChoice - removed:
export ElectraForPreTraining - removed:
export ElectraForQuestionAnswering - removed:
export ElectraForSequenceClassification - removed:
export ElectraForTokenClassification - removed:
export ElectraModel - removed:
export ElectraPreTrainedModel - removed:
export ElectraTokenizer - removed:
export ElectraTokenizerFast - removed:
export EncoderDecoderConfig - removed:
export EncoderDecoderModel - removed:
export EvaluationStrategy - removed:
export FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export FSMTConfig - removed:
export FSMTForConditionalGeneration - removed:
export FSMTModel - removed:
export FSMTTokenizer - removed:
export FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export FlaubertConfig - removed:
export FlaubertForMultipleChoice - removed:
export FlaubertForQuestionAnswering - removed:
export FlaubertForQuestionAnsweringSimple - removed:
export FlaubertForSequenceClassification - removed:
export FlaubertForTokenClassification - removed:
export FlaubertModel - removed:
export FlaubertTokenizer - removed:
export FlaubertWithLMHeadModel - removed:
export FlaxBertModel - removed:
export FlaxRobertaModel - removed:
export FunnelBaseModel - removed:
export FunnelConfig - removed:
export FunnelForMaskedLM - removed:
export FunnelForMultipleChoice - removed:
export FunnelForPreTraining - removed:
export FunnelForQuestionAnswering - removed:
export FunnelForSequenceClassification - removed:
export FunnelForTokenClassification - removed:
export FunnelModel - removed:
export FunnelTokenizer - removed:
export FunnelTokenizerFast - removed:
export GPT2Config - removed:
export GPT2DoubleHeadsModel - removed:
export GPT2ForSequenceClassification - removed:
export GPT2LMHeadModel - removed:
export GPT2Model - removed:
export GPT2PreTrainedModel - removed:
export GPT2Tokenizer - removed:
export GPT2TokenizerFast - removed:
export GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export HerbertTokenizer - removed:
export HerbertTokenizerFast - removed:
export LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export LayoutLMConfig - removed:
export LayoutLMForMaskedLM - removed:
export LayoutLMForTokenClassification - removed:
export LayoutLMModel - removed:
export LayoutLMTokenizer - removed:
export LayoutLMTokenizerFast - removed:
export LogitsWarper - removed:
export LongformerConfig - removed:
export LongformerForMaskedLM - removed:
export LongformerForMultipleChoice - removed:
export LongformerForQuestionAnswering - removed:
export LongformerForSequenceClassification - removed:
export LongformerForTokenClassification - removed:
export LongformerModel - removed:
export LongformerSelfAttention - removed:
export LongformerTokenizer - removed:
export LongformerTokenizerFast - removed:
export LxmertConfig - removed:
export LxmertEncoder - removed:
export LxmertForPreTraining - removed:
export LxmertForQuestionAnswering - removed:
export LxmertModel - removed:
export LxmertPreTrainedModel - removed:
export LxmertTokenizer - removed:
export LxmertTokenizerFast - removed:
export LxmertVisualFeatureEncoder - removed:
export LxmertXLayer - removed:
export MBartConfig - removed:
export MBartForConditionalGeneration - removed:
export MBartTokenizer - removed:
export MBartTokenizerFast - removed:
export MMBTConfig - removed:
export MMBTForClassification - removed:
export MMBTModel - removed:
export MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export MODEL_FOR_CAUSAL_LM_MAPPING - removed:
export MODEL_FOR_MASKED_LM_MAPPING - removed:
export MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
export MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - removed:
export MODEL_FOR_PRETRAINING_MAPPING - removed:
export MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
export MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
export MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
export MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
export MODEL_MAPPING - removed:
export MODEL_WITH_LM_HEAD_MAPPING - removed:
export MarianConfig - removed:
export MarianMTModel - removed:
export MarianTokenizer - removed:
export MecabTokenizer - removed:
export MobileBertConfig - removed:
export MobileBertForMaskedLM - removed:
export MobileBertForMultipleChoice - removed:
export MobileBertForNextSentencePrediction - removed:
export MobileBertForPreTraining - removed:
export MobileBertForQuestionAnswering - removed:
export MobileBertForSequenceClassification - removed:
export MobileBertForTokenClassification - removed:
export MobileBertLayer - removed:
export MobileBertModel - removed:
export MobileBertPreTrainedModel - removed:
export MobileBertTokenizer - removed:
export MobileBertTokenizerFast - removed:
export ModalEmbeddings - removed:
export OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export OpenAIGPTConfig - removed:
export OpenAIGPTDoubleHeadsModel - removed:
export OpenAIGPTForSequenceClassification - removed:
export OpenAIGPTLMHeadModel - removed:
export OpenAIGPTModel - removed:
export OpenAIGPTPreTrainedModel - removed:
export OpenAIGPTTokenizer - removed:
export OpenAIGPTTokenizerFast - removed:
export PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export PegasusConfig - removed:
export PegasusForConditionalGeneration - removed:
export PegasusTokenizer - removed:
export PegasusTokenizerFast - removed:
export PhobertTokenizer - removed:
export PretrainedBartModel - removed:
export PretrainedFSMTModel - removed:
export ProphetNetConfig - removed:
export ProphetNetDecoder - removed:
export ProphetNetEncoder - removed:
export ProphetNetForCausalLM - removed:
export ProphetNetForConditionalGeneration - removed:
export ProphetNetModel - removed:
export ProphetNetPreTrainedModel - removed:
export ProphetNetTokenizer - removed:
export PyTorchBenchmark - removed:
export PyTorchBenchmarkArguments - removed:
export REFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export RagConfig - removed:
export RagModel - removed:
export RagRetriever - removed:
export RagSequenceForGeneration - removed:
export RagTokenForGeneration - removed:
export RagTokenizer - removed:
export ReformerAttention - removed:
export ReformerConfig - removed:
export ReformerForMaskedLM - removed:
export ReformerForQuestionAnswering - removed:
export ReformerForSequenceClassification - removed:
export ReformerLayer - removed:
export ReformerModel - removed:
export ReformerModelWithLMHead - removed:
export ReformerTokenizer - removed:
export ReformerTokenizerFast - removed:
export RetriBertConfig - removed:
export RetriBertModel - removed:
export RetriBertPreTrainedModel - removed:
export RetriBertTokenizer - removed:
export RetriBertTokenizerFast - removed:
export RobertaConfig - removed:
export RobertaForCausalLM - removed:
export RobertaForMaskedLM - removed:
export RobertaForMultipleChoice - removed:
export RobertaForQuestionAnswering - removed:
export RobertaForSequenceClassification - removed:
export RobertaForTokenClassification - removed:
export RobertaModel - removed:
export RobertaTokenizer - removed:
export RobertaTokenizerFast - removed:
export SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export SqueezeBertConfig - removed:
export SqueezeBertForMaskedLM - removed:
export SqueezeBertForMultipleChoice - removed:
export SqueezeBertForQuestionAnswering - removed:
export SqueezeBertForSequenceClassification - removed:
export SqueezeBertForTokenClassification - removed:
export SqueezeBertModel - removed:
export SqueezeBertModule - removed:
export SqueezeBertPreTrainedModel - removed:
export SqueezeBertTokenizer - removed:
export SqueezeBertTokenizerFast - removed:
export T5Config - removed:
export T5ForConditionalGeneration - removed:
export T5Model - removed:
export T5PreTrainedModel - removed:
export T5Tokenizer - removed:
export T5TokenizerFast - removed:
export T5_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TFAdaptiveEmbedding - removed:
export TFAlbertForMaskedLM - removed:
export TFAlbertForMultipleChoice - removed:
export TFAlbertForPreTraining - removed:
export TFAlbertForQuestionAnswering - removed:
export TFAlbertForSequenceClassification - removed:
export TFAlbertForTokenClassification - removed:
export TFAlbertMainLayer - removed:
export TFAlbertModel - removed:
export TFAlbertPreTrainedModel - removed:
export TFAutoModel - removed:
export TFAutoModelForCausalLM - removed:
export TFAutoModelForMaskedLM - removed:
export TFAutoModelForMultipleChoice - removed:
export TFAutoModelForPreTraining - removed:
export TFAutoModelForQuestionAnswering - removed:
export TFAutoModelForSeq2SeqLM - removed:
export TFAutoModelForSequenceClassification - removed:
export TFAutoModelForTokenClassification - removed:
export TFAutoModelWithLMHead - removed:
export TFBartForConditionalGeneration - removed:
export TFBartModel - removed:
export TFBertEmbeddings - removed:
export TFBertForMaskedLM - removed:
export TFBertForMultipleChoice - removed:
export TFBertForNextSentencePrediction - removed:
export TFBertForPreTraining - removed:
export TFBertForQuestionAnswering - removed:
export TFBertForSequenceClassification - removed:
export TFBertForTokenClassification - removed:
export TFBertLMHeadModel - removed:
export TFBertMainLayer - removed:
export TFBertModel - removed:
export TFBertPreTrainedModel - removed:
export TFBlenderbotForConditionalGeneration - removed:
export TFCTRLLMHeadModel - removed:
export TFCTRLModel - removed:
export TFCTRLPreTrainedModel - removed:
export TFCamembertForMaskedLM - removed:
export TFCamembertForMultipleChoice - removed:
export TFCamembertForQuestionAnswering - removed:
export TFCamembertForSequenceClassification - removed:
export TFCamembertForTokenClassification - removed:
export TFCamembertModel - removed:
export TFDistilBertForMaskedLM - removed:
export TFDistilBertForMultipleChoice - removed:
export TFDistilBertForQuestionAnswering - removed:
export TFDistilBertForSequenceClassification - removed:
export TFDistilBertForTokenClassification - removed:
export TFDistilBertMainLayer - removed:
export TFDistilBertModel - removed:
export TFDistilBertPreTrainedModel - removed:
export TFElectraForMaskedLM - removed:
export TFElectraForMultipleChoice - removed:
export TFElectraForPreTraining - removed:
export TFElectraForQuestionAnswering - removed:
export TFElectraForSequenceClassification - removed:
export TFElectraForTokenClassification - removed:
export TFElectraModel - removed:
export TFElectraPreTrainedModel - removed:
export TFFlaubertForMultipleChoice - removed:
export TFFlaubertForQuestionAnsweringSimple - removed:
export TFFlaubertForSequenceClassification - removed:
export TFFlaubertForTokenClassification - removed:
export TFFlaubertModel - removed:
export TFFlaubertWithLMHeadModel - removed:
export TFFunnelBaseModel - removed:
export TFFunnelForMaskedLM - removed:
export TFFunnelForMultipleChoice - removed:
export TFFunnelForPreTraining - removed:
export TFFunnelForQuestionAnswering - removed:
export TFFunnelForSequenceClassification - removed:
export TFFunnelForTokenClassification - removed:
export TFFunnelModel - removed:
export TFGPT2DoubleHeadsModel - removed:
export TFGPT2LMHeadModel - removed:
export TFGPT2MainLayer - removed:
export TFGPT2Model - removed:
export TFGPT2PreTrainedModel - removed:
export TFLongformerForMaskedLM - removed:
export TFLongformerForQuestionAnswering - removed:
export TFLongformerModel - removed:
export TFLongformerSelfAttention - removed:
export TFLxmertForPreTraining - removed:
export TFLxmertMainLayer - removed:
export TFLxmertModel - removed:
export TFLxmertPreTrainedModel - removed:
export TFLxmertVisualFeatureEncoder - removed:
export TFMBartForConditionalGeneration - removed:
export TFMarianMTModel - removed:
export TFMobileBertForMaskedLM - removed:
export TFMobileBertForMultipleChoice - removed:
export TFMobileBertForNextSentencePrediction - removed:
export TFMobileBertForPreTraining - removed:
export TFMobileBertForQuestionAnswering - removed:
export TFMobileBertForSequenceClassification - removed:
export TFMobileBertForTokenClassification - removed:
export TFMobileBertMainLayer - removed:
export TFMobileBertModel - removed:
export TFMobileBertPreTrainedModel - removed:
export TFOpenAIGPTDoubleHeadsModel - removed:
export TFOpenAIGPTLMHeadModel - removed:
export TFOpenAIGPTMainLayer - removed:
export TFOpenAIGPTModel - removed:
export TFOpenAIGPTPreTrainedModel - removed:
export TFPegasusForConditionalGeneration - removed:
export TFRobertaForMaskedLM - removed:
export TFRobertaForMultipleChoice - removed:
export TFRobertaForQuestionAnswering - removed:
export TFRobertaForSequenceClassification - removed:
export TFRobertaForTokenClassification - removed:
export TFRobertaMainLayer - removed:
export TFRobertaModel - removed:
export TFRobertaPreTrainedModel - removed:
export TFT5ForConditionalGeneration - removed:
export TFT5Model - removed:
export TFT5PreTrainedModel - removed:
export TFTrainer - removed:
export TFTransfoXLLMHeadModel - removed:
export TFTransfoXLMainLayer - removed:
export TFTransfoXLModel - removed:
export TFTransfoXLPreTrainedModel - removed:
export TFXLMForMultipleChoice - removed:
export TFXLMForQuestionAnsweringSimple - removed:
export TFXLMForSequenceClassification - removed:
export TFXLMForTokenClassification - removed:
export TFXLMMainLayer - removed:
export TFXLMModel - removed:
export TFXLMPreTrainedModel - removed:
export TFXLMRobertaForMaskedLM - removed:
export TFXLMRobertaForMultipleChoice - removed:
export TFXLMRobertaForQuestionAnswering - removed:
export TFXLMRobertaForSequenceClassification - removed:
export TFXLMRobertaForTokenClassification - removed:
export TFXLMRobertaModel - removed:
export TFXLMWithLMHeadModel - removed:
export TFXLNetForMultipleChoice - removed:
export TFXLNetForQuestionAnsweringSimple - removed:
export TFXLNetForSequenceClassification - removed:
export TFXLNetForTokenClassification - removed:
export TFXLNetLMHeadModel - removed:
export TFXLNetMainLayer - removed:
export TFXLNetModel - removed:
export TFXLNetPreTrainedModel - removed:
export TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_MODEL_FOR_CAUSAL_LM_MAPPING - removed:
export TF_MODEL_FOR_MASKED_LM_MAPPING - removed:
export TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
export TF_MODEL_FOR_PRETRAINING_MAPPING - removed:
export TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
export TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
export TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
export TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
export TF_MODEL_MAPPING - removed:
export TF_MODEL_WITH_LM_HEAD_MAPPING - removed:
export TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TOKENIZER_MAPPING - removed:
export TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export TensorFlowBenchmark - removed:
export TensorFlowBenchmarkArguments - removed:
export TransfoXLConfig - removed:
export TransfoXLCorpus - removed:
export TransfoXLLMHeadModel - removed:
export TransfoXLModel - removed:
export TransfoXLPreTrainedModel - removed:
export TransfoXLTokenizer - removed:
export WordpieceTokenizer - removed:
export XLMConfig - removed:
export XLMForMultipleChoice - removed:
export XLMForQuestionAnswering - removed:
export XLMForQuestionAnsweringSimple - removed:
export XLMForSequenceClassification - removed:
export XLMForTokenClassification - removed:
export XLMModel - removed:
export XLMPreTrainedModel - removed:
export XLMProphetNetConfig - removed:
export XLMProphetNetDecoder - removed:
export XLMProphetNetEncoder - removed:
export XLMProphetNetForCausalLM - removed:
export XLMProphetNetForConditionalGeneration - removed:
export XLMProphetNetModel - removed:
export XLMProphetNetTokenizer - removed:
export XLMRobertaConfig - removed:
export XLMRobertaForCausalLM - removed:
export XLMRobertaForMaskedLM - removed:
export XLMRobertaForMultipleChoice - removed:
export XLMRobertaForQuestionAnswering - removed:
export XLMRobertaForSequenceClassification - removed:
export XLMRobertaForTokenClassification - removed:
export XLMRobertaModel - removed:
export XLMRobertaTokenizer - removed:
export XLMRobertaTokenizerFast - removed:
export XLMTokenizer - removed:
export XLMWithLMHeadModel - removed:
export XLM_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
export XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
export XLNetConfig - removed:
export XLNetForMultipleChoice - removed:
export XLNetForQuestionAnswering - removed:
export XLNetForQuestionAnsweringSimple - removed:
export XLNetForSequenceClassification - removed:
export XLNetForTokenClassification - removed:
export XLNetLMHeadModel - removed:
export XLNetModel - removed:
export XLNetPreTrainedModel - removed:
export XLNetTokenizer - removed:
export XLNetTokenizerFast - removed:
export cached_path - removed:
export is_torch_tpu_available - removed:
export load_tf_weights_in_albert - removed:
export load_tf_weights_in_bert - removed:
export load_tf_weights_in_bert_generation - removed:
export load_tf_weights_in_electra - removed:
export load_tf_weights_in_funnel - removed:
export load_tf_weights_in_gpt2 - removed:
export load_tf_weights_in_mobilebert - removed:
export load_tf_weights_in_openai_gpt - removed:
export load_tf_weights_in_t5 - removed:
export load_tf_weights_in_transfo_xl - removed:
export load_tf_weights_in_xlnet - removed:
export tf_top_k_top_p_filtering - removed:
export top_k_top_p_filtering - added:
export AlternatingCodebooksLogitsProcessor - added:
export AqlmConfig - added:
export AsyncTextIteratorStreamer - added:
export AttentionInterface - added:
export AttentionMaskInterface - added:
export AudioClassificationPipeline - added:
export AutoRoundConfig - added:
export AutomaticSpeechRecognitionPipeline - added:
export AwqConfig - added:
export BaseImageProcessor - added:
export BaseImageProcessorFast - added:
export BaseVideoProcessor - added:
export BatchFeature - added:
export BayesianDetectorConfig - added:
export BayesianDetectorModel - added:
export BitNetQuantConfig - added:
export BitsAndBytesConfig - added:
export Cache - added:
export ChunkedSlidingLayer - added:
export ClassifierFreeGuidanceLogitsProcessor - added:
export CompileConfig - added:
export CompressedTensorsConfig - added:
export ConstrainedBeamSearchScorer - added:
export Constraint - added:
export ConstraintListState - added:
export DataCollatorForMultipleChoice - added:
export DataCollatorForSeq2Seq - added:
export DataCollatorWithFlattening - added:
export DefaultDataCollator - added:
export DepthEstimationPipeline - added:
export DisjunctiveConstraint - added:
export DocumentQuestionAnsweringPipeline - added:
export DynamicCache - added:
export DynamicLayer - added:
export EarlyStoppingCallback - added:
export EetqConfig - added:
export EncoderDecoderCache - added:
export EncoderNoRepeatNGramLogitsProcessor - added:
export EncoderRepetitionPenaltyLogitsProcessor - added:
export EosTokenCriteria - added:
export EpsilonLogitsWarper - added:
export EtaLogitsWarper - added:
export ExponentialDecayLengthPenalty - added:
export FPQuantConfig - added:
export FbgemmFp8Config - added:
export FeatureExtractionMixin - added:
export FineGrainedFP8Config - added:
export FlaxForceTokensLogitsProcessor - added:
export FlaxForcedBOSTokenLogitsProcessor - added:
export FlaxForcedEOSTokenLogitsProcessor - added:
export FlaxGenerationMixin - added:
export FlaxLogitsProcessor - added:
export FlaxLogitsProcessorList - added:
export FlaxLogitsWarper - added:
export FlaxMinLengthLogitsProcessor - added:
export FlaxPreTrainedModel - added:
export FlaxSuppressTokensAtBeginLogitsProcessor - added:
export FlaxSuppressTokensLogitsProcessor - added:
export FlaxTemperatureLogitsWarper - added:
export FlaxTopKLogitsWarper - added:
export FlaxTopPLogitsWarper - added:
export FlaxWhisperTimeStampLogitsProcessor - added:
export ForcedBOSTokenLogitsProcessor - added:
export ForcedEOSTokenLogitsProcessor - added:
export GPTQConfig - added:
export GenerationConfig - added:
export GenerationMixin - added:
export GradientCheckpointingLayer - added:
export HQQQuantizedCache - added:
export HQQQuantizedLayer - added:
export HiggsConfig - added:
export HqqConfig - added:
export HybridCache - added:
export ImageClassificationPipeline - added:
export ImageFeatureExtractionMixin - added:
export ImageFeatureExtractionPipeline - added:
export ImageProcessingMixin - added:
export ImageSegmentationPipeline - added:
export ImageTextToTextPipeline - added:
export ImageToImagePipeline - added:
export ImageToTextPipeline - added:
export InfNanRemoveLogitsProcessor - added:
export IntervalStrategy - added:
export KerasMetricCallback - added:
export KeypointMatchingPipeline - added:
export LogitNormalization - added:
export MambaCache - added:
export MaskGenerationPipeline - added:
export MaxLengthCriteria - added:
export MaxTimeCriteria - added:
export MinNewTokensLengthLogitsProcessor - added:
export MinPLogitsWarper - added:
export ObjectDetectionPipeline - added:
export OffloadedCache - added:
export OffloadedStaticCache - added:
export OptionalDependencyNotAvailable - added:
export PhrasalConstraint - added:
export PrefixConstrainedLogitsProcessor - added:
export ProcessorMixin - added:
export PushToHubCallback - added:
export QuantizedCache - added:
export QuantoConfig - added:
export QuantoQuantizedCache - added:
export QuantoQuantizedLayer - added:
export QuarkConfig - added:
export ROPE_INIT_FUNCTIONS - added:
export SchedulerType - added:
export Seq2SeqTrainer - added:
export Seq2SeqTrainingArguments - added:
export SequenceBiasLogitsProcessor - added:
export SequenceFeatureExtractor - added:
export SinkCache - added:
export SlidingWindowCache - added:
export SlidingWindowLayer - added:
export SpQRConfig - added:
export StaticCache - added:
export StaticLayer - added:
export StaticSlidingWindowLayer - added:
export StopStringCriteria - added:
export StoppingCriteria - added:
export StoppingCriteriaList - added:
export SuppressTokensAtBeginLogitsProcessor - added:
export SuppressTokensLogitsProcessor - added:
export SynthIDTextWatermarkDetector - added:
export SynthIDTextWatermarkLogitsProcessor - added:
export SynthIDTextWatermarkingConfig - added:
export TFForceTokensLogitsProcessor - added:
export TFForcedBOSTokenLogitsProcessor - added:
export TFForcedEOSTokenLogitsProcessor - added:
export TFGenerationMixin - added:
export TFLogitsProcessor - added:
export TFLogitsProcessorList - added:
export TFLogitsWarper - added:
export TFMinLengthLogitsProcessor - added:
export TFNoBadWordsLogitsProcessor - added:
export TFNoRepeatNGramLogitsProcessor - added:
export TFRepetitionPenaltyLogitsProcessor - added:
export TFSuppressTokensAtBeginLogitsProcessor - added:
export TFSuppressTokensLogitsProcessor - added:
export TFTemperatureLogitsWarper - added:
export TFTopKLogitsWarper - added:
export TFTopPLogitsWarper - added:
export TableQuestionAnsweringPipeline - added:
export TextIteratorStreamer - added:
export TextStreamer - added:
export TextToAudioPipeline - added:
export TimmWrapperImageProcessor - added:
export TorchAoConfig - added:
export TorchExportableModuleWithStaticCache - added:
export TypicalLogitsWarper - added:
export UnbatchedClassifierFreeGuidanceLogitsProcessor - added:
export VideoClassificationPipeline - added:
export VisualQuestionAnsweringPipeline - added:
export VptqConfig - added:
export WatermarkDetector - added:
export WatermarkLogitsProcessor - added:
export WatermarkingConfig - added:
export WhisperTimeStampLogitsProcessor - added:
export ZeroShotAudioClassificationPipeline - added:
export ZeroShotImageClassificationPipeline - added:
export ZeroShotObjectDetectionPipeline - added:
export convert_and_export_with_cache - added:
export define_import_structure - added:
export dependency_versions_check - added:
export dummy_flax_objects - added:
export dummy_mistral_common_objects - added:
export dummy_pt_objects - added:
export dummy_sentencepiece_and_tokenizers_objects - added:
export dummy_tf_objects - added:
export dummy_tokenizers_objects - added:
export dummy_torchvision_objects - added:
export dummy_vision_objects - added:
export dynamic_rope_update - added:
export enable_full_determinism - added:
export get_cosine_with_min_lr_schedule_with_warmup - added:
export get_cosine_with_min_lr_schedule_with_warmup_lr_rate - added:
export get_inverse_sqrt_schedule - added:
export get_scheduler - added:
export get_wsd_schedule - added:
export is_av_available - added:
export is_bitsandbytes_available - added:
export is_clearml_available - added:
export is_dvclive_available - added:
export is_essentia_available - added:
export is_g2p_en_available - added:
export is_keras_nlp_available - added:
export is_librosa_available - added:
export is_matplotlib_available - added:
export is_mistral_common_available - added:
export is_mlx_available - added:
export is_neptune_available - added:
export is_phonemizer_available - added:
export is_pretty_midi_available - added:
export is_pyctcdecode_available - added:
export is_ray_tune_available - added:
export is_sacremoses_available - added:
export is_safetensors_available - added:
export is_scipy_available - added:
export is_sigopt_available - added:
export is_speech_available - added:
export is_swanlab_available - added:
export is_tensorflow_text_available - added:
export is_timm_available - added:
export is_torch_hpu_available - added:
export is_torch_mlu_available - added:
export is_torch_musa_available - added:
export is_torch_neuroncore_available - added:
export is_torch_npu_available - added:
export is_torch_xla_available - added:
export is_torch_xpu_available - added:
export is_torchaudio_available - added:
export is_torchvision_available - added:
export is_trackio_available - added:
export is_vision_available - added:
export model_addition_debugger_context - added:
variable import_structure
- removed:
function gelu_fast(x) - removed:
function gelu_new(x) - removed:
function linear_act(x) - removed:
function mish(x) - added:
class AccurateGELUActivation(nn.Module) - added:
class ClassInstantier(OrderedDict) - added:
class ClippedGELUActivation(nn.Module) - added:
class FastGELUActivation(nn.Module) - added:
class GELUActivation(nn.Module) - added:
class GELUTanh(nn.Module) - added:
class LaplaceActivation(nn.Module) - added:
class LinearActivation(nn.Module) - added:
class MishActivation(nn.Module) - added:
class NewGELUActivation(nn.Module) - added:
class QuickGELUActivation(nn.Module) - added:
class ReLUSquaredActivation(nn.Module) - added:
class SiLUActivation(nn.Module) - added:
class XIELUActivation(nn.Module) - added:
constructor AccurateGELUActivation() - added:
constructor ClippedGELUActivation(min: float, max: float) - added:
constructor GELUActivation(use_gelu_python: bool=…) - added:
constructor GELUTanh(use_gelu_tanh_python: bool=…) - added:
constructor MishActivation() - added:
constructor XIELUActivation(alpha_p_init=…, alpha_n_init=…, beta=…, eps=…, dtype=…, with_vector_loads=…) - added:
method AccurateGELUActivation.forward(input: Tensor) -> Tensor - added:
method ClippedGELUActivation.forward(x: Tensor) -> Tensor - added:
method FastGELUActivation.forward(input: Tensor) -> Tensor - added:
method GELUActivation.forward(input: Tensor) -> Tensor - added:
method GELUTanh.forward(input: Tensor) -> Tensor - added:
method LaplaceActivation.forward(input, mu=…, sigma=…) - added:
method LinearActivation.forward(input: Tensor) -> Tensor - added:
method MishActivation.forward(input: Tensor) -> Tensor - added:
method NewGELUActivation.forward(input: Tensor) -> Tensor - added:
method QuickGELUActivation.forward(input: Tensor) -> Tensor - added:
method ReLUSquaredActivation.forward(input) - added:
method SiLUActivation.forward(input: Tensor) -> Tensor - added:
method XIELUActivation.forward(input: Tensor) -> Tensor - added:
variable ACT2CLS - added:
variable gelu_fast - added:
variable gelu_new - added:
variable gelu_python - added:
variable linear_act - added:
variable mish - added:
variable quick_gelu
- removed:
function gelu(x) - removed:
function gelu_new(x) - added:
function approximate_gelu_wrap(x) - added:
function gelu_10(x) - added:
function glu(x, axis=…) - added:
function quick_gelu(x) - added:
variable gelu - added:
variable gelu_new
- added:
function amplitude_to_db(spectrogram: np.ndarray, reference: float=…, min_value: float=…, db_range: Optional[float]=…) -> np.ndarray - added:
function amplitude_to_db_batch(spectrogram: np.ndarray, reference: float=…, min_value: float=…, db_range: Optional[float]=…) -> np.ndarray - added:
function chroma_filter_bank(num_frequency_bins: int, num_chroma: int, sampling_rate: int, tuning: float=…, power: Optional[float]=…, weighting_parameters: Optional[tuple[float, float]]=…, start_at_c_chroma: bool=…) - added:
function hertz_to_mel(freq: Union[float, np.ndarray], mel_scale: str=…) -> Union[float, np.ndarray] - added:
function hertz_to_octave(freq: Union[float, np.ndarray], tuning: float=…, bins_per_octave: int=…) - added:
function is_valid_audio(audio) - added:
function is_valid_list_of_audio(audio) - added:
function load_audio(audio: Union[str, np.ndarray], sampling_rate=…, timeout=…) -> np.ndarray - added:
function load_audio_as(audio: str, return_format: str, timeout: Optional[int]=…, force_mono: bool=…, sampling_rate: Optional[int]=…) -> Union[str, dict[str, Any], io.BytesIO, None] - added:
function load_audio_librosa(audio: Union[str, np.ndarray], sampling_rate=…, timeout=…) -> np.ndarray - added:
function load_audio_torchcodec(audio: Union[str, np.ndarray], sampling_rate=…) -> np.ndarray - added:
function make_list_of_audio(audio: Union[list[AudioInput], AudioInput]) -> AudioInput - added:
function mel_filter_bank(num_frequency_bins: int, num_mel_filters: int, min_frequency: float, max_frequency: float, sampling_rate: int, norm: Optional[str]=…, mel_scale: str=…, triangularize_in_mel_space: bool=…) -> np.ndarray - added:
function mel_to_hertz(mels: Union[float, np.ndarray], mel_scale: str=…) -> Union[float, np.ndarray] - added:
function optimal_fft_length(window_length: int) -> int - added:
function power_to_db(spectrogram: np.ndarray, reference: float=…, min_value: float=…, db_range: Optional[float]=…) -> np.ndarray - added:
function power_to_db_batch(spectrogram: np.ndarray, reference: float=…, min_value: float=…, db_range: Optional[float]=…) -> np.ndarray - added:
function spectrogram(waveform: np.ndarray, window: np.ndarray, frame_length: int, hop_length: int, fft_length: Optional[int]=…, power: Optional[float]=…, center: bool=…, pad_mode: str=…, onesided: bool=…, dither: float=…, preemphasis: Optional[float]=…, mel_filters: Optional[np.ndarray]=…, mel_floor: float=…, log_mel: Optional[str]=…, reference: float=…, min_value: float=…, db_range: Optional[flo… - added:
function spectrogram_batch(waveform_list: list[np.ndarray], window: np.ndarray, frame_length: int, hop_length: int, fft_length: Optional[int]=…, power: Optional[float]=…, center: bool=…, pad_mode: str=…, onesided: bool=…, dither: float=…, preemphasis: Optional[float]=…, mel_filters: Optional[np.ndarray]=…, mel_floor: float=…, log_mel: Optional[str]=…, reference: float=…, min_value: float=…, db_ra… - added:
function window_function(window_length: int, name: str=…, periodic: bool=…, frame_length: Optional[int]=…, center: bool=…) -> np.ndarray - added:
variable AudioInput - added:
variable TORCHCODEC_VERSION
- added:
class Cache - added:
class CacheLayerMixin(ABC) - added:
class ChunkedSlidingLayer(StaticSlidingWindowLayer) - added:
class DynamicCache(Cache) - added:
class DynamicLayer(CacheLayerMixin) - added:
class DynamicSlidingWindowLayer(DynamicLayer) - added:
class EncoderDecoderCache(Cache) - added:
class HQQQuantizedCache(QuantizedCache) - added:
class HQQQuantizedLayer(QuantizedLayer) - added:
class HybridCache(StaticCache) - added:
class HybridChunkedCache(StaticCache) - added:
class OffloadedCache(DynamicCache) - added:
class OffloadedHybridCache(StaticCache) - added:
class OffloadedStaticCache(StaticCache) - added:
class QuantizedCache(Cache) - added:
class QuantizedLayer(DynamicLayer) - added:
class QuantoQuantizedCache(QuantizedCache) - added:
class QuantoQuantizedLayer(QuantizedLayer) - added:
class SinkCache(Cache) - added:
class SlidingWindowCache(StaticCache) - added:
class SlidingWindowLayer(StaticSlidingWindowLayer) - added:
class StaticCache(Cache) - added:
class StaticLayer(CacheLayerMixin) - added:
class StaticSlidingWindowLayer(StaticLayer) - added:
constructor Cache(layers: Optional[list[CacheLayerMixin]]=…, layer_class_to_replicate: Optional[type[CacheLayerMixin]]=…, offloading: bool=…, offload_only_non_sliding: bool=…) - added:
constructor CacheLayerMixin() - added:
constructor ChunkedSlidingLayer(max_cache_len: int, sliding_window: int) - added:
constructor DynamicCache(ddp_cache_data: Optional[Iterable[tuple[torch.Tensor, torch.Tensor]]]=…, config: Optional[PretrainedConfig]=…, offloading: bool=…, offload_only_non_sliding: bool=…) - added:
constructor DynamicSlidingWindowLayer(sliding_window: int) - added:
constructor EncoderDecoderCache(*caches) -> None - added:
constructor HQQQuantizedCache(config: PretrainedConfig, nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor HQQQuantizedLayer(nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor HybridCache(config: PretrainedConfig, max_cache_len: int, *args, **kwargs) - added:
constructor HybridChunkedCache(config: PretrainedConfig, max_cache_len: int, *args, **kwargs) - added:
constructor OffloadedCache() -> None - added:
constructor OffloadedHybridCache(config: PretrainedConfig, max_cache_len: int, *args, **kwargs) - added:
constructor OffloadedStaticCache(config: PretrainedConfig, max_cache_len: int, *args, **kwargs) - added:
constructor QuantizedCache(backend: str, config: PretrainedConfig, nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor QuantizedLayer(nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor QuantoQuantizedCache(config: PretrainedConfig, nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor QuantoQuantizedLayer(nbits: int=…, axis_key: int=…, axis_value: int=…, q_group_size: int=…, residual_length: int=…) - added:
constructor SinkCache(**kwargs) -> None - added:
constructor SlidingWindowCache(config: PretrainedConfig, max_cache_len: int, *args, **kwargs) - added:
constructor SlidingWindowLayer(max_cache_len: int, sliding_window: int) - added:
constructor StaticCache(config: PretrainedConfig, max_cache_len: int, offloading: bool=…, offload_only_non_sliding: bool=…, **kwargs) - added:
constructor StaticLayer(max_cache_len: int) - added:
constructor StaticSlidingWindowLayer(max_cache_len: int, sliding_window: int) - added:
field CacheLayerMixin.is_compileable - added:
field DynamicLayer.is_sliding - added:
field DynamicSlidingWindowLayer.is_sliding - added:
field StaticLayer.is_compileable - added:
field StaticLayer.is_sliding - added:
field StaticSlidingWindowLayer.is_sliding - added:
method Cache.batch_repeat_interleave(repeats: int) - added:
method Cache.batch_select_indices(indices: torch.Tensor) - added:
method Cache.crop(max_length: int) - added:
method Cache.early_initialization(batch_size: int, num_heads: int, head_dim: int, dtype: torch.dtype, device: torch.device) - added:
method Cache.get_mask_sizes(cache_position: torch.Tensor, layer_idx: int) -> tuple[int, int] - added:
method Cache.get_max_cache_shape(layer_idx: int=…) -> int - added:
method Cache.get_seq_length(layer_idx: int=…) -> int - added:
method Cache.offload(layer_idx: int, only_non_sliding: bool=…) - added:
method Cache.prefetch(layer_idx: int, only_non_sliding: bool=…) - added:
method Cache.reorder_cache(beam_idx: torch.LongTensor) - added:
method Cache.reset() - added:
method Cache.update(key_states: torch.Tensor, value_states: torch.Tensor, layer_idx: int, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method CacheLayerMixin.get_mask_sizes(cache_position: torch.Tensor) -> tuple[int, int] - added:
method CacheLayerMixin.get_max_cache_shape() -> int - added:
method CacheLayerMixin.get_seq_length() -> int - added:
method CacheLayerMixin.lazy_initialization(key_states: torch.Tensor) - added:
method CacheLayerMixin.offload() - added:
method CacheLayerMixin.prefetch() - added:
method CacheLayerMixin.reorder_cache(beam_idx: torch.LongTensor) -> None - added:
method CacheLayerMixin.reset() -> None - added:
method CacheLayerMixin.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method DynamicCache.to_legacy_cache() -> tuple[tuple[torch.Tensor, torch.Tensor]] - added:
method DynamicLayer.batch_repeat_interleave(repeats: int) -> None - added:
method DynamicLayer.batch_select_indices(indices: torch.Tensor) -> None - added:
method DynamicLayer.crop(max_length: int) -> None - added:
method DynamicLayer.get_mask_sizes(cache_position: torch.Tensor) -> tuple[int, int] - added:
method DynamicLayer.get_max_cache_shape() -> int - added:
method DynamicLayer.get_seq_length() -> int - added:
method DynamicLayer.lazy_initialization(key_states: torch.Tensor) - added:
method DynamicLayer.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method DynamicSlidingWindowLayer.crop(max_length: int) -> None - added:
method DynamicSlidingWindowLayer.get_mask_sizes(cache_position: torch.Tensor) -> tuple[int, int] - added:
method DynamicSlidingWindowLayer.get_max_cache_shape() -> int - added:
method DynamicSlidingWindowLayer.get_seq_length() -> int - added:
method DynamicSlidingWindowLayer.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method EncoderDecoderCache.batch_repeat_interleave(repeats: int) - added:
method EncoderDecoderCache.batch_select_indices(indices: torch.Tensor) - added:
method EncoderDecoderCache.batch_split(full_batch_size: int, split_size: int) -> list[EncoderDecoderCache] - added:
method EncoderDecoderCache.check_dynamic_cache(method: str) - added:
method EncoderDecoderCache.crop(maximum_length: int) - added:
method EncoderDecoderCache.get_mask_sizes(cache_position: torch.Tensor, layer_idx: int) -> tuple[int, int] - added:
method EncoderDecoderCache.get_max_cache_shape() -> int - added:
method EncoderDecoderCache.get_seq_length(layer_idx: int=…) -> int - added:
method EncoderDecoderCache.reorder_cache(beam_idx: torch.LongTensor) - added:
method EncoderDecoderCache.reset() - added:
method EncoderDecoderCache.to_legacy_cache() -> tuple[tuple[torch.Tensor]] - added:
method QuantizedLayer.get_seq_length() -> int - added:
method QuantizedLayer.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method StaticLayer.get_mask_sizes(cache_position: torch.Tensor) -> tuple[int, int] - added:
method StaticLayer.get_max_cache_shape() -> int - added:
method StaticLayer.get_seq_length() -> int - added:
method StaticLayer.lazy_initialization(key_states: torch.Tensor) - added:
method StaticLayer.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method StaticSlidingWindowLayer.get_mask_sizes(cache_position: torch.Tensor) -> tuple[int, int] - added:
method StaticSlidingWindowLayer.get_seq_length() -> int - added:
method StaticSlidingWindowLayer.update(key_states: torch.Tensor, value_states: torch.Tensor, cache_kwargs: Optional[dict[str, Any]]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
property Cache.is_compileable -> bool - added:
property Cache.is_initialized -> bool - added:
property Cache.is_sliding -> list[bool] - added:
property Cache.max_batch_size -> int - added:
property Cache.max_cache_len -> int - added:
property EncoderDecoderCache.is_compileable -> bool - added:
property EncoderDecoderCache.is_sliding - added:
static method DynamicCache.from_legacy_cache(past_key_values: tuple[tuple[torch.Tensor, torch.Tensor]]) -> DynamicCache - added:
static method EncoderDecoderCache.from_legacy_cache(past_key_values: Optional[Iterable[tuple[torch.FloatTensor, ...]]]) -> EncoderDecoderCache - added:
variable logger
- added:
class AddFastImageProcessorCommand(BaseTransformersCLICommand) - added:
constructor AddFastImageProcessorCommand(model_name: str, *args) - added:
function add_fast_image_processor(model_name: str) - added:
function add_fast_image_processor_file(fast_image_processing_module_file: str, fast_image_processor_name: str, content_base_file: str) - added:
function add_fast_image_processor_to_auto(image_processor_name: str, fast_image_processor_name: str) - added:
function add_fast_image_processor_to_doc(fast_image_processor_name: str, model_name: str) - added:
function add_fast_image_processor_to_model_init(fast_image_processing_module_file: str, fast_image_processor_name, model_name: str) - added:
function add_fast_image_processor_to_tests(fast_image_processor_name: str, model_name: str) - added:
function add_new_model_like_command_factory(args: Namespace) - added:
function get_fast_image_processing_content_header(content: str) -> str - added:
function write_default_fast_image_processor_file(fast_image_processing_module_file: str, fast_image_processor_name: str, content_base_file: str) - added:
method AddFastImageProcessorCommand.run() - added:
static method AddFastImageProcessorCommand.register_subcommand(parser: ArgumentParser) - added:
variable CURRENT_YEAR - added:
variable REPO_PATH - added:
variable TRANSFORMERS_PATH - added:
variable logger
- added:
class AddNewModelLikeCommand(BaseTransformersCLICommand) - added:
class ClassFinder(CSTVisitor) - added:
class ModelInfos - added:
constructor AddNewModelLikeCommand(path_to_repo=…, **kwargs) - added:
constructor ClassFinder() - added:
constructor ModelInfos(lowercase_name: str) - added:
function add_content_to_file(file_name: Union[str, os.PathLike], new_content: str, add_after: str) - added:
function add_model_to_auto_mappings(old_model_infos: ModelInfos, new_lowercase_name: str, new_model_paper_name: str, filenames_to_add: list[tuple[str, bool]]) - added:
function add_new_model_like_command_factory(args: Namespace) - added:
function convert_to_bool(x: str) -> bool - added:
function create_doc_file(new_paper_name: str, public_classes: list[str]) - added:
function create_init_file(old_lowercase_name: str, new_lowercase_name: str, filenames_to_add: list[tuple[str, bool]]) - added:
function create_modular_file(old_model_infos: ModelInfos, new_lowercase_name: str, filenames_to_add: list[tuple[str, bool]]) -> str - added:
function create_new_model_like(old_model_infos: ModelInfos, new_lowercase_name: str, new_model_paper_name: str, filenames_to_add: list[tuple[str, bool]], create_fast_image_processor: bool) - added:
function create_test_files(old_model_infos: ModelInfos, new_lowercase_name, filenames_to_add: list[tuple[str, bool]]) - added:
function find_all_classes_from_file(module_name: str) -> set - added:
function find_modular_structure(module_name: str, old_model_infos: ModelInfos, new_cased_name: str) -> tuple[str, str, list] - added:
function get_user_field(question: str, default_value: Optional[str]=…, convert_to: Optional[Callable]=…, fallback_message: Optional[str]=…) -> Any - added:
function get_user_input() - added:
function insert_model_in_doc_toc(old_lowercase_name: str, new_lowercase_name: str, new_model_paper_name: str) - added:
method AddNewModelLikeCommand.run() - added:
method ClassFinder.leave_ClassDef(node: cst.ClassDef) - added:
method ClassFinder.visit_ClassDef(node: cst.ClassDef) -> None - added:
method ClassFinder.visit_SimpleStatementLine(node: cst.SimpleStatementLine) - added:
static method AddNewModelLikeCommand.register_subcommand(parser: ArgumentParser) - added:
variable COPYRIGHT - added:
variable CURRENT_YEAR - added:
variable REPO_PATH - added:
variable TRANSFORMERS_PATH
- added:
async method RichInterface.stream_output(stream: AsyncIterator[ChatCompletionStreamOutput]) -> tuple[str, int] - added:
class ChatArguments - added:
class ChatCommand(BaseTransformersCLICommand) - added:
class RichInterface - added:
constructor ChatCommand(args) - added:
constructor RichInterface(model_name: Optional[str]=…, user_name: Optional[str]=…) - added:
field ChatArguments.attn_implementation: Optional[str] - added:
field ChatArguments.bnb_4bit_quant_type: str - added:
field ChatArguments.device: str - added:
field ChatArguments.dtype: Optional[str] - added:
field ChatArguments.examples_path: Optional[str] - added:
field ChatArguments.generation_config: Optional[str] - added:
field ChatArguments.host: str - added:
field ChatArguments.load_in_4bit: bool - added:
field ChatArguments.load_in_8bit: bool - added:
field ChatArguments.model_name_or_path: Optional[str] - added:
field ChatArguments.model_revision: str - added:
field ChatArguments.port: int - added:
field ChatArguments.save_folder: str - added:
field ChatArguments.system_prompt: Optional[str] - added:
field ChatArguments.torch_dtype: Optional[str] - added:
field ChatArguments.trust_remote_code: bool - added:
field ChatArguments.use_bnb_nested_quant: bool - added:
field ChatArguments.user: Optional[str] - added:
field ChatArguments.verbose: bool - added:
function chat_command_factory(args: Namespace) - added:
method ChatCommand.get_generation_parameterization(args: ChatArguments, model_generation_config: GenerationConfig) -> tuple[GenerationConfig, dict] - added:
method ChatCommand.handle_non_exit_user_commands(user_input: str, args: ChatArguments, interface: RichInterface, examples: dict[str, dict[str, str]], generation_config: GenerationConfig, model_kwargs: dict, chat: list[dict]) -> tuple[list[dict], GenerationConfig, dict] - added:
method ChatCommand.load_model_and_tokenizer(args: ChatArguments) -> tuple[AutoModelForCausalLM, AutoTokenizer] - added:
method ChatCommand.parse_generate_flags(generate_flags: list[str]) -> dict - added:
method ChatCommand.run() - added:
method RichInterface.clear() - added:
method RichInterface.input() -> str - added:
method RichInterface.print_color(text: str, color: str) - added:
method RichInterface.print_help(minimal: bool=…) - added:
method RichInterface.print_status(model_name: str, generation_config: GenerationConfig, model_kwargs: dict) - added:
method RichInterface.print_user_message(text: str) - added:
static method ChatCommand.clear_chat_history(system_prompt: Optional[str]=…) -> list[dict] - added:
static method ChatCommand.get_quantization_config(model_args: ChatArguments) -> Optional[BitsAndBytesConfig] - added:
static method ChatCommand.get_username() -> str - added:
static method ChatCommand.parse_eos_tokens(tokenizer: PreTrainedTokenizer, generation_config: GenerationConfig, eos_tokens: Optional[str], eos_token_ids: Optional[str]) -> tuple[int, list[int]] - added:
static method ChatCommand.register_subcommand(parser: ArgumentParser) - added:
static method ChatCommand.save_chat(chat, args: ChatArguments, filename: Optional[str]=…) -> str - added:
variable ALLOWED_KEY_CHARS - added:
variable ALLOWED_VALUE_CHARS - added:
variable DEFAULT_EXAMPLES - added:
variable HELP_STRING - added:
variable HELP_STRING_MINIMAL - added:
variable args - added:
variable chat
- changed:
constructor DownloadCommand(model: str, cache: str, force: bool) → constructor DownloadCommand(model: str, cache: str, force: bool, trust_remote_code: bool)
- added:
constructor EnvironmentCommand(accelerate_config_file, *args) -> None - added:
function download_command_factory(args)
- removed:
async method ServeCommand.forward(inputs=…) - removed:
class ServeDeTokenizeResult(BaseModel) - removed:
class ServeForwardResult(BaseModel) - removed:
class ServeModelInfoResult(BaseModel) - removed:
class ServeTokenizeResult(BaseModel) - removed:
field ServeDeTokenizeResult.text: str - removed:
field ServeForwardResult.output: Any - removed:
field ServeModelInfoResult.infos: dict - removed:
field ServeTokenizeResult.tokens: List[str] - removed:
field ServeTokenizeResult.tokens_ids: Optional[List[int]] - removed:
function Body(*x, **y) - removed:
method ServeCommand.detokenize(tokens_ids: List[int]=…, skip_special_tokens: bool=…, cleanup_tokenization_spaces: bool=…) - removed:
method ServeCommand.model_info() - removed:
method ServeCommand.tokenize(text_input: str=…, return_ids: bool=…) - removed:
variable BaseModel - changed:
constructor ServeCommand(pipeline: Pipeline, host: str, port: int, workers: int) → constructor ServeCommand(args: ServeArguments) - added:
class Modality(enum.Enum) - added:
class ServeArguments - added:
class TimedModel - added:
class ToolState - added:
class TransformersCompletionCreateParamsStreaming(CompletionCreateParamsStreaming) - added:
class TransformersResponseCreateParamsStreaming(ResponseCreateParamsStreaming) - added:
class TransformersTranscriptionCreateParams(TranscriptionCreateParamsBase) - added:
constructor TimedModel(model: PreTrainedModel, timeout_seconds: int, processor: Optional[Union[ProcessorMixin, PreTrainedTokenizerFast]]=…) - added:
constructor ToolState() - added:
field Modality.LLM - added:
field Modality.STT - added:
field Modality.TTS - added:
field Modality.VLM - added:
field ServeArguments.attn_implementation: Optional[str] - added:
field ServeArguments.bnb_4bit_quant_type: str - added:
field ServeArguments.continuous_batching: bool - added:
field ServeArguments.default_seed: Optional[int] - added:
field ServeArguments.device: str - added:
field ServeArguments.dtype: Optional[str] - added:
field ServeArguments.enable_cors: bool - added:
field ServeArguments.force_model: Optional[str] - added:
field ServeArguments.host: str - added:
field ServeArguments.input_validation: bool - added:
field ServeArguments.load_in_4bit: bool - added:
field ServeArguments.load_in_8bit: bool - added:
field ServeArguments.log_level: str - added:
field ServeArguments.model_timeout: int - added:
field ServeArguments.port: int - added:
field ServeArguments.torch_dtype: Optional[str] - added:
field ServeArguments.trust_remote_code: bool - added:
field ServeArguments.use_bnb_nested_quant: bool - added:
field TransformersCompletionCreateParamsStreaming.generation_config: str - added:
field TransformersResponseCreateParamsStreaming.generation_config: str - added:
field TransformersTranscriptionCreateParams.file: bytes - added:
field TransformersTranscriptionCreateParams.generation_config: str - added:
field TransformersTranscriptionCreateParams.stream: bool - added:
function create_generation_config_from_req(req: dict, model_generation_config: GenerationConfig, **kwargs) -> GenerationConfig - added:
method ServeCommand.build_chat_completion_chunk(request_id: str=…, content: Optional[int]=…, model: Optional[str]=…, role: Optional[str]=…, finish_reason: Optional[str]=…, tool_calls: Optional[list[ChoiceDeltaToolCall]]=…, decode_stream: Optional[DecodeStream]=…, tokenizer: Optional[PreTrainedTokenizerFast]=…) -> str - added:
method ServeCommand.build_response_event(response: BaseModel) -> str - added:
method ServeCommand.continuous_batching_chat_completion(req: dict, request_id: str) -> AsyncGenerator[str, None] - added:
method ServeCommand.generate_chat_completion(req: dict) -> Generator[str, None, None] - added:
method ServeCommand.generate_response(req: dict) -> Generator[str, None, None] - added:
method ServeCommand.generate_transcription(req: dict) -> Generator[str, None, None] - added:
method ServeCommand.get_gen_models() -> list[dict[str, any]] - added:
method ServeCommand.is_continuation(req: dict) -> bool - added:
method ServeCommand.load_audio_model_and_processor(model_id_and_revision: str) -> tuple[PreTrainedModel, ProcessorMixin] - added:
method ServeCommand.load_model_and_processor(model_id_and_revision: str) -> tuple[PreTrainedModel, PreTrainedTokenizerFast] - added:
method ServeCommand.process_model_name(model_id: str) -> str - added:
method ServeCommand.validate_chat_completion_request(request: dict) - added:
method ServeCommand.validate_response_request(request: dict) - added:
method ServeCommand.validate_transcription_request(request: dict) - added:
method TimedModel.delete_model() - added:
method TimedModel.is_deleted() - added:
method TimedModel.reset_timer() - added:
method TimedModel.timeout_reached() - added:
method ToolState.reset() - added:
static method ServeCommand.get_model_modality(model: PreTrainedModel) -> Modality - added:
static method ServeCommand.get_processor_inputs_from_inbound_messages(messages, modality: Modality) - added:
static method ServeCommand.get_quantization_config(args: ServeArguments) -> Optional[BitsAndBytesConfig] - added:
variable UNUSED_CHAT_COMPLETION_FIELDS - added:
variable UNUSED_RESPONSE_FIELDS - added:
variable UNUSED_TRANSCRIPTION_FIELDS - added:
variable X_REQUEST_ID - added:
variable completion_validator - added:
variable response_validator - added:
variable serve - added:
variable serve_dependencies_available - added:
variable transcription_validator
- added:
function main_cli()
- removed:
class ANSI - removed:
class BaseUserCommand - removed:
class DeleteObjCommand(BaseUserCommand) - removed:
class DeprecatedUploadCommand(BaseUserCommand) - removed:
class ListObjsCommand(BaseUserCommand) - removed:
class ListReposObjsCommand(BaseUserCommand) - removed:
class LoginCommand(BaseUserCommand) - removed:
class LogoutCommand(BaseUserCommand) - removed:
class RepoCreateCommand(BaseUserCommand) - removed:
class UploadCommand(BaseUserCommand) - removed:
class UserCommands(BaseTransformersCLICommand) - removed:
class WhoamiCommand(BaseUserCommand) - removed:
constructor BaseUserCommand(args) - removed:
function tabulate(rows: List[List[Union[str, int]]], headers: List[str]) -> str - removed:
method DeleteObjCommand.run() - removed:
method DeprecatedUploadCommand.run() - removed:
method ListObjsCommand.run() - removed:
method ListReposObjsCommand.run() - removed:
method LoginCommand.run() - removed:
method LogoutCommand.run() - removed:
method RepoCreateCommand.run() - removed:
method UploadCommand.run() - removed:
method UploadCommand.walk_dir(rel_path) - removed:
method WhoamiCommand.run() - removed:
static method ANSI.bold(s) - removed:
static method ANSI.gray(s) - removed:
static method ANSI.red(s) - removed:
static method UserCommands.register_subcommand(parser: ArgumentParser) - removed:
variable UPLOAD_MAX_FILES
- removed:
class AlbertConfig(PretrainedConfig) - removed:
constructor AlbertConfig(vocab_size=…, embedding_size=…, hidden_size=…, num_hidden_layers=…, num_hidden_groups=…, num_attention_heads=…, intermediate_size=…, inner_group_num=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, classifier_dropout_prob=…, pad_token_id=…, bos_token_id=…, eos_token… - removed:
field AlbertConfig.model_type - removed:
variable ALBERT_PRETRAINED_CONFIG_ARCHIVE_MAP
- removed:
class AutoConfig - removed:
constructor AutoConfig() - removed:
function replace_list_option_in_docstrings(config_to_class=…, use_model_types=…) - removed:
static method AutoConfig.for_model(model_type: str, *args, **kwargs) - removed:
static method AutoConfig.from_pretrained(pretrained_model_name_or_path, **kwargs) - removed:
variable ALL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable CONFIG_MAPPING - removed:
variable MODEL_NAMES_MAPPING
- removed:
class BartConfig(PretrainedConfig) - removed:
constructor BartConfig(activation_dropout=…, extra_pos_embeddings=…, activation_function=…, vocab_size=…, d_model=…, encoder_ffn_dim=…, encoder_layers=…, encoder_attention_heads=…, decoder_ffn_dim=…, decoder_layers=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, attention_dropout=…, dropout=…, max_position_embeddings=…, init_std=…, classifier_dropout=…, num_labels=…, is_en… - removed:
field BartConfig.model_type - removed:
method BartConfig.is_valid_mbart() -> bool - removed:
property BartConfig.hidden_size -> int - removed:
property BartConfig.num_attention_heads -> int - removed:
variable BART_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class BertConfig(PretrainedConfig) - removed:
constructor BertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, gradient_checkpointing=…, **kwargs) - removed:
field BertConfig.model_type - removed:
variable BERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class BertGenerationConfig(PretrainedConfig) - removed:
constructor BertGenerationConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, gradient_checkpointing=…, **kwargs) - removed:
field BertGenerationConfig.model_type
- removed:
class BlenderbotConfig(BartConfig) - removed:
constructor BlenderbotConfig(activation_dropout=…, extra_pos_embeddings=…, activation_function=…, vocab_size=…, d_model=…, encoder_ffn_dim=…, encoder_layers=…, encoder_attention_heads=…, decoder_ffn_dim=…, decoder_layers=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, attention_dropout=…, dropout=…, max_position_embeddings=…, classifier_dropout=…, is_encoder_decoder=…, pad… - removed:
field BlenderbotConfig.model_type - removed:
variable BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP
- removed:
class CamembertConfig(RobertaConfig) - removed:
field CamembertConfig.model_type - removed:
variable CAMEMBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class CTRLConfig(PretrainedConfig) - removed:
constructor CTRLConfig(vocab_size=…, n_positions=…, n_ctx=…, n_embd=…, dff=…, n_layer=…, n_head=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dropout=…, **kwargs) - removed:
field CTRLConfig.model_type - removed:
property CTRLConfig.hidden_size - removed:
property CTRLConfig.max_position_embeddings - removed:
property CTRLConfig.num_attention_heads - removed:
property CTRLConfig.num_hidden_layers - removed:
variable CTRL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class DebertaConfig(PretrainedConfig) - removed:
constructor DebertaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, relative_attention=…, max_relative_positions=…, pad_token_id=…, position_biased_input=…, pos_att_type=…, pooler_dropout=…, po… - removed:
field DebertaConfig.model_type - removed:
variable DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class DistilBertConfig(PretrainedConfig) - removed:
constructor DistilBertConfig(vocab_size=…, max_position_embeddings=…, sinusoidal_pos_embds=…, n_layers=…, n_heads=…, dim=…, hidden_dim=…, dropout=…, attention_dropout=…, activation=…, initializer_range=…, qa_dropout=…, seq_classif_dropout=…, pad_token_id=…, **kwargs) - removed:
field DistilBertConfig.model_type - removed:
property DistilBertConfig.hidden_size - removed:
property DistilBertConfig.num_attention_heads - removed:
property DistilBertConfig.num_hidden_layers - removed:
variable DISTILBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class DPRConfig(PretrainedConfig) - removed:
constructor DPRConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, gradient_checkpointing=…, projection_dim: int=…, **kwargs) - removed:
field DPRConfig.model_type - removed:
variable DPR_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class ElectraConfig(PretrainedConfig) - removed:
constructor ElectraConfig(vocab_size=…, embedding_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_last_dropout=…, pad_token_id=…, **kwargs) - removed:
field ElectraConfig.model_type - removed:
variable ELECTRA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class EncoderDecoderConfig(PretrainedConfig) - removed:
constructor EncoderDecoderConfig(**kwargs) - removed:
field EncoderDecoderConfig.is_composition - removed:
field EncoderDecoderConfig.model_type - removed:
method EncoderDecoderConfig.to_dict() - removed:
static method EncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs) -> PretrainedConfig - removed:
variable logger
- removed:
class FlaubertConfig(XLMConfig) - removed:
constructor FlaubertConfig(layerdrop=…, pre_norm=…, pad_token_id=…, bos_token_id=…, **kwargs) - removed:
field FlaubertConfig.model_type - removed:
variable FLAUBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class DecoderConfig(PretrainedConfig) - removed:
class FSMTConfig(PretrainedConfig) - removed:
constructor DecoderConfig(vocab_size=…, bos_token_id=…) - removed:
constructor FSMTConfig(langs=…, src_vocab_size=…, tgt_vocab_size=…, activation_function=…, d_model=…, max_length=…, max_position_embeddings=…, encoder_ffn_dim=…, encoder_layers=…, encoder_attention_heads=…, encoder_layerdrop=…, decoder_ffn_dim=…, decoder_layers=…, decoder_attention_heads=…, decoder_layerdrop=…, attention_dropout=…, dropout=…, activation_dropout=…, init_std=…, pad_token_id=…, bos_… - removed:
field DecoderConfig.model_type - removed:
field FSMTConfig.model_type - removed:
method FSMTConfig.to_dict() - removed:
property FSMTConfig.hidden_size -> int - removed:
property FSMTConfig.num_attention_heads -> int - removed:
variable FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class FunnelConfig(PretrainedConfig) - removed:
constructor FunnelConfig(vocab_size=…, block_sizes=…, block_repeats=…, num_decoder_layers=…, d_model=…, n_head=…, d_head=…, d_inner=…, hidden_act=…, hidden_dropout=…, attention_dropout=…, activation_dropout=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, initializer_std=…, layer_norm_eps=…, pooling_type=…, attention_type=…, separate_cls=…, truncate_seq=…, pool_q_only=…, **kw… - removed:
field FunnelConfig.model_type - removed:
property FunnelConfig.hidden_size - removed:
property FunnelConfig.num_attention_heads - removed:
property FunnelConfig.num_blocks - removed:
property FunnelConfig.num_hidden_layers - removed:
variable FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class GPT2Config(PretrainedConfig) - removed:
constructor GPT2Config(vocab_size=…, n_positions=…, n_ctx=…, n_embd=…, n_layer=…, n_head=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dropout=…, bos_token_id=…, eos_token_id=…, gradient_checkpointing=…, **kwargs) - removed:
field GPT2Config.model_type - removed:
property GPT2Config.hidden_size - removed:
property GPT2Config.max_position_embeddings - removed:
property GPT2Config.num_attention_heads - removed:
property GPT2Config.num_hidden_layers - removed:
variable GPT2_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class LayoutLMConfig(BertConfig) - removed:
constructor LayoutLMConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, gradient_checkpointing=…, max_2d_position_embeddings=…, **kwargs) - removed:
field LayoutLMConfig.model_type - removed:
variable LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class LongformerConfig(RobertaConfig) - removed:
constructor LongformerConfig(attention_window: Union[List[int], int]=…, sep_token_id: int=…, **kwargs) - removed:
field LongformerConfig.model_type - removed:
variable LONGFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class LxmertConfig(PretrainedConfig) - removed:
constructor LxmertConfig(vocab_size=…, hidden_size=…, num_attention_heads=…, num_labels=…, num_qa_labels=…, num_object_labels=…, num_attr_labels=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, l_layers=…, x_layers=…, r_layers=…, visual_feat_dim=…, visua… - removed:
field LxmertConfig.model_type - removed:
variable LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class MarianConfig(BartConfig) - removed:
field MarianConfig.model_type - removed:
variable PRETRAINED_CONFIG_ARCHIVE_MAP
- removed:
class MBartConfig(BartConfig) - removed:
field MBartConfig.model_type - removed:
variable MBART_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class MMBTConfig(object) - removed:
constructor MMBTConfig(config, num_labels=…, modal_hidden_size=…) - removed:
variable logger
- removed:
class MobileBertConfig(PretrainedConfig) - removed:
constructor MobileBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, embedding_size=…, trigram_input=…, use_bottleneck=…, intra_bottleneck_size=…, use_bottleneck_attention=…,… - removed:
field MobileBertConfig.model_type - removed:
field MobileBertConfig.pretrained_config_archive_map - removed:
variable MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class OpenAIGPTConfig(PretrainedConfig) - removed:
constructor OpenAIGPTConfig(vocab_size=…, n_positions=…, n_ctx=…, n_embd=…, n_layer=…, n_head=…, afn=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, predict_special_tokens=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dropout=…, **kwargs) - removed:
field OpenAIGPTConfig.model_type - removed:
property OpenAIGPTConfig.hidden_size - removed:
property OpenAIGPTConfig.max_position_embeddings - removed:
property OpenAIGPTConfig.num_attention_heads - removed:
property OpenAIGPTConfig.num_hidden_layers - removed:
variable OPENAI_GPT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class PegasusConfig(BartConfig) - removed:
field PegasusConfig.model_type - removed:
variable DEFAULTS - removed:
variable logger - removed:
variable task_specific_params
- removed:
class ProphetNetConfig(PretrainedConfig) - removed:
constructor ProphetNetConfig(activation_dropout=…, activation_function=…, vocab_size=…, hidden_size=…, encoder_ffn_dim=…, num_encoder_layers=…, num_encoder_attention_heads=…, decoder_ffn_dim=…, num_decoder_layers=…, num_decoder_attention_heads=…, attention_dropout=…, dropout=…, max_position_embeddings=…, init_std=…, is_encoder_decoder=…, add_cross_attention=…, pad_token_id=…, bos_token_id=…, eos_… - removed:
field ProphetNetConfig.model_type - removed:
property ProphetNetConfig.num_attention_heads -> int - removed:
property ProphetNetConfig.num_hidden_layers -> int - removed:
variable PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class RagConfig(PretrainedConfig) - removed:
constructor RagConfig(vocab_size=…, is_encoder_decoder=…, prefix=…, bos_token_id=…, pad_token_id=…, eos_token_id=…, decoder_start_token_id=…, title_sep=…, doc_sep=…, n_docs=…, max_combined_length=…, retrieval_vector_size=…, retrieval_batch_size=…, dataset=…, dataset_split=…, index_name=…, index_path=…, passages_path=…, use_dummy_dataset=…, reduce_loss=…, label_smoothing=…, do_deduplication=…, exc… - removed:
field RagConfig.is_composition - removed:
field RagConfig.model_type - removed:
method RagConfig.to_dict() - removed:
static method RagConfig.from_question_encoder_generator_configs(question_encoder_config: PretrainedConfig, generator_config: PretrainedConfig, **kwargs) -> PretrainedConfig - removed:
variable RAG_CONFIG_DOC
- removed:
class ReformerConfig(PretrainedConfig) - removed:
constructor ReformerConfig(attention_head_size=…, attn_layers=…, axial_norm_std=…, axial_pos_embds=…, axial_pos_shape=…, axial_pos_embds_dim=…, chunk_size_lm_head=…, eos_token_id=…, feed_forward_size=…, hash_seed=…, hidden_act=…, hidden_dropout_prob=…, hidden_size=…, initializer_range=…, is_decoder=…, layer_norm_eps=…, local_num_chunks_before=…, local_num_chunks_after=…, local_attention_probs_dro… - removed:
field ReformerConfig.model_type - removed:
variable REFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class RetriBertConfig(PretrainedConfig) - removed:
constructor RetriBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, share_encoders=…, projection_dim=…, pad_token_id=…, **kwargs) - removed:
field RetriBertConfig.model_type - removed:
variable RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class RobertaConfig(BertConfig) - removed:
constructor RobertaConfig(pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - removed:
field RobertaConfig.model_type - removed:
variable ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class SqueezeBertConfig(PretrainedConfig) - removed:
constructor SqueezeBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, embedding_size=…, q_groups=…, k_groups=…, v_groups=…, post_attention_groups=…, intermediate_groups=…, out… - removed:
field SqueezeBertConfig.model_type - removed:
field SqueezeBertConfig.pretrained_config_archive_map - removed:
variable SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class T5Config(PretrainedConfig) - removed:
constructor T5Config(vocab_size=…, n_positions=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, is_encoder_decoder=…, pad_token_id=…, eos_token_id=…, **kwargs) - removed:
field T5Config.model_type - removed:
property T5Config.hidden_size - removed:
property T5Config.max_position_embeddings - removed:
property T5Config.num_attention_heads - removed:
property T5Config.num_hidden_layers - removed:
variable T5_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class TransfoXLConfig(PretrainedConfig) - removed:
constructor TransfoXLConfig(vocab_size=…, cutoffs=…, d_model=…, d_embed=…, n_head=…, d_head=…, d_inner=…, div_val=…, pre_lnorm=…, n_layer=…, mem_len=…, clamp_len=…, same_length=…, proj_share_all_but_first=…, attn_type=…, sample_softmax=…, adaptive=…, dropout=…, dropatt=…, untie_r=…, init=…, init_range=…, proj_init_std=…, init_std=…, layer_norm_epsilon=…, eos_token_id=…, **kwargs) - removed:
field TransfoXLConfig.model_type - removed:
property TransfoXLConfig.hidden_size - removed:
property TransfoXLConfig.max_position_embeddings - removed:
property TransfoXLConfig.n_token - removed:
property TransfoXLConfig.num_attention_heads - removed:
property TransfoXLConfig.num_hidden_layers - removed:
variable TRANSFO_XL_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
field PretrainedConfig.is_composition: bool - removed:
property PretrainedConfig.name_or_path -> str - changed:
class PretrainedConfig(object) → class PretrainedConfig(PushToHubMixin) - changed:
constructor PretrainedConfig(**kwargs) → constructor PretrainedConfig(*, output_hidden_states: bool=…, output_attentions: bool=…, return_dict: bool=…, torchscript: bool=…, dtype: Optional[Union[str, torch.dtype]]=…, pruned_heads: Optional[dict[int, list[int]]]=…, tie_word_embeddings: bool=…, chunk_size_feed_forward: int=…, is_encoder_decoder: bool=…, is_decoder: bool=…, cross_attention_hidden_size: Optional[int]=…, add_cross_attention:… - changed:
method PretrainedConfig.save_pretrained(save_directory: str) → method PretrainedConfig.save_pretrained(save_directory: Union[str, os.PathLike], push_to_hub: bool=…, **kwargs) - changed:
method PretrainedConfig.to_dict() -> Dict[str, Any] → method PretrainedConfig.to_dict() -> dict[str, Any] - changed:
method PretrainedConfig.to_diff_dict() -> Dict[str, Any] → method PretrainedConfig.to_diff_dict() -> dict[str, Any] - changed:
method PretrainedConfig.to_json_file(json_file_path: str, use_diff: bool=…) → method PretrainedConfig.to_json_file(json_file_path: Union[str, os.PathLike], use_diff: bool=…) - changed:
method PretrainedConfig.update(config_dict: Dict[str, Any]) → method PretrainedConfig.update(config_dict: dict[str, Any]) - changed:
static method PretrainedConfig.from_dict(config_dict: Dict[str, Any], **kwargs) -> PretrainedConfig → static method PretrainedConfig.from_dict(config_dict: dict[str, Any], **kwargs) -> SpecificPretrainedConfigType - changed:
static method PretrainedConfig.from_json_file(json_file: str) -> PretrainedConfig → static method PretrainedConfig.from_json_file(json_file: Union[str, os.PathLike]) -> SpecificPretrainedConfigType - changed:
static method PretrainedConfig.from_pretrained(pretrained_model_name_or_path: str, **kwargs) -> PretrainedConfig → static method PretrainedConfig.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) -> SpecificPretrainedConfigType - changed:
static method PretrainedConfig.get_config_dict(pretrained_model_name_or_path: str, **kwargs) -> Tuple[Dict[str, Any], Dict[str, Any]] → static method PretrainedConfig.get_config_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]] - added:
field PretrainedConfig.attribute_map: dict[str, str] - added:
field PretrainedConfig.base_config_key: str - added:
field PretrainedConfig.base_model_ep_plan: Optional[dict[str, tuple[list[str]]]] - added:
field PretrainedConfig.base_model_pp_plan: Optional[dict[str, tuple[list[str]]]] - added:
field PretrainedConfig.base_model_tp_plan: Optional[dict[str, Any]] - added:
field PretrainedConfig.has_no_defaults_at_init: bool - added:
field PretrainedConfig.sub_configs: dict[str, type[PretrainedConfig]] - added:
function get_configuration_file(configuration_files: list[str]) -> str - added:
function layer_type_validation(layer_types: list[str], num_hidden_layers: Optional[int]=…) - added:
function recursive_diff_dict(dict_a, dict_b, config_obj=…) - added:
method PretrainedConfig.dict_dtype_to_str(d: dict[str, Any]) -> None - added:
method PretrainedConfig.get_text_config(decoder=…, encoder=…) -> PretrainedConfig - added:
method PretrainedConfig.update_from_string(update_str: str) - added:
property PretrainedConfig.name_or_path -> Optional[str] - added:
property PretrainedConfig.output_attentions - added:
property PretrainedConfig.torch_dtype - added:
static method PretrainedConfig.from_text_audio_configs(text_config, audio_config, **kwargs) - added:
static method PretrainedConfig.from_text_vision_configs(text_config, vision_config, **kwargs) - added:
static method PretrainedConfig.register_for_auto_class(auto_class=…) - added:
variable ALLOWED_LAYER_TYPES - added:
variable SpecificPretrainedConfigType
- removed:
class XLMConfig(PretrainedConfig) - removed:
constructor XLMConfig(vocab_size=…, emb_dim=…, n_layers=…, n_heads=…, dropout=…, attention_dropout=…, gelu_activation=…, sinusoidal_embeddings=…, causal=…, asm=…, n_langs=…, use_lang_emb=…, max_position_embeddings=…, embed_init_std=…, layer_norm_eps=…, init_std=…, bos_index=…, eos_index=…, pad_index=…, unk_index=…, mask_index=…, is_encoder=…, summary_type=…, summary_use_proj=…, summary_activation… - removed:
field XLMConfig.model_type - removed:
property XLMConfig.hidden_size - removed:
property XLMConfig.n_words - removed:
property XLMConfig.num_attention_heads - removed:
property XLMConfig.num_hidden_layers - removed:
variable XLM_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class XLMProphetNetConfig(ProphetNetConfig) - removed:
field XLMProphetNetConfig.model_type - removed:
variable XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class XLMRobertaConfig(RobertaConfig) - removed:
field XLMRobertaConfig.model_type - removed:
variable XLM_ROBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
class XLNetConfig(PretrainedConfig) - removed:
constructor XLNetConfig(vocab_size=…, d_model=…, n_layer=…, n_head=…, d_inner=…, ff_activation=…, untie_r=…, attn_type=…, initializer_range=…, layer_norm_eps=…, dropout=…, mem_len=…, reuse_len=…, bi_data=…, clamp_len=…, same_length=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_last_dropout=…, start_n_top=…, end_n_top=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwar… - removed:
field XLNetConfig.model_type - removed:
property XLNetConfig.hidden_size - removed:
property XLNetConfig.max_position_embeddings - removed:
property XLNetConfig.n_token - removed:
property XLNetConfig.num_attention_heads - removed:
property XLNetConfig.num_hidden_layers - removed:
variable XLNET_PRETRAINED_CONFIG_ARCHIVE_MAP - removed:
variable logger
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, albert_config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_bart_checkpoint(checkpoint_path, pytorch_dump_folder_path, hf_checkpoint_name=…) - removed:
function load_xsum_checkpoint(checkpoint_path) - removed:
function remove_ignore_keys_(state_dict) - removed:
function rename_key(dct, old, new) - removed:
variable FAIRSEQ_MODELS - removed:
variable SAMPLE_TEXT - removed:
variable args - removed:
variable extra_arch - removed:
variable logger - removed:
variable mnli_rename_keys - removed:
variable parser
- removed:
function convert_tf2_checkpoint_to_pytorch(tf_checkpoint_path, config_path, pytorch_dump_path) - removed:
function load_tf2_weights_in_bert(model, tf_checkpoint_path, config) - removed:
variable args - removed:
variable logger - removed:
variable parser
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_pytorch_checkpoint_to_tf(model: BertModel, ckpt_dir: str, model_name: str) - removed:
function main(raw_args=…)
- removed:
function convert_parlai_checkpoint(checkpoint_path, pytorch_dump_folder_path, config_json_path) - removed:
function rename_layernorm_keys(sd) - removed:
function rename_state_dict_key(k) - removed:
variable IGNORE_KEYS - removed:
variable PATTERNS - removed:
variable args - removed:
variable logger - removed:
variable parser
- removed:
function convert_dialogpt_checkpoint(checkpoint_path: str, pytorch_dump_folder_path: str) - removed:
variable DIALOGPT_MODELS - removed:
variable NEW_KEY - removed:
variable OLD_KEY - removed:
variable args - removed:
variable parser
- removed:
class DPRContextEncoderState(DPRState) - removed:
class DPRQuestionEncoderState(DPRState) - removed:
class DPRReaderState(DPRState) - removed:
class DPRState - removed:
constructor DPRState(src_file: Path) - removed:
function convert(comp_type: str, src_file: Path, dest_dir: Path) - removed:
function load_states_from_checkpoint(model_file: str) -> CheckpointState - removed:
method DPRContextEncoderState.load_dpr_model() - removed:
method DPRQuestionEncoderState.load_dpr_model() - removed:
method DPRReaderState.load_dpr_model() - removed:
method DPRState.load_dpr_model() - removed:
static method DPRState.from_type(comp_type: str, *args, **kwargs) -> DPRState - removed:
variable CheckpointState - removed:
variable args - removed:
variable dest_dir - removed:
variable parser - removed:
variable src_file
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path, discriminator_or_generator) - removed:
variable args - removed:
variable parser
- removed:
function convert_fsmt_checkpoint_to_pytorch(fsmt_checkpoint_path, pytorch_dump_folder_path) - removed:
function rewrite_dict_keys(d) - removed:
variable args - removed:
variable best_score_hparams - removed:
variable json_indent - removed:
variable org_names - removed:
variable parser
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_gpt2_checkpoint_to_pytorch(gpt2_checkpoint_path, gpt2_config_file, pytorch_dump_folder_path) - removed:
variable args - removed:
variable parser
- changed:
function convert(framework: str, model: str, output: Path, opset: int, tokenizer: Optional[str]=…, use_external_format: bool=…, pipeline_name: str=…) → function convert(framework: str, model: str, output: Path, opset: int, tokenizer: Optional[str]=…, use_external_format: bool=…, pipeline_name: str=…, **model_kwargs) - changed:
function infer_shapes(nlp: Pipeline, framework: str) -> Tuple[List[str], List[str], Dict, BatchEncoding] → function infer_shapes(nlp: Pipeline, framework: str) -> tuple[list[str], list[str], dict, BatchEncoding] - changed:
function load_graph_from_args(pipeline_name: str, framework: str, model: str, tokenizer: Optional[str]=…) -> Pipeline → function load_graph_from_args(pipeline_name: str, framework: str, model: str, tokenizer: Optional[str]=…, **models_kwargs) -> Pipeline
- removed:
class LightningModel(pl.LightningModule) - removed:
constructor LightningModel(model) - removed:
function convert_longformer_qa_checkpoint_to_pytorch(longformer_model: str, longformer_question_answering_ckpt_path: str, pytorch_dump_folder_path: str) - removed:
method LightningModel.forward() - removed:
variable args - removed:
variable parser
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
class TatoebaConverter - removed:
constructor TatoebaConverter(save_dir=…) - removed:
function dedup(lst) - removed:
function l2front_matter(langs) - removed:
method TatoebaConverter.convert_models(tatoeba_ids, dry_run=…) - removed:
method TatoebaConverter.download_metadata() - removed:
method TatoebaConverter.expand_group_to_two_letter_codes(grp_name) - removed:
method TatoebaConverter.get_tags(code, ref_name) - removed:
method TatoebaConverter.get_two_letter_code(three_letter_code) - removed:
method TatoebaConverter.resolve_lang_code(r) -> Tuple[List[str], str, str] - removed:
method TatoebaConverter.write_model_card(hf_model_id: str, repo_root=…, dry_run=…) -> str - removed:
static method TatoebaConverter.make_tatoeba_registry(repo_path=…) - removed:
variable DEFAULT_MODEL_DIR - removed:
variable DEFAULT_REPO - removed:
variable GROUP_MEMBERS - removed:
variable ISO_PATH - removed:
variable ISO_URL - removed:
variable LANG_CODE_PATH - removed:
variable LANG_CODE_URL - removed:
variable args - removed:
variable parser - removed:
variable resolver
- removed:
class OpusState - removed:
constructor OpusState(source_dir) - removed:
function add_emb_entries(wemb, final_bias, n_special_tokens=…) - removed:
function add_special_tokens_to_vocab(model_dir: Path) -> None - removed:
function add_to_vocab_(vocab: Dict[str, int], special_tokens: List[str]) - removed:
function cast_marian_config(raw_cfg: Dict[str, str]) -> Dict - removed:
function check_equal(marian_cfg, k1, k2) - removed:
function check_marian_cfg_assumptions(marian_cfg) - removed:
function convert(source_dir: Path, dest_dir) - removed:
function convert_all_sentencepiece_models(model_list=…, repo_path=…, dest_dir=…) - removed:
function convert_encoder_layer(opus_dict, layer_prefix: str, converter: dict) - removed:
function convert_hf_name_to_opus_name(hf_model_name) - removed:
function convert_opus_name_to_hf_name(x) - removed:
function convert_whole_dir(path=…) - removed:
function download_and_unzip(url, dest_dir) - removed:
function fetch_test_set(test_set_url) - removed:
function find_model_file(dest_dir) - removed:
function find_pretrained_model(src_lang: str, tgt_lang: str) -> List[str] - removed:
function find_vocab_file(model_dir) - removed:
function get_system_metadata(repo_root) - removed:
function lmap(f, x) -> List - removed:
function load_config_from_state_dict(opus_dict) - removed:
function load_layers_(layer_lst: torch.nn.ModuleList, opus_state: dict, converter, is_decoder=…) - removed:
function load_yaml(path) - removed:
function make_registry(repo_path=…) - removed:
function remove_prefix(text: str, prefix: str) - removed:
function remove_suffix(text: str, suffix: str) - removed:
function save_json(content: Union[Dict, List], path: str) -> None - removed:
function save_tokenizer_config(dest_dir: Path) - removed:
function unzip(zip_path: str, dest_dir: str) -> None - removed:
function write_model_card(hf_model_name: str, repo_root=…, save_dir=…, dry_run=…, extra_metadata=…) -> str - removed:
method OpusState.load_marian_model() -> MarianMTModel - removed:
method OpusState.sub_keys(layer_prefix) - removed:
property OpusState.extra_keys - removed:
variable BART_CONVERTER - removed:
variable BIAS_KEY - removed:
variable CONFIG_KEY - removed:
variable DEFAULT_MODEL_DIR - removed:
variable DEFAULT_REPO - removed:
variable FRONT_MATTER_TEMPLATE - removed:
variable GROUPS - removed:
variable GROUP_TO_OPUS_NAME - removed:
variable OPUS_GITHUB_URL - removed:
variable ORG_NAME - removed:
variable ROM_GROUP - removed:
variable args - removed:
variable dest_dir - removed:
variable parser - removed:
variable source_dir
- removed:
function convert_fairseq_mbart_checkpoint_from_disk(checkpoint_path, hf_config_path=…) - removed:
variable args - removed:
variable model - removed:
variable parser
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, mobilebert_config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_openai_checkpoint_to_pytorch(openai_checkpoint_folder_path, openai_config_file, pytorch_dump_folder_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_pegasus(tf_weights: dict, cfg_updates: dict) -> PegasusForConditionalGeneration - removed:
function convert_pegasus_ckpt_to_pytorch(ckpt_path: str, save_dir: str) - removed:
function get_tf_weights_as_numpy(path=…) -> Dict - removed:
function rename_state_dict_key(k) - removed:
variable PATTERNS - removed:
variable args - removed:
variable dataset - removed:
variable parser
- removed:
function convert_prophetnet_checkpoint_to_pytorch(prophetnet_checkpoint_path: str, pytorch_dump_folder_path: str) - removed:
variable args - removed:
variable logger - removed:
variable parser
- removed:
function convert_all_pt_checkpoints_to_tf(args_model_type, tf_dump_path, model_shortcut_names_or_path=…, config_shortcut_names_or_path=…, compare_with_pt_model=…, use_cached_models=…, remove_cached_files=…, only_convert_finetuned_models=…) - removed:
function convert_pt_checkpoint_to_tf(model_type, pytorch_checkpoint_path, config_file, tf_dump_path, compare_with_pt_model=…, use_cached_models=…) - removed:
variable MODEL_CLASSES - removed:
variable args - removed:
variable parser
- removed:
function convert_trax_checkpoint_to_pytorch(trax_model_pkl_path, config_file, pytorch_dump_path) - removed:
function set_block_weights_in_torch(weights, torch_block, hidden_size) - removed:
function set_layer_weights_in_torch_local(weights, torch_layer, hidden_size) - removed:
function set_layer_weights_in_torch_lsh(weights, torch_layer, hidden_size) - removed:
function set_model_weights_in_torch(weights, torch_model, hidden_size) - removed:
function set_param(torch_layer, weight, bias=…) - removed:
variable args - removed:
variable parser
- removed:
function convert_roberta_checkpoint_to_pytorch(roberta_checkpoint_path: str, pytorch_dump_folder_path: str, classification_head: bool) - removed:
variable SAMPLE_TEXT - removed:
variable args - removed:
variable logger - removed:
variable parser
- removed:
function get_proto(filename: str) - changed:
function convert_slow_tokenizer(transformer_tokenizer) -> Tokenizer → function convert_slow_tokenizer(transformer_tokenizer, from_tiktoken=…) -> Tokenizer - changed:
method GPT2Converter.converted() -> Tokenizer → method GPT2Converter.converted(vocab: Optional[dict[str, int]]=…, merges: Optional[list[tuple[str, str]]]=…) -> Tokenizer - changed:
method SentencePieceExtractor.extract() -> Tuple[Dict[str, int], List[Tuple]] → method SentencePieceExtractor.extract(vocab_scores=…) -> tuple[dict[str, int], list[tuple]] - added:
class BarthezConverter(SpmConverter) - added:
class BigBirdConverter(SpmConverter) - added:
class BlenderbotConverter(Converter) - added:
class CLIPConverter(Converter) - added:
class DebertaConverter(Converter) - added:
class DebertaV2Converter(SpmConverter) - added:
class GemmaConverter(SpmConverter) - added:
class GemmaSentencePieceExtractor(SentencePieceExtractor) - added:
class HeliumConverter(SpmConverter) - added:
class LayoutLMv2Converter(Converter) - added:
class LlamaConverter(SpmConverter) - added:
class MBart50Converter(SpmConverter) - added:
class MPNetConverter(Converter) - added:
class MarkupLMConverter(Converter) - added:
class MistralConverter - added:
class MoshiConverter(SpmConverter) - added:
class NllbConverter(SpmConverter) - added:
class ParakeetConverter(SpmConverter) - added:
class Qwen2Converter(Converter) - added:
class RemBertConverter(SpmConverter) - added:
class RoFormerConverter(Converter) - added:
class SeamlessM4TConverter(SpmConverter) - added:
class SplinterConverter(Converter) - added:
class TikTokenConverter - added:
class UdopConverter(SpmConverter) - added:
class WhisperConverter(Converter) - added:
class XGLMConverter(SpmConverter) - added:
constructor HeliumConverter(vocab_file=…, **kwargs) - added:
constructor MistralConverter(vocab_file=…, pattern=…, add_prefix_space=…, additional_special_tokens=…, **kwargs) - added:
constructor MoshiConverter(vocab_file, model_max_length=…, **kwargs) - added:
constructor ParakeetConverter(vocab_file=…, *args) - added:
constructor TikTokenConverter(vocab_file=…, pattern=…, add_prefix_space=…, additional_special_tokens=…, **kwargs) - added:
field GemmaConverter.SpmExtractor - added:
field GemmaConverter.handle_byte_fallback - added:
field GemmaConverter.special_tokens - added:
field HeliumConverter.handle_byte_fallback - added:
field LlamaConverter.handle_byte_fallback - added:
field MoshiConverter.handle_byte_fallback - added:
field ParakeetConverter.handle_byte_fallback - added:
field SpmConverter.SpmExtractor - added:
field SpmConverter.handle_byte_fallback - added:
field SpmConverter.special_tokens - added:
function bytes_to_unicode() - added:
function generate_merges(vocab, vocab_scores) - added:
function import_protobuf(error_message=…) - added:
method BarthezConverter.post_processor() - added:
method BarthezConverter.unk_id(proto) - added:
method BigBirdConverter.post_processor() - added:
method BlenderbotConverter.converted() -> Tokenizer - added:
method CLIPConverter.converted() -> Tokenizer - added:
method DebertaConverter.converted() -> Tokenizer - added:
method DebertaV2Converter.normalizer(proto) - added:
method DebertaV2Converter.post_processor() - added:
method DebertaV2Converter.pre_tokenizer(replacement, add_prefix_space) - added:
method GemmaConverter.decoder(replacement, add_prefix_space) - added:
method GemmaConverter.normalizer(proto) - added:
method GemmaConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method GemmaConverter.unk_id(proto) - added:
method GemmaConverter.vocab(proto) - added:
method GemmaSentencePieceExtractor.extract(vocab_scores=…) -> tuple[dict[str, int], list[tuple]] - added:
method HeliumConverter.decoder(replacement, add_prefix_space) - added:
method HeliumConverter.normalizer(proto) - added:
method HeliumConverter.post_processor() - added:
method HeliumConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method HeliumConverter.tokenizer(proto) - added:
method HeliumConverter.unk_id(proto) - added:
method HeliumConverter.vocab(proto) - added:
method LayoutLMv2Converter.converted() -> Tokenizer - added:
method LlamaConverter.decoder(replacement, add_prefix_space) - added:
method LlamaConverter.normalizer(proto) - added:
method LlamaConverter.post_processor() - added:
method LlamaConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method LlamaConverter.unk_id(proto) - added:
method LlamaConverter.vocab(proto) - added:
method MBart50Converter.post_processor() - added:
method MBart50Converter.unk_id(proto) - added:
method MBart50Converter.vocab(proto) - added:
method MPNetConverter.converted() -> Tokenizer - added:
method MarkupLMConverter.converted() -> Tokenizer - added:
method MistralConverter.converted() -> Tokenizer - added:
method MistralConverter.extract_vocab_merges_from_model(tiktoken_url: str) - added:
method MistralConverter.tokenizer() - added:
method MoshiConverter.decoder(replacement, add_prefix_space) - added:
method MoshiConverter.normalizer(proto) - added:
method MoshiConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method NllbConverter.post_processor() - added:
method NllbConverter.unk_id(proto) - added:
method NllbConverter.vocab(proto) - added:
method ParakeetConverter.tokenizer(proto) - added:
method PegasusConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method Qwen2Converter.converted(vocab: Optional[dict[str, int]]=…, merges: Optional[list[tuple[str, str]]]=…) -> Tokenizer - added:
method RemBertConverter.normalizer(proto) - added:
method RemBertConverter.post_processor() - added:
method RoFormerConverter.converted() -> Tokenizer - added:
method SeamlessM4TConverter.post_processor() - added:
method SeamlessM4TConverter.unk_id(proto) - added:
method SeamlessM4TConverter.vocab(proto) - added:
method SplinterConverter.converted() -> Tokenizer - added:
method SpmConverter.decoder(replacement, add_prefix_space) - added:
method SpmConverter.pre_tokenizer(replacement, add_prefix_space) - added:
method TikTokenConverter.converted() -> Tokenizer - added:
method TikTokenConverter.extract_vocab_merges_from_model(tiktoken_url: str) - added:
method TikTokenConverter.tokenizer() - added:
method UdopConverter.post_processor() - added:
method WhisperConverter.converted() -> Tokenizer - added:
method XGLMConverter.post_processor() - added:
method XGLMConverter.unk_id(proto) - added:
method XGLMConverter.vocab(proto) - added:
variable logger
- removed:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_transfo_xl_checkpoint_to_pytorch(tf_checkpoint_path, transfo_xl_config_file, pytorch_dump_folder_path, transfo_xl_dataset_file) - removed:
variable args - removed:
variable parser
- removed:
function convert_xlm_checkpoint_to_pytorch(xlm_checkpoint_path, pytorch_dump_folder_path) - removed:
variable args - removed:
variable parser
- removed:
function convert_xlnet_checkpoint_to_pytorch(tf_checkpoint_path, bert_config_file, pytorch_dump_folder_path, finetuning_task=…) - removed:
variable GLUE_TASKS_NUM_LABELS - removed:
variable args - removed:
variable parser
- added:
export DataCollatorForLanguageModeling - added:
export DataCollatorForMultipleChoice - added:
export DataCollatorForPermutationLanguageModeling - added:
export DataCollatorForSOP - added:
export DataCollatorForSeq2Seq - added:
export DataCollatorForTokenClassification - added:
export DataCollatorForWholeWordMask - added:
export DataCollatorWithFlattening - added:
export DataCollatorWithPadding - added:
export DefaultDataCollator - added:
export default_data_collator
- removed:
method DataCollatorForLanguageModeling.mask_tokens(inputs: torch.Tensor, special_tokens_mask: Optional[torch.Tensor]=…) -> Tuple[torch.Tensor, torch.Tensor] - removed:
method DataCollatorForPermutationLanguageModeling.mask_tokens(inputs: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor] - removed:
method DataCollatorForWholeWordMask.mask_tokens(inputs: torch.Tensor, mask_labels: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor] - changed:
class DataCollatorForLanguageModeling → class DataCollatorForLanguageModeling(DataCollatorMixin) - changed:
class DataCollatorForPermutationLanguageModeling → class DataCollatorForPermutationLanguageModeling(DataCollatorMixin) - changed:
class DataCollatorForTokenClassification → class DataCollatorForTokenClassification(DataCollatorMixin) - changed:
field DataCollatorForLanguageModeling.mlm_probability: float → field DataCollatorForLanguageModeling.mlm_probability: Optional[float] - changed:
function default_data_collator(features: List[InputDataClass]) -> Dict[str, torch.Tensor] → function default_data_collator(features: list[InputDataClass], return_tensors=…) -> dict[str, Any] - changed:
function tolist(x: Union[List[Any], torch.Tensor]) → function tolist(x) -> list[Any] - changed:
method DataCollatorForSOP.mask_tokens(inputs: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor] → method DataCollatorForSOP.mask_tokens(inputs: Any) -> tuple[Any, Any, Any] - added:
class DataCollatorForMultipleChoice(DataCollatorMixin) - added:
class DataCollatorForSeq2Seq - added:
class DataCollatorMixin - added:
class DataCollatorWithFlattening(DefaultDataCollator) - added:
class DefaultDataCollator(DataCollatorMixin) - added:
constructor DataCollatorForWholeWordMask(*args, **kwargs) - added:
constructor DataCollatorWithFlattening(*args, return_position_ids=…, separator_id=…, return_flash_attn_kwargs=…, return_seq_idx=…, **kwargs) - added:
field DataCollatorForLanguageModeling.mask_replace_prob: float - added:
field DataCollatorForLanguageModeling.pad_to_multiple_of: Optional[int] - added:
field DataCollatorForLanguageModeling.random_replace_prob: float - added:
field DataCollatorForLanguageModeling.return_tensors: str - added:
field DataCollatorForLanguageModeling.seed: Optional[int] - added:
field DataCollatorForLanguageModeling.tf_experimental_compile: bool - added:
field DataCollatorForLanguageModeling.whole_word_mask: bool - added:
field DataCollatorForMultipleChoice.max_length: Optional[int] - added:
field DataCollatorForMultipleChoice.pad_to_multiple_of: Optional[int] - added:
field DataCollatorForMultipleChoice.padding: Union[bool, str, PaddingStrategy] - added:
field DataCollatorForMultipleChoice.return_tensors: str - added:
field DataCollatorForMultipleChoice.tokenizer: PreTrainedTokenizerBase - added:
field DataCollatorForPermutationLanguageModeling.return_tensors: str - added:
field DataCollatorForSeq2Seq.label_pad_token_id: int - added:
field DataCollatorForSeq2Seq.max_length: Optional[int] - added:
field DataCollatorForSeq2Seq.model: Optional[Any] - added:
field DataCollatorForSeq2Seq.pad_to_multiple_of: Optional[int] - added:
field DataCollatorForSeq2Seq.padding: Union[bool, str, PaddingStrategy] - added:
field DataCollatorForSeq2Seq.return_tensors: str - added:
field DataCollatorForSeq2Seq.tokenizer: PreTrainedTokenizerBase - added:
field DataCollatorForTokenClassification.return_tensors: str - added:
field DataCollatorWithPadding.return_tensors: str - added:
field DefaultDataCollator.return_tensors: str - added:
function numpy_default_data_collator(features: list[InputDataClass]) -> dict[str, Any] - added:
function pad_without_fast_tokenizer_warning(tokenizer, *pad_args, **pad_kwargs) - added:
function tf_default_data_collator(features: list[InputDataClass]) -> dict[str, Any] - added:
function to_numpy(x) -> np.ndarray[Any] - added:
function torch_default_data_collator(features: list[InputDataClass]) -> dict[str, Any] - added:
method DataCollatorForLanguageModeling.create_rng() - added:
method DataCollatorForLanguageModeling.get_generator(seed) - added:
method DataCollatorForLanguageModeling.numpy_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForLanguageModeling.numpy_mask_tokens(inputs: Any, special_tokens_mask: Optional[Any]=…, offset_mapping: Optional[Any]=…) -> tuple[Any, Any] - added:
method DataCollatorForLanguageModeling.tf_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForLanguageModeling.tf_mask_tokens(inputs: Any, vocab_size, mask_token_id, special_tokens_mask: Optional[Any]=…) -> tuple[Any, Any] - added:
method DataCollatorForLanguageModeling.torch_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForLanguageModeling.torch_mask_tokens(inputs: Any, special_tokens_mask: Optional[Any]=…, offset_mapping: Optional[Any]=…) -> tuple[Any, Any] - added:
method DataCollatorForMultipleChoice.tf_call(features) - added:
method DataCollatorForMultipleChoice.torch_call(examples: list[dict[str, Any]]) - added:
method DataCollatorForPermutationLanguageModeling.numpy_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForPermutationLanguageModeling.numpy_mask_tokens(inputs: Any) -> tuple[Any, Any, Any, Any] - added:
method DataCollatorForPermutationLanguageModeling.tf_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForPermutationLanguageModeling.tf_mask_tokens(inputs: Any) -> tuple[Any, Any, Any, Any] - added:
method DataCollatorForPermutationLanguageModeling.torch_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForPermutationLanguageModeling.torch_mask_tokens(inputs: Any) -> tuple[Any, Any, Any, Any] - added:
method DataCollatorForTokenClassification.numpy_call(features) - added:
method DataCollatorForTokenClassification.tf_call(features) - added:
method DataCollatorForTokenClassification.torch_call(features) - added:
method DataCollatorForWholeWordMask.numpy_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForWholeWordMask.numpy_mask_tokens(inputs: Any, mask_labels: Any) -> tuple[Any, Any] - added:
method DataCollatorForWholeWordMask.tf_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForWholeWordMask.tf_mask_tokens(inputs: Any, mask_labels: Any) -> tuple[Any, Any] - added:
method DataCollatorForWholeWordMask.torch_call(examples: list[Union[list[int], Any, dict[str, Any]]]) -> dict[str, Any] - added:
method DataCollatorForWholeWordMask.torch_mask_tokens(inputs: Any, mask_labels: Any) -> tuple[Any, Any] - added:
static method DataCollatorForLanguageModeling.tf_bernoulli(shape, probability, generator=…)
- changed:
field GlueDataset.features: List[InputFeatures] → field GlueDataset.features: list[InputFeatures]
- changed:
method TextDatasetForNextSentencePrediction.create_examples_from_document(document: List[List[int]], doc_index: int) → method TextDatasetForNextSentencePrediction.create_examples_from_document(document: list[list[int]], doc_index: int, block_size: int)
- changed:
constructor SquadDataset(args: SquadDataTrainingArguments, tokenizer: PreTrainedTokenizer, limit_length: Optional[int]=…, mode: Union[str, Split]=…, is_language_sensitive: Optional[bool]=…, cache_dir: Optional[str]=…, dataset_format: Optional[str]=…) → constructor SquadDataset(args: SquadDataTrainingArguments, tokenizer: PreTrainedTokenizer, limit_length: Optional[int]=…, mode: Union[str, Split]=…, is_language_sensitive: bool=…, cache_dir: Optional[str]=…, dataset_format: str=…) - changed:
field SquadDataset.features: List[SquadFeatures] → field SquadDataset.features: list[SquadFeatures]
- removed:
export requires_sklearn - added:
export requires_backends
- changed:
function glue_convert_examples_to_features(examples: Union[List[InputExample], tf.data.Dataset], tokenizer: PreTrainedTokenizer, max_length: Optional[int]=…, task=…, label_list=…, output_mode=…) → function glue_convert_examples_to_features(examples: Union[list[InputExample], tf.data.Dataset], tokenizer: PreTrainedTokenizer, max_length: Optional[int]=…, task=…, label_list=…, output_mode=…)
- changed:
constructor SquadFeatures(input_ids, attention_mask, token_type_ids, cls_index, p_mask, example_index, unique_id, paragraph_len, token_is_max_context, tokens, token_to_orig_map, start_position, end_position, is_impossible, qas_id: str=…) → constructor SquadFeatures(input_ids, attention_mask, token_type_ids, cls_index, p_mask, example_index, unique_id, paragraph_len, token_is_max_context, tokens, token_to_orig_map, start_position, end_position, is_impossible, qas_id: Optional[str]=…, encoding: Optional[BatchEncoding]=…)
- changed:
field InputFeatures.attention_mask: Optional[List[int]] → field InputFeatures.attention_mask: Optional[list[int]] - changed:
field InputFeatures.input_ids: List[int] → field InputFeatures.input_ids: list[int] - changed:
field InputFeatures.token_type_ids: Optional[List[int]] → field InputFeatures.token_type_ids: Optional[list[int]]
- removed:
class GenerationUtilsTest(unittest.TestCase) - removed:
method GenerationUtilsTest.test_postprocess_next_token_scores() - removed:
method GenerationUtilsTest.test_postprocess_next_token_scores_large_bad_words_list() - removed:
property GenerationUtilsTest.config - removed:
property GenerationUtilsTest.model
- added:
class DebugOption(ExplicitEnum) - added:
class DebugUnderflowOverflow - added:
constructor DebugUnderflowOverflow(model, max_frames_to_save=…, trace_batch_nums=…, abort_after_batch_num=…) - added:
field DebugOption.TPU_METRICS_DEBUG - added:
field DebugOption.UNDERFLOW_OVERFLOW - added:
function detect_overflow(var, ctx) - added:
function get_abs_min_max(var, ctx) - added:
method DebugUnderflowOverflow.analyse_model() - added:
method DebugUnderflowOverflow.analyse_variable(var, ctx) - added:
method DebugUnderflowOverflow.batch_end_frame() - added:
method DebugUnderflowOverflow.batch_start_frame() - added:
method DebugUnderflowOverflow.create_frame(module, input, output) - added:
method DebugUnderflowOverflow.dump_saved_frames() - added:
method DebugUnderflowOverflow.expand_frame(line) - added:
method DebugUnderflowOverflow.forward_hook(module, input, output) - added:
method DebugUnderflowOverflow.register_forward_hook() - added:
method DebugUnderflowOverflow.reset_saved_frames() - added:
method DebugUnderflowOverflow.save_frame(frame=…) - added:
method DebugUnderflowOverflow.trace_frames() - added:
variable logger
- added:
function dep_version_check(pkg, hint=…) - added:
variable pkgs_to_check_at_runtime
- added:
variable deps
- added:
export DistributedConfig
- added:
class DistributedConfig - added:
field DistributedConfig.enable_expert_parallel: bool - added:
method DistributedConfig.to_dict() -> dict[str, Any] - added:
method DistributedConfig.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method DistributedConfig.to_json_string() - added:
method DistributedConfig.update(**kwargs) - added:
static method DistributedConfig.from_dict(config_dict, **kwargs)
- added:
function check_imports(filename: Union[str, os.PathLike]) -> list[str] - added:
function check_python_requirements(path_or_repo_id, requirements_file=…, **kwargs) - added:
function create_dynamic_module(name: Union[str, os.PathLike]) -> None - added:
function custom_object_save(obj: Any, folder: Union[str, os.PathLike], config: Optional[dict]=…) -> list[str] - added:
function get_cached_module_file(pretrained_model_name_or_path: Union[str, os.PathLike], module_file: str, cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, resume_download: Optional[bool]=…, proxies: Optional[dict[str, str]]=…, token: Optional[Union[bool, str]]=…, revision: Optional[str]=…, local_files_only: bool=…, repo_type: Optional[str]=…, _commit_hash: Optional[str]=…,… - added:
function get_class_from_dynamic_module(class_reference: str, pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, resume_download: Optional[bool]=…, proxies: Optional[dict[str, str]]=…, token: Optional[Union[bool, str]]=…, revision: Optional[str]=…, local_files_only: bool=…, repo_type: Optional[str]=…, code_revision: Optio… - added:
function get_class_in_module(class_name: str, module_path: Union[str, os.PathLike], *, force_reload: bool=…) -> type - added:
function get_imports(filename: Union[str, os.PathLike]) -> list[str] - added:
function get_relative_import_files(module_file: Union[str, os.PathLike]) -> list[str] - added:
function get_relative_imports(module_file: Union[str, os.PathLike]) -> list[str] - added:
function init_hf_modules() - added:
function resolve_trust_remote_code(trust_remote_code, model_name, has_local_code, has_remote_code, error_message=…, upstream_repo=…) - added:
variable TIME_OUT_REMOTE_CODE - added:
variable logger
- added:
class SequenceFeatureExtractor(FeatureExtractionMixin) - added:
constructor SequenceFeatureExtractor(feature_size: int, sampling_rate: int, padding_value: float, **kwargs) - added:
method SequenceFeatureExtractor.pad(processed_features: Union[BatchFeature, list[BatchFeature], dict[str, BatchFeature], dict[str, list[BatchFeature]], list[dict[str, BatchFeature]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, truncation: bool=…, pad_to_multiple_of: Optional[int]=…, return_attention_mask: Optional[bool]=…, return_tensors: Optional[Union[str, Tensor… - added:
variable logger
- added:
class BatchFeature(UserDict) - added:
class FeatureExtractionMixin(PushToHubMixin) - added:
constructor BatchFeature(data: Optional[dict[str, Any]]=…, tensor_type: Union[None, str, TensorType]=…) - added:
constructor FeatureExtractionMixin(**kwargs) - added:
method BatchFeature.convert_to_tensors(tensor_type: Optional[Union[str, TensorType]]=…) - added:
method BatchFeature.to(*args, **kwargs) -> BatchFeature - added:
method FeatureExtractionMixin.save_pretrained(save_directory: Union[str, os.PathLike], push_to_hub: bool=…, **kwargs) - added:
method FeatureExtractionMixin.to_dict() -> dict[str, Any] - added:
method FeatureExtractionMixin.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method FeatureExtractionMixin.to_json_string() -> str - added:
static method FeatureExtractionMixin.from_dict(feature_extractor_dict: dict[str, Any], **kwargs) -> Union[FeatureExtractionMixin, tuple[FeatureExtractionMixin, dict[str, Any]]] - added:
static method FeatureExtractionMixin.from_json_file(json_file: Union[str, os.PathLike]) -> FeatureExtractionMixin - added:
static method FeatureExtractionMixin.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) -> SpecificFeatureExtractorType - added:
static method FeatureExtractionMixin.get_feature_extractor_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]] - added:
static method FeatureExtractionMixin.register_for_auto_class(auto_class=…) - added:
variable PreTrainedFeatureExtractor - added:
variable SpecificFeatureExtractorType - added:
variable logger
- removed:
class ModelOutput(OrderedDict) - removed:
class cached_property(property) - removed:
function add_code_sample_docstrings(*docstr, tokenizer_class=…, checkpoint=…, output_type=…, config_class=…, mask=…) - removed:
function add_end_docstrings(*docstr) - removed:
function add_start_docstrings(*docstr) - removed:
function add_start_docstrings_to_model_forward(*docstr) - removed:
function cached_path(url_or_filename, cache_dir=…, force_download=…, proxies=…, resume_download=…, user_agent: Union[Dict, str, None]=…, extract_compressed_file=…, force_extract=…, local_files_only=…) -> Optional[str] - removed:
function filename_to_url(filename, cache_dir=…) - removed:
function get_from_cache(url: str, cache_dir=…, force_download=…, proxies=…, etag_timeout=…, resume_download=…, user_agent: Union[Dict, str, None]=…, local_files_only=…) -> Optional[str] - removed:
function hf_bucket_url(model_id: str, filename: str, revision: Optional[str]=…, mirror=…) -> str - removed:
function http_get(url: str, temp_file: BinaryIO, proxies=…, resume_size=…, user_agent: Union[Dict, str, None]=…) - removed:
function http_user_agent(user_agent: Union[Dict, str, None]=…) -> str - removed:
function is_apex_available() - removed:
function is_datasets_available() - removed:
function is_faiss_available() - removed:
function is_flax_available() - removed:
function is_in_notebook() - removed:
function is_psutil_available() - removed:
function is_py3nvml_available() - removed:
function is_remote_url(url_or_filename) - removed:
function is_sentencepiece_available() - removed:
function is_sklearn_available() - removed:
function is_tensor(x) - removed:
function is_tf_available() - removed:
function is_tokenizers_available() - removed:
function is_torch_available() - removed:
function is_torch_tpu_available() - removed:
function replace_return_docstrings(output_type=…, config_class=…) - removed:
function requires_datasets(obj) - removed:
function requires_faiss(obj) - removed:
function requires_flax(obj) - removed:
function requires_pytorch(obj) - removed:
function requires_sentencepiece(obj) - removed:
function requires_sklearn(obj) - removed:
function requires_tf(obj) - removed:
function requires_tokenizers(obj) - removed:
function tf_required(func) - removed:
function torch_only_method(fn) - removed:
function torch_required(func) - removed:
function url_to_filename(url: str, etag: Optional[str]=…) -> str - removed:
method ModelOutput.pop(*args, **kwargs) - removed:
method ModelOutput.setdefault(*args, **kwargs) - removed:
method ModelOutput.to_tuple() -> Tuple[Any] - removed:
method ModelOutput.update(*args, **kwargs) - removed:
variable CLOUDFRONT_DISTRIB_PREFIX - removed:
variable CONFIG_NAME - removed:
variable DATASETS_IMPORT_ERROR - removed:
variable DUMMY_INPUTS - removed:
variable DUMMY_MASK - removed:
variable ENV_VARS_TRUE_AND_AUTO_VALUES - removed:
variable ENV_VARS_TRUE_VALUES - removed:
variable FAISS_IMPORT_ERROR - removed:
variable FLAX_IMPORT_ERROR - removed:
variable HUGGINGFACE_CO_PREFIX - removed:
variable MODEL_CARD_NAME - removed:
variable MULTIPLE_CHOICE_DUMMY_INPUTS - removed:
variable PRESET_MIRROR_DICT - removed:
variable PT_BASE_MODEL_SAMPLE - removed:
variable PT_CAUSAL_LM_SAMPLE - removed:
variable PT_MASKED_LM_SAMPLE - removed:
variable PT_MULTIPLE_CHOICE_SAMPLE - removed:
variable PT_QUESTION_ANSWERING_SAMPLE - removed:
variable PT_RETURN_INTRODUCTION - removed:
variable PT_SEQUENCE_CLASSIFICATION_SAMPLE - removed:
variable PT_TOKEN_CLASSIFICATION_SAMPLE - removed:
variable PYTORCH_IMPORT_ERROR - removed:
variable PYTORCH_PRETRAINED_BERT_CACHE - removed:
variable PYTORCH_TRANSFORMERS_CACHE - removed:
variable S3_BUCKET_PREFIX - removed:
variable SENTENCEPIECE_IMPORT_ERROR - removed:
variable SENTENCEPIECE_UNDERLINE - removed:
variable SKLEARN_IMPORT_ERROR - removed:
variable SPIECE_UNDERLINE - removed:
variable TENSORFLOW_IMPORT_ERROR - removed:
variable TF2_WEIGHTS_NAME - removed:
variable TF_BASE_MODEL_SAMPLE - removed:
variable TF_CAUSAL_LM_SAMPLE - removed:
variable TF_MASKED_LM_SAMPLE - removed:
variable TF_MULTIPLE_CHOICE_SAMPLE - removed:
variable TF_QUESTION_ANSWERING_SAMPLE - removed:
variable TF_RETURN_INTRODUCTION - removed:
variable TF_SEQUENCE_CLASSIFICATION_SAMPLE - removed:
variable TF_TOKEN_CLASSIFICATION_SAMPLE - removed:
variable TF_WEIGHTS_NAME - removed:
variable TOKENIZERS_IMPORT_ERROR - removed:
variable TRANSFORMERS_CACHE - removed:
variable USE_JAX - removed:
variable USE_TF - removed:
variable USE_TORCH - removed:
variable WEIGHTS_NAME - removed:
variable default_cache_path - removed:
variable get_ipython - removed:
variable logger - removed:
variable torch_cache_home
- added:
export AlternatingCodebooksLogitsProcessor - added:
export AssistedCandidateGenerator - added:
export AsyncTextIteratorStreamer - added:
export BaseStreamer - added:
export BaseWatermarkingConfig - added:
export BayesianDetectorConfig - added:
export BayesianDetectorModel - added:
export BeamHypotheses - added:
export BeamSampleDecoderOnlyOutput - added:
export BeamSampleEncoderDecoderOutput - added:
export BeamScorer - added:
export BeamSearchDecoderOnlyOutput - added:
export BeamSearchEncoderDecoderOutput - added:
export CandidateGenerator - added:
export ClassifierFreeGuidanceLogitsProcessor - added:
export CompileConfig - added:
export ConfidenceCriteria - added:
export ConstrainedBeamSearchScorer - added:
export Constraint - added:
export ConstraintListState - added:
export ContinuousMixin - added:
export ContrastiveSearchDecoderOnlyOutput - added:
export ContrastiveSearchEncoderDecoderOutput - added:
export DisjunctiveConstraint - added:
export EarlyExitCandidateGenerator - added:
export EncoderNoRepeatNGramLogitsProcessor - added:
export EncoderRepetitionPenaltyLogitsProcessor - added:
export EosTokenCriteria - added:
export EpsilonLogitsWarper - added:
export EtaLogitsWarper - added:
export ExponentialDecayLengthPenalty - added:
export FlaxBeamSearchOutput - added:
export FlaxForceTokensLogitsProcessor - added:
export FlaxForcedBOSTokenLogitsProcessor - added:
export FlaxForcedEOSTokenLogitsProcessor - added:
export FlaxGenerationMixin - added:
export FlaxGreedySearchOutput - added:
export FlaxLogitsProcessor - added:
export FlaxLogitsProcessorList - added:
export FlaxLogitsWarper - added:
export FlaxMinLengthLogitsProcessor - added:
export FlaxNoRepeatNGramLogitsProcessor - added:
export FlaxSampleOutput - added:
export FlaxSuppressTokensAtBeginLogitsProcessor - added:
export FlaxSuppressTokensLogitsProcessor - added:
export FlaxTemperatureLogitsWarper - added:
export FlaxTopKLogitsWarper - added:
export FlaxTopPLogitsWarper - added:
export FlaxWhisperTimeStampLogitsProcessor - added:
export ForcedBOSTokenLogitsProcessor - added:
export ForcedEOSTokenLogitsProcessor - added:
export GenerateBeamDecoderOnlyOutput - added:
export GenerateBeamEncoderDecoderOutput - added:
export GenerateDecoderOnlyOutput - added:
export GenerateEncoderDecoderOutput - added:
export GenerationConfig - added:
export GenerationMixin - added:
export GenerationMode - added:
export GreedySearchDecoderOnlyOutput - added:
export GreedySearchEncoderDecoderOutput - added:
export InfNanRemoveLogitsProcessor - added:
export LogitNormalization - added:
export LogitsProcessor - added:
export LogitsProcessorList - added:
export MaxLengthCriteria - added:
export MaxTimeCriteria - added:
export MinLengthLogitsProcessor - added:
export MinNewTokensLengthLogitsProcessor - added:
export MinPLogitsWarper - added:
export NoBadWordsLogitsProcessor - added:
export NoRepeatNGramLogitsProcessor - added:
export OptionalDependencyNotAvailable - added:
export PhrasalConstraint - added:
export PrefixConstrainedLogitsProcessor - added:
export PromptLookupCandidateGenerator - added:
export RepetitionPenaltyLogitsProcessor - added:
export SampleDecoderOnlyOutput - added:
export SampleEncoderDecoderOutput - added:
export SequenceBiasLogitsProcessor - added:
export StopStringCriteria - added:
export StoppingCriteria - added:
export StoppingCriteriaList - added:
export SuppressTokensAtBeginLogitsProcessor - added:
export SuppressTokensLogitsProcessor - added:
export SynthIDTextWatermarkDetector - added:
export SynthIDTextWatermarkLogitsProcessor - added:
export SynthIDTextWatermarkingConfig - added:
export TFBeamSampleDecoderOnlyOutput - added:
export TFBeamSampleEncoderDecoderOutput - added:
export TFBeamSearchDecoderOnlyOutput - added:
export TFBeamSearchEncoderDecoderOutput - added:
export TFContrastiveSearchDecoderOnlyOutput - added:
export TFContrastiveSearchEncoderDecoderOutput - added:
export TFForceTokensLogitsProcessor - added:
export TFForcedBOSTokenLogitsProcessor - added:
export TFForcedEOSTokenLogitsProcessor - added:
export TFGenerationMixin - added:
export TFGreedySearchDecoderOnlyOutput - added:
export TFGreedySearchEncoderDecoderOutput - added:
export TFLogitsProcessor - added:
export TFLogitsProcessorList - added:
export TFLogitsWarper - added:
export TFMinLengthLogitsProcessor - added:
export TFNoBadWordsLogitsProcessor - added:
export TFNoRepeatNGramLogitsProcessor - added:
export TFRepetitionPenaltyLogitsProcessor - added:
export TFSampleDecoderOnlyOutput - added:
export TFSampleEncoderDecoderOutput - added:
export TFSuppressTokensAtBeginLogitsProcessor - added:
export TFSuppressTokensLogitsProcessor - added:
export TFTemperatureLogitsWarper - added:
export TFTopKLogitsWarper - added:
export TFTopPLogitsWarper - added:
export TemperatureLogitsWarper - added:
export TextIteratorStreamer - added:
export TextStreamer - added:
export TopKLogitsWarper - added:
export TopPLogitsWarper - added:
export TypicalLogitsWarper - added:
export UnbatchedClassifierFreeGuidanceLogitsProcessor - added:
export WatermarkDetector - added:
export WatermarkDetectorOutput - added:
export WatermarkLogitsProcessor - added:
export WatermarkingConfig - added:
export WhisperTimeStampLogitsProcessor - added:
export is_flax_available - added:
export is_tf_available - added:
export is_torch_available - added:
export validate_stopping_criteria
- added:
class Constraint(ABC) - added:
class ConstraintListState - added:
class DisjunctiveConstraint(Constraint) - added:
class DisjunctiveTrie - added:
class PhrasalConstraint(Constraint) - added:
constructor Constraint() - added:
constructor ConstraintListState(constraints: list[Constraint]) - added:
constructor DisjunctiveConstraint(nested_token_ids: list[list[int]]) - added:
constructor DisjunctiveTrie(nested_token_ids: list[list[int]], no_subsets=…) - added:
constructor PhrasalConstraint(token_ids: list[int]) - added:
method Constraint.advance() - added:
method Constraint.copy(stateful=…) - added:
method Constraint.does_advance(token_id: int) - added:
method Constraint.remaining() - added:
method Constraint.reset() - added:
method Constraint.test() - added:
method Constraint.update(token_id: int) - added:
method ConstraintListState.add(token_id: int) - added:
method ConstraintListState.advance() - added:
method ConstraintListState.copy(stateful=…) - added:
method ConstraintListState.get_bank() - added:
method ConstraintListState.init_state() - added:
method ConstraintListState.reset(token_ids: Optional[list[int]]) - added:
method DisjunctiveConstraint.advance() - added:
method DisjunctiveConstraint.copy(stateful=…) - added:
method DisjunctiveConstraint.does_advance(token_id: int) - added:
method DisjunctiveConstraint.remaining() - added:
method DisjunctiveConstraint.reset() - added:
method DisjunctiveConstraint.update(token_id: int) - added:
method DisjunctiveTrie.count_leaves(root) - added:
method DisjunctiveTrie.has_subsets(trie, nested_token_ids) - added:
method DisjunctiveTrie.next_tokens(current_seq) - added:
method DisjunctiveTrie.reached_leaf(current_seq) - added:
method PhrasalConstraint.advance() - added:
method PhrasalConstraint.copy(stateful=…) - added:
method PhrasalConstraint.does_advance(token_id: int) - added:
method PhrasalConstraint.remaining() - added:
method PhrasalConstraint.reset() - added:
method PhrasalConstraint.update(token_id: int) - added:
variable logger
- added:
class BeamHypotheses - added:
class BeamScorer(ABC) - added:
class BeamSearchScorer(BeamScorer) - added:
class ConstrainedBeamSearchScorer(BeamScorer) - added:
constructor BeamHypotheses(num_beams: int, length_penalty: float, early_stopping: Union[bool, str], max_length: Optional[int]=…) - added:
constructor BeamSearchScorer(batch_size: int, num_beams: int, device: torch.device, length_penalty: float=…, do_early_stopping: Union[bool, str]=…, num_beam_hyps_to_keep: int=…, num_beam_groups: int=…, max_length: Optional[int]=…) - added:
constructor ConstrainedBeamSearchScorer(batch_size: int, num_beams: int, constraints: list[Constraint], device: torch.device, length_penalty: float=…, do_early_stopping: Union[bool, str]=…, num_beam_hyps_to_keep: int=…, max_length: Optional[int]=…) - added:
method BeamHypotheses.add(hyp: torch.LongTensor, sum_logprobs: float, beam_indices: Optional[torch.LongTensor]=…, generated_len: Optional[int]=…) - added:
method BeamHypotheses.is_done(best_sum_logprobs: float, cur_len: int, decoder_prompt_len: int=…) -> bool - added:
method BeamScorer.finalize(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, max_length: int, **kwargs) -> torch.LongTensor - added:
method BeamScorer.process(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, **kwargs) -> tuple[torch.Tensor] - added:
method BeamSearchScorer.finalize(input_ids: torch.LongTensor, final_beam_scores: torch.FloatTensor, final_beam_tokens: torch.LongTensor, final_beam_indices: torch.LongTensor, max_length: int, pad_token_id: Optional[Union[int, torch.Tensor]]=…, eos_token_id: Optional[Union[int, list[int], torch.Tensor]]=…, beam_indices: Optional[torch.LongTensor]=…, decoder_prompt_len: int=…) -> tuple[torch.LongTe… - added:
method BeamSearchScorer.process(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, pad_token_id: Optional[Union[int, torch.Tensor]]=…, eos_token_id: Optional[Union[int, list[int], torch.Tensor]]=…, beam_indices: Optional[torch.LongTensor]=…, group_index: int=…, decoder_prompt_len: int=…) -> dict[str, torch.Tensor] - added:
method ConstrainedBeamSearchScorer.check_completes_constraints(sequence) - added:
method ConstrainedBeamSearchScorer.finalize(input_ids: torch.LongTensor, final_beam_scores: torch.FloatTensor, final_beam_tokens: torch.LongTensor, final_beam_indices: torch.LongTensor, max_length: int, pad_token_id: Optional[Union[int, torch.Tensor]]=…, eos_token_id: Optional[Union[int, list[int], torch.Tensor]]=…, beam_indices: Optional[torch.LongTensor]=…, decoder_prompt_len: int=…) -> tuple[t… - added:
method ConstrainedBeamSearchScorer.make_constraint_states(n) - added:
method ConstrainedBeamSearchScorer.process(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, scores_for_all_vocab: torch.FloatTensor, pad_token_id: Optional[Union[int, torch.Tensor]]=…, eos_token_id: Optional[Union[int, list[int], torch.Tensor]]=…, beam_indices: Optional[torch.LongTensor]=…, decoder_prompt_len: int=…) -> tu… - added:
method ConstrainedBeamSearchScorer.step_sentence_constraint(batch_idx: int, input_ids: torch.LongTensor, vocab_scores: torch.FloatTensor, sent_beam_scores: torch.FloatTensor, sent_beam_tokens: torch.LongTensor, sent_beam_indices: torch.LongTensor, push_progress: bool=…) - added:
property BeamSearchScorer.is_done -> bool - added:
property ConstrainedBeamSearchScorer.is_done -> bool - added:
variable FINALIZE_INPUTS_DOCSTRING - added:
variable PROCESS_INPUTS_DOCSTRING - added:
variable logger
- added:
class AssistantToTargetTranslator - added:
class AssistantVocabTranslatorCache - added:
class AssistedCandidateGenerator(CandidateGenerator) - added:
class AssistedCandidateGeneratorDifferentTokenizers(AssistedCandidateGenerator) - added:
class CandidateGenerator - added:
class EarlyExitCandidateGenerator(AssistedCandidateGenerator) - added:
class PromptLookupCandidateGenerator(CandidateGenerator) - added:
class UniversalSpeculativeDecodingGenerator(AssistedCandidateGeneratorDifferentTokenizers) - added:
constructor AssistantToTargetTranslator(target_tokenizer: PreTrainedTokenizerBase, assistant_tokenizer: PreTrainedTokenizerBase, target_vocab_size: int, assistant_model: Optional[PreTrainedModel]=…, assistant_prune_lm_head: bool=…) - added:
constructor AssistedCandidateGenerator(input_ids: torch.LongTensor, assistant_model: PreTrainedModel, generation_config: GenerationConfig, model_kwargs: dict, inputs_tensor: Optional[torch.Tensor]=…, logits_processor: Optional[LogitsProcessorList]=…) - added:
constructor AssistedCandidateGeneratorDifferentTokenizers(input_ids: torch.LongTensor, assistant_model: PreTrainedModel, target_tokenizer: PreTrainedTokenizerBase, assistant_tokenizer: PreTrainedTokenizerBase, generation_config: GenerationConfig, model_kwargs: dict, inputs_tensor: Optional[torch.Tensor]=…, logits_processor: Optional[LogitsProcessorList]=…) - added:
constructor EarlyExitCandidateGenerator(input_ids: torch.LongTensor, assistant_model: PreTrainedModel, generation_config: GenerationConfig, model_kwargs: dict, inputs_tensor: Optional[torch.Tensor]=…, logits_processor: Optional[LogitsProcessorList]=…) - added:
constructor PromptLookupCandidateGenerator(eos_token_id: Optional[torch.Tensor]=…, num_output_tokens: int=…, max_matching_ngram_size: int=…, max_length: int=…, logits_processor: Optional[LogitsProcessorList]=…, vocab_size: Optional[int]=…) - added:
constructor UniversalSpeculativeDecodingGenerator(input_ids: torch.LongTensor, assistant_model: PreTrainedModel, target_tokenizer: PreTrainedTokenizerBase, assistant_tokenizer: PreTrainedTokenizerBase, generation_config: GenerationConfig, model_kwargs: dict, atm_translator: AssistantToTargetTranslator, inputs_tensor: Optional[torch.Tensor]=…, logits_processor: Optional[LogitsProcessorList]=…) - added:
field AssistantToTargetTranslator.FILTER_VALUE: float - added:
field AssistantToTargetTranslator.SUPPRESS_TOKEN_ID: int - added:
method AssistantToTargetTranslator.get_target_ids(assistant_input_ids, target_input_ids, assistant_candidate_ids: torch.LongTensor) -> torch.LongTensor - added:
method AssistantToTargetTranslator.get_target_logits(assistant_logits: torch.FloatTensor) -> torch.FloatTensor - added:
method AssistantToTargetTranslator.unmap_input_ids() - added:
method AssistedCandidateGenerator.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
method AssistedCandidateGenerator.update_candidate_strategy(input_ids: torch.LongTensor, scores: torch.FloatTensor, num_matches: int) - added:
method AssistedCandidateGeneratorDifferentTokenizers.convert_source_tokens_to_target_tokens(input_ids, source_tokenizer, destination_tokenizer) - added:
method AssistedCandidateGeneratorDifferentTokenizers.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
method CandidateGenerator.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
method CandidateGenerator.update_candidate_strategy(input_ids: torch.LongTensor, scores: torch.FloatTensor, num_matches: int) - added:
method EarlyExitCandidateGenerator.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
method PromptLookupCandidateGenerator.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
method PromptLookupCandidateGenerator.update_candidate_strategy(input_ids: torch.LongTensor, scores: torch.FloatTensor, num_matches: int) - added:
method UniversalSpeculativeDecodingGenerator.get_candidates(input_ids: torch.LongTensor) -> tuple[torch.LongTensor, Optional[torch.FloatTensor]] - added:
static method AssistantVocabTranslatorCache.cleanup() - added:
static method AssistantVocabTranslatorCache.get_translator(target_tokenizer: PreTrainedTokenizerBase, assistant_tokenizer: PreTrainedTokenizerBase, target_vocab_size: int, assistant_model: Optional[PreTrainedModel]=…, assistant_prune_lm_head: bool=…) -> AssistantToTargetTranslator
- added:
class BaseWatermarkingConfig(ABC) - added:
class CompileConfig - added:
class GenerationConfig(PushToHubMixin) - added:
class GenerationMode(ExplicitEnum) - added:
class SynthIDTextWatermarkingConfig(BaseWatermarkingConfig) - added:
class WatermarkingConfig(BaseWatermarkingConfig) - added:
constructor GenerationConfig(**kwargs) - added:
constructor SynthIDTextWatermarkingConfig(ngram_len: int, keys: list[int], context_history_size: int=…, sampling_table_seed: int=…, sampling_table_size: int=…, skip_first_ngram_calls: bool=…, debug_mode: bool=…) - added:
constructor WatermarkingConfig(greenlist_ratio: float=…, bias: float=…, hashing_key: int=…, seeding_scheme: str=…, context_width: int=…) - added:
field CompileConfig.backend: Union[str, Callable] - added:
field CompileConfig.dynamic: Optional[bool] - added:
field CompileConfig.fullgraph: bool - added:
field CompileConfig.mode: str - added:
field CompileConfig.options: Optional[dict] - added:
field GenerationConfig.extra_output_flags - added:
field GenerationMode.ASSISTED_GENERATION - added:
field GenerationMode.BEAM_SAMPLE - added:
field GenerationMode.BEAM_SEARCH - added:
field GenerationMode.CONSTRAINED_BEAM_SEARCH - added:
field GenerationMode.CONTRASTIVE_SEARCH - added:
field GenerationMode.DOLA_GENERATION - added:
field GenerationMode.GREEDY_SEARCH - added:
field GenerationMode.GROUP_BEAM_SEARCH - added:
field GenerationMode.SAMPLE - added:
method BaseWatermarkingConfig.construct_processor(vocab_size) - added:
method BaseWatermarkingConfig.to_dict() -> dict[str, Any] - added:
method BaseWatermarkingConfig.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method BaseWatermarkingConfig.to_json_string() - added:
method BaseWatermarkingConfig.update(**kwargs) - added:
method BaseWatermarkingConfig.validate() - added:
method CompileConfig.to_dict() -> dict[str, Any] - added:
method GenerationConfig.dict_dtype_to_str(d: dict[str, Any]) -> None - added:
method GenerationConfig.get_generation_mode(assistant_model: Optional[PreTrainedModel]=…) -> GenerationMode - added:
method GenerationConfig.save_pretrained(save_directory: Union[str, os.PathLike], config_file_name: Optional[Union[str, os.PathLike]]=…, push_to_hub: bool=…, **kwargs) - added:
method GenerationConfig.to_dict() -> dict[str, Any] - added:
method GenerationConfig.to_diff_dict() -> dict[str, Any] - added:
method GenerationConfig.to_json_file(json_file_path: Union[str, os.PathLike], use_diff: bool=…) - added:
method GenerationConfig.to_json_string(use_diff: bool=…, ignore_metadata: bool=…) -> str - added:
method GenerationConfig.update(**kwargs) - added:
method GenerationConfig.validate(strict=…) - added:
method SynthIDTextWatermarkingConfig.construct_processor(vocab_size: int, device) -> WatermarkLogitsProcessor - added:
method SynthIDTextWatermarkingConfig.validate() - added:
method WatermarkingConfig.construct_processor(vocab_size: int, device) -> WatermarkLogitsProcessor - added:
method WatermarkingConfig.validate() - added:
static method BaseWatermarkingConfig.from_dict(config_dict, **kwargs) - added:
static method GenerationConfig.from_dict(config_dict: dict[str, Any], **kwargs) -> GenerationConfig - added:
static method GenerationConfig.from_model_config(model_config: PretrainedConfig) -> GenerationConfig - added:
static method GenerationConfig.from_pretrained(pretrained_model_name: Union[str, os.PathLike], config_file_name: Optional[Union[str, os.PathLike]]=…, cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) -> GenerationConfig - added:
variable ALL_CACHE_IMPLEMENTATIONS - added:
variable ALL_STATIC_CACHE_IMPLEMENTATIONS - added:
variable DEPRECATED_STATIC_CACHE_IMPLEMENTATIONS - added:
variable DYNAMIC_CACHE_IMPLEMENTATIONS - added:
variable METADATA_FIELDS - added:
variable STATIC_CACHE_IMPLEMENTATIONS - added:
variable logger
- added:
export ContinuousBatchingManager - added:
export ContinuousMixin - added:
export PagedAttentionCache - added:
export RequestState - added:
export RequestStatus
- added:
class PagedAttentionCache - added:
class PagedAttentionMemoryHandler - added:
constructor PagedAttentionCache(config: PretrainedConfig, generation_config: GenerationConfig, device: torch.device, dtype: torch.dtype=…, layer_device_map: Optional[dict[int, Union[str, torch.device, int]]]=…, tp_size: Optional[int]=…) -> None - added:
constructor PagedAttentionMemoryHandler(block_size: int, page_size: int, num_groups: int, group_size: int, peak_activation_per_token: int, num_attention_masks: int) -> None - added:
function group_layers_by_attn_type(config: PretrainedConfig) -> tuple[list[list[int]], list[str]] - added:
method PagedAttentionCache.allocate_blocks(n_blocks: int, request_id: str) -> int - added:
method PagedAttentionCache.extend_read_indices(request_id: str, past_length: int, query_length: int, read_index: list[list[int]]) -> None - added:
method PagedAttentionCache.extend_write_indices(request_id: str, past_length: int, query_length: int, write_index: list[list[int]]) -> None - added:
method PagedAttentionCache.free_blocks(request_id: str) -> None - added:
method PagedAttentionCache.get_num_free_blocks() -> int - added:
method PagedAttentionCache.get_seqlens_k(request_id: str, past_length: int, query_length: int) -> dict[str, int] - added:
method PagedAttentionCache.update(key_states: torch.Tensor, value_states: torch.Tensor, layer_idx: int, read_index: list[torch.Tensor], write_index: list[torch.Tensor], **kwargs) -> tuple[torch.Tensor, torch.Tensor] - added:
method PagedAttentionMemoryHandler.compute_max_batch_tokens(num_blocks: int, max_memory_percent: float=…, cache_dtype: torch.dtype=…) -> int - added:
method PagedAttentionMemoryHandler.compute_memory_footprint(num_blocks: Optional[int]=…, max_batch_tokens: Optional[int]=…, cache_dtype: torch.dtype=…) -> tuple[int, int, int] - added:
method PagedAttentionMemoryHandler.compute_num_blocks(max_batch_tokens: int, max_memory_percent: float=…, cache_dtype: torch.dtype=…) -> int - added:
method PagedAttentionMemoryHandler.compute_num_blocks_and_max_batch_tokens(max_memory_percent: float=…, cache_dtype: torch.dtype=…, m: float=…) -> tuple[int, int] - added:
method PagedAttentionMemoryHandler.infer_num_blocks_and_max_batch_tokens(num_blocks: Optional[int]=…, max_batch_tokens: Optional[int]=…, max_memory_percent: float=…, cache_dtype: torch.dtype=…) -> tuple[int, int] - added:
static method PagedAttentionMemoryHandler.get_available_memory(max_memory_percent: float=…) -> int
- added:
class CacheAllocator(ABC) - added:
class FullAttentionCacheAllocator(CacheAllocator) - added:
class SlidingAttentionCacheAllocator(CacheAllocator) - added:
constructor FullAttentionCacheAllocator(index: int, block_size: int) -> None - added:
constructor SlidingAttentionCacheAllocator(index: int, block_size: int, sliding_window: int) -> None - added:
method CacheAllocator.allocate_blocks(n_blocks: int, request_id: str, free_blocks: deque[int]) -> Optional[int] - added:
method CacheAllocator.free_blocks(request_id: str, free_blocks: deque[int]) -> None - added:
method CacheAllocator.get_read_indices(request_id: str, past_length: int, query_length: int) -> list[int] - added:
method CacheAllocator.get_seqlens_k(request_id: str, past_length: int, query_length: int) -> tuple[str, int] - added:
method CacheAllocator.get_write_indices(request_id: str, past_length: int, query_length: int) -> list[int] - added:
method FullAttentionCacheAllocator.allocate_blocks(n_blocks: int, request_id: str, free_blocks: deque[int]) -> Optional[int] - added:
method FullAttentionCacheAllocator.get_read_indices(request_id: str, past_length: int, query_length: int) -> list[int] - added:
method FullAttentionCacheAllocator.get_seqlens_k(request_id: str, past_length: int, query_length: int) -> tuple[str, int] - added:
method FullAttentionCacheAllocator.get_write_indices(request_id: str, past_length: int, query_length: int) -> list[int] - added:
method SlidingAttentionCacheAllocator.allocate_blocks(n_blocks: int, request_id: str, free_blocks: deque[int]) -> Optional[int] - added:
method SlidingAttentionCacheAllocator.get_read_indices(request_id: str, past_length: int, query_length: int) -> list[int] - added:
method SlidingAttentionCacheAllocator.get_seqlens_k(request_id: str, past_length: int, query_length: int) -> tuple[str, int] - added:
method SlidingAttentionCacheAllocator.get_write_indices(request_id: str, past_length: int, query_length: int) -> list[int]
- added:
class ContinuousBatchProcessor - added:
class ContinuousBatchingManager - added:
class ContinuousMixin - added:
class PagedAttentionArgs - added:
constructor ContinuousBatchProcessor(cache: PagedAttentionCache, config: PretrainedConfig, generation_config: GenerationConfig, input_queue: queue.Queue, output_queue: queue.Queue, stop_event: threading.Event, model_device: torch.device, model_dtype: torch.dtype, scheduler: Scheduler, streaming: bool=…, manual_eviction: bool=…, slice_inputs: bool=…) -> None - added:
constructor ContinuousBatchingManager(model, generation_config: GenerationConfig, manual_eviction: bool=…, max_queue_size=…, streaming: bool=…, slice_inputs: bool=…) - added:
field PagedAttentionArgs.attention_mask: Optional[torch.Tensor] - added:
field PagedAttentionArgs.cache: PagedAttentionCache - added:
field PagedAttentionArgs.cumulative_seqlens_k: torch.Tensor - added:
field PagedAttentionArgs.cumulative_seqlens_q: torch.Tensor - added:
field PagedAttentionArgs.input_ids: torch.Tensor - added:
field PagedAttentionArgs.logits_indices: torch.Tensor - added:
field PagedAttentionArgs.max_seqlen_k: int - added:
field PagedAttentionArgs.max_seqlen_q: int - added:
field PagedAttentionArgs.position_ids: torch.Tensor - added:
field PagedAttentionArgs.read_index: list[torch.Tensor] - added:
field PagedAttentionArgs.use_cache: bool - added:
field PagedAttentionArgs.write_index: list[torch.Tensor] - added:
function build_attention_mask(attention_mask: torch.Tensor, cumulative_seqlens_q: torch.Tensor, cumulative_seqlens_k: torch.Tensor, sliding_window: int=…) -> None - added:
method ContinuousBatchProcessor.fail_all_requests(error) - added:
method ContinuousBatchProcessor.get_model_kwargs() -> PagedAttentionArgs - added:
method ContinuousBatchProcessor.handle_batch_error(error) - added:
method ContinuousBatchProcessor.has_pending_requests() -> bool - added:
method ContinuousBatchProcessor.prepare_next_batch() -> bool - added:
method ContinuousBatchProcessor.reset_static_tensors(full_reset: bool=…) - added:
method ContinuousBatchProcessor.return_attention_mask() -> bool - added:
method ContinuousBatchProcessor.setup_static_tensors(num_groups: int) -> None - added:
method ContinuousBatchProcessor.update_batch() - added:
method ContinuousBatchingManager.add_request(input_ids: list[int], request_id: Optional[str]=…, max_new_tokens: Optional[int]=…) -> str - added:
method ContinuousBatchingManager.add_requests(inputs: list[list[int]], **kwargs) - added:
method ContinuousBatchingManager.cancel_request(request_id: str) - added:
method ContinuousBatchingManager.evict_request_from_cache(request_id: str) - added:
method ContinuousBatchingManager.get_result(request_id=…, timeout=…) -> Optional[GenerationOutput] - added:
method ContinuousBatchingManager.is_running() - added:
method ContinuousBatchingManager.join(timeout: Optional[float]=…) - added:
method ContinuousBatchingManager.request_id_iter(request_id) - added:
method ContinuousBatchingManager.start() - added:
method ContinuousBatchingManager.stop(block: bool=…, timeout: Optional[float]=…) - added:
method ContinuousBatchingManager.warmup(batch_processor) - added:
method ContinuousMixin.generate_batch(inputs: list[list[int]], generation_config: Optional[GenerationConfig]=…, progress_bar: bool=…, slice_inputs: bool=…, **kwargs) -> list[list[int]] - added:
method ContinuousMixin.init_continuous_batching(generation_config: Optional[GenerationConfig]=…, manual_eviction: bool=…, max_queue_size: int=…, streaming: bool=…, slice_inputs: bool=…) -> ContinuousBatchingManager - added:
static method ContinuousBatchingManager.default_attention_implementation() -> str - added:
static method ContinuousBatchingManager.supported_attention_implementations() -> set[str]
- added:
class GenerationOutput - added:
class RequestState - added:
class RequestStatus(Enum) - added:
field GenerationOutput.created_time: float - added:
field GenerationOutput.error: Optional[str] - added:
field GenerationOutput.generated_tokens: list[int] - added:
field GenerationOutput.logprobs: list[float] - added:
field GenerationOutput.prompt_ids: list[int] - added:
field GenerationOutput.request_id: str - added:
field GenerationOutput.status: RequestStatus - added:
field RequestState.allocated_blocks: int - added:
field RequestState.created_time: float - added:
field RequestState.eos_token_id: int - added:
field RequestState.error: Optional[str] - added:
field RequestState.full_prompt_ids: Optional[list[int]] - added:
field RequestState.lifespan: tuple[float, float] - added:
field RequestState.max_new_tokens: int - added:
field RequestState.position_offset: int - added:
field RequestState.prompt_ids: Optional[list[int]] - added:
field RequestState.remaining_prompt_ids: list[int] - added:
field RequestState.request_id: str - added:
field RequestState.static_outputs: list[int] - added:
field RequestStatus.DECODING - added:
field RequestStatus.FAILED - added:
field RequestStatus.FINISHED - added:
field RequestStatus.PENDING - added:
field RequestStatus.PREFILLING - added:
field RequestStatus.PREFILLING_SPLIT - added:
field RequestStatus.SPLIT_PENDING_REMAINDER - added:
function get_device_and_memory_breakdown() -> tuple[torch.device, int, int, int] - added:
method RequestState.current_len() -> int - added:
method RequestState.generated_len() -> int - added:
method RequestState.log_end_of_request() - added:
method RequestState.to_generation_output() - added:
method RequestState.update_with_token(token_id: int) -> bool - added:
property RequestState.status -> RequestStatus - added:
variable logger
- added:
class FIFOScheduler(Scheduler) - added:
class PrefillFirstScheduler(Scheduler) - added:
class Scheduler(ABC) - added:
constructor FIFOScheduler(cache: PagedAttentionCache, retain_cache_on_finish: bool=…, safety_margin: float=…) - added:
constructor Scheduler(cache: PagedAttentionCache, retain_cache_on_finish: bool=…) - added:
method FIFOScheduler.schedule_batch(token_budget: int) -> list[RequestState] - added:
method PrefillFirstScheduler.schedule_batch(token_budget: int) -> list[RequestState] - added:
method Scheduler.add_waiting_request(state: RequestState) - added:
method Scheduler.clear_cancelled_requests() - added:
method Scheduler.finish_request(request_id: str, evict_from_cache: bool=…) - added:
method Scheduler.get_active_request_static_outputs(request_id: str) -> list[int] - added:
method Scheduler.has_pending_requests() -> bool - added:
method Scheduler.request_is_cancelled(request_id: str) -> bool - added:
method Scheduler.schedule_batch(token_budget: int) -> list[RequestState] - added:
method Scheduler.set_request_cancellation(request_id: str) - added:
variable SCHEDULER_MAPPING
- added:
class FlaxForceTokensLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxForcedBOSTokenLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxForcedEOSTokenLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxLogitsProcessor - added:
class FlaxLogitsProcessorList(list) - added:
class FlaxLogitsWarper - added:
class FlaxMinLengthLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxNoRepeatNGramLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxSuppressTokensAtBeginLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxSuppressTokensLogitsProcessor(FlaxLogitsProcessor) - added:
class FlaxTemperatureLogitsWarper(FlaxLogitsWarper) - added:
class FlaxTopKLogitsWarper(FlaxLogitsWarper) - added:
class FlaxTopPLogitsWarper(FlaxLogitsWarper) - added:
class FlaxWhisperTimeStampLogitsProcessor(FlaxLogitsProcessor) - added:
constructor FlaxForceTokensLogitsProcessor(force_token_map) - added:
constructor FlaxForcedBOSTokenLogitsProcessor(bos_token_id: int) - added:
constructor FlaxForcedEOSTokenLogitsProcessor(max_length: int, eos_token_id: int) - added:
constructor FlaxMinLengthLogitsProcessor(min_length: int, eos_token_id: int) - added:
constructor FlaxNoRepeatNGramLogitsProcessor(ngram_size: int) - added:
constructor FlaxSuppressTokensAtBeginLogitsProcessor(begin_suppress_tokens, begin_index) - added:
constructor FlaxSuppressTokensLogitsProcessor(suppress_tokens: list) - added:
constructor FlaxTemperatureLogitsWarper(temperature: float) - added:
constructor FlaxTopKLogitsWarper(top_k: int, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor FlaxTopPLogitsWarper(top_p: float, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor FlaxWhisperTimeStampLogitsProcessor(generate_config, model_config, decoder_input_length) - added:
method FlaxNoRepeatNGramLogitsProcessor.get_banned_tokens_mask(latest_tokens: jnp.ndarray, previous_ngrams) -> jnp.ndarray - added:
method FlaxNoRepeatNGramLogitsProcessor.get_previous_ngrams(input_ids: jnp.ndarray, vocab_size: int, cur_len: int) - added:
variable LOGITS_PROCESSOR_INPUTS_DOCSTRING - added:
variable logger
- added:
class BeamSearchState - added:
class FlaxBeamSearchOutput(ModelOutput) - added:
class FlaxGenerationMixin - added:
class FlaxGreedySearchOutput(ModelOutput) - added:
class FlaxSampleOutput(ModelOutput) - added:
class GreedyState - added:
class SampleState - added:
field BeamSearchState.cur_len: jnp.ndarray - added:
field BeamSearchState.is_sent_finished: jnp.ndarray - added:
field BeamSearchState.model_kwargs: dict[str, jnp.ndarray] - added:
field BeamSearchState.running_scores: jnp.ndarray - added:
field BeamSearchState.running_sequences: jnp.ndarray - added:
field BeamSearchState.scores: jnp.ndarray - added:
field BeamSearchState.sequences: jnp.ndarray - added:
field FlaxBeamSearchOutput.scores: Optional[jnp.ndarray] - added:
field FlaxBeamSearchOutput.sequences: Optional[jnp.ndarray] - added:
field FlaxGreedySearchOutput.sequences: Optional[jnp.ndarray] - added:
field FlaxSampleOutput.sequences: Optional[jnp.ndarray] - added:
field GreedyState.cur_len: jnp.ndarray - added:
field GreedyState.is_sent_finished: jnp.ndarray - added:
field GreedyState.model_kwargs: dict[str, jnp.ndarray] - added:
field GreedyState.running_token: jnp.ndarray - added:
field GreedyState.sequences: jnp.ndarray - added:
field SampleState.cur_len: jnp.ndarray - added:
field SampleState.is_sent_finished: jnp.ndarray - added:
field SampleState.model_kwargs: dict[str, jnp.ndarray] - added:
field SampleState.prng_key: jnp.ndarray - added:
field SampleState.running_token: jnp.ndarray - added:
field SampleState.sequences: jnp.ndarray - added:
method FlaxGenerationMixin.generate(input_ids: jnp.ndarray, generation_config: Optional[GenerationConfig]=…, prng_key: Optional[jnp.ndarray]=…, trace: bool=…, params: Optional[dict[str, jnp.ndarray]]=…, logits_processor: Optional[FlaxLogitsProcessorList]=…, **kwargs) - added:
method FlaxGenerationMixin.prepare_inputs_for_generation(*args, **kwargs) - added:
variable logger
- added:
class AlternatingCodebooksLogitsProcessor(LogitsProcessor) - added:
class BarkEosPrioritizerLogitsProcessor(LogitsProcessor) - added:
class ClassifierFreeGuidanceLogitsProcessor(LogitsProcessor) - added:
class DiaClassifierFreeGuidanceLogitsProcessor(LogitsProcessor) - added:
class DiaEOSChannelFilterLogitsProcessor(LogitsProcessor) - added:
class DiaEOSDelayPatternLogitsProcessor(LogitsProcessor) - added:
class EncoderNoRepeatNGramLogitsProcessor(LogitsProcessor) - added:
class EncoderRepetitionPenaltyLogitsProcessor(LogitsProcessor) - added:
class EpsilonLogitsWarper(LogitsProcessor) - added:
class EtaLogitsWarper(LogitsProcessor) - added:
class ExponentialDecayLengthPenalty(LogitsProcessor) - added:
class ForcedBOSTokenLogitsProcessor(LogitsProcessor) - added:
class ForcedEOSTokenLogitsProcessor(LogitsProcessor) - added:
class HammingDiversityLogitsProcessor(LogitsProcessor) - added:
class InfNanRemoveLogitsProcessor(LogitsProcessor) - added:
class LogitNormalization(LogitsProcessor) - added:
class LogitsProcessor - added:
class LogitsProcessorList(list) - added:
class MinLengthLogitsProcessor(LogitsProcessor) - added:
class MinNewTokensLengthLogitsProcessor(LogitsProcessor) - added:
class MinPLogitsWarper(LogitsProcessor) - added:
class NoBadWordsLogitsProcessor(SequenceBiasLogitsProcessor) - added:
class NoRepeatNGramLogitsProcessor(LogitsProcessor) - added:
class PrefixConstrainedLogitsProcessor(LogitsProcessor) - added:
class RepetitionPenaltyLogitsProcessor(LogitsProcessor) - added:
class SequenceBiasLogitsProcessor(LogitsProcessor) - added:
class SuppressTokensAtBeginLogitsProcessor(LogitsProcessor) - added:
class SuppressTokensLogitsProcessor(LogitsProcessor) - added:
class SynthIDTextWatermarkLogitsProcessor(LogitsProcessor) - added:
class SynthIDTextWatermarkState - added:
class TemperatureLogitsWarper(LogitsProcessor) - added:
class TopKLogitsWarper(LogitsProcessor) - added:
class TopPLogitsWarper(LogitsProcessor) - added:
class TypicalLogitsWarper(LogitsProcessor) - added:
class UnbatchedClassifierFreeGuidanceLogitsProcessor(LogitsProcessor) - added:
class WatermarkLogitsProcessor(LogitsProcessor) - added:
class WhisperNoSpeechDetection(LogitsProcessor) - added:
class WhisperTimeStampLogitsProcessor(LogitsProcessor) - added:
constructor AlternatingCodebooksLogitsProcessor(input_start_len: int, semantic_vocab_size: int, codebook_size: int) - added:
constructor BarkEosPrioritizerLogitsProcessor(eos_token_id: Union[int, list[int], torch.Tensor], min_eos_p: float, device: str=…) - added:
constructor ClassifierFreeGuidanceLogitsProcessor(guidance_scale) - added:
constructor DiaClassifierFreeGuidanceLogitsProcessor(guidance_scale: float, guidance_top_k: Optional[int]=…) - added:
constructor DiaEOSChannelFilterLogitsProcessor(num_channels: int, eos_token_id: int) - added:
constructor DiaEOSDelayPatternLogitsProcessor(delay_pattern: list[int], eos_token_id: int, max_generation_len: int, device: str=…) - added:
constructor EncoderNoRepeatNGramLogitsProcessor(encoder_ngram_size: int, encoder_input_ids: torch.LongTensor) - added:
constructor EncoderRepetitionPenaltyLogitsProcessor(penalty: float, encoder_input_ids: torch.LongTensor) - added:
constructor EpsilonLogitsWarper(epsilon: float, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor EtaLogitsWarper(epsilon: float, filter_value: float=…, min_tokens_to_keep: int=…, device: str=…) - added:
constructor ExponentialDecayLengthPenalty(exponential_decay_length_penalty: tuple[int, float], eos_token_id: Union[int, list[int], torch.Tensor], input_ids_seq_length: int) - added:
constructor ForcedBOSTokenLogitsProcessor(bos_token_id: int) - added:
constructor ForcedEOSTokenLogitsProcessor(max_length: int, eos_token_id: Union[int, list[int], torch.Tensor], device: str=…) - added:
constructor HammingDiversityLogitsProcessor(diversity_penalty: float, num_beams: int, num_beam_groups: int) - added:
constructor MinLengthLogitsProcessor(min_length: int, eos_token_id: Union[int, list[int], torch.Tensor], device: str=…) - added:
constructor MinNewTokensLengthLogitsProcessor(prompt_length_to_skip: int, min_new_tokens: int, eos_token_id: Union[int, list[int], torch.Tensor], device: str=…) - added:
constructor MinPLogitsWarper(min_p: float, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor NoBadWordsLogitsProcessor(bad_words_ids: list[list[int]], eos_token_id: Optional[Union[int, list[int], torch.Tensor]]=…) - added:
constructor NoRepeatNGramLogitsProcessor(ngram_size: int) - added:
constructor PrefixConstrainedLogitsProcessor(prefix_allowed_tokens_fn: Callable[[int, torch.Tensor], list[int]], num_beams: int) - added:
constructor RepetitionPenaltyLogitsProcessor(penalty: float, prompt_ignore_length: Optional[int]=…) - added:
constructor SequenceBiasLogitsProcessor(sequence_bias: list[list[Union[list[int], float]]]) - added:
constructor SuppressTokensAtBeginLogitsProcessor(begin_suppress_tokens, begin_index, device: str=…) - added:
constructor SuppressTokensLogitsProcessor(suppress_tokens, device: str=…) - added:
constructor SynthIDTextWatermarkLogitsProcessor(ngram_len: int, keys: list[int], sampling_table_size: int, sampling_table_seed: int, context_history_size: int, device: torch.device, skip_first_ngram_calls: bool=…, debug_mode: bool=…) - added:
constructor SynthIDTextWatermarkState(batch_size: int, ngram_len: int, context_history_size: int, device: torch.device) - added:
constructor TemperatureLogitsWarper(temperature: float) - added:
constructor TopKLogitsWarper(top_k: int, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor TopPLogitsWarper(top_p: float, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor TypicalLogitsWarper(mass: float=…, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor UnbatchedClassifierFreeGuidanceLogitsProcessor(guidance_scale: float, model, unconditional_ids: Optional[torch.LongTensor]=…, unconditional_attention_mask: Optional[torch.LongTensor]=…, use_cache: bool=…) - added:
constructor WatermarkLogitsProcessor(vocab_size, device, greenlist_ratio: float=…, bias: float=…, hashing_key: int=…, seeding_scheme: str=…, context_width: int=…) - added:
constructor WhisperNoSpeechDetection(no_speech_token: int, begin_index: int, scores_is_logprobs: bool=…) - added:
constructor WhisperTimeStampLogitsProcessor(generate_config: GenerationConfig, begin_index: int, _detect_timestamp_from_logprob: Optional[bool]=…) - added:
method RepetitionPenaltyLogitsProcessor.set_continuous_batching_context(logits_indices: torch.Tensor, cu_seq_lens_q: torch.Tensor) - added:
method SuppressTokensAtBeginLogitsProcessor.set_begin_index(begin_index) - added:
method SynthIDTextWatermarkLogitsProcessor.accumulate_hash(current_hash: torch.LongTensor, data: torch.LongTensor, multiplier: int=…, increment: int=…) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.compute_context_repetition_mask(input_ids: torch.LongTensor) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.compute_eos_token_mask(input_ids: torch.LongTensor, eos_token_id: int) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.compute_g_values(input_ids: torch.LongTensor) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.compute_ngram_keys(ngrams: torch.LongTensor) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.expected_mean_g_value(vocab_size: int, coinflip_prob: float=…) -> float - added:
method SynthIDTextWatermarkLogitsProcessor.sample_g_values(ngram_keys: torch.LongTensor) -> torch.LongTensor - added:
method SynthIDTextWatermarkLogitsProcessor.update_scores(scores: torch.FloatTensor, g_values: torch.FloatTensor) -> torch.FloatTensor - added:
method UnbatchedClassifierFreeGuidanceLogitsProcessor.get_unconditional_logits(input_ids) - added:
method WatermarkLogitsProcessor.set_seed(input_seq: torch.LongTensor) - added:
method WhisperNoSpeechDetection.set_begin_index(begin_index) - added:
method WhisperNoSpeechDetection.set_inputs(inputs) - added:
method WhisperNoSpeechDetection.set_model(model) - added:
method WhisperTimeStampLogitsProcessor.set_begin_index(begin_index) - added:
property WhisperNoSpeechDetection.no_speech_prob - added:
variable LOGITS_PROCESSOR_INPUTS_DOCSTRING - added:
variable logger
- added:
class ConfidenceCriteria(StoppingCriteria) - added:
class EosTokenCriteria(StoppingCriteria) - added:
class MaxLengthCriteria(StoppingCriteria) - added:
class MaxTimeCriteria(StoppingCriteria) - added:
class StopStringCriteria(StoppingCriteria) - added:
class StoppingCriteria(ABC) - added:
class StoppingCriteriaList(list) - added:
constructor ConfidenceCriteria(assistant_confidence_threshold) - added:
constructor EosTokenCriteria(eos_token_id: Union[int, list[int], torch.Tensor]) - added:
constructor MaxLengthCriteria(max_length: int, max_position_embeddings: Optional[int]=…) - added:
constructor MaxTimeCriteria(max_time: float, initial_timestamp: Optional[float]=…) - added:
constructor StopStringCriteria(tokenizer: PreTrainedTokenizerBase, stop_strings: Union[str, list[str]]) - added:
function validate_stopping_criteria(stopping_criteria: StoppingCriteriaList, max_length: int) -> StoppingCriteriaList - added:
method StopStringCriteria.clean_and_embed_tokens_with_cache(token_list, token_indices, tokenizer) - added:
property StoppingCriteriaList.max_length -> Optional[int] - added:
static method StopStringCriteria.clean_tokenizer_vocab(tokenizer, static_prefix=…) - added:
variable STOPPING_CRITERIA_INPUTS_DOCSTRING - added:
variable STOP_STRING_EMBEDDING_CACHE - added:
variable logger
- added:
class AsyncTextIteratorStreamer(TextStreamer) - added:
class BaseStreamer - added:
class TextIteratorStreamer(TextStreamer) - added:
class TextStreamer(BaseStreamer) - added:
constructor AsyncTextIteratorStreamer(tokenizer: AutoTokenizer, skip_prompt: bool=…, timeout: float | None=…, **decode_kwargs) - added:
constructor TextIteratorStreamer(tokenizer: AutoTokenizer, skip_prompt: bool=…, timeout: float | None=…, **decode_kwargs) - added:
constructor TextStreamer(tokenizer: AutoTokenizer, skip_prompt: bool=…, **decode_kwargs) - added:
method AsyncTextIteratorStreamer.on_finalized_text(text: str, stream_end: bool=…) - added:
method BaseStreamer.end() - added:
method BaseStreamer.put(value) - added:
method TextIteratorStreamer.on_finalized_text(text: str, stream_end: bool=…) - added:
method TextStreamer.end() - added:
method TextStreamer.on_finalized_text(text: str, stream_end: bool=…) - added:
method TextStreamer.put(value)
- added:
class TFForceTokensLogitsProcessor(TFLogitsProcessor) - added:
class TFForcedBOSTokenLogitsProcessor(TFLogitsProcessor) - added:
class TFForcedEOSTokenLogitsProcessor(TFLogitsProcessor) - added:
class TFLogitsProcessor - added:
class TFLogitsProcessorList(list) - added:
class TFLogitsWarper - added:
class TFMinLengthLogitsProcessor(TFLogitsProcessor) - added:
class TFNoBadWordsLogitsProcessor(TFLogitsProcessor) - added:
class TFNoRepeatNGramLogitsProcessor(TFLogitsProcessor) - added:
class TFRepetitionPenaltyLogitsProcessor(TFLogitsProcessor) - added:
class TFSuppressTokensAtBeginLogitsProcessor(TFLogitsProcessor) - added:
class TFSuppressTokensLogitsProcessor(TFLogitsProcessor) - added:
class TFTemperatureLogitsWarper(TFLogitsWarper) - added:
class TFTopKLogitsWarper(TFLogitsWarper) - added:
class TFTopPLogitsWarper(TFLogitsWarper) - added:
constructor TFForceTokensLogitsProcessor(force_token_map: list[list[int]]) - added:
constructor TFForcedBOSTokenLogitsProcessor(bos_token_id: int) - added:
constructor TFForcedEOSTokenLogitsProcessor(max_length: int, eos_token_id: int) - added:
constructor TFMinLengthLogitsProcessor(min_length: int, eos_token_id: int) - added:
constructor TFNoBadWordsLogitsProcessor(bad_words_ids: list[list[int]], eos_token_id: int) - added:
constructor TFNoRepeatNGramLogitsProcessor(ngram_size: int) - added:
constructor TFRepetitionPenaltyLogitsProcessor(penalty: float) - added:
constructor TFSuppressTokensAtBeginLogitsProcessor(begin_suppress_tokens, begin_index) - added:
constructor TFSuppressTokensLogitsProcessor(suppress_tokens) - added:
constructor TFTemperatureLogitsWarper(temperature: float) - added:
constructor TFTopKLogitsWarper(top_k: int, filter_value: float=…, min_tokens_to_keep: int=…) - added:
constructor TFTopPLogitsWarper(top_p: float, filter_value: float=…, min_tokens_to_keep: int=…) - added:
method TFNoRepeatNGramLogitsProcessor.calc_banned_ngram_tokens(input_ids, num_hypos, cur_len) - added:
variable TF_LOGITS_PROCESSOR_INPUTS_DOCSTRING - added:
variable logger
- added:
class TFBeamSampleDecoderOnlyOutput(ModelOutput) - added:
class TFBeamSampleEncoderDecoderOutput(ModelOutput) - added:
class TFBeamSearchDecoderOnlyOutput(ModelOutput) - added:
class TFBeamSearchEncoderDecoderOutput(ModelOutput) - added:
class TFContrastiveSearchDecoderOnlyOutput(ModelOutput) - added:
class TFContrastiveSearchEncoderDecoderOutput(ModelOutput) - added:
class TFGenerationMixin - added:
class TFGreedySearchDecoderOnlyOutput(ModelOutput) - added:
class TFGreedySearchEncoderDecoderOutput(ModelOutput) - added:
class TFSampleDecoderOnlyOutput(ModelOutput) - added:
class TFSampleEncoderDecoderOutput(ModelOutput) - added:
field TFBeamSampleDecoderOnlyOutput.attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSampleDecoderOnlyOutput.beam_indices: Optional[tf.Tensor] - added:
field TFBeamSampleDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSampleDecoderOnlyOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFBeamSampleDecoderOnlyOutput.sequences: Optional[tf.Tensor] - added:
field TFBeamSampleDecoderOnlyOutput.sequences_scores: Optional[tf.Tensor] - added:
field TFBeamSampleEncoderDecoderOutput.beam_indices: Optional[tf.Tensor] - added:
field TFBeamSampleEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSampleEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSampleEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSampleEncoderDecoderOutput.encoder_attentions: Optional[tuple[tf.Tensor]] - added:
field TFBeamSampleEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFBeamSampleEncoderDecoderOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFBeamSampleEncoderDecoderOutput.sequences: Optional[tf.Tensor] - added:
field TFBeamSampleEncoderDecoderOutput.sequences_scores: Optional[tf.Tensor] - added:
field TFBeamSearchDecoderOnlyOutput.attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSearchDecoderOnlyOutput.beam_indices: Optional[tf.Tensor] - added:
field TFBeamSearchDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSearchDecoderOnlyOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFBeamSearchDecoderOnlyOutput.sequences: Optional[tf.Tensor] - added:
field TFBeamSearchDecoderOnlyOutput.sequences_scores: Optional[tf.Tensor] - added:
field TFBeamSearchEncoderDecoderOutput.beam_indices: Optional[tf.Tensor] - added:
field TFBeamSearchEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSearchEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSearchEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFBeamSearchEncoderDecoderOutput.encoder_attentions: Optional[tuple[tf.Tensor]] - added:
field TFBeamSearchEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFBeamSearchEncoderDecoderOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFBeamSearchEncoderDecoderOutput.sequences: Optional[tf.Tensor] - added:
field TFBeamSearchEncoderDecoderOutput.sequences_scores: Optional[tf.Tensor] - added:
field TFContrastiveSearchDecoderOnlyOutput.attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFContrastiveSearchDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFContrastiveSearchDecoderOnlyOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFContrastiveSearchDecoderOnlyOutput.sequences: Optional[tf.Tensor] - added:
field TFContrastiveSearchEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFContrastiveSearchEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFContrastiveSearchEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFContrastiveSearchEncoderDecoderOutput.encoder_attentions: Optional[tuple[tf.Tensor]] - added:
field TFContrastiveSearchEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFContrastiveSearchEncoderDecoderOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFContrastiveSearchEncoderDecoderOutput.sequences: Optional[tf.Tensor] - added:
field TFGenerationMixin.supports_xla_generation - added:
field TFGreedySearchDecoderOnlyOutput.attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFGreedySearchDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFGreedySearchDecoderOnlyOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFGreedySearchDecoderOnlyOutput.sequences: Optional[tf.Tensor] - added:
field TFGreedySearchEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFGreedySearchEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFGreedySearchEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFGreedySearchEncoderDecoderOutput.encoder_attentions: Optional[tuple[tf.Tensor]] - added:
field TFGreedySearchEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFGreedySearchEncoderDecoderOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFGreedySearchEncoderDecoderOutput.sequences: Optional[tf.Tensor] - added:
field TFSampleDecoderOnlyOutput.attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFSampleDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFSampleDecoderOnlyOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFSampleDecoderOnlyOutput.sequences: Optional[tf.Tensor] - added:
field TFSampleEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFSampleEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFSampleEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[tf.Tensor]]] - added:
field TFSampleEncoderDecoderOutput.encoder_attentions: Optional[tuple[tf.Tensor]] - added:
field TFSampleEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFSampleEncoderDecoderOutput.scores: Optional[tuple[tf.Tensor]] - added:
field TFSampleEncoderDecoderOutput.sequences: Optional[tf.Tensor] - added:
function sample_without_replacement(logits, num_samples) - added:
function scatter_values_on_batch_indices(values, batch_indices) - added:
method TFGenerationMixin.beam_search(input_ids: tf.Tensor, do_sample: bool=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, length_penalty: Optional[float]=…, early_stopping: Optional[Union[bool, str]]=…, logits_processor: Optional[TFLogitsProcessorList]=…, logits_warper: Optional[TFLogitsProcessorList]=…, num_return_sequences: Optional[int]=…, output_… - added:
method TFGenerationMixin.compute_transition_scores(sequences: tf.Tensor, scores: tuple[tf.Tensor], beam_indices: Optional[tf.Tensor]=…, normalize_logits: bool=…) -> tf.Tensor - added:
method TFGenerationMixin.contrastive_search(input_ids: tf.Tensor, top_k: Optional[int]=…, penalty_alpha: Optional[float]=…, logits_processor: Optional[TFLogitsProcessorList]=…, logits_warper: Optional[TFLogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, outp… - added:
method TFGenerationMixin.generate(inputs: Optional[tf.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[TFLogitsProcessorList]=…, seed=…, **kwargs) -> Union[TFGenerateOutput, tf.Tensor] - added:
method TFGenerationMixin.greedy_search(input_ids: tf.Tensor, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, logits_processor: Optional[TFLogitsProcessorList]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_scores: Optional[bool]=…, return_dict_in_generate: Optional[bool]=…, **model_kwargs) -> Union[TFGreedySearchOu… - added:
method TFGenerationMixin.prepare_inputs_for_generation(*args, **kwargs) - added:
method TFGenerationMixin.sample(input_ids: tf.Tensor, logits_processor: Optional[TFLogitsProcessorList]=…, logits_warper: Optional[TFLogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, seed: Optional[tuple[int, int]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_scores: Optional[bool]=…, return… - added:
property TFGenerationMixin.seed_generator - added:
variable TFBeamSampleOutput - added:
variable TFBeamSearchOutput - added:
variable TFContrastiveSearchOutput - added:
variable TFGenerateOutput - added:
variable TFGreedySearchOutput - added:
variable TFSampleOutput - added:
variable logger
- added:
class GenerateBeamDecoderOnlyOutput(ModelOutput) - added:
class GenerateBeamEncoderDecoderOutput(ModelOutput) - added:
class GenerateDecoderOnlyOutput(ModelOutput) - added:
class GenerateEncoderDecoderOutput(ModelOutput) - added:
class GenerationMixin(ContinuousMixin) - added:
field GenerateBeamDecoderOnlyOutput.attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateBeamDecoderOnlyOutput.beam_indices: Optional[torch.LongTensor] - added:
field GenerateBeamDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateBeamDecoderOnlyOutput.logits: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamDecoderOnlyOutput.past_key_values: Optional[Cache] - added:
field GenerateBeamDecoderOnlyOutput.scores: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamDecoderOnlyOutput.sequences: torch.LongTensor - added:
field GenerateBeamDecoderOnlyOutput.sequences_scores: Optional[torch.FloatTensor] - added:
field GenerateBeamEncoderDecoderOutput.beam_indices: Optional[torch.LongTensor] - added:
field GenerateBeamEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateBeamEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateBeamEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateBeamEncoderDecoderOutput.encoder_attentions: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamEncoderDecoderOutput.logits: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamEncoderDecoderOutput.past_key_values: Optional[Cache] - added:
field GenerateBeamEncoderDecoderOutput.scores: Optional[tuple[torch.FloatTensor]] - added:
field GenerateBeamEncoderDecoderOutput.sequences: torch.LongTensor - added:
field GenerateBeamEncoderDecoderOutput.sequences_scores: Optional[torch.FloatTensor] - added:
field GenerateDecoderOnlyOutput.attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateDecoderOnlyOutput.hidden_states: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateDecoderOnlyOutput.logits: Optional[tuple[torch.FloatTensor]] - added:
field GenerateDecoderOnlyOutput.past_key_values: Optional[Cache] - added:
field GenerateDecoderOnlyOutput.scores: Optional[tuple[torch.FloatTensor]] - added:
field GenerateDecoderOnlyOutput.sequences: torch.LongTensor - added:
field GenerateEncoderDecoderOutput.cross_attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateEncoderDecoderOutput.decoder_attentions: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateEncoderDecoderOutput.decoder_hidden_states: Optional[tuple[tuple[torch.FloatTensor]]] - added:
field GenerateEncoderDecoderOutput.encoder_attentions: Optional[tuple[torch.FloatTensor]] - added:
field GenerateEncoderDecoderOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor]] - added:
field GenerateEncoderDecoderOutput.logits: Optional[tuple[torch.FloatTensor]] - added:
field GenerateEncoderDecoderOutput.past_key_values: Optional[Cache] - added:
field GenerateEncoderDecoderOutput.scores: Optional[tuple[torch.FloatTensor]] - added:
field GenerateEncoderDecoderOutput.sequences: torch.LongTensor - added:
method GenerationMixin.compute_transition_scores(sequences: torch.Tensor, scores: tuple[torch.Tensor], beam_indices: Optional[torch.Tensor]=…, normalize_logits: bool=…) -> torch.Tensor - added:
method GenerationMixin.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], list[int]]]=…, synced_gpus: Optional[bool]=…, assistant_model: Optional[PreTrainedModel]=…, streamer: Optional[Base… - added:
method GenerationMixin.heal_tokens(input_ids: torch.LongTensor, tokenizer: Optional[PreTrainedTokenizerBase]=…) -> torch.LongTensor - added:
method GenerationMixin.load_custom_generate(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, trust_remote_code: Optional[bool]=…, **kwargs) -> Callable - added:
method GenerationMixin.prepare_inputs_for_generation(input_ids: torch.LongTensor, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) - added:
variable ALL_CACHE_NAMES - added:
variable BeamSampleDecoderOnlyOutput - added:
variable BeamSampleEncoderDecoderOutput - added:
variable BeamSampleOutput - added:
variable BeamSearchDecoderOnlyOutput - added:
variable BeamSearchEncoderDecoderOutput - added:
variable BeamSearchOutput - added:
variable ContrastiveSearchDecoderOnlyOutput - added:
variable ContrastiveSearchEncoderDecoderOutput - added:
variable ContrastiveSearchOutput - added:
variable GENERATION_MODES_MAPPING - added:
variable GenerateBeamOutput - added:
variable GenerateNonBeamOutput - added:
variable GenerateOutput - added:
variable GreedySearchDecoderOnlyOutput - added:
variable GreedySearchEncoderDecoderOutput - added:
variable GreedySearchOutput - added:
variable SampleDecoderOnlyOutput - added:
variable SampleEncoderDecoderOutput - added:
variable SampleOutput - added:
variable logger
- added:
class BayesianDetectorConfig(PretrainedConfig) - added:
class BayesianDetectorModel(PreTrainedModel) - added:
class BayesianDetectorWatermarkedLikelihood(nn.Module) - added:
class BayesianWatermarkDetectorModelOutput(ModelOutput) - added:
class SynthIDTextWatermarkDetector - added:
class WatermarkDetector - added:
class WatermarkDetectorOutput - added:
constructor BayesianDetectorConfig(watermarking_depth: Optional[int]=…, base_rate: float=…, **kwargs) - added:
constructor BayesianDetectorModel(config) - added:
constructor BayesianDetectorWatermarkedLikelihood(watermarking_depth: int) - added:
constructor SynthIDTextWatermarkDetector(detector_module: BayesianDetectorModel, logits_processor: SynthIDTextWatermarkLogitsProcessor, tokenizer: Any) - added:
constructor WatermarkDetector(model_config: PretrainedConfig, device: str, watermarking_config: Union[WatermarkingConfig, dict], ignore_repeated_ngrams: bool=…, max_cache_size: int=…) - added:
field BayesianDetectorModel.base_model_prefix - added:
field BayesianDetectorModel.config: BayesianDetectorConfig - added:
field BayesianWatermarkDetectorModelOutput.loss: Optional[torch.FloatTensor] - added:
field BayesianWatermarkDetectorModelOutput.posterior_probabilities: Optional[torch.FloatTensor] - added:
field WatermarkDetectorOutput.confidence: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.green_fraction: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.num_green_tokens: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.num_tokens_scored: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.p_value: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.prediction: Optional[np.ndarray] - added:
field WatermarkDetectorOutput.z_score: Optional[np.ndarray] - added:
method BayesianDetectorConfig.set_detector_information(model_name, watermarking_config) - added:
method BayesianDetectorModel.forward(g_values: torch.Tensor, mask: torch.Tensor, labels: Optional[torch.Tensor]=…, loss_batch_weight=…, return_dict=…) -> BayesianWatermarkDetectorModelOutput - added:
method BayesianDetectorWatermarkedLikelihood.forward(g_values: torch.Tensor) -> torch.Tensor - added:
variable logger
- removed:
class BeamHypotheses - removed:
class BeamScorer(ABC) - removed:
class BeamSearchScorer(BeamScorer) - removed:
constructor BeamHypotheses(num_beams: int, max_length: int, length_penalty: float, early_stopping: bool) - removed:
constructor BeamSearchScorer(batch_size: int, max_length: int, num_beams: int, device: torch.device, length_penalty: Optional[float]=…, do_early_stopping: Optional[bool]=…, num_beam_hyps_to_keep: Optional[int]=…) - removed:
method BeamHypotheses.add(hyp: torch.LongTensor, sum_logprobs: float) - removed:
method BeamHypotheses.is_done(best_sum_logprobs: float, cur_len: int) -> bool - removed:
method BeamScorer.finalize(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, **kwargs) -> torch.LongTensor - removed:
method BeamScorer.process(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, **kwargs) -> Tuple[torch.Tensor] - removed:
method BeamSearchScorer.finalize(input_ids: torch.LongTensor, final_beam_scores: torch.FloatTensor, final_beam_tokens: torch.LongTensor, final_beam_indices: torch.LongTensor, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…) -> torch.LongTensor - removed:
method BeamSearchScorer.process(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…) -> Tuple[torch.Tensor] - removed:
property BeamSearchScorer.is_done -> bool - removed:
variable FINALIZE_INPUTS_DOCSTRING - removed:
variable PROCESS_INPUTS_DOCSTRING
- removed:
class LogitsProcessor(ABC) - removed:
class LogitsProcessorList(list) - removed:
class LogitsWarper(ABC) - removed:
class MinLengthLogitsProcessor(LogitsProcessor) - removed:
class NoBadWordsLogitsProcessor(LogitsProcessor) - removed:
class NoRepeatNGramLogitsProcessor(LogitsProcessor) - removed:
class RepetitionPenaltyLogitsProcessor(LogitsProcessor) - removed:
class TemperatureLogitsWarper(LogitsWarper) - removed:
class TopKLogitsWarper(LogitsWarper) - removed:
class TopPLogitsWarper(LogitsWarper) - removed:
constructor MinLengthLogitsProcessor(min_length: int, eos_token_id: int) - removed:
constructor NoBadWordsLogitsProcessor(bad_words_ids: Iterable[Iterable[int]], eos_token_id: int) - removed:
constructor NoRepeatNGramLogitsProcessor(ngram_size: int) - removed:
constructor RepetitionPenaltyLogitsProcessor(penalty: float) - removed:
constructor TemperatureLogitsWarper(temperature: float) - removed:
constructor TopKLogitsWarper(top_k: int, filter_value: float=…, min_tokens_to_keep: int=…) - removed:
constructor TopPLogitsWarper(top_p: float, filter_value: float=…, min_tokens_to_keep: int=…) - removed:
variable LOGITS_PROCESSOR_INPUTS_DOCSTRING
- removed:
class BeamHypotheses(object) - removed:
class TFGenerationMixin - removed:
constructor BeamHypotheses(num_beams, max_length, length_penalty, early_stopping) - removed:
function calc_banned_bad_words_ids(prev_input_ids, bad_words_ids) - removed:
function calc_banned_ngram_tokens(prev_input_ids, num_hypos, no_repeat_ngram_size, cur_len) - removed:
function sample_without_replacement(logits, num_samples) - removed:
function scatter_values_on_batch_indices(values, batch_indices) - removed:
function set_tensor_by_indices_to_value(tensor, indices, value) - removed:
function shape_list(x) - removed:
function tf_top_k_top_p_filtering(logits, top_k=…, top_p=…, filter_value=…, min_tokens_to_keep=…) - removed:
method BeamHypotheses.add(hyp, sum_logprobs) - removed:
method BeamHypotheses.is_done(best_sum_logprobs, cur_len) - removed:
method TFGenerationMixin.adjust_logits_during_generation(logits, **kwargs) - removed:
method TFGenerationMixin.generate(input_ids=…, max_length=…, min_length=…, do_sample=…, early_stopping=…, num_beams=…, temperature=…, top_k=…, top_p=…, repetition_penalty=…, bad_words_ids=…, bos_token_id=…, pad_token_id=…, eos_token_id=…, length_penalty=…, no_repeat_ngram_size=…, num_return_sequences=…, attention_mask=…, decoder_start_token_id=…, use_cache=…) - removed:
method TFGenerationMixin.prepare_inputs_for_generation(inputs, **kwargs) - removed:
variable logger
- removed:
class GenerationMixin - removed:
function top_k_top_p_filtering(logits: torch.FloatTensor, top_k: int=…, top_p: float=…, filter_value: float=…, min_tokens_to_keep: int=…) -> torch.FloatTensor - removed:
method GenerationMixin.adjust_logits_during_generation(logits: torch.FloatTensor, **kwargs) -> torch.FloatTensor - removed:
method GenerationMixin.beam_sample(input_ids: torch.LongTensor, beam_scorer: BeamScorer, logits_processor: Optional[LogitsProcessorList]=…, logits_warper: Optional[LogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, **model_kwargs) - removed:
method GenerationMixin.beam_search(input_ids: torch.LongTensor, beam_scorer: BeamScorer, logits_processor: Optional[LogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, **model_kwargs) - removed:
method GenerationMixin.generate(input_ids: Optional[torch.LongTensor]=…, max_length: Optional[int]=…, min_length: Optional[int]=…, do_sample: Optional[bool]=…, early_stopping: Optional[bool]=…, num_beams: Optional[int]=…, temperature: Optional[float]=…, top_k: Optional[int]=…, top_p: Optional[float]=…, repetition_penalty: Optional[float]=…, bad_words_ids: Optional[Iterable[int]]=…, bos_token_id:… - removed:
method GenerationMixin.greedy_search(input_ids: torch.LongTensor, logits_processor: Optional[LogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, **model_kwargs) - removed:
method GenerationMixin.prepare_inputs_for_generation(input_ids: torch.LongTensor, **kwargs) -> Dict[str, Any] - removed:
method GenerationMixin.sample(input_ids: torch.LongTensor, logits_processor: Optional[LogitsProcessorList]=…, logits_warper: Optional[LogitsProcessorList]=…, max_length: Optional[int]=…, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…, **model_kwargs) - removed:
variable logger
- removed:
class HfApi - removed:
class HfFolder - removed:
class ModelInfo - removed:
class ModelSibling - removed:
class PresignedUrl - removed:
class RepoObj - removed:
class S3Obj - removed:
class TqdmProgressFileReader - removed:
constructor HfApi(endpoint=…) - removed:
constructor ModelInfo(modelId: Optional[str]=…, author: Optional[str]=…, downloads: Optional[int]=…, tags: List[str]=…, pipeline_tag: Optional[str]=…, siblings: Optional[List[Dict]]=…, **kwargs) - removed:
constructor ModelSibling(rfilename: str, **kwargs) - removed:
constructor PresignedUrl(write: str, access: str, type: str, **kwargs) - removed:
constructor RepoObj(filename: str, lastModified: str, commit: str, size: int, **kwargs) - removed:
constructor S3Obj(filename: str, LastModified: str, ETag: str, Size: int, **kwargs) - removed:
constructor TqdmProgressFileReader(f: io.BufferedReader) - removed:
field HfFolder.path_token - removed:
method HfApi.create_repo(token: str, name: str, organization: Optional[str]=…) -> str - removed:
method HfApi.delete_obj(token: str, filename: str, organization: Optional[str]=…) - removed:
method HfApi.delete_repo(token: str, name: str, organization: Optional[str]=…) - removed:
method HfApi.list_objs(token: str, organization: Optional[str]=…) -> List[S3Obj] - removed:
method HfApi.list_repos_objs(token: str, organization: Optional[str]=…) -> List[S3Obj] - removed:
method HfApi.login(username: str, password: str) -> str - removed:
method HfApi.logout(token: str) -> None - removed:
method HfApi.model_list() -> List[ModelInfo] - removed:
method HfApi.presign(token: str, filename: str, organization: Optional[str]=…) -> PresignedUrl - removed:
method HfApi.presign_and_upload(token: str, filename: str, filepath: str, organization: Optional[str]=…) -> str - removed:
method HfApi.whoami(token: str) -> Tuple[str, List[str]] - removed:
method TqdmProgressFileReader.close() - removed:
static method HfFolder.delete_token() - removed:
static method HfFolder.get_token() - removed:
static method HfFolder.save_token(token) - removed:
variable ENDPOINT
- changed:
constructor HfArgumentParser(dataclass_types: Union[DataClassType, Iterable[DataClassType]], **kwargs) → constructor HfArgumentParser(dataclass_types: Optional[Union[DataClassType, Iterable[DataClassType]]]=…, **kwargs) - changed:
method HfArgumentParser.parse_args_into_dataclasses(args=…, return_remaining_strings=…, look_for_args_file=…, args_filename=…) -> Tuple[DataClass, ...] → method HfArgumentParser.parse_args_into_dataclasses(args=…, return_remaining_strings=…, look_for_args_file=…, args_filename=…, args_file_flag=…) -> tuple[DataClass, ...] - changed:
method HfArgumentParser.parse_dict(args: dict) -> Tuple[DataClass, ...] → method HfArgumentParser.parse_dict(args: dict[str, Any], allow_extra_keys: bool=…) -> tuple[DataClass, ...] - changed:
method HfArgumentParser.parse_json_file(json_file: str) -> Tuple[DataClass, ...] → method HfArgumentParser.parse_json_file(json_file: Union[str, os.PathLike], allow_extra_keys: bool=…) -> tuple[DataClass, ...] - added:
function HfArg(*, aliases: Optional[Union[str, list[str]]]=…, help: Optional[str]=…, default: Any=…, default_factory: Callable[[], Any]=…, metadata: Optional[dict]=…, **kwargs) -> dataclasses.Field - added:
function make_choice_type_function(choices: list) -> Callable[[str], Any] - added:
function string_to_bool(v) - added:
method HfArgumentParser.parse_yaml_file(yaml_file: Union[str, os.PathLike], allow_extra_keys: bool=…) -> tuple[DataClass, ...]
- added:
class HyperParamSearchBackendBase - added:
class OptunaBackend(HyperParamSearchBackendBase) - added:
class RayTuneBackend(HyperParamSearchBackendBase) - added:
class SigOptBackend(HyperParamSearchBackendBase) - added:
class WandbBackend(HyperParamSearchBackendBase) - added:
field HyperParamSearchBackendBase.name: str - added:
field HyperParamSearchBackendBase.pip_package: Optional[str] - added:
field OptunaBackend.name - added:
field RayTuneBackend.name - added:
field RayTuneBackend.pip_package - added:
field SigOptBackend.name - added:
field WandbBackend.name - added:
function default_hp_search_backend() -> str - added:
method HyperParamSearchBackendBase.default_hp_space(trial) - added:
method HyperParamSearchBackendBase.ensure_available() - added:
method HyperParamSearchBackendBase.run(trainer, n_trials: int, direction: str, **kwargs) - added:
method OptunaBackend.default_hp_space(trial) - added:
method OptunaBackend.run(trainer, n_trials: int, direction: str, **kwargs) - added:
method RayTuneBackend.default_hp_space(trial) - added:
method RayTuneBackend.run(trainer, n_trials: int, direction: str, **kwargs) - added:
method SigOptBackend.default_hp_space(trial) - added:
method SigOptBackend.run(trainer, n_trials: int, direction: str, **kwargs) - added:
method WandbBackend.default_hp_space(trial) - added:
method WandbBackend.run(trainer, n_trials: int, direction: str, **kwargs) - added:
static method HyperParamSearchBackendBase.is_available() - added:
static method HyperParamSearchBackendBase.pip_install() - added:
static method OptunaBackend.is_available() - added:
static method RayTuneBackend.is_available() - added:
static method SigOptBackend.is_available() - added:
static method WandbBackend.is_available() - added:
variable ALL_HYPERPARAMETER_SEARCH_BACKENDS - added:
variable logger
- added:
class BatchFeature(BaseBatchFeature) - added:
class ImageProcessingMixin(PushToHubMixin) - added:
constructor ImageProcessingMixin(**kwargs) - added:
method ImageProcessingMixin.fetch_images(image_url_or_urls: Union[str, list[str], list[list[str]]]) - added:
method ImageProcessingMixin.save_pretrained(save_directory: Union[str, os.PathLike], push_to_hub: bool=…, **kwargs) - added:
method ImageProcessingMixin.to_dict() -> dict[str, Any] - added:
method ImageProcessingMixin.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method ImageProcessingMixin.to_json_string() -> str - added:
static method ImageProcessingMixin.from_dict(image_processor_dict: dict[str, Any], **kwargs) - added:
static method ImageProcessingMixin.from_json_file(json_file: Union[str, os.PathLike]) - added:
static method ImageProcessingMixin.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) -> ImageProcessorType - added:
static method ImageProcessingMixin.get_image_processor_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]] - added:
static method ImageProcessingMixin.register_for_auto_class(auto_class=…) - added:
variable ImageProcessorType - added:
variable logger
- added:
class BaseImageProcessor(ImageProcessingMixin) - added:
constructor BaseImageProcessor(**kwargs) - added:
function convert_to_size_dict(size, max_size: Optional[int]=…, default_to_square: bool=…, height_width_order: bool=…) - added:
function get_patch_output_size(image, target_resolution, input_data_format) - added:
function get_size_dict(size: Optional[Union[int, Iterable[int], dict[str, int]]]=…, max_size: Optional[int]=…, height_width_order: bool=…, default_to_square: bool=…, param_name=…) -> dict - added:
function is_valid_size_dict(size_dict) - added:
function select_best_resolution(original_size: tuple, possible_resolutions: list) -> tuple - added:
method BaseImageProcessor.center_crop(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method BaseImageProcessor.normalize(image: np.ndarray, mean: Union[float, Iterable[float]], std: Union[float, Iterable[float]], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method BaseImageProcessor.preprocess(images, **kwargs) -> BatchFeature - added:
method BaseImageProcessor.rescale(image: np.ndarray, scale: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method BaseImageProcessor.to_dict() - added:
property BaseImageProcessor.is_fast -> bool - added:
variable INIT_SERVICE_KWARGS - added:
variable VALID_SIZE_DICT_KEYS - added:
variable logger
- added:
class BaseImageProcessorFast(BaseImageProcessor) - added:
class DefaultFastImageProcessorKwargs(TypedDict) - added:
constructor BaseImageProcessorFast(**kwargs: Unpack[DefaultFastImageProcessorKwargs]) - added:
field BaseImageProcessorFast.crop_size - added:
field BaseImageProcessorFast.data_format - added:
field BaseImageProcessorFast.default_to_square - added:
field BaseImageProcessorFast.device - added:
field BaseImageProcessorFast.do_center_crop - added:
field BaseImageProcessorFast.do_convert_rgb - added:
field BaseImageProcessorFast.do_normalize - added:
field BaseImageProcessorFast.do_pad - added:
field BaseImageProcessorFast.do_rescale - added:
field BaseImageProcessorFast.do_resize - added:
field BaseImageProcessorFast.image_mean - added:
field BaseImageProcessorFast.image_std - added:
field BaseImageProcessorFast.input_data_format - added:
field BaseImageProcessorFast.model_input_names - added:
field BaseImageProcessorFast.pad_size - added:
field BaseImageProcessorFast.resample - added:
field BaseImageProcessorFast.rescale_factor - added:
field BaseImageProcessorFast.return_tensors - added:
field BaseImageProcessorFast.size - added:
field BaseImageProcessorFast.unused_kwargs - added:
field BaseImageProcessorFast.valid_kwargs - added:
field DefaultFastImageProcessorKwargs.crop_size: Optional[dict[str, int]] - added:
field DefaultFastImageProcessorKwargs.data_format: Optional[ChannelDimension] - added:
field DefaultFastImageProcessorKwargs.default_to_square: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.device: Optional[torch.device] - added:
field DefaultFastImageProcessorKwargs.disable_grouping: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_center_crop: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_convert_rgb: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_normalize: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_pad: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_rescale: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.do_resize: Optional[bool] - added:
field DefaultFastImageProcessorKwargs.image_mean: Optional[Union[float, list[float]]] - added:
field DefaultFastImageProcessorKwargs.image_std: Optional[Union[float, list[float]]] - added:
field DefaultFastImageProcessorKwargs.input_data_format: Optional[Union[str, ChannelDimension]] - added:
field DefaultFastImageProcessorKwargs.pad_size: Optional[dict[str, int]] - added:
field DefaultFastImageProcessorKwargs.resample: Optional[Union[PILImageResampling, F.InterpolationMode]] - added:
field DefaultFastImageProcessorKwargs.rescale_factor: Optional[Union[int, float]] - added:
field DefaultFastImageProcessorKwargs.return_tensors: Optional[Union[str, TensorType]] - added:
field DefaultFastImageProcessorKwargs.size: Optional[dict[str, int]] - added:
function divide_to_patches(image: Union[np.ndarray, torch.Tensor], patch_size: int) -> list[Union[np.ndarray, torch.Tensor]] - added:
function get_max_height_width(images: list[torch.Tensor]) -> tuple[int, ...] - added:
function max_across_indices(values: Iterable[Any]) -> list[Any] - added:
function safe_squeeze(tensor: torch.Tensor, axis: Optional[int]=…) -> torch.Tensor - added:
function validate_fast_preprocess_arguments(do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[SizeDict]=…, do_resize: Optional[bool]=…, size: Optional[SizeDict]=…, interpolation: Optional[F.Interpo… - added:
method BaseImageProcessorFast.center_crop(image: torch.Tensor, size: SizeDict, **kwargs) -> torch.Tensor - added:
method BaseImageProcessorFast.convert_to_rgb(image: ImageInput) -> ImageInput - added:
method BaseImageProcessorFast.filter_out_unused_kwargs(kwargs: dict) - added:
method BaseImageProcessorFast.normalize(image: torch.Tensor, mean: Union[float, Iterable[float]], std: Union[float, Iterable[float]], **kwargs) -> torch.Tensor - added:
method BaseImageProcessorFast.pad(images: torch.Tensor, pad_size: SizeDict=…, fill_value: Optional[int]=…, padding_mode: Optional[str]=…, return_mask: bool=…, disable_grouping: Optional[bool]=…, **kwargs) -> torch.Tensor - added:
method BaseImageProcessorFast.preprocess(images: ImageInput, *args, **kwargs: Unpack[DefaultFastImageProcessorKwargs]) -> BatchFeature - added:
method BaseImageProcessorFast.rescale(image: torch.Tensor, scale: float, **kwargs) -> torch.Tensor - added:
method BaseImageProcessorFast.rescale_and_normalize(images: torch.Tensor, do_rescale: bool, rescale_factor: float, do_normalize: bool, image_mean: Union[float, list[float]], image_std: Union[float, list[float]]) -> torch.Tensor - added:
method BaseImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method BaseImageProcessorFast.to_dict() - added:
property BaseImageProcessorFast.is_fast -> bool - added:
static method BaseImageProcessorFast.compile_friendly_resize(image: torch.Tensor, new_size: tuple[int, int], interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…) -> torch.Tensor - added:
variable logger - added:
variable pil_torch_interpolation_mapping
- added:
class NumpyToTensor - added:
class PaddingMode(ExplicitEnum) - added:
field PaddingMode.CONSTANT - added:
field PaddingMode.REFLECT - added:
field PaddingMode.REPLICATE - added:
field PaddingMode.SYMMETRIC - added:
function center_crop(image: np.ndarray, size: tuple[int, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function center_to_corners_format(bboxes_center: TensorType) -> TensorType - added:
function convert_to_rgb(image: ImageInput) -> ImageInput - added:
function corners_to_center_format(bboxes_corners: TensorType) -> TensorType - added:
function flip_channel_order(image: np.ndarray, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function get_resize_output_image_size(input_image: np.ndarray, size: Union[int, tuple[int, int], list[int], tuple[int, ...]], default_to_square: bool=…, max_size: Optional[int]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> tuple - added:
function get_size_with_aspect_ratio(image_size, size, max_size=…) -> tuple[int, int] - added:
function group_images_by_shape(images: Union[list[torch.Tensor], torch.Tensor], disable_grouping: bool, is_nested: bool=…) -> tuple[dict[tuple[int, int], list[torch.Tensor]], dict[Union[int, tuple[int, int]], tuple[tuple[int, int], int]]] - added:
function id_to_rgb(id_map) - added:
function normalize(image: np.ndarray, mean: Union[float, Collection[float]], std: Union[float, Collection[float]], data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function pad(image: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function reorder_images(processed_images: dict[tuple[int, int], torch.Tensor], grouped_images_index: dict[Union[int, tuple[int, int]], tuple[tuple[int, int], int]], is_nested: bool=…) -> Union[list[torch.Tensor], torch.Tensor] - added:
function rescale(image: np.ndarray, scale: float, data_format: Optional[ChannelDimension]=…, dtype: np.dtype=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function resize(image: np.ndarray, size: tuple[int, int], resample: Optional[PILImageResampling]=…, reducing_gap: Optional[int]=…, data_format: Optional[ChannelDimension]=…, return_numpy: bool=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function rgb_to_id(color) - added:
function to_channel_dimension_format(image: np.ndarray, channel_dim: Union[ChannelDimension, str], input_channel_dim: Optional[Union[ChannelDimension, str]]=…) -> np.ndarray - added:
function to_pil_image(image: Union[np.ndarray, PIL.Image.Image, torch.Tensor, tf.Tensor, jnp.ndarray], do_rescale: Optional[bool]=…, image_mode: Optional[str]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> PIL.Image.Image
- added:
class AnnotationFormat(ExplicitEnum) - added:
class AnnotionFormat(ExplicitEnum) - added:
class ChannelDimension(ExplicitEnum) - added:
class ImageFeatureExtractionMixin - added:
class ImageType(ExplicitEnum) - added:
class SizeDict - added:
field AnnotationFormat.COCO_DETECTION - added:
field AnnotationFormat.COCO_PANOPTIC - added:
field AnnotionFormat.COCO_DETECTION - added:
field AnnotionFormat.COCO_PANOPTIC - added:
field ChannelDimension.FIRST - added:
field ChannelDimension.LAST - added:
field ImageType.JAX - added:
field ImageType.NUMPY - added:
field ImageType.PIL - added:
field ImageType.TENSORFLOW - added:
field ImageType.TORCH - added:
field SizeDict.height: Optional[int] - added:
field SizeDict.longest_edge: Optional[int] - added:
field SizeDict.max_height: Optional[int] - added:
field SizeDict.max_width: Optional[int] - added:
field SizeDict.shortest_edge: Optional[int] - added:
field SizeDict.width: Optional[int] - added:
function concatenate_list(input_list) - added:
function get_channel_dimension_axis(image: np.ndarray, input_data_format: Optional[Union[ChannelDimension, str]]=…) -> int - added:
function get_image_size(image: np.ndarray, channel_dim: Optional[ChannelDimension]=…) -> tuple[int, int] - added:
function get_image_size_for_max_height_width(image_size: tuple[int, int], max_height: int, max_width: int) -> tuple[int, int] - added:
function get_image_type(image) - added:
function infer_channel_dimension_format(image: np.ndarray, num_channels: Optional[Union[int, tuple[int, ...]]]=…) -> ChannelDimension - added:
function is_batched(img) - added:
function is_pil_image(img) - added:
function is_scaled_image(image: np.ndarray) -> bool - added:
function is_valid_annotation_coco_detection(annotation: dict[str, Union[list, tuple]]) -> bool - added:
function is_valid_annotation_coco_panoptic(annotation: dict[str, Union[list, tuple]]) -> bool - added:
function is_valid_image(img) - added:
function is_valid_list_of_images(images: list) - added:
function load_image(image: Union[str, PIL.Image.Image], timeout: Optional[float]=…) -> PIL.Image.Image - added:
function load_images(images: Union[list, tuple, str, PIL.Image.Image], timeout: Optional[float]=…) -> Union[PIL.Image.Image, list[PIL.Image.Image], list[list[PIL.Image.Image]]] - added:
function make_flat_list_of_images(images: Union[list[ImageInput], ImageInput], expected_ndims: int=…) -> ImageInput - added:
function make_list_of_images(images, expected_ndims: int=…) -> list[ImageInput] - added:
function make_nested_list_of_images(images: Union[list[ImageInput], ImageInput], expected_ndims: int=…) -> list[ImageInput] - added:
function to_numpy_array(img) -> np.ndarray - added:
function valid_coco_detection_annotations(annotations: Iterable[dict[str, Union[list, tuple]]]) -> bool - added:
function valid_coco_panoptic_annotations(annotations: Iterable[dict[str, Union[list, tuple]]]) -> bool - added:
function valid_images(imgs) - added:
function validate_annotations(annotation_format: AnnotationFormat, supported_annotation_formats: tuple[AnnotationFormat, ...], annotations: list[dict]) -> None - added:
function validate_kwargs(valid_processor_keys: list[str], captured_kwargs: list[str]) - added:
function validate_preprocess_arguments(do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: Optional[bool]=…, pad_size: Optional[Union[dict[str, int], int]]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_resize: Op… - added:
method ImageFeatureExtractionMixin.center_crop(image, size) - added:
method ImageFeatureExtractionMixin.convert_rgb(image) - added:
method ImageFeatureExtractionMixin.expand_dims(image) - added:
method ImageFeatureExtractionMixin.flip_channel_order(image) - added:
method ImageFeatureExtractionMixin.normalize(image, mean, std, rescale=…) - added:
method ImageFeatureExtractionMixin.rescale(image: np.ndarray, scale: Union[float, int]) -> np.ndarray - added:
method ImageFeatureExtractionMixin.resize(image, size, resample=…, default_to_square=…, max_size=…) - added:
method ImageFeatureExtractionMixin.rotate(image, angle, resample=…, expand=…, center=…, translate=…, fillcolor=…) - added:
method ImageFeatureExtractionMixin.to_numpy_array(image, rescale=…, channel_first=…) - added:
method ImageFeatureExtractionMixin.to_pil_image(image, rescale=…) - added:
variable AnnotationType - added:
variable ImageInput - added:
variable PILImageResampling - added:
variable logger - added:
variable pil_torch_interpolation_mapping
- removed:
class AzureMLCallback(TrainerCallback) - removed:
class CometCallback(TrainerCallback) - removed:
class MLflowCallback(TrainerCallback) - removed:
class TensorBoardCallback(TrainerCallback) - removed:
class WandbCallback(TrainerCallback) - removed:
constructor AzureMLCallback(azureml_run=…) - removed:
constructor CometCallback() - removed:
constructor MLflowCallback() - removed:
constructor TensorBoardCallback(tb_writer=…) - removed:
constructor WandbCallback() - removed:
field MLflowCallback.MAX_LOG_SIZE - removed:
function default_hp_search_backend() - removed:
function hp_params(trial) - removed:
function is_azureml_available() - removed:
function is_comet_available() - removed:
function is_mlflow_available() - removed:
function is_optuna_available() - removed:
function is_ray_available() - removed:
function is_tensorboard_available() - removed:
function is_wandb_available() - removed:
function rewrite_logs(d) - removed:
function run_hp_search_optuna(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - removed:
function run_hp_search_ray(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - removed:
method AzureMLCallback.on_init_end(args, state, control, **kwargs) - removed:
method AzureMLCallback.on_log(args, state, control, logs=…, **kwargs) - removed:
method CometCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - removed:
method CometCallback.on_train_begin(args, state, control, model=…, **kwargs) - removed:
method CometCallback.setup(args, state, model) - removed:
method MLflowCallback.on_log(args, state, control, logs, model=…, **kwargs) - removed:
method MLflowCallback.on_train_begin(args, state, control, model=…, **kwargs) - removed:
method MLflowCallback.on_train_end(args, state, control, **kwargs) - removed:
method MLflowCallback.setup(args, state, model) - removed:
method TensorBoardCallback.on_log(args, state, control, logs=…, **kwargs) - removed:
method TensorBoardCallback.on_train_begin(args, state, control, **kwargs) - removed:
method TensorBoardCallback.on_train_end(args, state, control, **kwargs) - removed:
method WandbCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - removed:
method WandbCallback.on_train_begin(args, state, control, model=…, **kwargs) - removed:
method WandbCallback.setup(args, state, model, reinit, **kwargs) - removed:
variable logger - added:
export ALL_PARALLEL_STYLES - added:
export AzureMLCallback - added:
export BitLinear - added:
export ClearMLCallback - added:
export CodeCarbonCallback - added:
export CometCallback - added:
export DVCLiveCallback - added:
export DagsHubCallback - added:
export FP8Linear - added:
export FbgemmFp8Linear - added:
export FbgemmFp8Llama4TextExperts - added:
export FlyteCallback - added:
export GGUF_CONFIG_MAPPING - added:
export GGUF_TOKENIZER_MAPPING - added:
export HfDeepSpeedConfig - added:
export HfTrainerDeepSpeedConfig - added:
export HiggsLinear - added:
export INTEGRATION_TO_CALLBACK - added:
export LayerRepository - added:
export MLflowCallback - added:
export Mxfp4GptOssExperts - added:
export NeptuneCallback - added:
export NeptuneMissingConfiguration - added:
export OptionalDependencyNotAvailable - added:
export PeftAdapterMixin - added:
export SwanLabCallback - added:
export TensorBoardCallback - added:
export TorchExportableModuleWithStaticCache - added:
export TrackioCallback - added:
export WandbCallback - added:
export convert_and_export_with_cache - added:
export deepspeed_config - added:
export deepspeed_init - added:
export deepspeed_load_checkpoint - added:
export deepspeed_optim_sched - added:
export dequantize - added:
export dequantize_and_replace - added:
export dequantize_higgs - added:
export fuse_awq_modules - added:
export get_available_reporting_integrations - added:
export get_keys_to_not_convert - added:
export get_reporting_integration_callbacks - added:
export hp_params - added:
export is_azureml_available - added:
export is_clearml_available - added:
export is_codecarbon_available - added:
export is_comet_available - added:
export is_dagshub_available - added:
export is_deepspeed_available - added:
export is_deepspeed_zero3_enabled - added:
export is_dvclive_available - added:
export is_flyte_deck_standard_available - added:
export is_flytekit_available - added:
export is_fsdp_enabled - added:
export is_fsdp_managed_module - added:
export is_mlflow_available - added:
export is_neptune_available - added:
export is_optuna_available - added:
export is_ray_available - added:
export is_ray_tune_available - added:
export is_sigopt_available - added:
export is_swanlab_available - added:
export is_tensorboard_available - added:
export is_torch_available - added:
export is_torch_greater_or_equal - added:
export is_trackio_available - added:
export is_wandb_available - added:
export load_and_swizzle_mxfp4 - added:
export load_dequant_gguf_tensor - added:
export load_gguf - added:
export make_flex_block_causal_mask - added:
export pack_weights - added:
export post_init_awq_exllama_modules - added:
export post_init_awq_ipex_modules - added:
export prepare_for_hqq_linear - added:
export quantize_to_mxfp4 - added:
export quantize_with_higgs - added:
export register_kernel_mapping - added:
export replace_8bit_linear - added:
export replace_kernel_forward_from_hub - added:
export replace_quantization_scales - added:
export replace_with_aqlm_linear - added:
export replace_with_awq_linear - added:
export replace_with_bitnet_linear - added:
export replace_with_bnb_linear - added:
export replace_with_eetq_linear - added:
export replace_with_fbgemm_fp8_linear - added:
export replace_with_fp8_linear - added:
export replace_with_higgs_linear - added:
export replace_with_mxfp4_linear - added:
export replace_with_quanto_layers - added:
export replace_with_spqr_linear - added:
export replace_with_vptq_linear - added:
export rewrite_logs - added:
export run_hp_search_optuna - added:
export run_hp_search_ray - added:
export run_hp_search_sigopt - added:
export run_hp_search_wandb - added:
export set_hf_deepspeed_config - added:
export set_module_8bit_tensor_to_device - added:
export set_module_quantized_tensor_to_device - added:
export shard_and_distribute_module - added:
export swizzle_mxfp4 - added:
export translate_to_torch_parallel_style - added:
export unpack_weights - added:
export unset_hf_deepspeed_config - added:
export use_kernel_forward_from_hub - added:
export validate_bnb_backend_availability
- added:
function find_tied_parameters(model: nn.Module, **kwargs) - added:
function init_empty_weights(include_buffers: bool=…) - added:
function init_on_device(device: torch.device, include_buffers: bool=…) - added:
variable logger
- added:
function replace_with_aqlm_linear(model, quantization_config=…, linear_weights_not_to_quantize=…, current_key_name=…, has_been_replaced=…)
- added:
function fuse_awq_modules(model, quantization_config) - added:
function get_modules_to_fuse(model, quantization_config) - added:
function post_init_awq_exllama_modules(model, exllama_config) - added:
function post_init_awq_ipex_modules(model) - added:
function replace_quantization_scales(model, model_type) - added:
function replace_with_awq_linear(model, modules_to_not_convert=…, quantization_config=…, current_key_name=…, has_been_replaced=…) -> bool - added:
variable AWQ_FUSED_MAPPINGS - added:
variable AWQ_SCALES_MAPPINGS - added:
variable logger
- added:
class ActQuant(torch.autograd.Function) - added:
class AutoBitLinear(nn.Linear) - added:
class BitLinear(nn.Module) - added:
class WeightQuant(torch.autograd.Function) - added:
constructor AutoBitLinear(in_features: int, out_features: int, bias: bool=…, device=…, dtype=…, online_quant: bool=…, use_rms_norm: bool=…, rms_norm_eps: float=…) - added:
constructor BitLinear(in_features: int, out_features: int, bias: bool, device=…, dtype=…, use_rms_norm: bool=…, rms_norm_eps: float=…) - added:
function pack_weights(quantized_weights: torch.Tensor) -> torch.Tensor - added:
function replace_with_bitnet_linear(model, modules_to_not_convert=…, current_key_name=…, quantization_config=…, pre_quantized=…) - added:
function unpack_weights(packed: torch.Tensor, dtype: torch.dtype) -> torch.Tensor - added:
method AutoBitLinear.forward(input) - added:
method AutoBitLinear.load_hook(state_dict, prefix, *args, **kwargs) - added:
method BitLinear.activation_quant(input, num_bits=…) - added:
method BitLinear.forward(input) - added:
method BitLinear.post_quant_process(input, input_scale, weight_scale) - added:
static method ActQuant.backward(ctx, grad_output) - added:
static method ActQuant.forward(ctx, activation) - added:
static method WeightQuant.backward(ctx, grad_output) - added:
static method WeightQuant.forward(ctx, weight) - added:
variable VALUES_PER_ITEM - added:
variable logger
- added:
function dequantize_and_replace(model, modules_to_not_convert=…, quantization_config=…) - added:
function dequantize_bnb_weight(weight: torch.nn.Parameter, dtype: torch.dtype, state=…) - added:
function get_keys_to_not_convert(model) - added:
function replace_8bit_linear(*args, **kwargs) - added:
function replace_with_bnb_linear(model, modules_to_not_convert=…, current_key_name=…, quantization_config=…) - added:
function set_module_8bit_tensor_to_device(*args, **kwargs) - added:
function set_module_quantized_tensor_to_device(module, tensor_name, device, value=…, quantized_stats=…) - added:
function validate_bnb_backend_availability(raise_exception=…) - added:
variable logger
- added:
class HfDeepSpeedConfig(DeepSpeedConfig) - added:
class HfTrainerDeepSpeedConfig(HfDeepSpeedConfig) - added:
constructor HfDeepSpeedConfig(config_file_or_dict) - added:
constructor HfTrainerDeepSpeedConfig(config_file_or_dict) - added:
field HfTrainerDeepSpeedConfig.fill_only - added:
function deepspeed_config() - added:
function deepspeed_init(trainer, num_training_steps, inference=…) - added:
function deepspeed_load_checkpoint(deepspeed_engine, checkpoint_path, load_module_strict=…) - added:
function deepspeed_optim_sched(trainer, hf_deepspeed_config, args, num_training_steps, model_parameters) - added:
function is_deepspeed_available() - added:
function is_deepspeed_zero3_enabled() - added:
function set_hf_deepspeed_config(hf_deepspeed_config_obj) - added:
function unset_hf_deepspeed_config() - added:
method HfTrainerDeepSpeedConfig.dtype() - added:
method HfTrainerDeepSpeedConfig.fill_match(ds_key_long, hf_val, hf_key=…, must_match=…) - added:
method HfTrainerDeepSpeedConfig.is_auto(ds_key_long) - added:
method HfTrainerDeepSpeedConfig.trainer_config_finalize(args, model, num_training_steps) - added:
method HfTrainerDeepSpeedConfig.trainer_config_process(args, auto_find_batch_size=…) - added:
variable logger
- added:
function eager_paged_attention_forward(module: nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Optional[torch.Tensor], scaling: float, **kwargs) - added:
function repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor
- added:
function replace_with_eetq_linear(model, modules_to_not_convert=…, current_key_name=…, quantization_config=…, pre_quantized=…) - added:
variable logger
- added:
class Seq2SeqLMDecoderExportableModuleWithStaticCache(torch.nn.Module) - added:
class Seq2SeqLMEncoderExportableModule(torch.nn.Module) - added:
class Seq2SeqLMExportableModule(torch.nn.Module) - added:
class TorchExportableModuleForDecoderOnlyLM(torch.nn.Module) - added:
class TorchExportableModuleForVLM - added:
class TorchExportableModuleWithHybridCache(torch.nn.Module) - added:
class TorchExportableModuleWithStaticCache(torch.nn.Module) - added:
constructor Seq2SeqLMDecoderExportableModuleWithStaticCache(model, max_static_cache_length, batch_size) - added:
constructor Seq2SeqLMEncoderExportableModule(encoder_model) - added:
constructor Seq2SeqLMExportableModule(model, batch_size=…, max_hidden_seq_length=…, cache_implementation=…, max_cache_length=…) - added:
constructor TorchExportableModuleForDecoderOnlyLM(model: PreTrainedModel, batch_size: Optional[int]=…, max_cache_len: Optional[int]=…, device: Optional[torch.device]=…) -> None - added:
constructor TorchExportableModuleForVLM(model, max_batch_size: int=…, max_cache_len: int=…) - added:
constructor TorchExportableModuleWithHybridCache(model: PreTrainedModel, batch_size: Optional[int]=…, max_cache_len: Optional[int]=…, device: Optional[torch.device]=…) -> None - added:
constructor TorchExportableModuleWithStaticCache(model: PreTrainedModel, batch_size: Optional[int]=…, max_cache_len: Optional[int]=…, device: Optional[torch.device]=…) -> None - added:
function convert_and_export_with_cache(model: PreTrainedModel, example_input_ids: Optional[torch.Tensor]=…, example_cache_position: Optional[torch.Tensor]=…, dynamic_shapes: Optional[dict]=…, strict: Optional[bool]=…) - added:
function export_with_dynamic_cache(model: PreTrainedModel, example_input_ids: Optional[torch.Tensor]=…, example_attention_mask: Optional[torch.Tensor]=…) - added:
function register_dynamic_cache_export_support() - added:
function sdpa_mask_without_vmap(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Optional[Callable]=…, attention_mask: Optional[torch.Tensor]=…, local_size: Optional[int]=…, allow_is_causal_skip: bool=…, allow_torch_fix: bool=…, **kwargs) -> Optional[torch.Tensor] - added:
method Seq2SeqLMDecoderExportableModuleWithStaticCache.forward(decoder_input_ids, encoder_hidden_states, cache_position) - added:
method Seq2SeqLMEncoderExportableModule.forward(input_ids) - added:
method Seq2SeqLMExportableModule.export(encoder_input_ids=…, decoder_input_ids=…, encoder_hidden_states=…, cache_position=…) - added:
method Seq2SeqLMExportableModule.generate(prompt_token_ids, max_new_tokens) - added:
method TorchExportableModuleForDecoderOnlyLM.export(input_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, cache_position: Optional[torch.Tensor]=…, dynamic_shapes: Optional[dict]=…, strict: Optional[bool]=…) -> torch.export.ExportedProgram - added:
method TorchExportableModuleForDecoderOnlyLM.forward(input_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, cache_position: Optional[torch.Tensor]=…) -> torch.Tensor - added:
method TorchExportableModuleForVLM.export(**kwargs) - added:
method TorchExportableModuleForVLM.export_connector() - added:
method TorchExportableModuleForVLM.export_text_decoder() - added:
method TorchExportableModuleForVLM.export_vision_encoder() - added:
method TorchExportableModuleForVLM.forward(pixel_values, input_ids, cache_position) - added:
method TorchExportableModuleForVLM.generate(pixel_values=…, input_ids=…, max_new_tokens=…, do_sample=…, temperature=…, **kwargs) - added:
method TorchExportableModuleWithHybridCache.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, cache_position: Optional[torch.Tensor]=…) -> torch.Tensor - added:
method TorchExportableModuleWithStaticCache.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, cache_position: Optional[torch.Tensor]=…) - added:
static method TorchExportableModuleForDecoderOnlyLM.generate(exported_program: torch.export.ExportedProgram, tokenizer, prompt: str, max_new_tokens: int=…, do_sample: bool=…, temperature: float=…, top_k: int=…, top_p: float=…, device: str=…) -> str - added:
static method TorchExportableModuleWithStaticCache.generate(exported_program: torch.export.ExportedProgram, prompt_token_ids: torch.Tensor, max_new_tokens: int) -> torch.Tensor
- added:
class FbgemmFp8Linear(torch.nn.Linear) - added:
class FbgemmFp8Llama4TextExperts(nn.Module) - added:
constructor FbgemmFp8Linear(in_features, out_features, bias, weight_dtype=…) - added:
constructor FbgemmFp8Llama4TextExperts(config, dtype=…) - added:
function replace_with_fbgemm_fp8_linear(model, modules_to_not_convert=…, current_key_name=…, quantization_config=…, pre_quantized=…, config=…, tp_plan=…) - added:
method FbgemmFp8Linear.forward(x) - added:
method FbgemmFp8Llama4TextExperts.forward(hidden_states) - added:
variable logger
- added:
class FP8Linear(nn.Linear) - added:
constructor FP8Linear(in_features: int, out_features: int, bias: bool=…, dtype=…, block_size: Optional[tuple[int, int]]=…, device=…, activation_scheme=…) - added:
field FP8Linear.dtype - added:
function act_quant(x: torch.Tensor, block_size: int=…) -> tuple[torch.Tensor, torch.Tensor] - added:
function act_quant_kernel(x_ptr, y_ptr, s_ptr, BLOCK_SIZE: tl.constexpr) - added:
function replace_with_fp8_linear(model, modules_to_not_convert=…, quantization_config=…) - added:
function w8a8_block_fp8_matmul_compile(input_q: torch.Tensor, weight_q: torch.Tensor, input_scale: torch.Tensor, weight_scale: torch.Tensor, block_size: Optional[tuple[int, int]]=…, output_dtype: torch.dtype=…) -> torch.Tensor - added:
function w8a8_block_fp8_matmul_triton(A: torch.Tensor, B: torch.Tensor, As: torch.Tensor, Bs: torch.Tensor, block_size: list[int], output_dtype: torch.dtype=…) -> torch.Tensor - added:
method FP8Linear.forward(input: torch.Tensor) -> torch.Tensor - added:
variable logger
- added:
function flash_attention_forward(module: torch.nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Optional[torch.Tensor], dropout: float=…, scaling: Optional[float]=…, sliding_window: Optional[int]=…, softcap: Optional[float]=…, **kwargs) -> tuple[torch.Tensor, None] - added:
variable logger
- added:
function FLASH_ATTN_VARLEN_FUNC(*args, **kwargs) - added:
function paged_attention_forward(module: torch.nn.Module, q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, cache: PagedAttentionCache=…, cu_seq_lens_q=…, cu_seq_lens_k=…, max_seqlen_q=…, max_seqlen_k=…, implementation=…, **kwargs) -> torch.Tensor - added:
variable FLASH_ATTN_VARLEN_FUNC - added:
variable msg
- added:
class WrappedFlexAttention - added:
constructor WrappedFlexAttention(training) - added:
function compile_friendly_flex_attention(query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, training=…, **kwargs) -> Union[torch.Tensor, tuple[torch.Tensor, torch.Tensor]] - added:
function flex_attention_forward(module: torch.nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Union[torch.Tensor, BlockMask], scaling: Optional[float]=…, softcap: Optional[float]=…, head_mask: Optional[torch.Tensor]=…, s_aux: Optional[torch.Tensor]=…, **kwargs) -> tuple[torch.Tensor, Optional[torch.Tensor]] - added:
function make_flex_block_causal_mask(attention_mask_2d: torch.Tensor, attention_chunk_size: Optional[int]=…, query_length=…, key_length=…, offsets: Optional[tuple[Offset, Offset]]=…, is_causal: Optional[bool]=…) -> BlockMask - added:
function repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor - added:
variable Offset - added:
variable logger
- added:
function adapt_fp_quant_config(config: FPQuantConfig)
- added:
function is_fsdp_enabled() - added:
function is_fsdp_managed_module(module: nn.Module) -> bool
- added:
class GGUFGPTConverter(GPT2Converter) - added:
class GGUFGemmaConverter(GemmaConverter) - added:
class GGUFLlamaConverter(LlamaConverter) - added:
class GGUFPhi3Converter(LlamaConverter) - added:
class GGUFQwen2Converter(Qwen2Converter) - added:
class GGUFT5Converter(T5Converter) - added:
class GGUFTokenizerSkeleton - added:
constructor GGUFGPTConverter(tokenizer_dict) - added:
constructor GGUFGemmaConverter(tokenizer_dict) - added:
constructor GGUFLlamaConverter(tokenizer_dict) - added:
constructor GGUFPhi3Converter(tokenizer_dict) - added:
constructor GGUFQwen2Converter(tokenizer_dict) - added:
constructor GGUFT5Converter(tokenizer_dict) - added:
constructor GGUFTokenizerSkeleton(dict_) - added:
function convert_gguf_tokenizer(architecture: str, tokenizer_dict) -> tuple[Tokenizer, dict] - added:
method GGUFGPTConverter.converted() -> Tokenizer - added:
method GGUFGemmaConverter.converted() -> Tokenizer - added:
method GGUFGemmaConverter.decoder(replacement, add_prefix_space) - added:
method GGUFGemmaConverter.normalizer(proto) - added:
method GGUFGemmaConverter.vocab(proto) - added:
method GGUFLlamaConverter.converted() - added:
method GGUFLlamaConverter.decoder(replacement, add_prefix_space) - added:
method GGUFLlamaConverter.merges(proto) - added:
method GGUFLlamaConverter.tokenizer(proto) - added:
method GGUFLlamaConverter.vocab(proto) - added:
method GGUFPhi3Converter.converted() -> Tokenizer - added:
method GGUFPhi3Converter.decoder(replacement, add_prefix_space) - added:
method GGUFPhi3Converter.merges(proto) - added:
method GGUFPhi3Converter.tokenizer(proto) - added:
method GGUFPhi3Converter.vocab(proto) - added:
method GGUFQwen2Converter.converted() -> Tokenizer - added:
method GGUFT5Converter.converted() -> Tokenizer - added:
method GGUFT5Converter.normalizer(proto) - added:
method GGUFT5Converter.post_processor() - added:
method GGUFT5Converter.vocab(proto) - added:
variable GGUF_CONFIG_MAPPING - added:
variable GGUF_TOKENIZER_MAPPING - added:
variable GGUF_TO_FAST_CONVERTERS - added:
variable logger
- added:
class HiggsLinear(torch.nn.Module) - added:
constructor HiggsLinear(in_features: int, out_features: int, num_bits: int, bias=…, dtype: Optional[torch.dtype]=…, device: Optional[torch.device]=…, group_size: int=…, hadamard_size: int=…) - added:
function dequantize_higgs(model, current_key_name=…) - added:
function get_higgs_grid(p: int, n: int) -> torch.Tensor - added:
function pad_to_block(tensor, dims, had_block_size, value=…) - added:
function quantize_with_higgs(weight, bits: int=…, p: int=…, group_size: int=…, hadamard_size: int=…) - added:
function replace_with_higgs_linear(model, quantization_config=…, current_key_name=…, has_been_replaced=…, modules_to_not_convert=…) - added:
method HiggsLinear.forward(x)
- added:
function autoname_modules(model) - added:
function get_linear_tags(model) - added:
function name_to_linear_tag(name) - added:
function prepare_for_hqq_linear(model, quantization_config=…, modules_to_not_convert=…, has_been_replaced=…) - added:
variable logger
- added:
class LayerRepository - added:
constructor LayerRepository(*args, **kwargs) - added:
export LayerRepository - added:
export register_kernel_mapping - added:
export replace_kernel_forward_from_hub - added:
export use_kernel_forward_from_hub - added:
function register_kernel_mapping(*args, **kwargs) - added:
function replace_kernel_forward_from_hub(*args, **kwargs) - added:
function use_kernel_forward_from_hub(*args, **kwargs)
- added:
class AzureMLCallback(TrainerCallback) - added:
class ClearMLCallback(TrainerCallback) - added:
class CodeCarbonCallback(TrainerCallback) - added:
class CometCallback(TrainerCallback) - added:
class DVCLiveCallback(TrainerCallback) - added:
class DagsHubCallback(MLflowCallback) - added:
class FlyteCallback(TrainerCallback) - added:
class MLflowCallback(TrainerCallback) - added:
class NeptuneCallback(TrainerCallback) - added:
class NeptuneMissingConfiguration(Exception) - added:
class SwanLabCallback(TrainerCallback) - added:
class TensorBoardCallback(TrainerCallback) - added:
class TrackioCallback(TrainerCallback) - added:
class WandbCallback(TrainerCallback) - added:
class WandbLogModel(str, Enum) - added:
constructor AzureMLCallback(azureml_run=…) - added:
constructor ClearMLCallback() - added:
constructor CodeCarbonCallback() - added:
constructor CometCallback() - added:
constructor DVCLiveCallback(live: Optional[Any]=…, log_model: Optional[Union[Literal['all'], bool]]=…, **kwargs) - added:
constructor DagsHubCallback() - added:
constructor FlyteCallback(save_log_history: bool=…, sync_checkpoints: bool=…) - added:
constructor MLflowCallback() - added:
constructor NeptuneCallback(*, api_token: Optional[str]=…, project: Optional[str]=…, name: Optional[str]=…, base_namespace: str=…, run=…, log_parameters: bool=…, log_checkpoints: Optional[str]=…, **neptune_run_kwargs) - added:
constructor NeptuneMissingConfiguration() - added:
constructor SwanLabCallback() - added:
constructor TensorBoardCallback(tb_writer=…) - added:
constructor TrackioCallback() - added:
constructor WandbCallback() - added:
field ClearMLCallback.log_suffix - added:
field NeptuneCallback.flat_metrics - added:
field NeptuneCallback.integration_version_key - added:
field NeptuneCallback.model_parameters_key - added:
field NeptuneCallback.trainer_parameters_key - added:
field NeptuneCallback.trial_name_key - added:
field NeptuneCallback.trial_params_key - added:
field WandbLogModel.CHECKPOINT - added:
field WandbLogModel.END - added:
field WandbLogModel.FALSE - added:
function get_available_reporting_integrations() - added:
function get_reporting_integration_callbacks(report_to) - added:
function hp_params(trial) - added:
function is_azureml_available() - added:
function is_clearml_available() - added:
function is_codecarbon_available() - added:
function is_comet_available() - added:
function is_dagshub_available() - added:
function is_dvclive_available() - added:
function is_flyte_deck_standard_available() - added:
function is_flytekit_available() - added:
function is_mlflow_available() - added:
function is_neptune_available() - added:
function is_optuna_available() - added:
function is_ray_available() - added:
function is_ray_tune_available() - added:
function is_sigopt_available() - added:
function is_swanlab_available() - added:
function is_tensorboard_available() - added:
function is_trackio_available() - added:
function is_wandb_available() - added:
function rewrite_logs(d) - added:
function run_hp_search_optuna(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - added:
function run_hp_search_ray(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - added:
function run_hp_search_sigopt(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - added:
function run_hp_search_wandb(trainer, n_trials: int, direction: str, **kwargs) -> BestRun - added:
function save_model_architecture_to_file(model: Any, output_dir: str) - added:
method AzureMLCallback.on_init_end(args, state, control, **kwargs) - added:
method AzureMLCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method ClearMLCallback.on_log(args, state, control, model=…, processing_class=…, logs=…, **kwargs) - added:
method ClearMLCallback.on_save(args, state, control, **kwargs) - added:
method ClearMLCallback.on_train_begin(args, state, control, model=…, processing_class=…, **kwargs) - added:
method ClearMLCallback.on_train_end(args, state, control, **kwargs) - added:
method ClearMLCallback.setup(args, state, model, processing_class, **kwargs) - added:
method CodeCarbonCallback.on_init_end(args, state, control, **kwargs) - added:
method CodeCarbonCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method CodeCarbonCallback.on_train_end(args, state, control, **kwargs) - added:
method CometCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method CometCallback.on_predict(args, state, control, metrics, **kwargs) - added:
method CometCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method CometCallback.on_train_end(args, state, control, **kwargs) - added:
method CometCallback.setup(args, state, model) - added:
method DVCLiveCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method DVCLiveCallback.on_save(args, state, control, **kwargs) - added:
method DVCLiveCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method DVCLiveCallback.on_train_end(args, state, control, **kwargs) - added:
method DVCLiveCallback.setup(args, state, model) - added:
method DagsHubCallback.on_train_end(args, state, control, **kwargs) - added:
method DagsHubCallback.setup(*args, **kwargs) - added:
method FlyteCallback.on_save(args, state, control, **kwargs) - added:
method FlyteCallback.on_train_end(args, state, control, **kwargs) - added:
method MLflowCallback.on_log(args, state, control, logs, model=…, **kwargs) - added:
method MLflowCallback.on_save(args, state, control, **kwargs) - added:
method MLflowCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method MLflowCallback.on_train_end(args, state, control, **kwargs) - added:
method MLflowCallback.setup(args, state, model) - added:
method NeptuneCallback.on_evaluate(args, state, control, metrics=…, **kwargs) - added:
method NeptuneCallback.on_init_end(args, state, control, **kwargs) - added:
method NeptuneCallback.on_log(args, state, control, logs: Optional[dict[str, float]]=…, **kwargs) - added:
method NeptuneCallback.on_save(args, state, control, **kwargs) - added:
method NeptuneCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method NeptuneCallback.on_train_end(args, state, control, **kwargs) - added:
method SwanLabCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method SwanLabCallback.on_predict(args, state, control, metrics, **kwargs) - added:
method SwanLabCallback.on_save(args, state, control, **kwargs) - added:
method SwanLabCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method SwanLabCallback.on_train_end(args, state, control, model=…, processing_class=…, **kwargs) - added:
method SwanLabCallback.setup(args, state, model, **kwargs) - added:
method TensorBoardCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method TensorBoardCallback.on_train_begin(args, state, control, **kwargs) - added:
method TensorBoardCallback.on_train_end(args, state, control, **kwargs) - added:
method TrackioCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method TrackioCallback.on_predict(args, state, control, metrics, **kwargs) - added:
method TrackioCallback.on_save(args, state, control, **kwargs) - added:
method TrackioCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method TrackioCallback.on_train_end(args: TrainingArguments, state, control, model=…, processing_class=…, **kwargs) - added:
method TrackioCallback.setup(args, state, model, **kwargs) - added:
method WandbCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method WandbCallback.on_predict(args, state, control, metrics, **kwargs) - added:
method WandbCallback.on_save(args, state, control, **kwargs) - added:
method WandbCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method WandbCallback.on_train_end(args: TrainingArguments, state, control, model=…, processing_class=…, **kwargs) - added:
method WandbCallback.setup(args, state, model, **kwargs) - added:
property NeptuneCallback.run - added:
property WandbLogModel.is_enabled -> bool - added:
static method NeptuneCallback.get_run(trainer) - added:
variable INTEGRATION_TO_CALLBACK - added:
variable logger
- added:
class MistralConverter - added:
constructor MistralConverter(vocab=…, pattern=…, add_prefix_space=…, additional_special_tokens=…, **kwargs) - added:
function convert_tekken_tokenizer(tokenizer_file: str) - added:
method MistralConverter.converted() -> Tokenizer - added:
method MistralConverter.extract_vocab_merges_from_model(vocab: str) - added:
method MistralConverter.tokenizer()
- added:
class Mxfp4GptOssExperts(nn.Module) - added:
constructor Mxfp4GptOssExperts(config) - added:
function convert_moe_packed_tensors(blocks, scales, *, dtype: torch.dtype=…, rows_per_chunk: int=…) -> torch.Tensor - added:
function dequantize(module, param_name, param_value, target_device, dq_param_name, **kwargs) - added:
function load_and_swizzle_mxfp4(module, param_name, param_value, target_device, triton_kernels_hub, **kwargs) - added:
function mlp_forward(self, hidden_states) - added:
function on_device(dev) - added:
function quantize_to_mxfp4(w, triton_kernels_hub) - added:
function replace_with_mxfp4_linear(model, modules_to_not_convert=…, current_key_name=…, quantization_config=…, config=…) - added:
function routing_torch_dist(logits, n_expts_act) - added:
function should_convert_module(current_key_name, patterns) - added:
function swizzle_mxfp4(w, w_scale, triton_kernels_hub) - added:
method Mxfp4GptOssExperts.forward(hidden_states: torch.Tensor, routing_data, gather_idx, scatter_idx) -> torch.Tensor - added:
variable FP4_VALUES - added:
variable logger
- added:
function get_attn_mask_npu(device) - added:
function is_npu_fa2_top_left_aligned_causal_mask() - added:
function npu_flash_attn_func(q, k, v, dropout_p=…, softmax_scale=…, causal=…, **kwargs) - added:
function npu_flash_attn_varlen_func(q, k, v, cu_seqlens_q, cu_seqlens_k, max_seqlen_q=…, max_seqlen_k=…, dropout_p=…, softmax_scale=…, causal=…, **kwargs) - added:
variable ATTN_MASK_NPU_CACHE - added:
variable DOWN_RIGHT_ALIGNED_CAUSAL_MASK_MODE - added:
variable SPARSE_MODE - added:
variable TOP_LEFT_ALIGNED_CAUSAL_MASK_MODE
- added:
class PeftAdapterMixin - added:
method PeftAdapterMixin.active_adapters() -> list[str] - added:
method PeftAdapterMixin.add_adapter(adapter_config, adapter_name: Optional[str]=…) -> None - added:
method PeftAdapterMixin.delete_adapter(adapter_names: Union[list[str], str]) -> None - added:
method PeftAdapterMixin.disable_adapters() -> None - added:
method PeftAdapterMixin.enable_adapters() -> None - added:
method PeftAdapterMixin.get_adapter_state_dict(adapter_name: Optional[str]=…, state_dict: Optional[dict]=…) -> dict - added:
method PeftAdapterMixin.load_adapter(peft_model_id: Optional[str]=…, adapter_name: Optional[str]=…, revision: Optional[str]=…, token: Optional[str]=…, device_map: str=…, max_memory: Optional[str]=…, offload_folder: Optional[str]=…, offload_index: Optional[int]=…, peft_config: Optional[dict[str, Any]]=…, adapter_state_dict: Optional[dict[str, torch.Tensor]]=…, low_cpu_mem_usage: bool=…, is_trainab… - added:
method PeftAdapterMixin.set_adapter(adapter_name: Union[list[str], str]) -> None - added:
variable MIN_PEFT_VERSION - added:
variable VLMS - added:
variable logger
- added:
function replace_with_quanto_layers(model, quantization_config=…, modules_to_not_convert=…, current_key_name=…, has_been_replaced=…) - added:
variable logger
- added:
function repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor - added:
function sdpa_attention_forward(module: torch.nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Optional[torch.Tensor], dropout: float=…, scaling: Optional[float]=…, is_causal: Optional[bool]=…, **kwargs) -> tuple[torch.Tensor, None] - added:
function use_gqa_in_sdpa(attention_mask: Optional[torch.Tensor], key: torch.Tensor) -> bool - added:
variable logger
- added:
function repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor - added:
function sdpa_attention_paged_forward(module: torch.nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Optional[torch.Tensor], dropout: float=…, scaling: Optional[float]=…, **kwargs) -> tuple[torch.Tensor, None]
- added:
function replace_with_spqr_linear(model, quantization_config=…, modules_to_not_convert=…, current_key_name=…, has_been_replaced=…)
- added:
class ColwiseParallel(TensorParallelLayer) - added:
class GatherParallel(TensorParallelLayer) - added:
class GroupedGemmParallel(TensorParallelLayer) - added:
class IsolatedParallel(TensorParallelLayer) - added:
class PackedColwiseParallel(ColwiseParallel) - added:
class PackedRowwiseParallel(RowwiseParallel) - added:
class ParallelInterface(GeneralInterface) - added:
class ReplicateParallel(TensorParallelLayer) - added:
class RouterParallel(TensorParallelLayer) - added:
class RowwiseParallel(TensorParallelLayer) - added:
class SequenceParallel(TensorParallelLayer) - added:
class TensorParallelLayer - added:
constructor ColwiseParallel(*, input_layouts: Placement | None=…, output_layouts: Placement | None=…, use_local_output: bool=…, use_dtensor=…) - added:
constructor GatherParallel(*, input_layouts: Placement | None=…, output_layouts: Placement | None=…, use_local_output: bool=…) - added:
constructor GroupedGemmParallel() - added:
constructor ReplicateParallel(*, use_dtensor=…, use_local_output=…) - added:
constructor RouterParallel(*args, **kwargs) - added:
constructor RowwiseParallel(*, input_layouts: Placement | None=…, output_layouts: Placement | None=…, use_local_output: bool=…, use_dtensor=…) - added:
constructor SequenceParallel(*, sequence_dim: int=…, use_local_output: bool=…, use_dtensor=…) - added:
field TensorParallelLayer.use_dtensor - added:
function add_tensor_parallel_hooks_to_module(model, module, tp_plan, layer_name, current_module_plan, device_mesh, parameter_name=…) - added:
function convert_local_tensor_to_dtensor(parameter: torch.Tensor, parameter_name: str, device_mesh, tp_plan: dict[str, str]) -> DTensor - added:
function distribute_model(model, distributed_config, device_mesh, tp_size) - added:
function distribute_module(module: nn.Module, device_mesh=…, input_fn=…, output_fn=…) -> nn.Module - added:
function get_packed_weights(param, empty_param, device_mesh, rank, dim) - added:
function get_tensor_shard(param, empty_param, device_mesh, rank, dim) - added:
function initialize_tensor_parallelism(tp_plan, tp_size=…) - added:
function repack_weights(packed_parameter: torch.Tensor, sharded_dim: int, world_size: int, num_blocks: int=…) -> torch.Tensor - added:
function replace_state_dict_local_with_dtensor(state_dict: dict[str, torch.Tensor], tp_plan: dict[str, str], device_mesh) -> dict[str, torch.Tensor] - added:
function shard_and_distribute_module(model, param, empty_param, parameter_name, param_casting_dtype, is_contiguous, rank, device_mesh) - added:
function verify_tp_plan(expected_keys: list[str], tp_plan: dict[str, str] | None) - added:
method ColwiseParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method GatherParallel.prepare_module_tp(module: nn.Module, device_mesh) -> nn.Module - added:
method GroupedGemmParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method IsolatedParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method IsolatedParallel.prepare_module_tp(module: nn.Module, device_mesh) -> nn.Module - added:
method PackedColwiseParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method PackedRowwiseParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method ReplicateParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method RouterParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method RouterParallel.prepare_module_tp(module: nn.Module, device_mesh) -> nn.Module - added:
method RowwiseParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method RowwiseParallel.prepare_module_tp(module: nn.Module, device_mesh) -> nn.Module - added:
method SequenceParallel.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method TensorParallelLayer.partition_tensor(param, empty_param, param_type, param_casting_dtype, to_contiguous, rank, device_mesh) - added:
method TensorParallelLayer.prepare_module_tp(module: nn.Module, device_mesh) -> nn.Module - added:
variable ALL_PARALLEL_STYLES: ParallelInterface - added:
variable logger - added:
variable str_to_dtype
- added:
function convert_tiktoken_to_fast(encoding: Any, output_dir: str)
- added:
function tpu_spmd_dataloader(dataloader: DataLoader)
- added:
function replace_with_vptq_linear(model, quantization_config=…, modules_to_not_convert=…, current_key_name=…, has_been_replaced=…)
- added:
class KerasMetricCallback(keras.callbacks.Callback) - added:
class PushToHubCallback(keras.callbacks.Callback) - added:
constructor KerasMetricCallback(metric_fn: Callable, eval_dataset: Union[tf.data.Dataset, np.ndarray, tf.Tensor, tuple, dict], output_cols: Optional[list[str]]=…, label_cols: Optional[list[str]]=…, batch_size: Optional[int]=…, predict_with_generate: bool=…, use_xla_generation: bool=…, generate_kwargs: Optional[dict]=…) - added:
constructor PushToHubCallback(output_dir: Union[str, Path], save_strategy: Union[str, IntervalStrategy]=…, save_steps: Optional[int]=…, tokenizer: Optional[PreTrainedTokenizerBase]=…, hub_model_id: Optional[str]=…, hub_token: Optional[str]=…, checkpoint: bool=…, **model_card_args) - added:
method KerasMetricCallback.on_epoch_end(epoch, logs=…) - added:
method PushToHubCallback.on_epoch_end(epoch, logs=…) - added:
method PushToHubCallback.on_train_batch_end(batch, logs=…) - added:
method PushToHubCallback.on_train_begin(logs=…) - added:
method PushToHubCallback.on_train_end(logs=…) - added:
variable logger
- added:
export mamba_inner_fn
- added:
class MambaInnerFn(torch.autograd.Function) - added:
class SelectiveScanFn(torch.autograd.Function) - added:
function mamba_inner_fn(xz, conv1d_weight, conv1d_bias, x_proj_weight, delta_proj_weight, out_proj_weight, out_proj_bias, A, B=…, C=…, D=…, delta_bias=…, B_proj_bias=…, C_proj_bias=…, delta_softplus=…, checkpoint_lvl=…, b_rms_weight=…, c_rms_weight=…, dt_rms_weight=…, b_c_dt_rms_eps=…) - added:
function rms_norm_forward(x, weight, bias, eps=…, is_rms_norm=…) - added:
function selective_scan_fn(u, delta, A, B, C, D=…, z=…, delta_bias=…, delta_softplus=…, return_last_state=…) - added:
function selective_scan_ref(u, delta, A, B, C, D=…, z=…, delta_bias=…, delta_softplus=…, return_last_state=…) - added:
static method MambaInnerFn.backward(ctx, dout) - added:
static method MambaInnerFn.forward(ctx, xz, conv1d_weight, conv1d_bias, x_proj_weight, delta_proj_weight, out_proj_weight, out_proj_bias, A, B=…, C=…, D=…, delta_bias=…, B_proj_bias=…, C_proj_bias=…, delta_softplus=…, checkpoint_lvl=…, b_rms_weight=…, c_rms_weight=…, dt_rms_weight=…, b_c_dt_rms_eps=…) - added:
static method SelectiveScanFn.backward(ctx, dout, *args) - added:
static method SelectiveScanFn.forward(ctx, u, delta, A, B, C, D=…, z=…, delta_bias=…, delta_softplus=…, return_last_state=…) - added:
variable causal_conv1d_cuda
- added:
class DFineLoss(RTDetrLoss) - added:
constructor DFineLoss(config) - added:
function DFineForObjectDetectionLoss(logits, labels, device, pred_boxes, config, outputs_class=…, outputs_coord=…, enc_topk_logits=…, enc_topk_bboxes=…, denoising_meta_values=…, predicted_corners=…, initial_reference_points=…, **kwargs) - added:
function bbox2distance(points, bbox, max_num_bins, reg_scale, up, eps=…) - added:
function translate_gt(gt: torch.Tensor, max_num_bins: int, reg_scale: int, up: torch.Tensor) - added:
function weighting_function(max_num_bins: int, up: torch.Tensor, reg_scale: int) -> torch.Tensor - added:
method DFineLoss.get_loss(loss, outputs, targets, indices, num_boxes) - added:
method DFineLoss.loss_local(outputs, targets, indices, num_boxes, T=…) - added:
method DFineLoss.unimodal_distribution_focal_loss(pred, label, weight_right, weight_left, weight=…, reduction=…, avg_factor=…)
- added:
class DeformableDetrHungarianMatcher(HungarianMatcher) - added:
class DeformableDetrImageLoss(ImageLoss) - added:
constructor DeformableDetrImageLoss(matcher, num_classes, focal_alpha, losses) - added:
function DeformableDetrForObjectDetectionLoss(logits, labels, device, pred_boxes, config, outputs_class=…, outputs_coord=…, **kwargs) - added:
function DeformableDetrForSegmentationLoss(logits, labels, device, pred_boxes, pred_masks, config, outputs_class=…, outputs_coord=…, **kwargs) - added:
method DeformableDetrHungarianMatcher.forward(outputs, targets) - added:
method DeformableDetrImageLoss.loss_labels(outputs, targets, indices, num_boxes)
- added:
class HungarianMatcher(nn.Module) - added:
class ImageLoss(nn.Module) - added:
class NestedTensor - added:
constructor HungarianMatcher(class_cost: float=…, bbox_cost: float=…, giou_cost: float=…) - added:
constructor ImageLoss(matcher, num_classes, eos_coef, losses) - added:
constructor NestedTensor(tensors, mask: Optional[Tensor]) - added:
function ForObjectDetectionLoss(logits, labels, device, pred_boxes, config, outputs_class=…, outputs_coord=…, **kwargs) - added:
function ForSegmentationLoss(logits, labels, device, pred_boxes, pred_masks, config, outputs_class=…, outputs_coord=…, **kwargs) - added:
function box_area(boxes: Tensor) -> Tensor - added:
function box_iou(boxes1, boxes2) - added:
function dice_loss(inputs, targets, num_boxes) - added:
function generalized_box_iou(boxes1, boxes2) - added:
function nested_tensor_from_tensor_list(tensor_list: list[Tensor]) - added:
function sigmoid_focal_loss(inputs, targets, num_boxes, alpha: float=…, gamma: float=…) - added:
method HungarianMatcher.forward(outputs, targets) - added:
method ImageLoss.forward(outputs, targets) - added:
method ImageLoss.get_loss(loss, outputs, targets, indices, num_boxes) - added:
method ImageLoss.loss_boxes(outputs, targets, indices, num_boxes) - added:
method ImageLoss.loss_cardinality(outputs, targets, indices, num_boxes) - added:
method ImageLoss.loss_labels(outputs, targets, indices, num_boxes) - added:
method ImageLoss.loss_masks(outputs, targets, indices, num_boxes) - added:
method NestedTensor.decompose() - added:
method NestedTensor.to(device)
- added:
class GroundingDinoHungarianMatcher(HungarianMatcher) - added:
class GroundingDinoImageLoss(ImageLoss) - added:
constructor GroundingDinoImageLoss(matcher, focal_alpha, losses) - added:
function GroundingDinoForObjectDetectionLoss(logits, labels, device, pred_boxes, config, label_maps, text_mask, outputs_class=…, outputs_coord=…, encoder_logits=…, encoder_pred_boxes=…) - added:
function sigmoid_focal_loss(inputs: torch.Tensor, targets: torch.Tensor, num_boxes: int, alpha: float=…, gamma: float=…) - added:
method GroundingDinoHungarianMatcher.forward(outputs, targets) - added:
method GroundingDinoImageLoss.loss_labels(outputs, targets, indices, num_boxes)
- added:
class RTDetrHungarianMatcher(nn.Module) - added:
class RTDetrLoss(nn.Module) - added:
constructor RTDetrHungarianMatcher(config) - added:
constructor RTDetrLoss(config) - added:
function RTDetrForObjectDetectionLoss(logits, labels, device, pred_boxes, config, outputs_class=…, outputs_coord=…, enc_topk_logits=…, enc_topk_bboxes=…, denoising_meta_values=…, **kwargs) - added:
method RTDetrHungarianMatcher.forward(outputs, targets) - added:
method RTDetrLoss.forward(outputs, targets) - added:
method RTDetrLoss.get_loss(loss, outputs, targets, indices, num_boxes) - added:
method RTDetrLoss.loss_boxes(outputs, targets, indices, num_boxes) - added:
method RTDetrLoss.loss_cardinality(outputs, targets, indices, num_boxes) - added:
method RTDetrLoss.loss_labels(outputs, targets, indices, num_boxes, log=…) - added:
method RTDetrLoss.loss_labels_bce(outputs, targets, indices, num_boxes, log=…) - added:
method RTDetrLoss.loss_labels_focal(outputs, targets, indices, num_boxes, log=…) - added:
method RTDetrLoss.loss_labels_vfl(outputs, targets, indices, num_boxes, log=…) - added:
method RTDetrLoss.loss_masks(outputs, targets, indices, num_boxes) - added:
static method RTDetrLoss.get_cdn_matched_indices(dn_meta, targets)
- added:
function ForCausalLMLoss(logits, labels, vocab_size: int, num_items_in_batch: Optional[torch.Tensor]=…, ignore_index: int=…, shift_labels: Optional[torch.Tensor]=…, **kwargs) -> torch.Tensor - added:
function ForMaskedLMLoss(logits: torch.Tensor, labels: torch.Tensor, vocab_size: int, num_items_in_batch: Optional[torch.Tensor]=…, ignore_index: int=…, **kwargs) - added:
function ForQuestionAnsweringLoss(start_logits, end_logits, start_positions, end_positions, **kwargs) - added:
function ForSequenceClassificationLoss(labels: torch.Tensor, pooled_logits: torch.Tensor, config, **kwargs) -> torch.Tensor - added:
function ForTokenClassification(logits: torch.Tensor, labels, config, **kwargs) - added:
function fixed_cross_entropy(source: torch.Tensor, target: torch.Tensor, num_items_in_batch: Optional[torch.Tensor]=…, ignore_index: int=…, **kwargs) -> torch.Tensor - added:
variable LOSS_MAPPING
- added:
class AttentionMask(torch.Tensor) - added:
class AttentionMaskInterface(GeneralInterface) - added:
constructor AttentionMask(data) - added:
function add_offsets_to_mask_function(mask_function: Callable, q_offset: int, kv_offset: int) -> Callable - added:
function and_masks(*mask_functions: Callable) -> Callable - added:
function causal_mask_function(batch_idx: int, head_idx: int, q_idx: int, kv_idx: int) -> bool - added:
function chunked_causal_mask_function(chunk_size: int, left_padding: torch.Tensor) -> Callable - added:
function chunked_overlay(chunk_size: int, left_padding: torch.Tensor) -> Callable - added:
function create_causal_mask(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor]=…, or_mask_function: Optional[Callable]=…, and_mask_function: Optional[Callable]=…) -> Optional[Union[torch.Tensor, BlockMask]] - added:
function create_chunked_causal_mask(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor]=…, or_mask_function: Optional[Callable]=…, and_mask_function: Optional[Callable]=…) -> Optional[Union[torch.Tensor, BlockMask]] - added:
function create_masks_for_generate(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor]=…, or_mask_function: Optional[Callable]=…, and_mask_function: Optional[Callable]=…, **kwargs) - added:
function create_sliding_window_causal_mask(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor]=…, or_mask_function: Optional[Callable]=…, and_mask_function: Optional[Callable]=…) -> Optional[Union[torch.Tensor, BlockMask]] - added:
function eager_mask(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Callable=…, attention_mask: Optional[torch.Tensor]=…, dtype: torch.dtype=…, **kwargs) -> torch.Tensor - added:
function find_packed_sequence_indices(position_ids: torch.Tensor) -> torch.Tensor - added:
function flash_attention_mask(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Callable=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) - added:
function flex_attention_mask(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Callable=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) -> BlockMask - added:
function get_style(style) - added:
function or_masks(*mask_functions: Callable) -> Callable - added:
function packed_sequence_mask_function(packed_sequence_mask: torch.Tensor) -> Callable - added:
function padding_mask_function(padding_mask: torch.Tensor) -> Callable - added:
function prepare_padding_mask(attention_mask: Optional[torch.Tensor], kv_length: int, kv_offset: int, _slice: bool=…) -> Optional[torch.Tensor] - added:
function sdpa_mask_older_torch(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Callable=…, attention_mask: Optional[torch.Tensor]=…, local_size: Optional[int]=…, allow_is_causal_skip: bool=…, allow_torch_fix: bool=…, **kwargs) -> Optional[torch.Tensor] - added:
function sdpa_mask_recent_torch(batch_size: int, cache_position: torch.Tensor, kv_length: int, kv_offset: int=…, mask_function: Callable=…, attention_mask: Optional[torch.Tensor]=…, local_size: Optional[int]=…, allow_is_causal_skip: bool=…, **kwargs) -> Optional[torch.Tensor] - added:
function sliding_window_causal_mask_function(sliding_window: int) -> Callable - added:
function sliding_window_overlay(sliding_window: int) -> Callable - added:
function tensor_to_mask_visual(original_tensor: torch.Tensor, grid_size=…, style=…) -> str - added:
method AttentionMask.to_string(grid_size=…, limit=…) - added:
static method AttentionMask.from_tensor(tensor: torch.Tensor, style: Optional[str]=…) -> AttentionMask - added:
variable ALL_MASK_ATTENTION_FUNCTIONS: AttentionMaskInterface - added:
variable BLACK_SQUARE - added:
variable BlockMask - added:
variable GREEN - added:
variable GREEN_SQUARE - added:
variable GREY_SQUARE - added:
variable LAYER_PATTERN_TO_MASK_FUNCTION_MAPPING - added:
variable LOW_TRIANGLE - added:
variable RESET - added:
variable UPPER_TRIANGLE - added:
variable WHITE_SQUARE - added:
variable YELLOW - added:
variable YELLOW_SQUARE - added:
variable logger - added:
variable sdpa_mask
- added:
function is_layer_block(node) - added:
function log_model_debug_trace(debug_path: Optional[str], model) - added:
function model_addition_debugger_context(model, debug_path: Optional[str]=…, do_prune_layers: bool=…, use_repr: bool=…) - added:
function prune_intermediate_layers(node) - added:
function prune_outputs_if_children(node) - added:
variable LAYER_SUFFIX_RE - added:
variable MEMORY_ADDRESS_REGEX - added:
variable logger
- added:
class TrainingSummary - added:
field TrainingSummary.dataset: Optional[Union[str, list[str]]] - added:
field TrainingSummary.dataset_args: Optional[Union[str, list[str]]] - added:
field TrainingSummary.dataset_metadata: Optional[dict[str, Any]] - added:
field TrainingSummary.dataset_tags: Optional[Union[str, list[str]]] - added:
field TrainingSummary.eval_lines: Optional[list[str]] - added:
field TrainingSummary.eval_results: Optional[dict[str, float]] - added:
field TrainingSummary.finetuned_from: Optional[str] - added:
field TrainingSummary.hyperparameters: Optional[dict[str, Any]] - added:
field TrainingSummary.language: Optional[Union[str, list[str]]] - added:
field TrainingSummary.license: Optional[str] - added:
field TrainingSummary.model_name: str - added:
field TrainingSummary.source: Optional[str] - added:
field TrainingSummary.tags: Optional[Union[str, list[str]]] - added:
field TrainingSummary.tasks: Optional[Union[str, list[str]]] - added:
function extract_hyperparameters_from_keras(model) - added:
function extract_hyperparameters_from_trainer(trainer) - added:
function infer_metric_tags_from_eval_results(eval_results) - added:
function is_hf_dataset(dataset) - added:
function make_markdown_table(lines) - added:
function parse_keras_history(logs) - added:
function parse_log_history(log_history) - added:
method TrainingSummary.create_metadata() - added:
method TrainingSummary.create_model_index(metric_mapping) - added:
method TrainingSummary.to_model_card() - added:
static method TrainingSummary.from_keras(model, model_name, keras_history=…, language=…, license=…, tags=…, finetuned_from=…, tasks=…, dataset_tags=…, dataset=…, dataset_args=…) - added:
static method TrainingSummary.from_trainer(trainer, language=…, license=…, tags=…, model_name=…, finetuned_from=…, tasks=…, dataset_tags=…, dataset_metadata=…, dataset=…, dataset_args=…) - added:
variable AUTOGENERATED_KERAS_COMMENT - added:
variable AUTOGENERATED_TRAINER_COMMENT - added:
variable METRIC_TAGS - added:
variable TASK_MAPPING - added:
variable TASK_TAG_TO_NAME_MAPPING
- removed:
class AlbertAttention(nn.Module) - removed:
class AlbertEmbeddings(nn.Module) - removed:
class AlbertForMaskedLM(AlbertPreTrainedModel) - removed:
class AlbertForMultipleChoice(AlbertPreTrainedModel) - removed:
class AlbertForPreTraining(AlbertPreTrainedModel) - removed:
class AlbertForPreTrainingOutput(ModelOutput) - removed:
class AlbertForQuestionAnswering(AlbertPreTrainedModel) - removed:
class AlbertForSequenceClassification(AlbertPreTrainedModel) - removed:
class AlbertForTokenClassification(AlbertPreTrainedModel) - removed:
class AlbertLayer(nn.Module) - removed:
class AlbertLayerGroup(nn.Module) - removed:
class AlbertMLMHead(nn.Module) - removed:
class AlbertModel(AlbertPreTrainedModel) - removed:
class AlbertPreTrainedModel(PreTrainedModel) - removed:
class AlbertSOPHead(nn.Module) - removed:
class AlbertTransformer(nn.Module) - removed:
constructor AlbertAttention(config) - removed:
constructor AlbertEmbeddings(config) - removed:
constructor AlbertForMaskedLM(config) - removed:
constructor AlbertForMultipleChoice(config) - removed:
constructor AlbertForPreTraining(config) - removed:
constructor AlbertForQuestionAnswering(config) - removed:
constructor AlbertForSequenceClassification(config) - removed:
constructor AlbertForTokenClassification(config) - removed:
constructor AlbertLayer(config) - removed:
constructor AlbertLayerGroup(config) - removed:
constructor AlbertMLMHead(config) - removed:
constructor AlbertModel(config, add_pooling_layer=…) - removed:
constructor AlbertSOPHead(config) - removed:
constructor AlbertTransformer(config) - removed:
field AlbertForMaskedLM.authorized_unexpected_keys - removed:
field AlbertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field AlbertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field AlbertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - removed:
field AlbertForPreTrainingOutput.prediction_logits: torch.FloatTensor - removed:
field AlbertForPreTrainingOutput.sop_logits: torch.FloatTensor - removed:
field AlbertForQuestionAnswering.authorized_unexpected_keys - removed:
field AlbertForTokenClassification.authorized_unexpected_keys - removed:
field AlbertModel.base_model_prefix - removed:
field AlbertModel.config_class - removed:
field AlbertModel.load_tf_weights - removed:
field AlbertPreTrainedModel.authorized_missing_keys - removed:
field AlbertPreTrainedModel.base_model_prefix - removed:
field AlbertPreTrainedModel.config_class - removed:
function load_tf_weights_in_albert(model, config, tf_checkpoint_path) - removed:
method AlbertAttention.forward(input_ids, attention_mask=…, head_mask=…, output_attentions=…) - removed:
method AlbertAttention.prune_heads(heads) - removed:
method AlbertAttention.transpose_for_scores(x) - removed:
method AlbertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method AlbertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method AlbertForMaskedLM.get_input_embeddings() - removed:
method AlbertForMaskedLM.get_output_embeddings() - removed:
method AlbertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method AlbertForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, sentence_order_label=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method AlbertForPreTraining.get_input_embeddings() - removed:
method AlbertForPreTraining.get_output_embeddings() - removed:
method AlbertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method AlbertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method AlbertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method AlbertLayer.ff_chunk(attention_output) - removed:
method AlbertLayer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…) - removed:
method AlbertLayerGroup.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…) - removed:
method AlbertMLMHead.forward(hidden_states) - removed:
method AlbertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method AlbertModel.get_input_embeddings() - removed:
method AlbertModel.set_input_embeddings(value) - removed:
method AlbertSOPHead.forward(pooled_output) - removed:
method AlbertTransformer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
variable ALBERT_INPUTS_DOCSTRING - removed:
variable ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable ALBERT_START_DOCSTRING - removed:
variable logger
- added:
class AttentionMaskConverter - added:
constructor AttentionMaskConverter(is_causal: bool, sliding_window: Optional[int]=…) - added:
field AttentionMaskConverter.is_causal: bool - added:
field AttentionMaskConverter.sliding_window: int - added:
method AttentionMaskConverter.to_4d(attention_mask_2d: torch.Tensor, query_length: int, dtype: torch.dtype, key_value_length: Optional[int]=…) -> torch.Tensor - added:
method AttentionMaskConverter.to_causal_4d(batch_size: int, query_length: int, key_value_length: int, dtype: torch.dtype, device: Union[torch.device, str]=…) -> Optional[torch.Tensor]
- removed:
class AutoModel - removed:
class AutoModelForCausalLM - removed:
class AutoModelForMaskedLM - removed:
class AutoModelForMultipleChoice - removed:
class AutoModelForNextSentencePrediction - removed:
class AutoModelForPreTraining - removed:
class AutoModelForQuestionAnswering - removed:
class AutoModelForSeq2SeqLM - removed:
class AutoModelForSequenceClassification - removed:
class AutoModelForTokenClassification - removed:
class AutoModelWithLMHead - removed:
constructor AutoModel() - removed:
constructor AutoModelForCausalLM() - removed:
constructor AutoModelForMaskedLM() - removed:
constructor AutoModelForMultipleChoice() - removed:
constructor AutoModelForNextSentencePrediction() - removed:
constructor AutoModelForPreTraining() - removed:
constructor AutoModelForQuestionAnswering() - removed:
constructor AutoModelForSeq2SeqLM() - removed:
constructor AutoModelForSequenceClassification() - removed:
constructor AutoModelForTokenClassification() - removed:
constructor AutoModelWithLMHead() - removed:
static method AutoModel.from_config(config) - removed:
static method AutoModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForCausalLM.from_config(config) - removed:
static method AutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForMaskedLM.from_config(config) - removed:
static method AutoModelForMaskedLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForMultipleChoice.from_config(config) - removed:
static method AutoModelForMultipleChoice.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForNextSentencePrediction.from_config(config) - removed:
static method AutoModelForNextSentencePrediction.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForPreTraining.from_config(config) - removed:
static method AutoModelForPreTraining.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForQuestionAnswering.from_config(config) - removed:
static method AutoModelForQuestionAnswering.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForSeq2SeqLM.from_config(config) - removed:
static method AutoModelForSeq2SeqLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForSequenceClassification.from_config(config) - removed:
static method AutoModelForSequenceClassification.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelForTokenClassification.from_config(config) - removed:
static method AutoModelForTokenClassification.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method AutoModelWithLMHead.from_config(config) - removed:
static method AutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
variable AUTO_MODEL_PRETRAINED_DOCSTRING - removed:
variable MODEL_FOR_CAUSAL_LM_MAPPING - removed:
variable MODEL_FOR_MASKED_LM_MAPPING - removed:
variable MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
variable MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - removed:
variable MODEL_FOR_PRETRAINING_MAPPING - removed:
variable MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
variable MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
variable MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
variable MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
variable MODEL_MAPPING - removed:
variable MODEL_WITH_LM_HEAD_MAPPING - removed:
variable logger
- removed:
class Attention(nn.Module) - removed:
class BartClassificationHead(nn.Module) - removed:
class BartDecoder(nn.Module) - removed:
class BartEncoder(nn.Module) - removed:
class BartForConditionalGeneration(PretrainedBartModel) - removed:
class BartForQuestionAnswering(PretrainedBartModel) - removed:
class BartForSequenceClassification(PretrainedBartModel) - removed:
class BartModel(PretrainedBartModel) - removed:
class DecoderLayer(nn.Module) - removed:
class EncoderLayer(nn.Module) - removed:
class LearnedPositionalEmbedding(nn.Embedding) - removed:
class PretrainedBartModel(PreTrainedModel) - removed:
class SinusoidalPositionalEmbedding(nn.Embedding) - removed:
constructor Attention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…) - removed:
constructor BartClassificationHead(input_dim, inner_dim, num_classes, pooler_dropout) - removed:
constructor BartDecoder(config: BartConfig, embed_tokens: nn.Embedding) - removed:
constructor BartEncoder(config: BartConfig, embed_tokens) - removed:
constructor BartForConditionalGeneration(config: BartConfig) - removed:
constructor BartForQuestionAnswering(config) - removed:
constructor BartForSequenceClassification(config: BartConfig, **kwargs) - removed:
constructor BartModel(config: BartConfig) - removed:
constructor DecoderLayer(config: BartConfig) - removed:
constructor EncoderLayer(config: BartConfig) - removed:
constructor LearnedPositionalEmbedding(num_embeddings: int, embedding_dim: int, padding_idx: int, offset) - removed:
constructor SinusoidalPositionalEmbedding(num_positions, embedding_dim, padding_idx=…) - removed:
field BartForConditionalGeneration.authorized_missing_keys - removed:
field BartForConditionalGeneration.base_model_prefix - removed:
field PretrainedBartModel.base_model_prefix - removed:
field PretrainedBartModel.config_class - removed:
function LayerNorm(normalized_shape, eps=…, elementwise_affine=…) - removed:
function fill_with_neg_inf(t) - removed:
function invert_mask(attention_mask) - removed:
function make_padding_mask(input_ids, padding_idx=…) - removed:
function shift_tokens_right(input_ids, pad_token_id) - removed:
method Attention.forward(query, key: Tensor, key_padding_mask: Optional[Tensor]=…, layer_state: Optional[Dict[str, Tensor]]=…, attn_mask: Optional[Tensor]=…, output_attentions=…) -> Tuple[Tensor, Optional[Tensor]] - removed:
method BartClassificationHead.forward(x) - removed:
method BartDecoder.forward(input_ids, encoder_hidden_states, encoder_padding_mask, decoder_padding_mask, decoder_causal_mask, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **unused) - removed:
method BartEncoder.forward(input_ids, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BartForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
method BartForConditionalGeneration.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **unused) - removed:
method BartForConditionalGeneration.get_encoder() - removed:
method BartForConditionalGeneration.get_output_embeddings() - removed:
method BartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - removed:
method BartForConditionalGeneration.resize_token_embeddings(new_num_tokens: int) -> nn.Embedding - removed:
method BartForQuestionAnswering.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, start_positions=…, end_positions=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BartForSequenceClassification.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BartModel.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs: Optional[Tuple]=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method BartModel.get_input_embeddings() - removed:
method BartModel.get_output_embeddings() - removed:
method BartModel.set_input_embeddings(value) - removed:
method DecoderLayer.forward(x, encoder_hidden_states, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…, output_attentions=…) - removed:
method EncoderLayer.forward(x, encoder_padding_mask, output_attentions=…) - removed:
method LearnedPositionalEmbedding.forward(input_ids, use_cache=…) - removed:
method SinusoidalPositionalEmbedding.forward(input_ids, use_cache=…) - removed:
property PretrainedBartModel.dummy_inputs - removed:
variable BART_GENERATION_EXAMPLE - removed:
variable BART_INPUTS_DOCSTRING - removed:
variable BART_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BART_START_DOCSTRING - removed:
variable logger
- removed:
class BertAttention(nn.Module) - removed:
class BertEmbeddings(nn.Module) - removed:
class BertEncoder(nn.Module) - removed:
class BertForMaskedLM(BertPreTrainedModel) - removed:
class BertForMultipleChoice(BertPreTrainedModel) - removed:
class BertForNextSentencePrediction(BertPreTrainedModel) - removed:
class BertForPreTraining(BertPreTrainedModel) - removed:
class BertForPreTrainingOutput(ModelOutput) - removed:
class BertForQuestionAnswering(BertPreTrainedModel) - removed:
class BertForSequenceClassification(BertPreTrainedModel) - removed:
class BertForTokenClassification(BertPreTrainedModel) - removed:
class BertIntermediate(nn.Module) - removed:
class BertLMHeadModel(BertPreTrainedModel) - removed:
class BertLMPredictionHead(nn.Module) - removed:
class BertLayer(nn.Module) - removed:
class BertModel(BertPreTrainedModel) - removed:
class BertOnlyMLMHead(nn.Module) - removed:
class BertOnlyNSPHead(nn.Module) - removed:
class BertOutput(nn.Module) - removed:
class BertPooler(nn.Module) - removed:
class BertPreTrainedModel(PreTrainedModel) - removed:
class BertPreTrainingHeads(nn.Module) - removed:
class BertPredictionHeadTransform(nn.Module) - removed:
class BertSelfAttention(nn.Module) - removed:
class BertSelfOutput(nn.Module) - removed:
constructor BertAttention(config) - removed:
constructor BertEmbeddings(config) - removed:
constructor BertEncoder(config) - removed:
constructor BertForMaskedLM(config) - removed:
constructor BertForMultipleChoice(config) - removed:
constructor BertForNextSentencePrediction(config) - removed:
constructor BertForPreTraining(config) - removed:
constructor BertForQuestionAnswering(config) - removed:
constructor BertForSequenceClassification(config) - removed:
constructor BertForTokenClassification(config) - removed:
constructor BertIntermediate(config) - removed:
constructor BertLMHeadModel(config) - removed:
constructor BertLMPredictionHead(config) - removed:
constructor BertLayer(config) - removed:
constructor BertModel(config, add_pooling_layer=…) - removed:
constructor BertOnlyMLMHead(config) - removed:
constructor BertOnlyNSPHead(config) - removed:
constructor BertOutput(config) - removed:
constructor BertPooler(config) - removed:
constructor BertPreTrainingHeads(config) - removed:
constructor BertPredictionHeadTransform(config) - removed:
constructor BertSelfAttention(config) - removed:
constructor BertSelfOutput(config) - removed:
field BertForMaskedLM.authorized_missing_keys - removed:
field BertForMaskedLM.authorized_unexpected_keys - removed:
field BertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field BertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field BertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - removed:
field BertForPreTrainingOutput.prediction_logits: torch.FloatTensor - removed:
field BertForPreTrainingOutput.seq_relationship_logits: torch.FloatTensor - removed:
field BertForQuestionAnswering.authorized_unexpected_keys - removed:
field BertForTokenClassification.authorized_unexpected_keys - removed:
field BertLMHeadModel.authorized_missing_keys - removed:
field BertLMHeadModel.authorized_unexpected_keys - removed:
field BertPreTrainedModel.authorized_missing_keys - removed:
field BertPreTrainedModel.base_model_prefix - removed:
field BertPreTrainedModel.config_class - removed:
field BertPreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_bert(model, config, tf_checkpoint_path) - removed:
method BertAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method BertAttention.prune_heads(heads) - removed:
method BertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method BertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method BertForMaskedLM.get_output_embeddings() - removed:
method BertForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - removed:
method BertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertForNextSentencePrediction.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method BertForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, next_sentence_label=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method BertForPreTraining.get_output_embeddings() - removed:
method BertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertIntermediate.forward(hidden_states) - removed:
method BertLMHeadModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertLMHeadModel.get_output_embeddings() - removed:
method BertLMHeadModel.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - removed:
method BertLMPredictionHead.forward(hidden_states) - removed:
method BertLayer.feed_forward_chunk(attention_output) - removed:
method BertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method BertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertModel.get_input_embeddings() - removed:
method BertModel.set_input_embeddings(value) - removed:
method BertOnlyMLMHead.forward(sequence_output) - removed:
method BertOnlyNSPHead.forward(pooled_output) - removed:
method BertOutput.forward(hidden_states, input_tensor) - removed:
method BertPooler.forward(hidden_states) - removed:
method BertPreTrainingHeads.forward(sequence_output, pooled_output) - removed:
method BertPredictionHeadTransform.forward(hidden_states) - removed:
method BertSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method BertSelfAttention.transpose_for_scores(x) - removed:
method BertSelfOutput.forward(hidden_states, input_tensor) - removed:
variable BERT_INPUTS_DOCSTRING - removed:
variable BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BERT_START_DOCSTRING - removed:
variable logger
- removed:
class BertGenerationDecoder(BertGenerationPreTrainedModel) - removed:
class BertGenerationEmbeddings(nn.Module) - removed:
class BertGenerationEncoder(BertGenerationPreTrainedModel) - removed:
class BertGenerationOnlyLMHead(nn.Module) - removed:
class BertGenerationPreTrainedModel(PreTrainedModel) - removed:
constructor BertGenerationDecoder(config) - removed:
constructor BertGenerationEmbeddings(config) - removed:
constructor BertGenerationEncoder(config) - removed:
constructor BertGenerationOnlyLMHead(config) - removed:
field BertGenerationPreTrainedModel.authorized_missing_keys - removed:
field BertGenerationPreTrainedModel.base_model_prefix - removed:
field BertGenerationPreTrainedModel.config_class - removed:
function load_tf_weights_in_bert_generation(model, tf_hub_path, model_class, is_encoder_named_decoder=…, is_encoder=…) - removed:
method BertGenerationDecoder.forward(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertGenerationDecoder.get_output_embeddings() - removed:
method BertGenerationDecoder.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - removed:
method BertGenerationEmbeddings.forward(input_ids=…, position_ids=…, inputs_embeds=…) - removed:
method BertGenerationEncoder.forward(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method BertGenerationEncoder.get_input_embeddings() - removed:
method BertGenerationEncoder.set_input_embeddings(value) - removed:
method BertGenerationOnlyLMHead.forward(hidden_states) - removed:
variable BERT_GENERATION_INPUTS_DOCSTRING - removed:
variable BERT_GENERATION_START_DOCSTRING - removed:
variable logger
- removed:
class BlenderbotForConditionalGeneration(BartForConditionalGeneration) - removed:
field BlenderbotForConditionalGeneration.config_class - removed:
method BlenderbotForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
variable BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BLENDER_START_DOCSTRING
- removed:
class CamembertForCausalLM(RobertaForCausalLM) - removed:
class CamembertForMaskedLM(RobertaForMaskedLM) - removed:
class CamembertForMultipleChoice(RobertaForMultipleChoice) - removed:
class CamembertForQuestionAnswering(RobertaForQuestionAnswering) - removed:
class CamembertForSequenceClassification(RobertaForSequenceClassification) - removed:
class CamembertForTokenClassification(RobertaForTokenClassification) - removed:
class CamembertModel(RobertaModel) - removed:
field CamembertForCausalLM.config_class - removed:
field CamembertForMaskedLM.config_class - removed:
field CamembertForMultipleChoice.config_class - removed:
field CamembertForQuestionAnswering.config_class - removed:
field CamembertForSequenceClassification.config_class - removed:
field CamembertForTokenClassification.config_class - removed:
field CamembertModel.config_class - removed:
variable CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable CAMEMBERT_START_DOCSTRING - removed:
variable logger
- removed:
class CTRLLMHeadModel(CTRLPreTrainedModel) - removed:
class CTRLModel(CTRLPreTrainedModel) - removed:
class CTRLPreTrainedModel(PreTrainedModel) - removed:
class EncoderLayer(torch.nn.Module) - removed:
class MultiHeadAttention(torch.nn.Module) - removed:
constructor CTRLLMHeadModel(config) - removed:
constructor CTRLModel(config) - removed:
constructor EncoderLayer(d_model_size, num_heads, dff, rate=…) - removed:
constructor MultiHeadAttention(d_model_size, num_heads) - removed:
field CTRLPreTrainedModel.base_model_prefix - removed:
field CTRLPreTrainedModel.config_class - removed:
function angle_defn(pos, i, d_model_size) - removed:
function point_wise_feed_forward_network(d_model_size, dff) - removed:
function positional_encoding(position, d_model_size, dtype) - removed:
function scaled_dot_product_attention(q, k, v, mask, attention_mask=…, head_mask=…) - removed:
method CTRLLMHeadModel.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method CTRLLMHeadModel.get_output_embeddings() - removed:
method CTRLLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, **kwargs) - removed:
method CTRLModel.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method CTRLModel.get_input_embeddings() - removed:
method CTRLModel.set_input_embeddings(new_embeddings) - removed:
method EncoderLayer.forward(x, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…, output_attentions=…) - removed:
method MultiHeadAttention.forward(v, k, q, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…, output_attentions=…) - removed:
method MultiHeadAttention.prune_heads(heads) - removed:
method MultiHeadAttention.split_into_heads(x, batch_size) - removed:
variable CTRL_INPUTS_DOCSTRING - removed:
variable CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable CTRL_START_DOCSTRING - removed:
variable logger
- removed:
class ContextPooler(nn.Module) - removed:
class DebertaAttention(nn.Module) - removed:
class DebertaEmbeddings(nn.Module) - removed:
class DebertaEncoder(nn.Module) - removed:
class DebertaForSequenceClassification(DebertaPreTrainedModel) - removed:
class DebertaIntermediate(nn.Module) - removed:
class DebertaLayer(nn.Module) - removed:
class DebertaLayerNorm(nn.Module) - removed:
class DebertaModel(DebertaPreTrainedModel) - removed:
class DebertaOutput(nn.Module) - removed:
class DebertaPreTrainedModel(PreTrainedModel) - removed:
class DebertaSelfOutput(nn.Module) - removed:
class DisentangledSelfAttention(torch.nn.Module) - removed:
class DropoutContext(object) - removed:
class StableDropout(torch.nn.Module) - removed:
class XDropout(torch.autograd.Function) - removed:
class XSoftmax(torch.autograd.Function) - removed:
constructor ContextPooler(config) - removed:
constructor DebertaAttention(config) - removed:
constructor DebertaEmbeddings(config) - removed:
constructor DebertaEncoder(config) - removed:
constructor DebertaForSequenceClassification(config) - removed:
constructor DebertaIntermediate(config) - removed:
constructor DebertaLayer(config) - removed:
constructor DebertaLayerNorm(size, eps=…) - removed:
constructor DebertaModel(config) - removed:
constructor DebertaOutput(config) - removed:
constructor DebertaSelfOutput(config) - removed:
constructor DisentangledSelfAttention(config) - removed:
constructor DropoutContext() - removed:
constructor StableDropout(drop_prob) - removed:
field DebertaPreTrainedModel.authorized_missing_keys - removed:
field DebertaPreTrainedModel.base_model_prefix - removed:
field DebertaPreTrainedModel.config_class - removed:
function build_relative_position(query_size, key_size, device) - removed:
function c2p_dynamic_expand(c2p_pos, query_layer, relative_pos) - removed:
function get_mask(input, local_context) - removed:
function p2c_dynamic_expand(c2p_pos, query_layer, key_layer) - removed:
function pos_dynamic_expand(pos_index, p2c_att, key_layer) - removed:
method ContextPooler.forward(hidden_states, mask=…) - removed:
method DebertaAttention.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - removed:
method DebertaEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, mask=…, inputs_embeds=…) - removed:
method DebertaEncoder.forward(hidden_states, attention_mask, output_hidden_states=…, output_attentions=…, query_states=…, relative_pos=…, return_dict=…) - removed:
method DebertaEncoder.get_attention_mask(attention_mask) - removed:
method DebertaEncoder.get_rel_embedding() - removed:
method DebertaEncoder.get_rel_pos(hidden_states, query_states=…, relative_pos=…) - removed:
method DebertaForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DebertaForSequenceClassification.get_input_embeddings() - removed:
method DebertaForSequenceClassification.set_input_embeddings(new_embeddings) - removed:
method DebertaIntermediate.forward(hidden_states) - removed:
method DebertaLayer.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - removed:
method DebertaLayerNorm.forward(hidden_states) - removed:
method DebertaModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DebertaModel.get_input_embeddings() - removed:
method DebertaModel.set_input_embeddings(new_embeddings) - removed:
method DebertaOutput.forward(hidden_states, input_tensor) - removed:
method DebertaSelfOutput.forward(hidden_states, input_tensor) - removed:
method DisentangledSelfAttention.disentangled_att_bias(query_layer, key_layer, relative_pos, rel_embeddings, scale_factor) - removed:
method DisentangledSelfAttention.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - removed:
method DisentangledSelfAttention.transpose_for_scores(x) - removed:
method StableDropout.clear_context() - removed:
method StableDropout.forward(x) - removed:
method StableDropout.get_context() - removed:
method StableDropout.init_context(reuse_mask=…, scale=…) - removed:
property ContextPooler.output_dim - removed:
static method XDropout.backward(ctx, grad_output) - removed:
static method XDropout.forward(ctx, input, local_ctx) - removed:
static method XSoftmax.backward(self, grad_output) - removed:
static method XSoftmax.forward(self, input, mask, dim) - removed:
variable DEBERTA_INPUTS_DOCSTRING - removed:
variable DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DEBERTA_START_DOCSTRING - removed:
variable logger
- removed:
class DistilBertForMaskedLM(DistilBertPreTrainedModel) - removed:
class DistilBertForMultipleChoice(DistilBertPreTrainedModel) - removed:
class DistilBertForQuestionAnswering(DistilBertPreTrainedModel) - removed:
class DistilBertForSequenceClassification(DistilBertPreTrainedModel) - removed:
class DistilBertForTokenClassification(DistilBertPreTrainedModel) - removed:
class DistilBertModel(DistilBertPreTrainedModel) - removed:
class DistilBertPreTrainedModel(PreTrainedModel) - removed:
class Embeddings(nn.Module) - removed:
class FFN(nn.Module) - removed:
class MultiHeadSelfAttention(nn.Module) - removed:
class Transformer(nn.Module) - removed:
class TransformerBlock(nn.Module) - removed:
constructor DistilBertForMaskedLM(config) - removed:
constructor DistilBertForMultipleChoice(config) - removed:
constructor DistilBertForQuestionAnswering(config) - removed:
constructor DistilBertForSequenceClassification(config) - removed:
constructor DistilBertForTokenClassification(config) - removed:
constructor DistilBertModel(config) - removed:
constructor Embeddings(config) - removed:
constructor FFN(config) - removed:
constructor MultiHeadSelfAttention(config) - removed:
constructor Transformer(config) - removed:
constructor TransformerBlock(config) - removed:
field DistilBertPreTrainedModel.base_model_prefix - removed:
field DistilBertPreTrainedModel.config_class - removed:
field DistilBertPreTrainedModel.load_tf_weights - removed:
function create_sinusoidal_embeddings(n_pos, dim, out) - removed:
method DistilBertForMaskedLM.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method DistilBertForMaskedLM.get_output_embeddings() - removed:
method DistilBertForMultipleChoice.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DistilBertForQuestionAnswering.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DistilBertForSequenceClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DistilBertForTokenClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DistilBertModel.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method DistilBertModel.get_input_embeddings() - removed:
method DistilBertModel.set_input_embeddings(new_embeddings) - removed:
method Embeddings.forward(input_ids) - removed:
method FFN.ff_chunk(input) - removed:
method FFN.forward(input) - removed:
method MultiHeadSelfAttention.forward(query, key, value, mask, head_mask=…, output_attentions=…) - removed:
method MultiHeadSelfAttention.prune_heads(heads) - removed:
method Transformer.forward(x, attn_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method TransformerBlock.forward(x, attn_mask=…, head_mask=…, output_attentions=…) - removed:
variable DISTILBERT_INPUTS_DOCSTRING - removed:
variable DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DISTILBERT_START_DOCSTRING - removed:
variable logger
- removed:
class DPRContextEncoder(DPRPretrainedContextEncoder) - removed:
class DPRContextEncoderOutput(ModelOutput) - removed:
class DPREncoder(PreTrainedModel) - removed:
class DPRPretrainedContextEncoder(PreTrainedModel) - removed:
class DPRPretrainedQuestionEncoder(PreTrainedModel) - removed:
class DPRPretrainedReader(PreTrainedModel) - removed:
class DPRQuestionEncoder(DPRPretrainedQuestionEncoder) - removed:
class DPRQuestionEncoderOutput(ModelOutput) - removed:
class DPRReader(DPRPretrainedReader) - removed:
class DPRReaderOutput(ModelOutput) - removed:
class DPRSpanPredictor(PreTrainedModel) - removed:
constructor DPRContextEncoder(config: DPRConfig) - removed:
constructor DPREncoder(config: DPRConfig) - removed:
constructor DPRQuestionEncoder(config: DPRConfig) - removed:
constructor DPRReader(config: DPRConfig) - removed:
constructor DPRSpanPredictor(config: DPRConfig) - removed:
field DPRContextEncoderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRContextEncoderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRContextEncoderOutput.pooler_output: torch.FloatTensor - removed:
field DPREncoder.base_model_prefix - removed:
field DPRPretrainedContextEncoder.authorized_missing_keys - removed:
field DPRPretrainedContextEncoder.base_model_prefix - removed:
field DPRPretrainedContextEncoder.config_class - removed:
field DPRPretrainedContextEncoder.load_tf_weights - removed:
field DPRPretrainedQuestionEncoder.authorized_missing_keys - removed:
field DPRPretrainedQuestionEncoder.base_model_prefix - removed:
field DPRPretrainedQuestionEncoder.config_class - removed:
field DPRPretrainedQuestionEncoder.load_tf_weights - removed:
field DPRPretrainedReader.authorized_missing_keys - removed:
field DPRPretrainedReader.base_model_prefix - removed:
field DPRPretrainedReader.config_class - removed:
field DPRPretrainedReader.load_tf_weights - removed:
field DPRQuestionEncoderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRQuestionEncoderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRQuestionEncoderOutput.pooler_output: torch.FloatTensor - removed:
field DPRReaderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRReaderOutput.end_logits: torch.FloatTensor - removed:
field DPRReaderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field DPRReaderOutput.relevance_logits: torch.FloatTensor - removed:
field DPRReaderOutput.start_logits: torch.FloatTensor - removed:
field DPRSpanPredictor.base_model_prefix - removed:
method DPRContextEncoder.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, token_type_ids: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions=…, output_hidden_states=…, return_dict=…) -> Union[DPRContextEncoderOutput, Tuple[Tensor, ...]] - removed:
method DPREncoder.forward(input_ids: Tensor, attention_mask: Optional[Tensor]=…, token_type_ids: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: bool=…, output_hidden_states: bool=…, return_dict: bool=…) -> Union[BaseModelOutputWithPooling, Tuple[Tensor, ...]] - removed:
method DPREncoder.init_weights() - removed:
method DPRPretrainedContextEncoder.init_weights() - removed:
method DPRPretrainedQuestionEncoder.init_weights() - removed:
method DPRPretrainedReader.init_weights() - removed:
method DPRQuestionEncoder.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, token_type_ids: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions=…, output_hidden_states=…, return_dict=…) -> Union[DPRQuestionEncoderOutput, Tuple[Tensor, ...]] - removed:
method DPRReader.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: bool=…, output_hidden_states: bool=…, return_dict=…) -> Union[DPRReaderOutput, Tuple[Tensor, ...]] - removed:
method DPRSpanPredictor.forward(input_ids: Tensor, attention_mask: Tensor, inputs_embeds: Optional[Tensor]=…, output_attentions: bool=…, output_hidden_states: bool=…, return_dict: bool=…) -> Union[DPRReaderOutput, Tuple[Tensor, ...]] - removed:
method DPRSpanPredictor.init_weights() - removed:
property DPREncoder.embeddings_size -> int - removed:
variable DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DPR_ENCODERS_INPUTS_DOCSTRING - removed:
variable DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DPR_READER_INPUTS_DOCSTRING - removed:
variable DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DPR_START_DOCSTRING - removed:
variable logger
- removed:
class ElectraAttention(nn.Module) - removed:
class ElectraClassificationHead(nn.Module) - removed:
class ElectraDiscriminatorPredictions(nn.Module) - removed:
class ElectraEmbeddings(nn.Module) - removed:
class ElectraEncoder(nn.Module) - removed:
class ElectraForMaskedLM(ElectraPreTrainedModel) - removed:
class ElectraForMultipleChoice(ElectraPreTrainedModel) - removed:
class ElectraForPreTraining(ElectraPreTrainedModel) - removed:
class ElectraForPreTrainingOutput(ModelOutput) - removed:
class ElectraForQuestionAnswering(ElectraPreTrainedModel) - removed:
class ElectraForSequenceClassification(ElectraPreTrainedModel) - removed:
class ElectraForTokenClassification(ElectraPreTrainedModel) - removed:
class ElectraGeneratorPredictions(nn.Module) - removed:
class ElectraIntermediate(nn.Module) - removed:
class ElectraLayer(nn.Module) - removed:
class ElectraModel(ElectraPreTrainedModel) - removed:
class ElectraOutput(nn.Module) - removed:
class ElectraPreTrainedModel(PreTrainedModel) - removed:
class ElectraSelfAttention(nn.Module) - removed:
class ElectraSelfOutput(nn.Module) - removed:
constructor ElectraAttention(config) - removed:
constructor ElectraClassificationHead(config) - removed:
constructor ElectraDiscriminatorPredictions(config) - removed:
constructor ElectraEmbeddings(config) - removed:
constructor ElectraEncoder(config) - removed:
constructor ElectraForMaskedLM(config) - removed:
constructor ElectraForMultipleChoice(config) - removed:
constructor ElectraForPreTraining(config) - removed:
constructor ElectraForQuestionAnswering(config) - removed:
constructor ElectraForSequenceClassification(config) - removed:
constructor ElectraForTokenClassification(config) - removed:
constructor ElectraGeneratorPredictions(config) - removed:
constructor ElectraIntermediate(config) - removed:
constructor ElectraLayer(config) - removed:
constructor ElectraModel(config) - removed:
constructor ElectraOutput(config) - removed:
constructor ElectraSelfAttention(config) - removed:
constructor ElectraSelfOutput(config) - removed:
field ElectraForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ElectraForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ElectraForPreTrainingOutput.logits: torch.FloatTensor - removed:
field ElectraForPreTrainingOutput.loss: Optional[torch.FloatTensor] - removed:
field ElectraForQuestionAnswering.base_model_prefix - removed:
field ElectraForQuestionAnswering.config_class - removed:
field ElectraPreTrainedModel.authorized_missing_keys - removed:
field ElectraPreTrainedModel.authorized_unexpected_keys - removed:
field ElectraPreTrainedModel.base_model_prefix - removed:
field ElectraPreTrainedModel.config_class - removed:
field ElectraPreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_electra(model, config, tf_checkpoint_path, discriminator_or_generator=…) - removed:
method ElectraAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method ElectraAttention.prune_heads(heads) - removed:
method ElectraClassificationHead.forward(features, **kwargs) - removed:
method ElectraDiscriminatorPredictions.forward(discriminator_hidden_states) - removed:
method ElectraEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method ElectraEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method ElectraForMaskedLM.get_output_embeddings() - removed:
method ElectraForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraGeneratorPredictions.forward(generator_hidden_states) - removed:
method ElectraIntermediate.forward(hidden_states) - removed:
method ElectraLayer.feed_forward_chunk(attention_output) - removed:
method ElectraLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method ElectraModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ElectraModel.get_input_embeddings() - removed:
method ElectraModel.set_input_embeddings(value) - removed:
method ElectraOutput.forward(hidden_states, input_tensor) - removed:
method ElectraSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method ElectraSelfAttention.transpose_for_scores(x) - removed:
method ElectraSelfOutput.forward(hidden_states, input_tensor) - removed:
variable ELECTRA_INPUTS_DOCSTRING - removed:
variable ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable ELECTRA_START_DOCSTRING - removed:
variable logger
- removed:
class EncoderDecoderModel(PreTrainedModel) - removed:
constructor EncoderDecoderModel(config: Optional[PretrainedConfig]=…, encoder: Optional[PreTrainedModel]=…, decoder: Optional[PreTrainedModel]=…) - removed:
field EncoderDecoderModel.base_model_prefix - removed:
field EncoderDecoderModel.config_class - removed:
method EncoderDecoderModel.forward(input_ids=…, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, inputs_embeds=…, decoder_inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method EncoderDecoderModel.get_decoder() - removed:
method EncoderDecoderModel.get_encoder() - removed:
method EncoderDecoderModel.get_input_embeddings() - removed:
method EncoderDecoderModel.get_output_embeddings() - removed:
method EncoderDecoderModel.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, encoder_outputs=…, **kwargs) - removed:
method EncoderDecoderModel.tie_weights() - removed:
static method EncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: str=…, decoder_pretrained_model_name_or_path: str=…, *model_args, **kwargs) -> PreTrainedModel - removed:
variable ENCODER_DECODER_INPUTS_DOCSTRING - removed:
variable ENCODER_DECODER_START_DOCSTRING - removed:
variable logger
- added:
class FlashAttentionKwargs(TypedDict) - added:
field FlashAttentionKwargs.cu_seq_lens_k: Optional[torch.LongTensor] - added:
field FlashAttentionKwargs.cu_seq_lens_q: Optional[torch.LongTensor] - added:
field FlashAttentionKwargs.max_length_k: Optional[int] - added:
field FlashAttentionKwargs.max_length_q: Optional[int] - added:
function fa_peft_integration_check(q: torch.Tensor, k: torch.Tensor, v: torch.Tensor, target_dtype: Optional[torch.dtype]=…) - added:
function flash_attn_supports_top_left_mask() - added:
function is_flash_attn_available() - added:
function lazy_import_flash_attention(implementation: Optional[str], force_import: Optional[bool]=…) - added:
function prepare_fa_kwargs_from_position_ids(position_ids) - added:
variable logger
- removed:
class FlaubertForMultipleChoice(XLMForMultipleChoice) - removed:
class FlaubertForQuestionAnswering(XLMForQuestionAnswering) - removed:
class FlaubertForQuestionAnsweringSimple(XLMForQuestionAnsweringSimple) - removed:
class FlaubertForSequenceClassification(XLMForSequenceClassification) - removed:
class FlaubertForTokenClassification(XLMForTokenClassification) - removed:
class FlaubertModel(XLMModel) - removed:
class FlaubertWithLMHeadModel(XLMWithLMHeadModel) - removed:
constructor FlaubertForMultipleChoice(config) - removed:
constructor FlaubertForQuestionAnswering(config) - removed:
constructor FlaubertForQuestionAnsweringSimple(config) - removed:
constructor FlaubertForSequenceClassification(config) - removed:
constructor FlaubertForTokenClassification(config) - removed:
constructor FlaubertModel(config) - removed:
constructor FlaubertWithLMHeadModel(config) - removed:
field FlaubertForMultipleChoice.config_class - removed:
field FlaubertForQuestionAnswering.config_class - removed:
field FlaubertForQuestionAnsweringSimple.config_class - removed:
field FlaubertForSequenceClassification.config_class - removed:
field FlaubertForTokenClassification.config_class - removed:
field FlaubertModel.config_class - removed:
field FlaubertWithLMHeadModel.config_class - removed:
method FlaubertModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
variable FLAUBERT_INPUTS_DOCSTRING - removed:
variable FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable FLAUBERT_START_DOCSTRING - removed:
variable logger
- removed:
class FlaxAutoModel(object) - removed:
constructor FlaxAutoModel() - removed:
static method FlaxAutoModel.from_config(config) - removed:
static method FlaxAutoModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
variable ALL_PRETRAINED_MODEL_ARCHIVE_MAP - removed:
variable MODEL_MAPPING - removed:
variable logger
- removed:
class FlaxBertAttention(nn.Module) - removed:
class FlaxBertEmbedding(nn.Module) - removed:
class FlaxBertEmbeddings(nn.Module) - removed:
class FlaxBertEncoder(nn.Module) - removed:
class FlaxBertIntermediate(nn.Module) - removed:
class FlaxBertLayer(nn.Module) - removed:
class FlaxBertLayerCollection(nn.Module) - removed:
class FlaxBertLayerNorm(nn.Module) - removed:
class FlaxBertModel(FlaxPreTrainedModel) - removed:
class FlaxBertModule(nn.Module) - removed:
class FlaxBertOutput(nn.Module) - removed:
class FlaxBertPooler(nn.Module) - removed:
constructor FlaxBertModel(config: BertConfig, state: dict, seed: int=…, **kwargs) - removed:
field FlaxBertAttention.head_size: int - removed:
field FlaxBertAttention.num_heads: int - removed:
field FlaxBertEmbedding.emb_init: Callable[..., np.ndarray] - removed:
field FlaxBertEmbedding.hidden_size: int - removed:
field FlaxBertEmbedding.vocab_size: int - removed:
field FlaxBertEmbeddings.hidden_size: int - removed:
field FlaxBertEmbeddings.max_length: int - removed:
field FlaxBertEmbeddings.type_vocab_size: int - removed:
field FlaxBertEmbeddings.vocab_size: int - removed:
field FlaxBertEncoder.head_size: int - removed:
field FlaxBertEncoder.intermediate_size: int - removed:
field FlaxBertEncoder.num_heads: int - removed:
field FlaxBertEncoder.num_layers: int - removed:
field FlaxBertIntermediate.output_size: int - removed:
field FlaxBertLayer.head_size: int - removed:
field FlaxBertLayer.intermediate_size: int - removed:
field FlaxBertLayer.num_heads: int - removed:
field FlaxBertLayerCollection.head_size: int - removed:
field FlaxBertLayerCollection.intermediate_size: int - removed:
field FlaxBertLayerCollection.num_heads: int - removed:
field FlaxBertLayerCollection.num_layers: int - removed:
field FlaxBertLayerNorm.bias: bool - removed:
field FlaxBertLayerNorm.bias_init: jnp.ndarray - removed:
field FlaxBertLayerNorm.dtype: jnp.dtype - removed:
field FlaxBertLayerNorm.epsilon: float - removed:
field FlaxBertLayerNorm.scale: bool - removed:
field FlaxBertLayerNorm.scale_init: jnp.ndarray - removed:
field FlaxBertModel.base_model_prefix - removed:
field FlaxBertModel.config_class - removed:
field FlaxBertModel.model_class - removed:
field FlaxBertModule.head_size: int - removed:
field FlaxBertModule.hidden_size: int - removed:
field FlaxBertModule.intermediate_size: int - removed:
field FlaxBertModule.max_length: int - removed:
field FlaxBertModule.num_encoder_layers: int - removed:
field FlaxBertModule.num_heads: int - removed:
field FlaxBertModule.type_vocab_size: int - removed:
field FlaxBertModule.vocab_size: int - removed:
property FlaxBertModel.module -> nn.Module - removed:
static method FlaxBertModel.convert_from_pytorch(pt_state: Dict, config: BertConfig) -> Dict - removed:
variable BERT_INPUTS_DOCSTRING - removed:
variable BERT_START_DOCSTRING - removed:
variable logger
- added:
class FlaxBaseModelOutput(ModelOutput) - added:
class FlaxBaseModelOutputWithNoAttention(ModelOutput) - added:
class FlaxBaseModelOutputWithPast(ModelOutput) - added:
class FlaxBaseModelOutputWithPastAndCrossAttentions(ModelOutput) - added:
class FlaxBaseModelOutputWithPooling(ModelOutput) - added:
class FlaxBaseModelOutputWithPoolingAndCrossAttentions(ModelOutput) - added:
class FlaxBaseModelOutputWithPoolingAndNoAttention(ModelOutput) - added:
class FlaxCausalLMOutputWithCrossAttentions(ModelOutput) - added:
class FlaxImageClassifierOutputWithNoAttention(ModelOutput) - added:
class FlaxMaskedLMOutput(ModelOutput) - added:
class FlaxMultipleChoiceModelOutput(ModelOutput) - added:
class FlaxNextSentencePredictorOutput(ModelOutput) - added:
class FlaxQuestionAnsweringModelOutput(ModelOutput) - added:
class FlaxSeq2SeqLMOutput(ModelOutput) - added:
class FlaxSeq2SeqModelOutput(ModelOutput) - added:
class FlaxSeq2SeqQuestionAnsweringModelOutput(ModelOutput) - added:
class FlaxSeq2SeqSequenceClassifierOutput(ModelOutput) - added:
class FlaxSequenceClassifierOutput(ModelOutput) - added:
class FlaxTokenClassifierOutput(ModelOutput) - added:
field FlaxBaseModelOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutput.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithNoAttention.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithNoAttention.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPast.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPast.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPast.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPast.past_key_values: Optional[dict[str, jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPastAndCrossAttentions.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPastAndCrossAttentions.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPastAndCrossAttentions.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPastAndCrossAttentions.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPastAndCrossAttentions.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxBaseModelOutputWithPooling.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPooling.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPooling.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPooling.pooler_output: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxBaseModelOutputWithPoolingAndCrossAttentions.pooler_output: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPoolingAndNoAttention.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxBaseModelOutputWithPoolingAndNoAttention.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxBaseModelOutputWithPoolingAndNoAttention.pooler_output: Optional[jnp.ndarray] - added:
field FlaxCausalLMOutputWithCrossAttentions.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxCausalLMOutputWithCrossAttentions.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxCausalLMOutputWithCrossAttentions.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxCausalLMOutputWithCrossAttentions.logits: Optional[jnp.ndarray] - added:
field FlaxCausalLMOutputWithCrossAttentions.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxImageClassifierOutputWithNoAttention.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxImageClassifierOutputWithNoAttention.logits: Optional[jnp.ndarray] - added:
field FlaxMaskedLMOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxMaskedLMOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxMaskedLMOutput.logits: Optional[jnp.ndarray] - added:
field FlaxMultipleChoiceModelOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxMultipleChoiceModelOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxMultipleChoiceModelOutput.logits: Optional[jnp.ndarray] - added:
field FlaxNextSentencePredictorOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxNextSentencePredictorOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxNextSentencePredictorOutput.logits: Optional[jnp.ndarray] - added:
field FlaxQuestionAnsweringModelOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxQuestionAnsweringModelOutput.end_logits: Optional[jnp.ndarray] - added:
field FlaxQuestionAnsweringModelOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxQuestionAnsweringModelOutput.start_logits: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqLMOutput.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqLMOutput.decoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqLMOutput.decoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqLMOutput.encoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqLMOutput.encoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqLMOutput.encoder_last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqLMOutput.logits: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqLMOutput.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxSeq2SeqModelOutput.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqModelOutput.decoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqModelOutput.decoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqModelOutput.encoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqModelOutput.encoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqModelOutput.encoder_last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqModelOutput.last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqModelOutput.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.encoder_last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.end_logits: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxSeq2SeqQuestionAnsweringModelOutput.start_logits: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqSequenceClassifierOutput.cross_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqSequenceClassifierOutput.decoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqSequenceClassifierOutput.encoder_attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSeq2SeqSequenceClassifierOutput.encoder_last_hidden_state: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqSequenceClassifierOutput.logits: Optional[jnp.ndarray] - added:
field FlaxSeq2SeqSequenceClassifierOutput.past_key_values: Optional[tuple[tuple[jnp.ndarray]]] - added:
field FlaxSequenceClassifierOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxSequenceClassifierOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxSequenceClassifierOutput.logits: Optional[jnp.ndarray] - added:
field FlaxTokenClassifierOutput.attentions: Optional[tuple[jnp.ndarray]] - added:
field FlaxTokenClassifierOutput.hidden_states: Optional[tuple[jnp.ndarray]] - added:
field FlaxTokenClassifierOutput.logits: Optional[jnp.ndarray] - added:
variable FlaxCausalLMOutput
- added:
function convert_pytorch_sharded_state_dict_to_flax(shard_filenames, flax_model) - added:
function convert_pytorch_state_dict_to_flax(pt_state_dict, flax_model) - added:
function load_flax_checkpoint_in_pytorch_model(model, flax_checkpoint_path) - added:
function load_flax_weights_in_pytorch_model(pt_model, flax_state) - added:
function load_pytorch_checkpoint_in_flax_state_dict(flax_model, pytorch_checkpoint_path, is_sharded, allow_missing_keys=…) - added:
function rename_key_and_reshape_tensor(pt_tuple_key: tuple[str], pt_tensor: np.ndarray, random_flax_state_dict: dict[str, jnp.ndarray], model_prefix: str) -> tuple[tuple[str], np.ndarray] - added:
variable logger
- removed:
class FlaxRobertaAttention(nn.Module) - removed:
class FlaxRobertaEmbedding(nn.Module) - removed:
class FlaxRobertaEmbeddings(nn.Module) - removed:
class FlaxRobertaEncoder(nn.Module) - removed:
class FlaxRobertaIntermediate(nn.Module) - removed:
class FlaxRobertaLayer(nn.Module) - removed:
class FlaxRobertaLayerCollection(nn.Module) - removed:
class FlaxRobertaLayerNorm(nn.Module) - removed:
class FlaxRobertaModel(FlaxPreTrainedModel) - removed:
class FlaxRobertaModule(nn.Module) - removed:
class FlaxRobertaOutput(nn.Module) - removed:
class FlaxRobertaPooler(nn.Module) - removed:
constructor FlaxRobertaModel(config: RobertaConfig, state: dict, seed: int=…, **kwargs) - removed:
field FlaxRobertaAttention.head_size: int - removed:
field FlaxRobertaAttention.num_heads: int - removed:
field FlaxRobertaEmbedding.emb_init: Callable[..., np.ndarray] - removed:
field FlaxRobertaEmbedding.hidden_size: int - removed:
field FlaxRobertaEmbedding.vocab_size: int - removed:
field FlaxRobertaEmbeddings.hidden_size: int - removed:
field FlaxRobertaEmbeddings.max_length: int - removed:
field FlaxRobertaEmbeddings.type_vocab_size: int - removed:
field FlaxRobertaEmbeddings.vocab_size: int - removed:
field FlaxRobertaEncoder.head_size: int - removed:
field FlaxRobertaEncoder.intermediate_size: int - removed:
field FlaxRobertaEncoder.num_heads: int - removed:
field FlaxRobertaEncoder.num_layers: int - removed:
field FlaxRobertaIntermediate.output_size: int - removed:
field FlaxRobertaLayer.head_size: int - removed:
field FlaxRobertaLayer.intermediate_size: int - removed:
field FlaxRobertaLayer.num_heads: int - removed:
field FlaxRobertaLayerCollection.head_size: int - removed:
field FlaxRobertaLayerCollection.intermediate_size: int - removed:
field FlaxRobertaLayerCollection.num_heads: int - removed:
field FlaxRobertaLayerCollection.num_layers: int - removed:
field FlaxRobertaLayerNorm.bias: bool - removed:
field FlaxRobertaLayerNorm.bias_init: jnp.ndarray - removed:
field FlaxRobertaLayerNorm.dtype: jnp.dtype - removed:
field FlaxRobertaLayerNorm.epsilon: float - removed:
field FlaxRobertaLayerNorm.scale: bool - removed:
field FlaxRobertaLayerNorm.scale_init: jnp.ndarray - removed:
field FlaxRobertaModel.base_model_prefix - removed:
field FlaxRobertaModel.config_class - removed:
field FlaxRobertaModel.model_class - removed:
field FlaxRobertaModule.head_size: int - removed:
field FlaxRobertaModule.hidden_size: int - removed:
field FlaxRobertaModule.intermediate_size: int - removed:
field FlaxRobertaModule.max_length: int - removed:
field FlaxRobertaModule.num_encoder_layers: int - removed:
field FlaxRobertaModule.num_heads: int - removed:
field FlaxRobertaModule.type_vocab_size: int - removed:
field FlaxRobertaModule.vocab_size: int - removed:
property FlaxRobertaModel.module -> nn.Module - removed:
static method FlaxRobertaModel.convert_from_pytorch(pt_state: Dict, config: RobertaConfig) -> Dict - removed:
variable ROBERTA_INPUTS_DOCSTRING - removed:
variable ROBERTA_START_DOCSTRING - removed:
variable logger
- removed:
field FlaxPreTrainedModel.model_class - removed:
field FlaxPreTrainedModel.pretrained_model_archive_map - removed:
function gelu(x) - removed:
static method FlaxPreTrainedModel.convert_from_pytorch(pt_state: Dict, config: PretrainedConfig) -> Dict - changed:
class FlaxPreTrainedModel(ABC) → class FlaxPreTrainedModel(PushToHubMixin, FlaxGenerationMixin) - changed:
constructor FlaxPreTrainedModel(config: PretrainedConfig, module: nn.Module, params: Dict, seed: int=…) → constructor FlaxPreTrainedModel(config: PretrainedConfig, module: nn.Module, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…) - changed:
method FlaxPreTrainedModel.save_pretrained(folder) → method FlaxPreTrainedModel.save_pretrained(save_directory: Union[str, os.PathLike], params=…, push_to_hub=…, max_shard_size=…, token: Optional[Union[str, bool]]=…, safe_serialization: bool=…, **kwargs) - changed:
static method FlaxPreTrainedModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) → static method FlaxPreTrainedModel.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], dtype: jnp.dtype=…, *model_args, config: Optional[Union[PretrainedConfig, str, os.PathLike]]=…, cache_dir: Optional[Union[str, os.PathLike]]=…, ignore_mismatched_sizes: bool=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) - added:
field FlaxPreTrainedModel.main_input_name - added:
function append_call_sample_docstring(model_class, checkpoint, output_type, config_class, mask=…, revision=…, real_checkpoint=…) - added:
function append_replace_return_docstrings(model_class, output_type, config_class) - added:
function flax_shard_checkpoint(params, max_shard_size=…) - added:
function overwrite_call_docstring(model_class, docstring) - added:
function quick_gelu(x) - added:
method FlaxPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> dict - added:
method FlaxPreTrainedModel.to_bf16(params: Union[dict, FrozenDict], mask: Any=…) - added:
method FlaxPreTrainedModel.to_fp16(params: Union[dict, FrozenDict], mask: Any=…) - added:
method FlaxPreTrainedModel.to_fp32(params: Union[dict, FrozenDict], mask: Any=…) - added:
property FlaxPreTrainedModel.framework -> str - added:
property FlaxPreTrainedModel.module -> nn.Module - added:
property FlaxPreTrainedModel.params -> Union[dict, FrozenDict] - added:
property FlaxPreTrainedModel.params_shape_tree -> dict - added:
property FlaxPreTrainedModel.required_params -> set - added:
static method FlaxPreTrainedModel.can_generate() -> bool - added:
static method FlaxPreTrainedModel.load_flax_sharded_weights(shard_files) - added:
static method FlaxPreTrainedModel.load_flax_weights(resolved_archive_file) - added:
static method FlaxPreTrainedModel.register_for_auto_class(auto_class=…)
- removed:
class Attention(nn.Module) - removed:
class DecoderLayer(nn.Module) - removed:
class EncoderLayer(nn.Module) - removed:
class FSMTDecoder(nn.Module) - removed:
class FSMTEncoder(nn.Module) - removed:
class FSMTForConditionalGeneration(PretrainedFSMTModel) - removed:
class FSMTModel(PretrainedFSMTModel) - removed:
class PretrainedFSMTModel(PreTrainedModel) - removed:
class SinusoidalPositionalEmbedding(nn.Embedding) - removed:
constructor Attention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…) - removed:
constructor DecoderLayer(config: FSMTConfig) - removed:
constructor EncoderLayer(config: FSMTConfig) - removed:
constructor FSMTDecoder(config: FSMTConfig, embed_tokens: nn.Embedding) - removed:
constructor FSMTEncoder(config: FSMTConfig, embed_tokens) - removed:
constructor FSMTForConditionalGeneration(config: FSMTConfig) - removed:
constructor FSMTModel(config: FSMTConfig) - removed:
constructor SinusoidalPositionalEmbedding(num_positions, embedding_dim, padding_idx) - removed:
field FSMTForConditionalGeneration.authorized_missing_keys - removed:
field FSMTForConditionalGeneration.base_model_prefix - removed:
field FSMTForConditionalGeneration.keys_to_never_save - removed:
field PretrainedFSMTModel.base_model_prefix - removed:
field PretrainedFSMTModel.config_class - removed:
function fill_with_neg_inf(t) - removed:
function invert_mask(attention_mask) - removed:
function make_padding_mask(input_ids, padding_idx=…) - removed:
function shift_tokens_right(input_ids, pad_token_id) - removed:
method Attention.forward(query, key: Optional[Tensor], key_padding_mask: Optional[Tensor]=…, layer_state: Optional[Dict[str, Optional[Tensor]]]=…, attn_mask: Optional[Tensor]=…, output_attentions=…) -> Tuple[Tensor, Optional[Tensor]] - removed:
method DecoderLayer.forward(x, encoder_hidden_states, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…, output_attentions=…) - removed:
method EncoderLayer.forward(x, encoder_padding_mask, output_attentions=…) - removed:
method FSMTDecoder.forward(input_ids, encoder_hidden_states, encoder_padding_mask, decoder_padding_mask, decoder_causal_mask, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **unused) - removed:
method FSMTEncoder.forward(input_ids, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FSMTForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
method FSMTForConditionalGeneration.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **unused) - removed:
method FSMTForConditionalGeneration.get_encoder() - removed:
method FSMTForConditionalGeneration.get_output_embeddings() - removed:
method FSMTForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - removed:
method FSMTForConditionalGeneration.resize_token_embeddings(new_num_tokens: int) -> nn.Embedding - removed:
method FSMTModel.forward(input_ids, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs: Optional[Tuple]=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method FSMTModel.get_input_embeddings() - removed:
method FSMTModel.get_output_embeddings() - removed:
method FSMTModel.set_input_embeddings(value) - removed:
method FSMTModel.set_output_embeddings(value) - removed:
method SinusoidalPositionalEmbedding.forward(input, incremental_state: Optional[Any]=…, timestep: Optional[Tensor]=…) - removed:
method SinusoidalPositionalEmbedding.make_weight(num_positions, embedding_dim, padding_idx) - removed:
property PretrainedFSMTModel.dummy_inputs - removed:
static method SinusoidalPositionalEmbedding.get_embedding(num_embeddings, embedding_dim, padding_idx) - removed:
static method SinusoidalPositionalEmbedding.make_positions(tensor, padding_idx: int) - removed:
variable FSMT_GENERATION_EXAMPLE - removed:
variable FSMT_INPUTS_DOCSTRING - removed:
variable FSMT_START_DOCSTRING - removed:
variable LayerNorm - removed:
variable have_fused_layer_norm - removed:
variable logger
- removed:
class FunnelAttentionStructure(nn.Module) - removed:
class FunnelBaseModel(FunnelPreTrainedModel) - removed:
class FunnelClassificationHead(nn.Module) - removed:
class FunnelDecoder(nn.Module) - removed:
class FunnelDiscriminatorPredictions(nn.Module) - removed:
class FunnelEmbeddings(nn.Module) - removed:
class FunnelEncoder(nn.Module) - removed:
class FunnelForMaskedLM(FunnelPreTrainedModel) - removed:
class FunnelForMultipleChoice(FunnelPreTrainedModel) - removed:
class FunnelForPreTraining(FunnelPreTrainedModel) - removed:
class FunnelForPreTrainingOutput(ModelOutput) - removed:
class FunnelForQuestionAnswering(FunnelPreTrainedModel) - removed:
class FunnelForSequenceClassification(FunnelPreTrainedModel) - removed:
class FunnelForTokenClassification(FunnelPreTrainedModel) - removed:
class FunnelLayer(nn.Module) - removed:
class FunnelModel(FunnelPreTrainedModel) - removed:
class FunnelPositionwiseFFN(nn.Module) - removed:
class FunnelPreTrainedModel(PreTrainedModel) - removed:
class FunnelRelMultiheadAttention(nn.Module) - removed:
constructor FunnelAttentionStructure(config) - removed:
constructor FunnelBaseModel(config) - removed:
constructor FunnelClassificationHead(config, n_labels) - removed:
constructor FunnelDecoder(config) - removed:
constructor FunnelDiscriminatorPredictions(config) - removed:
constructor FunnelEmbeddings(config) - removed:
constructor FunnelEncoder(config) - removed:
constructor FunnelForMaskedLM(config) - removed:
constructor FunnelForMultipleChoice(config) - removed:
constructor FunnelForPreTraining(config) - removed:
constructor FunnelForQuestionAnswering(config) - removed:
constructor FunnelForSequenceClassification(config) - removed:
constructor FunnelForTokenClassification(config) - removed:
constructor FunnelLayer(config, block_index) - removed:
constructor FunnelModel(config) - removed:
constructor FunnelPositionwiseFFN(config) - removed:
constructor FunnelRelMultiheadAttention(config, block_index) - removed:
field FunnelAttentionStructure.cls_token_type_id: int - removed:
field FunnelForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field FunnelForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field FunnelForPreTrainingOutput.logits: torch.FloatTensor - removed:
field FunnelForPreTrainingOutput.loss: Optional[torch.FloatTensor] - removed:
field FunnelPreTrainedModel.base_model_prefix - removed:
field FunnelPreTrainedModel.config_class - removed:
field FunnelPreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_funnel(model, config, tf_checkpoint_path) - removed:
function upsample(x, stride, target_len, separate_cls=…, truncate_seq=…) - removed:
method FunnelAttentionStructure.get_position_embeds(seq_len, dtype, device) - removed:
method FunnelAttentionStructure.init_attention_inputs(inputs_embeds, attention_mask=…, token_type_ids=…) - removed:
method FunnelAttentionStructure.pool_tensor(tensor, mode=…, stride=…) - removed:
method FunnelAttentionStructure.post_attention_pooling(attention_inputs) - removed:
method FunnelAttentionStructure.pre_attention_pooling(output, attention_inputs) - removed:
method FunnelAttentionStructure.relative_pos(pos, stride, pooled_pos=…, shift=…) - removed:
method FunnelAttentionStructure.stride_pool(tensor, axis) - removed:
method FunnelAttentionStructure.stride_pool_pos(pos_id, block_index) - removed:
method FunnelAttentionStructure.token_type_ids_to_mat(token_type_ids) - removed:
method FunnelBaseModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelBaseModel.get_input_embeddings() - removed:
method FunnelBaseModel.set_input_embeddings(new_embeddings) - removed:
method FunnelClassificationHead.forward(hidden) - removed:
method FunnelDecoder.forward(final_hidden, first_block_hidden, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelDiscriminatorPredictions.forward(discriminator_hidden_states) - removed:
method FunnelEmbeddings.forward(input_ids=…, inputs_embeds=…) - removed:
method FunnelEncoder.forward(inputs_embeds, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForMaskedLM.get_output_embeddings() - removed:
method FunnelForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelLayer.forward(query, key, value, attention_inputs, output_attentions=…) - removed:
method FunnelModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method FunnelModel.get_input_embeddings() - removed:
method FunnelModel.set_input_embeddings(new_embeddings) - removed:
method FunnelPositionwiseFFN.forward(hidden) - removed:
method FunnelRelMultiheadAttention.forward(query, key, value, attention_inputs, output_attentions=…) - removed:
method FunnelRelMultiheadAttention.relative_positional_attention(position_embeds, q_head, context_len, cls_mask=…) - removed:
method FunnelRelMultiheadAttention.relative_token_type_attention(token_type_mat, q_head, cls_mask=…) - removed:
variable FUNNEL_INPUTS_DOCSTRING - removed:
variable FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable FUNNEL_START_DOCSTRING - removed:
variable INF - removed:
variable logger
- added:
class BloomTensorProcessor(TensorProcessor) - added:
class GGUFTensor(NamedTuple) - added:
class GPT2TensorProcessor(TensorProcessor) - added:
class Gemma2TensorProcessor(TensorProcessor) - added:
class Lfm2TensorProcessor(TensorProcessor) - added:
class LlamaTensorProcessor(TensorProcessor) - added:
class MambaTensorProcessor(TensorProcessor) - added:
class NemotronTensorProcessor(TensorProcessor) - added:
class Qwen2MoeTensorProcessor(TensorProcessor) - added:
class T5TensorProcessor(TensorProcessor) - added:
class TensorProcessor - added:
constructor BloomTensorProcessor(config=…) - added:
constructor GPT2TensorProcessor(config=…) - added:
constructor Gemma2TensorProcessor(config=…) - added:
constructor Lfm2TensorProcessor(config=…) - added:
constructor LlamaTensorProcessor(config=…) - added:
constructor MambaTensorProcessor(config=…) - added:
constructor NemotronTensorProcessor(config=…) - added:
constructor Qwen2MoeTensorProcessor(config=…) - added:
constructor T5TensorProcessor(config=…) - added:
constructor TensorProcessor(config=…) - added:
field GGUFTensor.metadata: dict - added:
field GGUFTensor.name: str - added:
field GGUFTensor.weights: np.ndarray - added:
function get_gguf_hf_weights_map(hf_model, model_type: Optional[str]=…, num_layers: Optional[int]=…, qual_name: str=…) - added:
function load_gguf_checkpoint(gguf_checkpoint_path, return_tensors=…, model_to_load=…) - added:
function read_field(reader, field) - added:
method BloomTensorProcessor.process(weights, name, **kwargs) - added:
method GPT2TensorProcessor.process(weights, name, **kwargs) - added:
method Gemma2TensorProcessor.process(weights, name, **kwargs) - added:
method Lfm2TensorProcessor.process(weights, name, **kwargs) - added:
method LlamaTensorProcessor.process(weights, name, **kwargs) - added:
method MambaTensorProcessor.process(weights, name, **kwargs) - added:
method NemotronTensorProcessor.process(weights, name, **kwargs) - added:
method Qwen2MoeTensorProcessor.process(weights, name, **kwargs) - added:
method T5TensorProcessor.process(weights, name, **kwargs) - added:
method TensorProcessor.process(weights, name, **kwargs) - added:
variable GGUF_SUPPORTED_ARCHITECTURES - added:
variable GGUF_TO_TRANSFORMERS_MAPPING - added:
variable TENSOR_PROCESSORS - added:
variable logger
- removed:
class Attention(nn.Module) - removed:
class Block(nn.Module) - removed:
class GPT2DoubleHeadsModel(GPT2PreTrainedModel) - removed:
class GPT2DoubleHeadsModelOutput(ModelOutput) - removed:
class GPT2ForSequenceClassification(GPT2PreTrainedModel) - removed:
class GPT2LMHeadModel(GPT2PreTrainedModel) - removed:
class GPT2Model(GPT2PreTrainedModel) - removed:
class GPT2PreTrainedModel(PreTrainedModel) - removed:
class MLP(nn.Module) - removed:
constructor Attention(nx, n_ctx, config, scale=…, is_cross_attention=…) - removed:
constructor Block(n_ctx, config, scale=…) - removed:
constructor GPT2DoubleHeadsModel(config) - removed:
constructor GPT2ForSequenceClassification(config) - removed:
constructor GPT2LMHeadModel(config) - removed:
constructor GPT2Model(config) - removed:
constructor GPT2PreTrainedModel(*inputs, **kwargs) - removed:
constructor MLP(n_state, config) - removed:
field GPT2DoubleHeadsModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field GPT2DoubleHeadsModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field GPT2DoubleHeadsModelOutput.logits: torch.FloatTensor - removed:
field GPT2DoubleHeadsModelOutput.loss: Optional[torch.FloatTensor] - removed:
field GPT2DoubleHeadsModelOutput.mc_logits: torch.FloatTensor - removed:
field GPT2DoubleHeadsModelOutput.mc_loss: Optional[torch.FloatTensor] - removed:
field GPT2DoubleHeadsModelOutput.past_key_values: Optional[List[torch.FloatTensor]] - removed:
field GPT2ForSequenceClassification.authorized_missing_keys - removed:
field GPT2LMHeadModel.authorized_missing_keys - removed:
field GPT2PreTrainedModel.base_model_prefix - removed:
field GPT2PreTrainedModel.config_class - removed:
field GPT2PreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path) - removed:
method Attention.forward(hidden_states, layer_past=…, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, use_cache=…, output_attentions=…) - removed:
method Attention.merge_heads(x) - removed:
method Attention.prune_heads(heads) - removed:
method Attention.split_heads(x, k=…) - removed:
method Block.forward(hidden_states, layer_past=…, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, use_cache=…, output_attentions=…) - removed:
method GPT2DoubleHeadsModel.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, labels=…, mc_labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method GPT2DoubleHeadsModel.get_output_embeddings() - removed:
method GPT2DoubleHeadsModel.prepare_inputs_for_generation(input_ids, past=…, **kwargs) - removed:
method GPT2ForSequenceClassification.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method GPT2LMHeadModel.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method GPT2LMHeadModel.get_output_embeddings() - removed:
method GPT2LMHeadModel.prepare_inputs_for_generation(input_ids, past=…, **kwargs) - removed:
method GPT2Model.forward(input_ids=…, past_key_values=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method GPT2Model.get_input_embeddings() - removed:
method GPT2Model.set_input_embeddings(new_embeddings) - removed:
method MLP.forward(x) - removed:
variable GPT2_INPUTS_DOCSTRING - removed:
variable GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable GPT2_START_DOCSTRING - removed:
variable logger
- added:
class GenericForQuestionAnswering - added:
class GenericForSequenceClassification - added:
class GenericForTokenClassification - added:
class GradientCheckpointingLayer(nn.Module) - added:
constructor GenericForQuestionAnswering(config) - added:
constructor GenericForSequenceClassification(config) - added:
constructor GenericForTokenClassification(config) - added:
field GenericForQuestionAnswering.base_model_prefix - added:
field GenericForSequenceClassification.base_model_prefix - added:
field GenericForTokenClassification.base_model_prefix - added:
field GradientCheckpointingLayer.gradient_checkpointing - added:
method GenericForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> QuestionAnswer… - added:
method GenericForQuestionAnswering.get_input_embeddings() - added:
method GenericForQuestionAnswering.set_input_embeddings(value) - added:
method GenericForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> SequenceClassifierOutputWithPast - added:
method GenericForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> TokenClassifierOutput - added:
variable logger
- removed:
class LayoutLMAttention(nn.Module) - removed:
class LayoutLMEmbeddings(nn.Module) - removed:
class LayoutLMEncoder(nn.Module) - removed:
class LayoutLMForMaskedLM(LayoutLMPreTrainedModel) - removed:
class LayoutLMForTokenClassification(LayoutLMPreTrainedModel) - removed:
class LayoutLMIntermediate(nn.Module) - removed:
class LayoutLMLMPredictionHead(nn.Module) - removed:
class LayoutLMLayer(nn.Module) - removed:
class LayoutLMModel(LayoutLMPreTrainedModel) - removed:
class LayoutLMOnlyMLMHead(nn.Module) - removed:
class LayoutLMOutput(nn.Module) - removed:
class LayoutLMPooler(nn.Module) - removed:
class LayoutLMPreTrainedModel(PreTrainedModel) - removed:
class LayoutLMPredictionHeadTransform(nn.Module) - removed:
class LayoutLMSelfAttention(nn.Module) - removed:
class LayoutLMSelfOutput(nn.Module) - removed:
constructor LayoutLMAttention(config) - removed:
constructor LayoutLMEmbeddings(config) - removed:
constructor LayoutLMEncoder(config) - removed:
constructor LayoutLMForMaskedLM(config) - removed:
constructor LayoutLMForTokenClassification(config) - removed:
constructor LayoutLMIntermediate(config) - removed:
constructor LayoutLMLMPredictionHead(config) - removed:
constructor LayoutLMLayer(config) - removed:
constructor LayoutLMModel(config) - removed:
constructor LayoutLMOnlyMLMHead(config) - removed:
constructor LayoutLMOutput(config) - removed:
constructor LayoutLMPooler(config) - removed:
constructor LayoutLMPredictionHeadTransform(config) - removed:
constructor LayoutLMSelfAttention(config) - removed:
constructor LayoutLMSelfOutput(config) - removed:
field LayoutLMForMaskedLM.base_model_prefix - removed:
field LayoutLMForMaskedLM.config_class - removed:
field LayoutLMForMaskedLM.pretrained_model_archive_map - removed:
field LayoutLMForTokenClassification.base_model_prefix - removed:
field LayoutLMForTokenClassification.config_class - removed:
field LayoutLMForTokenClassification.pretrained_model_archive_map - removed:
field LayoutLMModel.base_model_prefix - removed:
field LayoutLMModel.config_class - removed:
field LayoutLMModel.pretrained_model_archive_map - removed:
field LayoutLMPreTrainedModel.authorized_missing_keys - removed:
field LayoutLMPreTrainedModel.base_model_prefix - removed:
field LayoutLMPreTrainedModel.config_class - removed:
method LayoutLMAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method LayoutLMAttention.prune_heads(heads) - removed:
method LayoutLMEmbeddings.forward(input_ids=…, bbox=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method LayoutLMEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LayoutLMForMaskedLM.forward(input_ids=…, bbox=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LayoutLMForMaskedLM.get_input_embeddings() - removed:
method LayoutLMForMaskedLM.get_output_embeddings() - removed:
method LayoutLMForTokenClassification.forward(input_ids=…, bbox=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LayoutLMForTokenClassification.get_input_embeddings() - removed:
method LayoutLMIntermediate.forward(hidden_states) - removed:
method LayoutLMLMPredictionHead.forward(hidden_states) - removed:
method LayoutLMLayer.feed_forward_chunk(attention_output) - removed:
method LayoutLMLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method LayoutLMModel.forward(input_ids=…, bbox=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LayoutLMModel.get_input_embeddings() - removed:
method LayoutLMModel.set_input_embeddings(value) - removed:
method LayoutLMOnlyMLMHead.forward(sequence_output) - removed:
method LayoutLMOutput.forward(hidden_states, input_tensor) - removed:
method LayoutLMPooler.forward(hidden_states) - removed:
method LayoutLMPredictionHeadTransform.forward(hidden_states) - removed:
method LayoutLMSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method LayoutLMSelfAttention.transpose_for_scores(x) - removed:
method LayoutLMSelfOutput.forward(hidden_states, input_tensor) - removed:
variable LAYOUTLM_INPUTS_DOCSTRING - removed:
variable LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable LAYOUTLM_START_DOCSTRING - removed:
variable LayoutLMLayerNorm - removed:
variable logger
- removed:
class LongformerAttention(nn.Module) - removed:
class LongformerBaseModelOutput(ModelOutput) - removed:
class LongformerBaseModelOutputWithPooling(ModelOutput) - removed:
class LongformerClassificationHead(nn.Module) - removed:
class LongformerEmbeddings(nn.Module) - removed:
class LongformerEncoder(nn.Module) - removed:
class LongformerForMaskedLM(LongformerPreTrainedModel) - removed:
class LongformerForMultipleChoice(LongformerPreTrainedModel) - removed:
class LongformerForQuestionAnswering(LongformerPreTrainedModel) - removed:
class LongformerForSequenceClassification(LongformerPreTrainedModel) - removed:
class LongformerForTokenClassification(LongformerPreTrainedModel) - removed:
class LongformerIntermediate(nn.Module) - removed:
class LongformerLMHead(nn.Module) - removed:
class LongformerLayer(nn.Module) - removed:
class LongformerModel(LongformerPreTrainedModel) - removed:
class LongformerMultipleChoiceModelOutput(ModelOutput) - removed:
class LongformerOutput(nn.Module) - removed:
class LongformerPooler(nn.Module) - removed:
class LongformerPreTrainedModel(PreTrainedModel) - removed:
class LongformerQuestionAnsweringModelOutput(ModelOutput) - removed:
class LongformerSelfAttention(nn.Module) - removed:
class LongformerSelfOutput(nn.Module) - removed:
constructor LongformerAttention(config, layer_id=…) - removed:
constructor LongformerClassificationHead(config) - removed:
constructor LongformerEmbeddings(config) - removed:
constructor LongformerEncoder(config) - removed:
constructor LongformerForMaskedLM(config) - removed:
constructor LongformerForMultipleChoice(config) - removed:
constructor LongformerForQuestionAnswering(config) - removed:
constructor LongformerForSequenceClassification(config) - removed:
constructor LongformerForTokenClassification(config) - removed:
constructor LongformerIntermediate(config) - removed:
constructor LongformerLMHead(config) - removed:
constructor LongformerLayer(config, layer_id=…) - removed:
constructor LongformerModel(config, add_pooling_layer=…) - removed:
constructor LongformerOutput(config) - removed:
constructor LongformerPooler(config) - removed:
constructor LongformerSelfAttention(config, layer_id) - removed:
constructor LongformerSelfOutput(config) - removed:
field LongformerBaseModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutput.last_hidden_state: torch.FloatTensor - removed:
field LongformerBaseModelOutputWithPooling.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutputWithPooling.global_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutputWithPooling.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerBaseModelOutputWithPooling.last_hidden_state: torch.FloatTensor - removed:
field LongformerBaseModelOutputWithPooling.pooler_output: torch.FloatTensor - removed:
field LongformerForMaskedLM.authorized_unexpected_keys - removed:
field LongformerForQuestionAnswering.authorized_unexpected_keys - removed:
field LongformerForSequenceClassification.authorized_unexpected_keys - removed:
field LongformerForTokenClassification.authorized_unexpected_keys - removed:
field LongformerMultipleChoiceModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerMultipleChoiceModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerMultipleChoiceModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerMultipleChoiceModelOutput.logits: torch.FloatTensor - removed:
field LongformerMultipleChoiceModelOutput.loss: Optional[torch.FloatTensor] - removed:
field LongformerPreTrainedModel.authorized_missing_keys - removed:
field LongformerPreTrainedModel.base_model_prefix - removed:
field LongformerPreTrainedModel.config_class - removed:
field LongformerQuestionAnsweringModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerQuestionAnsweringModelOutput.end_logits: torch.FloatTensor - removed:
field LongformerQuestionAnsweringModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LongformerQuestionAnsweringModelOutput.loss: Optional[torch.FloatTensor] - removed:
field LongformerQuestionAnsweringModelOutput.start_logits: torch.FloatTensor - removed:
function create_position_ids_from_input_ids(input_ids, padding_idx) - removed:
method LongformerAttention.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - removed:
method LongformerAttention.prune_heads(heads) - removed:
method LongformerClassificationHead.forward(hidden_states, **kwargs) - removed:
method LongformerEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - removed:
method LongformerEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method LongformerEncoder.forward(hidden_states, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerForMaskedLM.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method LongformerForMaskedLM.get_output_embeddings() - removed:
method LongformerForMultipleChoice.forward(input_ids=…, token_type_ids=…, attention_mask=…, global_attention_mask=…, labels=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerForQuestionAnswering.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerForSequenceClassification.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerForTokenClassification.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerIntermediate.forward(hidden_states) - removed:
method LongformerLMHead.forward(features, **kwargs) - removed:
method LongformerLayer.ff_chunk(attn_output) - removed:
method LongformerLayer.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - removed:
method LongformerModel.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LongformerModel.get_input_embeddings() - removed:
method LongformerModel.set_input_embeddings(value) - removed:
method LongformerOutput.forward(hidden_states, input_tensor) - removed:
method LongformerPooler.forward(hidden_states) - removed:
method LongformerSelfAttention.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - removed:
method LongformerSelfOutput.forward(hidden_states, input_tensor) - removed:
variable LONGFORMER_INPUTS_DOCSTRING - removed:
variable LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable LONGFORMER_START_DOCSTRING - removed:
variable logger
- removed:
class GeLU(nn.Module) - removed:
class LxmertAttention(nn.Module) - removed:
class LxmertAttentionOutput(nn.Module) - removed:
class LxmertCrossAttentionLayer(nn.Module) - removed:
class LxmertEmbeddings(nn.Module) - removed:
class LxmertEncoder(nn.Module) - removed:
class LxmertForPreTraining(LxmertPreTrainedModel) - removed:
class LxmertForPreTrainingOutput(ModelOutput) - removed:
class LxmertForQuestionAnswering(LxmertPreTrainedModel) - removed:
class LxmertForQuestionAnsweringOutput(ModelOutput) - removed:
class LxmertIntermediate(nn.Module) - removed:
class LxmertLMPredictionHead(nn.Module) - removed:
class LxmertLayer(nn.Module) - removed:
class LxmertModel(LxmertPreTrainedModel) - removed:
class LxmertModelOutput(ModelOutput) - removed:
class LxmertOutput(nn.Module) - removed:
class LxmertPooler(nn.Module) - removed:
class LxmertPreTrainedModel(PreTrainedModel) - removed:
class LxmertPreTrainingHeads(nn.Module) - removed:
class LxmertPredictionHeadTransform(nn.Module) - removed:
class LxmertSelfAttentionLayer(nn.Module) - removed:
class LxmertVisualAnswerHead(nn.Module) - removed:
class LxmertVisualFeatureEncoder(nn.Module) - removed:
class LxmertVisualObjHead(nn.Module) - removed:
class LxmertXLayer(nn.Module) - removed:
constructor GeLU() - removed:
constructor LxmertAttention(config, ctx_dim=…) - removed:
constructor LxmertAttentionOutput(config) - removed:
constructor LxmertCrossAttentionLayer(config) - removed:
constructor LxmertEmbeddings(config) - removed:
constructor LxmertEncoder(config) - removed:
constructor LxmertForPreTraining(config) - removed:
constructor LxmertForQuestionAnswering(config) - removed:
constructor LxmertIntermediate(config) - removed:
constructor LxmertLMPredictionHead(config, lxmert_model_embedding_weights) - removed:
constructor LxmertLayer(config) - removed:
constructor LxmertModel(config) - removed:
constructor LxmertOutput(config) - removed:
constructor LxmertPooler(config) - removed:
constructor LxmertPreTrainingHeads(config, lxmert_model_embedding_weights) - removed:
constructor LxmertPredictionHeadTransform(config) - removed:
constructor LxmertSelfAttentionLayer(config) - removed:
constructor LxmertVisualAnswerHead(config, num_labels) - removed:
constructor LxmertVisualFeatureEncoder(config) - removed:
constructor LxmertVisualObjHead(config) - removed:
constructor LxmertXLayer(config) - removed:
field LxmertForPreTrainingOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForPreTrainingOutput.cross_relationship_score: Optional[torch.FloatTensor] - removed:
field LxmertForPreTrainingOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForPreTrainingOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForPreTrainingOutput.loss: [torch.FloatTensor] - removed:
field LxmertForPreTrainingOutput.prediction_logits: Optional[torch.FloatTensor] - removed:
field LxmertForPreTrainingOutput.question_answering_score: Optional[torch.FloatTensor] - removed:
field LxmertForPreTrainingOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForPreTrainingOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForQuestionAnsweringOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForQuestionAnsweringOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForQuestionAnsweringOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - removed:
field LxmertForQuestionAnsweringOutput.question_answering_score: Optional[torch.FloatTensor] - removed:
field LxmertForQuestionAnsweringOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertForQuestionAnsweringOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.language_output: Optional[torch.FloatTensor] - removed:
field LxmertModelOutput.pooled_output: Optional[torch.FloatTensor] - removed:
field LxmertModelOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field LxmertModelOutput.vision_output: Optional[torch.FloatTensor] - removed:
field LxmertPreTrainedModel.base_model_prefix - removed:
field LxmertPreTrainedModel.config_class - removed:
field LxmertPreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_lxmert(model, config, tf_checkpoint_path) - removed:
method GeLU.forward(x) - removed:
method LxmertAttention.forward(hidden_states, context, attention_mask=…, output_attentions=…) - removed:
method LxmertAttention.transpose_for_scores(x) - removed:
method LxmertAttentionOutput.forward(hidden_states, input_tensor) - removed:
method LxmertCrossAttentionLayer.forward(input_tensor, ctx_tensor, ctx_att_mask=…, output_attentions=…) - removed:
method LxmertEmbeddings.forward(input_ids, token_type_ids=…, inputs_embeds=…) - removed:
method LxmertEncoder.forward(lang_feats, lang_attention_mask, visual_feats, visual_pos, visual_attention_mask=…, output_attentions=…) - removed:
method LxmertForPreTraining.forward(input_ids=…, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, obj_labels=…, matched_label=…, ans=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method LxmertForPreTraining.get_qa_logit_layer() -> nn.Module - removed:
method LxmertForPreTraining.resize_num_qa_labels(num_labels) - removed:
method LxmertForQuestionAnswering.forward(input_ids=…, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LxmertForQuestionAnswering.get_qa_logit_layer() -> nn.Module - removed:
method LxmertForQuestionAnswering.resize_num_qa_labels(num_labels) - removed:
method LxmertIntermediate.forward(hidden_states) - removed:
method LxmertLMPredictionHead.forward(hidden_states) - removed:
method LxmertLayer.forward(hidden_states, attention_mask=…, output_attentions=…) - removed:
method LxmertModel.forward(input_ids=…, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method LxmertModel.get_input_embeddings() - removed:
method LxmertModel.set_input_embeddings(new_embeddings) - removed:
method LxmertOutput.forward(hidden_states, input_tensor) - removed:
method LxmertPooler.forward(hidden_states) - removed:
method LxmertPreTrainingHeads.forward(sequence_output, pooled_output) - removed:
method LxmertPredictionHeadTransform.forward(hidden_states) - removed:
method LxmertSelfAttentionLayer.forward(input_tensor, attention_mask, output_attentions=…) - removed:
method LxmertVisualAnswerHead.forward(hidden_states) - removed:
method LxmertVisualFeatureEncoder.forward(visual_feats, visual_pos) - removed:
method LxmertVisualObjHead.forward(hidden_states) - removed:
method LxmertXLayer.cross_att(lang_input, lang_attention_mask, visual_input, visual_attention_mask, output_x_attentions=…) - removed:
method LxmertXLayer.forward(lang_feats, lang_attention_mask, visual_feats, visual_attention_mask, output_attentions=…) - removed:
method LxmertXLayer.output_fc(lang_input, visual_input) - removed:
method LxmertXLayer.self_att(lang_input, lang_attention_mask, visual_input, visual_attention_mask) - removed:
variable LXMERT_INPUTS_DOCSTRING - removed:
variable LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable LXMERT_START_DOCSTRING - removed:
variable logger
- removed:
class MarianMTModel(BartForConditionalGeneration) - removed:
field MarianMTModel.authorized_missing_keys - removed:
field MarianMTModel.config_class - removed:
field MarianMTModel.keys_to_never_save - removed:
method MarianMTModel.adjust_logits_during_generation(logits, cur_len, max_length)
- removed:
class MBartForConditionalGeneration(BartForConditionalGeneration) - removed:
field MBartForConditionalGeneration.authorized_missing_keys - removed:
field MBartForConditionalGeneration.config_class - removed:
field MBartForConditionalGeneration.keys_to_never_save - removed:
field MBartForConditionalGeneration.model_type - removed:
variable MBART_PRETRAINED_MODEL_ARCHIVE_LIST
- removed:
class MMBTForClassification(nn.Module) - removed:
class MMBTModel(nn.Module, ModuleUtilsMixin) - removed:
class ModalEmbeddings(nn.Module) - removed:
constructor MMBTForClassification(config, transformer, encoder) - removed:
constructor MMBTModel(config, transformer, encoder) - removed:
constructor ModalEmbeddings(config, encoder, embeddings) - removed:
method MMBTForClassification.forward(input_modal, input_ids=…, modal_start_tokens=…, modal_end_tokens=…, attention_mask=…, token_type_ids=…, modal_token_type_ids=…, position_ids=…, modal_position_ids=…, head_mask=…, inputs_embeds=…, labels=…, return_dict=…) - removed:
method MMBTModel.forward(input_modal, input_ids=…, modal_start_tokens=…, modal_end_tokens=…, attention_mask=…, token_type_ids=…, modal_token_type_ids=…, position_ids=…, modal_position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MMBTModel.get_input_embeddings() - removed:
method MMBTModel.set_input_embeddings(value) - removed:
method ModalEmbeddings.forward(input_modal, start_token=…, end_token=…, position_ids=…, token_type_ids=…) - removed:
variable MMBT_INPUTS_DOCSTRING - removed:
variable MMBT_START_DOCSTRING - removed:
variable logger
- removed:
class Bottleneck(nn.Module) - removed:
class BottleneckLayer(nn.Module) - removed:
class FFNLayer(nn.Module) - removed:
class FFNOutput(nn.Module) - removed:
class MobileBertAttention(nn.Module) - removed:
class MobileBertEmbeddings(nn.Module) - removed:
class MobileBertEncoder(nn.Module) - removed:
class MobileBertForMaskedLM(MobileBertPreTrainedModel) - removed:
class MobileBertForMultipleChoice(MobileBertPreTrainedModel) - removed:
class MobileBertForNextSentencePrediction(MobileBertPreTrainedModel) - removed:
class MobileBertForPreTraining(MobileBertPreTrainedModel) - removed:
class MobileBertForPreTrainingOutput(ModelOutput) - removed:
class MobileBertForQuestionAnswering(MobileBertPreTrainedModel) - removed:
class MobileBertForSequenceClassification(MobileBertPreTrainedModel) - removed:
class MobileBertForTokenClassification(MobileBertPreTrainedModel) - removed:
class MobileBertIntermediate(nn.Module) - removed:
class MobileBertLMPredictionHead(nn.Module) - removed:
class MobileBertLayer(nn.Module) - removed:
class MobileBertModel(MobileBertPreTrainedModel) - removed:
class MobileBertOnlyMLMHead(nn.Module) - removed:
class MobileBertOnlyNSPHead(nn.Module) - removed:
class MobileBertOutput(nn.Module) - removed:
class MobileBertPooler(nn.Module) - removed:
class MobileBertPreTrainedModel(PreTrainedModel) - removed:
class MobileBertPreTrainingHeads(nn.Module) - removed:
class MobileBertPredictionHeadTransform(nn.Module) - removed:
class MobileBertSelfAttention(nn.Module) - removed:
class MobileBertSelfOutput(nn.Module) - removed:
class NoNorm(nn.Module) - removed:
class OutputBottleneck(nn.Module) - removed:
constructor Bottleneck(config) - removed:
constructor BottleneckLayer(config) - removed:
constructor FFNLayer(config) - removed:
constructor FFNOutput(config) - removed:
constructor MobileBertAttention(config) - removed:
constructor MobileBertEmbeddings(config) - removed:
constructor MobileBertEncoder(config) - removed:
constructor MobileBertForMaskedLM(config) - removed:
constructor MobileBertForMultipleChoice(config) - removed:
constructor MobileBertForNextSentencePrediction(config) - removed:
constructor MobileBertForPreTraining(config) - removed:
constructor MobileBertForQuestionAnswering(config) - removed:
constructor MobileBertForSequenceClassification(config) - removed:
constructor MobileBertForTokenClassification(config) - removed:
constructor MobileBertIntermediate(config) - removed:
constructor MobileBertLMPredictionHead(config) - removed:
constructor MobileBertLayer(config) - removed:
constructor MobileBertModel(config, add_pooling_layer=…) - removed:
constructor MobileBertOnlyMLMHead(config) - removed:
constructor MobileBertOnlyNSPHead(config) - removed:
constructor MobileBertOutput(config) - removed:
constructor MobileBertPooler(config) - removed:
constructor MobileBertPreTrainingHeads(config) - removed:
constructor MobileBertPredictionHeadTransform(config) - removed:
constructor MobileBertSelfAttention(config) - removed:
constructor MobileBertSelfOutput(config) - removed:
constructor NoNorm(feat_size, eps=…) - removed:
constructor OutputBottleneck(config) - removed:
field MobileBertForMaskedLM.authorized_unexpected_keys - removed:
field MobileBertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field MobileBertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field MobileBertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - removed:
field MobileBertForPreTrainingOutput.prediction_logits: torch.FloatTensor - removed:
field MobileBertForPreTrainingOutput.seq_relationship_logits: torch.FloatTensor - removed:
field MobileBertForQuestionAnswering.authorized_unexpected_keys - removed:
field MobileBertForTokenClassification.authorized_unexpected_keys - removed:
field MobileBertPreTrainedModel.authorized_missing_keys - removed:
field MobileBertPreTrainedModel.base_model_prefix - removed:
field MobileBertPreTrainedModel.config_class - removed:
field MobileBertPreTrainedModel.load_tf_weights - removed:
field MobileBertPreTrainedModel.pretrained_model_archive_map - removed:
function load_tf_weights_in_mobilebert(model, config, tf_checkpoint_path) - removed:
function mish(x) - removed:
method Bottleneck.forward(hidden_states) - removed:
method BottleneckLayer.forward(hidden_states) - removed:
method FFNLayer.forward(hidden_states) - removed:
method FFNOutput.forward(hidden_states, residual_tensor) - removed:
method MobileBertAttention.forward(query_tensor, key_tensor, value_tensor, layer_input, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method MobileBertAttention.prune_heads(heads) - removed:
method MobileBertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method MobileBertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method MobileBertForMaskedLM.get_output_embeddings() - removed:
method MobileBertForMaskedLM.tie_weights() - removed:
method MobileBertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertForNextSentencePrediction.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method MobileBertForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, next_sentence_label=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertForPreTraining.get_output_embeddings() - removed:
method MobileBertForPreTraining.tie_weights() - removed:
method MobileBertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method MobileBertIntermediate.forward(hidden_states) - removed:
method MobileBertLMPredictionHead.forward(hidden_states) - removed:
method MobileBertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method MobileBertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_hidden_states=…, output_attentions=…, return_dict=…) - removed:
method MobileBertModel.get_input_embeddings() - removed:
method MobileBertModel.set_input_embeddings(value) - removed:
method MobileBertOnlyMLMHead.forward(sequence_output) - removed:
method MobileBertOnlyNSPHead.forward(pooled_output) - removed:
method MobileBertOutput.forward(intermediate_states, residual_tensor_1, residual_tensor_2) - removed:
method MobileBertPooler.forward(hidden_states) - removed:
method MobileBertPreTrainingHeads.forward(sequence_output, pooled_output) - removed:
method MobileBertPredictionHeadTransform.forward(hidden_states) - removed:
method MobileBertSelfAttention.forward(query_tensor, key_tensor, value_tensor, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method MobileBertSelfAttention.transpose_for_scores(x) - removed:
method MobileBertSelfOutput.forward(hidden_states, residual_tensor) - removed:
method NoNorm.forward(input_tensor) - removed:
method OutputBottleneck.forward(hidden_states, residual_tensor) - removed:
variable MOBILEBERT_INPUTS_DOCSTRING - removed:
variable MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable MOBILEBERT_START_DOCSTRING - removed:
variable NORM2FN - removed:
variable logger
- removed:
class Attention(nn.Module) - removed:
class Block(nn.Module) - removed:
class MLP(nn.Module) - removed:
class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel) - removed:
class OpenAIGPTDoubleHeadsModelOutput(ModelOutput) - removed:
class OpenAIGPTForSequenceClassification(OpenAIGPTPreTrainedModel) - removed:
class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel) - removed:
class OpenAIGPTModel(OpenAIGPTPreTrainedModel) - removed:
class OpenAIGPTPreTrainedModel(PreTrainedModel) - removed:
constructor Attention(nx, n_ctx, config, scale=…) - removed:
constructor Block(n_ctx, config, scale=…) - removed:
constructor MLP(n_state, config) - removed:
constructor OpenAIGPTDoubleHeadsModel(config) - removed:
constructor OpenAIGPTForSequenceClassification(config) - removed:
constructor OpenAIGPTLMHeadModel(config) - removed:
constructor OpenAIGPTModel(config) - removed:
field OpenAIGPTDoubleHeadsModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field OpenAIGPTDoubleHeadsModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field OpenAIGPTDoubleHeadsModelOutput.logits: torch.FloatTensor - removed:
field OpenAIGPTDoubleHeadsModelOutput.loss: Optional[torch.FloatTensor] - removed:
field OpenAIGPTDoubleHeadsModelOutput.mc_logits: torch.FloatTensor - removed:
field OpenAIGPTDoubleHeadsModelOutput.mc_loss: Optional[torch.FloatTensor] - removed:
field OpenAIGPTPreTrainedModel.authorized_missing_keys - removed:
field OpenAIGPTPreTrainedModel.base_model_prefix - removed:
field OpenAIGPTPreTrainedModel.config_class - removed:
field OpenAIGPTPreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_openai_gpt(model, config, openai_checkpoint_folder_path) - removed:
method Attention.forward(x, attention_mask=…, head_mask=…, output_attentions=…) - removed:
method Attention.merge_heads(x) - removed:
method Attention.prune_heads(heads) - removed:
method Attention.split_heads(x, k=…) - removed:
method Block.forward(x, attention_mask=…, head_mask=…, output_attentions=…) - removed:
method MLP.forward(x) - removed:
method OpenAIGPTDoubleHeadsModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, labels=…, mc_labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method OpenAIGPTDoubleHeadsModel.get_output_embeddings() - removed:
method OpenAIGPTForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method OpenAIGPTLMHeadModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method OpenAIGPTLMHeadModel.get_output_embeddings() - removed:
method OpenAIGPTModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method OpenAIGPTModel.get_input_embeddings() - removed:
method OpenAIGPTModel.set_input_embeddings(new_embeddings) - removed:
variable ACT_FNS - removed:
variable OPENAI_GPT_INPUTS_DOCSTRING - removed:
variable OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable OPENAI_GPT_START_DOCSTRING - removed:
variable logger
- removed:
class CausalLMOutputWithPastAndCrossAttentions(ModelOutput) - removed:
field CausalLMOutputWithPastAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field CausalLMOutputWithPastAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field CausalLMOutputWithPastAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field CausalLMOutputWithPastAndCrossAttentions.logits: torch.FloatTensor - removed:
field CausalLMOutputWithPastAndCrossAttentions.loss: Optional[torch.FloatTensor] - removed:
field CausalLMOutputWithPastAndCrossAttentions.past_key_values: Optional[List[torch.FloatTensor]] - changed:
field BaseModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutput.last_hidden_state: torch.FloatTensor → field BaseModelOutput.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithCrossAttentions.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithCrossAttentions.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithCrossAttentions.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithCrossAttentions.last_hidden_state: torch.FloatTensor → field BaseModelOutputWithCrossAttentions.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPast.last_hidden_state: torch.FloatTensor → field BaseModelOutputWithPast.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] → field BaseModelOutputWithPast.past_key_values: Optional[Cache] - changed:
field BaseModelOutputWithPastAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPastAndCrossAttentions.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPastAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPastAndCrossAttentions.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPastAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPastAndCrossAttentions.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPastAndCrossAttentions.last_hidden_state: torch.FloatTensor → field BaseModelOutputWithPastAndCrossAttentions.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPastAndCrossAttentions.past_key_values: Optional[List[torch.FloatTensor]] → field BaseModelOutputWithPastAndCrossAttentions.past_key_values: Optional[Cache] - changed:
field BaseModelOutputWithPooling.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPooling.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPooling.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPooling.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPooling.last_hidden_state: torch.FloatTensor → field BaseModelOutputWithPooling.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPooling.pooler_output: torch.FloatTensor → field BaseModelOutputWithPooling.pooler_output: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPoolingAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPoolingAndCrossAttentions.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPoolingAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPoolingAndCrossAttentions.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPoolingAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] → field BaseModelOutputWithPoolingAndCrossAttentions.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field BaseModelOutputWithPoolingAndCrossAttentions.last_hidden_state: torch.FloatTensor → field BaseModelOutputWithPoolingAndCrossAttentions.last_hidden_state: Optional[torch.FloatTensor] - changed:
field BaseModelOutputWithPoolingAndCrossAttentions.pooler_output: torch.FloatTensor → field BaseModelOutputWithPoolingAndCrossAttentions.pooler_output: Optional[torch.FloatTensor] - changed:
field CausalLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutput.logits: torch.FloatTensor → field CausalLMOutput.logits: Optional[torch.FloatTensor] - changed:
field CausalLMOutputWithCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutputWithCrossAttentions.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutputWithCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutputWithCrossAttentions.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutputWithCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutputWithCrossAttentions.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutputWithCrossAttentions.logits: torch.FloatTensor → field CausalLMOutputWithCrossAttentions.logits: Optional[torch.FloatTensor] - changed:
field CausalLMOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] → field CausalLMOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field CausalLMOutputWithPast.logits: torch.FloatTensor → field CausalLMOutputWithPast.logits: Optional[torch.FloatTensor] - changed:
field CausalLMOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] → field CausalLMOutputWithPast.past_key_values: Optional[Cache] - changed:
field MaskedLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field MaskedLMOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field MaskedLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field MaskedLMOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field MaskedLMOutput.logits: torch.FloatTensor → field MaskedLMOutput.logits: Optional[torch.FloatTensor] - changed:
field MultipleChoiceModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field MultipleChoiceModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field MultipleChoiceModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field MultipleChoiceModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field MultipleChoiceModelOutput.logits: torch.FloatTensor → field MultipleChoiceModelOutput.logits: Optional[torch.FloatTensor] - changed:
field NextSentencePredictorOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field NextSentencePredictorOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field NextSentencePredictorOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field NextSentencePredictorOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field NextSentencePredictorOutput.logits: torch.FloatTensor → field NextSentencePredictorOutput.logits: Optional[torch.FloatTensor] - changed:
field QuestionAnsweringModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field QuestionAnsweringModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field QuestionAnsweringModelOutput.end_logits: torch.FloatTensor → field QuestionAnsweringModelOutput.end_logits: Optional[torch.FloatTensor] - changed:
field QuestionAnsweringModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field QuestionAnsweringModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field QuestionAnsweringModelOutput.start_logits: torch.FloatTensor → field QuestionAnsweringModelOutput.start_logits: Optional[torch.FloatTensor] - changed:
field Seq2SeqLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqLMOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqLMOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqLMOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqLMOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqLMOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqLMOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqLMOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqLMOutput.logits: torch.FloatTensor → field Seq2SeqLMOutput.logits: Optional[torch.FloatTensor] - changed:
field Seq2SeqLMOutput.past_key_values: Optional[List[torch.FloatTensor]] → field Seq2SeqLMOutput.past_key_values: Optional[EncoderDecoderCache] - changed:
field Seq2SeqModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqModelOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqModelOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqModelOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqModelOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqModelOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqModelOutput.last_hidden_state: torch.FloatTensor → field Seq2SeqModelOutput.last_hidden_state: Optional[torch.FloatTensor] - changed:
field Seq2SeqModelOutput.past_key_values: Optional[List[torch.FloatTensor]] → field Seq2SeqModelOutput.past_key_values: Optional[EncoderDecoderCache] - changed:
field Seq2SeqQuestionAnsweringModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqQuestionAnsweringModelOutput.end_logits: torch.FloatTensor → field Seq2SeqQuestionAnsweringModelOutput.end_logits: Optional[torch.FloatTensor] - changed:
field Seq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[List[torch.FloatTensor]] → field Seq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[EncoderDecoderCache] - changed:
field Seq2SeqQuestionAnsweringModelOutput.start_logits: torch.FloatTensor → field Seq2SeqQuestionAnsweringModelOutput.start_logits: Optional[torch.FloatTensor] - changed:
field Seq2SeqSequenceClassifierOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqSequenceClassifierOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqSequenceClassifierOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field Seq2SeqSequenceClassifierOutput.logits: torch.FloatTensor → field Seq2SeqSequenceClassifierOutput.logits: Optional[torch.FloatTensor] - changed:
field Seq2SeqSequenceClassifierOutput.past_key_values: Optional[List[torch.FloatTensor]] → field Seq2SeqSequenceClassifierOutput.past_key_values: Optional[EncoderDecoderCache] - changed:
field SequenceClassifierOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field SequenceClassifierOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field SequenceClassifierOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field SequenceClassifierOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field SequenceClassifierOutput.logits: torch.FloatTensor → field SequenceClassifierOutput.logits: Optional[torch.FloatTensor] - changed:
field SequenceClassifierOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] → field SequenceClassifierOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field SequenceClassifierOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] → field SequenceClassifierOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field SequenceClassifierOutputWithPast.logits: torch.FloatTensor → field SequenceClassifierOutputWithPast.logits: Optional[torch.FloatTensor] - changed:
field SequenceClassifierOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] → field SequenceClassifierOutputWithPast.past_key_values: Optional[Cache] - changed:
field TokenClassifierOutput.attentions: Optional[Tuple[torch.FloatTensor]] → field TokenClassifierOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - changed:
field TokenClassifierOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] → field TokenClassifierOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - changed:
field TokenClassifierOutput.logits: torch.FloatTensor → field TokenClassifierOutput.logits: Optional[torch.FloatTensor] - added:
class BackboneOutput(ModelOutput) - added:
class BaseModelOutputWithNoAttention(ModelOutput) - added:
class BaseModelOutputWithPoolingAndNoAttention(ModelOutput) - added:
class BaseModelOutputWithPoolingAndProjection(ModelOutput) - added:
class DepthEstimatorOutput(ModelOutput) - added:
class ImageClassifierOutput(ModelOutput) - added:
class ImageClassifierOutputWithNoAttention(ModelOutput) - added:
class ImageSuperResolutionOutput(ModelOutput) - added:
class MaskedImageModelingOutput(ModelOutput) - added:
class MoECausalLMOutputWithPast(ModelOutput) - added:
class MoEModelOutput(ModelOutput) - added:
class MoEModelOutputWithPastAndCrossAttentions(ModelOutput) - added:
class MoeCausalLMOutputWithPast(ModelOutput) - added:
class MoeModelOutputWithPast(ModelOutput) - added:
class SampleTSPredictionOutput(ModelOutput) - added:
class SemanticSegmenterOutput(ModelOutput) - added:
class Seq2SeqMoEModelOutput(ModelOutput) - added:
class Seq2SeqMoEOutput(ModelOutput) - added:
class Seq2SeqSpectrogramOutput(ModelOutput) - added:
class Seq2SeqTSModelOutput(ModelOutput) - added:
class Seq2SeqTSPredictionOutput(ModelOutput) - added:
class Wav2Vec2BaseModelOutput(ModelOutput) - added:
class XVectorOutput(ModelOutput) - added:
field BackboneOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field BackboneOutput.feature_maps: Optional[tuple[torch.FloatTensor]] - added:
field BackboneOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field BaseModelOutputWithNoAttention.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field BaseModelOutputWithNoAttention.last_hidden_state: Optional[torch.FloatTensor] - added:
field BaseModelOutputWithPoolingAndCrossAttentions.past_key_values: Optional[Cache] - added:
field BaseModelOutputWithPoolingAndNoAttention.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field BaseModelOutputWithPoolingAndNoAttention.last_hidden_state: Optional[torch.FloatTensor] - added:
field BaseModelOutputWithPoolingAndNoAttention.pooler_output: Optional[torch.FloatTensor] - added:
field BaseModelOutputWithPoolingAndProjection.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field BaseModelOutputWithPoolingAndProjection.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field BaseModelOutputWithPoolingAndProjection.last_hidden_state: Optional[torch.FloatTensor] - added:
field BaseModelOutputWithPoolingAndProjection.pooler_output: Optional[torch.FloatTensor] - added:
field BaseModelOutputWithPoolingAndProjection.projection_state: Optional[tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithCrossAttentions.past_key_values: Optional[Cache] - added:
field DepthEstimatorOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field DepthEstimatorOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field DepthEstimatorOutput.loss: Optional[torch.FloatTensor] - added:
field DepthEstimatorOutput.predicted_depth: Optional[torch.FloatTensor] - added:
field ImageClassifierOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field ImageClassifierOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field ImageClassifierOutput.logits: Optional[torch.FloatTensor] - added:
field ImageClassifierOutput.loss: Optional[torch.FloatTensor] - added:
field ImageClassifierOutputWithNoAttention.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field ImageClassifierOutputWithNoAttention.logits: Optional[torch.FloatTensor] - added:
field ImageClassifierOutputWithNoAttention.loss: Optional[torch.FloatTensor] - added:
field ImageSuperResolutionOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field ImageSuperResolutionOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field ImageSuperResolutionOutput.loss: Optional[torch.FloatTensor] - added:
field ImageSuperResolutionOutput.reconstruction: Optional[torch.FloatTensor] - added:
field MaskedImageModelingOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MaskedImageModelingOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MaskedImageModelingOutput.loss: Optional[torch.FloatTensor] - added:
field MaskedImageModelingOutput.reconstruction: Optional[torch.FloatTensor] - added:
field MoECausalLMOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoECausalLMOutputWithPast.aux_loss: Optional[torch.FloatTensor] - added:
field MoECausalLMOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoECausalLMOutputWithPast.logits: Optional[torch.FloatTensor] - added:
field MoECausalLMOutputWithPast.loss: Optional[torch.FloatTensor] - added:
field MoECausalLMOutputWithPast.past_key_values: Optional[Cache] - added:
field MoECausalLMOutputWithPast.router_logits: Optional[tuple[torch.FloatTensor]] - added:
field MoECausalLMOutputWithPast.z_loss: Optional[torch.FloatTensor] - added:
field MoEModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoEModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoEModelOutput.last_hidden_state: Optional[torch.FloatTensor] - added:
field MoEModelOutput.router_probs: Optional[tuple[torch.FloatTensor]] - added:
field MoEModelOutputWithPastAndCrossAttentions.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoEModelOutputWithPastAndCrossAttentions.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoEModelOutputWithPastAndCrossAttentions.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoEModelOutputWithPastAndCrossAttentions.last_hidden_state: Optional[torch.FloatTensor] - added:
field MoEModelOutputWithPastAndCrossAttentions.past_key_values: Optional[Cache] - added:
field MoEModelOutputWithPastAndCrossAttentions.router_probs: Optional[tuple[torch.FloatTensor]] - added:
field MoeCausalLMOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoeCausalLMOutputWithPast.aux_loss: Optional[torch.FloatTensor] - added:
field MoeCausalLMOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoeCausalLMOutputWithPast.logits: Optional[torch.FloatTensor] - added:
field MoeCausalLMOutputWithPast.loss: Optional[torch.FloatTensor] - added:
field MoeCausalLMOutputWithPast.past_key_values: Optional[Cache] - added:
field MoeCausalLMOutputWithPast.router_logits: Optional[tuple[torch.FloatTensor]] - added:
field MoeModelOutputWithPast.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoeModelOutputWithPast.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field MoeModelOutputWithPast.last_hidden_state: Optional[torch.FloatTensor] - added:
field MoeModelOutputWithPast.past_key_values: Optional[Cache] - added:
field MoeModelOutputWithPast.router_logits: Optional[tuple[torch.FloatTensor]] - added:
field SampleTSPredictionOutput.sequences: Optional[torch.FloatTensor] - added:
field SemanticSegmenterOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field SemanticSegmenterOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field SemanticSegmenterOutput.logits: Optional[torch.FloatTensor] - added:
field SemanticSegmenterOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEModelOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEModelOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEModelOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEModelOutput.decoder_router_logits: Optional[tuple[torch.FloatTensor]] - added:
field Seq2SeqMoEModelOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEModelOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEModelOutput.encoder_router_logits: Optional[tuple[torch.FloatTensor]] - added:
field Seq2SeqMoEModelOutput.last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEModelOutput.past_key_values: Optional[EncoderDecoderCache] - added:
field Seq2SeqMoEOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEOutput.decoder_aux_loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEOutput.decoder_router_logits: Optional[tuple[torch.FloatTensor]] - added:
field Seq2SeqMoEOutput.decoder_z_loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEOutput.encoder_aux_loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqMoEOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.encoder_router_logits: Optional[tuple[torch.FloatTensor]] - added:
field Seq2SeqMoEOutput.encoder_z_loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.logits: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqMoEOutput.past_key_values: Optional[EncoderDecoderCache] - added:
field Seq2SeqSpectrogramOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqSpectrogramOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqSpectrogramOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqSpectrogramOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqSpectrogramOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqSpectrogramOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqSpectrogramOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqSpectrogramOutput.past_key_values: Optional[EncoderDecoderCache] - added:
field Seq2SeqSpectrogramOutput.spectrogram: Optional[torch.FloatTensor] - added:
field Seq2SeqTSModelOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSModelOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSModelOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSModelOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSModelOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqTSModelOutput.last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqTSModelOutput.loc: Optional[torch.FloatTensor] - added:
field Seq2SeqTSModelOutput.past_key_values: Optional[EncoderDecoderCache] - added:
field Seq2SeqTSModelOutput.scale: Optional[torch.FloatTensor] - added:
field Seq2SeqTSModelOutput.static_features: Optional[torch.FloatTensor] - added:
field Seq2SeqTSPredictionOutput.cross_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.decoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.decoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.encoder_attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.encoder_hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqTSPredictionOutput.loc: Optional[torch.FloatTensor] - added:
field Seq2SeqTSPredictionOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqTSPredictionOutput.params: Optional[tuple[torch.FloatTensor, ...]] - added:
field Seq2SeqTSPredictionOutput.past_key_values: Optional[EncoderDecoderCache] - added:
field Seq2SeqTSPredictionOutput.scale: Optional[torch.FloatTensor] - added:
field Seq2SeqTSPredictionOutput.static_features: Optional[torch.FloatTensor] - added:
field Wav2Vec2BaseModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field Wav2Vec2BaseModelOutput.extract_features: Optional[torch.FloatTensor] - added:
field Wav2Vec2BaseModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field Wav2Vec2BaseModelOutput.last_hidden_state: Optional[torch.FloatTensor] - added:
field XVectorOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field XVectorOutput.embeddings: Optional[torch.FloatTensor] - added:
field XVectorOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field XVectorOutput.logits: Optional[torch.FloatTensor] - added:
field XVectorOutput.loss: Optional[torch.FloatTensor] - added:
property MaskedImageModelingOutput.logits
- removed:
class PegasusForConditionalGeneration(BartForConditionalGeneration) - removed:
field PegasusForConditionalGeneration.authorized_missing_keys - removed:
field PegasusForConditionalGeneration.config_class - removed:
field PegasusForConditionalGeneration.keys_to_never_save
- removed:
class ProhpetNetFeedForward(nn.Module) - removed:
class ProhpetNetPositionalEmbeddings(nn.Embedding) - removed:
class ProphetNetDecoder(ProphetNetPreTrainedModel) - removed:
class ProphetNetDecoderLMOutput(ModelOutput) - removed:
class ProphetNetDecoderLayer(nn.Module) - removed:
class ProphetNetDecoderModelOutput(ModelOutput) - removed:
class ProphetNetEncoder(ProphetNetPreTrainedModel) - removed:
class ProphetNetEncoderLayer(nn.Module) - removed:
class ProphetNetForCausalLM(ProphetNetPreTrainedModel) - removed:
class ProphetNetForConditionalGeneration(ProphetNetPreTrainedModel) - removed:
class ProphetNetModel(ProphetNetPreTrainedModel) - removed:
class ProphetNetNgramProphetNetSelfAttention(nn.Module) - removed:
class ProphetNetPreTrainedModel(PreTrainedModel) - removed:
class ProphetNetSelfAttention(nn.Module) - removed:
class ProphetNetSeq2SeqLMOutput(ModelOutput) - removed:
class ProphetNetSeq2SeqModelOutput(ModelOutput) - removed:
constructor ProhpetNetFeedForward(config: ProphetNetConfig, ffn_dim: int) - removed:
constructor ProhpetNetPositionalEmbeddings(config: ProphetNetConfig) - removed:
constructor ProphetNetDecoder(config: ProphetNetConfig, word_embeddings: nn.Embedding=…) - removed:
constructor ProphetNetDecoderLayer(config: ProphetNetConfig) - removed:
constructor ProphetNetEncoder(config: ProphetNetConfig, word_embeddings: nn.Embedding=…) - removed:
constructor ProphetNetEncoderLayer(config: ProphetNetConfig) - removed:
constructor ProphetNetForCausalLM(config) - removed:
constructor ProphetNetForConditionalGeneration(config: ProphetNetConfig) - removed:
constructor ProphetNetModel(config) - removed:
constructor ProphetNetNgramProphetNetSelfAttention(config: ProphetNetConfig) - removed:
constructor ProphetNetSelfAttention(config: ProphetNetConfig, num_attn_heads: int) - removed:
field ProphetNetDecoderLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderLMOutput.hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderLMOutput.logits: torch.FloatTensor - removed:
field ProphetNetDecoderLMOutput.logits_ngram: Optional[torch.FloatTensor] - removed:
field ProphetNetDecoderLMOutput.loss: Optional[torch.FloatTensor] - removed:
field ProphetNetDecoderLMOutput.ngram_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderLMOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.last_hidden_state: torch.FloatTensor - removed:
field ProphetNetDecoderModelOutput.last_hidden_state_ngram: Optional[torch.FloatTensor] - removed:
field ProphetNetDecoderModelOutput.ngram_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetDecoderModelOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetPreTrainedModel.base_model_prefix - removed:
field ProphetNetPreTrainedModel.config_class - removed:
field ProphetNetSeq2SeqLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqLMOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - removed:
field ProphetNetSeq2SeqLMOutput.logits: torch.FloatTensor - removed:
field ProphetNetSeq2SeqLMOutput.logits_ngram: Optional[torch.FloatTensor] - removed:
field ProphetNetSeq2SeqLMOutput.loss: Optional[torch.FloatTensor] - removed:
field ProphetNetSeq2SeqLMOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ProphetNetSeq2SeqModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - removed:
field ProphetNetSeq2SeqModelOutput.last_hidden_state: torch.FloatTensor - removed:
field ProphetNetSeq2SeqModelOutput.last_hidden_state_ngram: Optional[torch.FloatTensor] - removed:
field ProphetNetSeq2SeqModelOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - removed:
function ProphetNetLayerNorm(normalized_shape, eps=…, elementwise_affine=…) - removed:
function compute_all_stream_relative_buckets(num_buckets, max_distance, position_ids) - removed:
function compute_relative_buckets(num_buckets, max_distance, relative_positions, is_bidirectional=…) - removed:
function ngram_attention_bias(sequence_length, ngram, device, dtype) - removed:
function softmax(hidden_state, dim, onnx_trace=…) - removed:
method ProhpetNetFeedForward.forward(hidden_states) - removed:
method ProhpetNetPositionalEmbeddings.forward(inputs_shape, device, attention_mask=…, past_key_values=…, position_ids=…) - removed:
method ProphetNetDecoder.compute_buffered_relative_buckets(position_ids) - removed:
method ProphetNetDecoder.forward(input_ids=…, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, past_key_values=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ProphetNetDecoder.get_input_embeddings() - removed:
method ProphetNetDecoder.prepare_attention_mask(hidden_states, attention_mask) - removed:
method ProphetNetDecoder.prepare_predict_attention_mask(hidden_states, attention_mask) - removed:
method ProphetNetDecoder.set_input_embeddings(value) - removed:
method ProphetNetDecoderLayer.forward(hidden_states, encoder_hidden_states=…, encoder_attn_mask=…, layer_state=…, attention_mask=…, extended_predict_attention_mask=…, main_relative_position_buckets=…, predict_relative_position_buckets=…, position_ids=…) - removed:
method ProphetNetEncoder.forward(input_ids=…, attention_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ProphetNetEncoder.get_input_embeddings() - removed:
method ProphetNetEncoder.set_input_embeddings(value) - removed:
method ProphetNetEncoderLayer.forward(hidden_states, attention_mask) - removed:
method ProphetNetForCausalLM.forward(input_ids=…, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, past_key_values=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ProphetNetForCausalLM.get_decoder() - removed:
method ProphetNetForCausalLM.get_input_embeddings() - removed:
method ProphetNetForCausalLM.get_output_embeddings() - removed:
method ProphetNetForCausalLM.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, use_cache=…, **kwargs) - removed:
method ProphetNetForCausalLM.set_decoder(decoder) - removed:
method ProphetNetForCausalLM.set_input_embeddings(value) - removed:
method ProphetNetForConditionalGeneration.forward(input_ids=…, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, inputs_embeds=…, decoder_inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ProphetNetForConditionalGeneration.get_decoder() - removed:
method ProphetNetForConditionalGeneration.get_encoder() - removed:
method ProphetNetForConditionalGeneration.get_input_embeddings() - removed:
method ProphetNetForConditionalGeneration.get_output_embeddings() - removed:
method ProphetNetForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - removed:
method ProphetNetModel.forward(input_ids=…, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs: Optional[Tuple]=…, past_key_values=…, inputs_embeds=…, decoder_inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method ProphetNetModel.get_decoder() - removed:
method ProphetNetModel.get_encoder() - removed:
method ProphetNetModel.get_input_embeddings() - removed:
method ProphetNetModel.set_input_embeddings(value) - removed:
method ProphetNetNgramProphetNetSelfAttention.forward(hidden_states, layer_state=…, attention_mask=…, extended_predict_attention_mask=…, main_relative_position_buckets=…, predict_relative_position_buckets=…, position_ids=…) - removed:
method ProphetNetNgramProphetNetSelfAttention.get_main_relative_pos_embeddings(hidden_states, attn_weights, position_ids, main_relative_position_buckets) - removed:
method ProphetNetNgramProphetNetSelfAttention.get_predict_relative_pos_embeddings(hidden_states, attn_weights, position_ids, predict_relative_position_buckets) - removed:
method ProphetNetNgramProphetNetSelfAttention.prepare_for_onnx_export_() - removed:
method ProphetNetSelfAttention.forward(hidden_states, key_value_states: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, layer_state: Optional[Dict[str, Optional[Tensor]]]=…) -> Tuple[Tensor, Optional[Tensor]] - removed:
property ProphetNetSeq2SeqLMOutput.decoder_cross_attentions - removed:
property ProphetNetSeq2SeqModelOutput.decoder_cross_attentions - removed:
variable PROPHETNET_INPUTS_DOCSTRING - removed:
variable PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable PROPHETNET_STANDALONE_INPUTS_DOCSTRING - removed:
variable PROPHETNET_START_DOCSTRING - removed:
variable logger
- removed:
class RagModel(RagPreTrainedModel) - removed:
class RagPreTrainedModel(PreTrainedModel) - removed:
class RagSequenceForGeneration(RagPreTrainedModel) - removed:
class RagTokenForGeneration(RagPreTrainedModel) - removed:
class RetrievAugLMMarginOutput(ModelOutput) - removed:
class RetrievAugLMOutput(ModelOutput) - removed:
constructor RagModel(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - removed:
constructor RagSequenceForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - removed:
constructor RagTokenForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - removed:
field RagPreTrainedModel.authorized_missing_keys - removed:
field RagPreTrainedModel.base_model_prefix - removed:
field RagPreTrainedModel.config_class - removed:
field RetrievAugLMMarginOutput.context_attention_mask: Optional[torch.LongTensor] - removed:
field RetrievAugLMMarginOutput.context_input_ids: Optional[torch.LongTensor] - removed:
field RetrievAugLMMarginOutput.doc_scores: torch.FloatTensor - removed:
field RetrievAugLMMarginOutput.generator_dec_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.generator_dec_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.generator_enc_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.generator_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.generator_enc_last_hidden_state: Optional[torch.FloatTensor] - removed:
field RetrievAugLMMarginOutput.logits: torch.FloatTensor - removed:
field RetrievAugLMMarginOutput.loss: Optional[torch.FloatTensor] - removed:
field RetrievAugLMMarginOutput.past_key_values: Optional[List[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.question_enc_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.question_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMMarginOutput.question_encoder_last_hidden_state: Optional[torch.FloatTensor] - removed:
field RetrievAugLMMarginOutput.retrieved_doc_embeds: Optional[torch.FloatTensor] - removed:
field RetrievAugLMMarginOutput.retrieved_doc_ids: Optional[torch.LongTensor] - removed:
field RetrievAugLMOutput.context_attention_mask: Optional[torch.LongTensor] - removed:
field RetrievAugLMOutput.context_input_ids: Optional[torch.LongTensor] - removed:
field RetrievAugLMOutput.doc_scores: torch.FloatTensor - removed:
field RetrievAugLMOutput.generator_dec_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.generator_dec_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.generator_enc_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.generator_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.generator_enc_last_hidden_state: Optional[torch.FloatTensor] - removed:
field RetrievAugLMOutput.logits: torch.FloatTensor - removed:
field RetrievAugLMOutput.past_key_values: Optional[List[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.question_enc_attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.question_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field RetrievAugLMOutput.question_encoder_last_hidden_state: Optional[torch.FloatTensor] - removed:
field RetrievAugLMOutput.retrieved_doc_embeds: Optional[torch.FloatTensor] - removed:
field RetrievAugLMOutput.retrieved_doc_ids: Optional[torch.LongTensor] - removed:
method RagModel.forward(input_ids=…, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, past_key_values=…, doc_scores=…, context_input_ids=…, context_attention_mask=…, use_cache=…, output_attentions=…, output_hidden_states=…, output_retrieved=…, n_docs=…) - removed:
method RagSequenceForGeneration.forward(input_ids=…, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, past_key_values=…, context_input_ids=…, context_attention_mask=…, doc_scores=…, use_cache=…, output_attentions=…, output_hidden_states=…, output_retrieved=…, exclude_bos_score=…, reduce_loss=…, labels=…, n_docs=…, **kwargs) - removed:
method RagSequenceForGeneration.generate(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, context_input_ids=…, do_deduplication=…, num_return_sequences=…, num_beams=…, n_docs=…, **model_kwargs) - removed:
method RagSequenceForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, exclude_bos_score=…, n_docs=…) - removed:
method RagSequenceForGeneration.set_retriever(retriever: RagRetriever) - removed:
method RagTokenForGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
method RagTokenForGeneration.forward(input_ids=…, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, past_key_values=…, context_input_ids=…, context_attention_mask=…, doc_scores=…, use_cache=…, output_attentions=…, output_hidden_states=…, output_retrieved=…, do_marginalize=…, reduce_loss=…, labels=…, n_docs=…, **kwargs) - removed:
method RagTokenForGeneration.generate(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, context_input_ids=…, context_attention_mask=…, doc_scores=…, max_length=…, min_length=…, early_stopping=…, use_cache=…, num_beams=…, bos_token_id=…, pad_token_id=…, eos_token_id=…, length_penalty=…, no_repeat_ngram_size=…, repetition_penalty=…, bad_words_ids=…, num_return_s… - removed:
method RagTokenForGeneration.get_input_embeddings() - removed:
method RagTokenForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, n_docs=…) - removed:
method RagTokenForGeneration.get_output_embeddings() - removed:
method RagTokenForGeneration.marginalize(seq_logits, doc_scores, n_docs=…) - removed:
method RagTokenForGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, doc_scores=…, n_docs=…, **kwargs) - removed:
method RagTokenForGeneration.set_retriever(retriever: RagRetriever) - removed:
method RagTokenForGeneration.shift_tokens_right(input_ids, start_token_id=…) - removed:
property RagSequenceForGeneration.generator - removed:
property RagSequenceForGeneration.question_encoder - removed:
property RagSequenceForGeneration.retriever - removed:
property RagTokenForGeneration.generator - removed:
property RagTokenForGeneration.question_encoder - removed:
property RagTokenForGeneration.retriever - removed:
static method RagPreTrainedModel.from_pretrained_question_encoder_generator(question_encoder_pretrained_model_name_or_path: str=…, generator_pretrained_model_name_or_path: str=…, retriever: RagRetriever=…, *model_args, **kwargs) -> PreTrainedModel - removed:
variable RAG_FORWARD_INPUTS_DOCSTRING - removed:
variable RAG_START_DOCSTRING - removed:
variable logger
- removed:
class AxialPositionEmbeddings(nn.Module) - removed:
class ChunkReformerFeedForward(nn.Module) - removed:
class EfficientAttentionMixin - removed:
class LSHSelfAttention(nn.Module, EfficientAttentionMixin) - removed:
class LocalSelfAttention(nn.Module, EfficientAttentionMixin) - removed:
class PositionEmbeddings(nn.Module) - removed:
class ReformerAttention(nn.Module) - removed:
class ReformerClassificationHead(nn.Module) - removed:
class ReformerEmbeddings(nn.Module) - removed:
class ReformerEncoder(nn.Module) - removed:
class ReformerFeedForwardDense(nn.Module) - removed:
class ReformerFeedForwardOutput(nn.Module) - removed:
class ReformerForMaskedLM(ReformerPreTrainedModel) - removed:
class ReformerForQuestionAnswering(ReformerPreTrainedModel) - removed:
class ReformerForSequenceClassification(ReformerPreTrainedModel) - removed:
class ReformerLayer(nn.Module) - removed:
class ReformerModel(ReformerPreTrainedModel) - removed:
class ReformerModelOutput(ModelOutput) - removed:
class ReformerModelWithLMHead(ReformerPreTrainedModel) - removed:
class ReformerModelWithLMHeadOutput(ModelOutput) - removed:
class ReformerOnlyLMHead(nn.Module) - removed:
class ReformerPreTrainedModel(PreTrainedModel) - removed:
class ReformerSelfOutput(nn.Module) - removed:
class ReverseSort(Function) - removed:
constructor AxialPositionEmbeddings(config) - removed:
constructor ChunkReformerFeedForward(config) - removed:
constructor LSHSelfAttention(config) - removed:
constructor LocalSelfAttention(config) - removed:
constructor PositionEmbeddings(config) - removed:
constructor ReformerAttention(config, layer_id=…) - removed:
constructor ReformerClassificationHead(config) - removed:
constructor ReformerEmbeddings(config) - removed:
constructor ReformerEncoder(config) - removed:
constructor ReformerFeedForwardDense(config) - removed:
constructor ReformerFeedForwardOutput(config) - removed:
constructor ReformerForMaskedLM(config) - removed:
constructor ReformerForQuestionAnswering(config) - removed:
constructor ReformerForSequenceClassification(config) - removed:
constructor ReformerLayer(config, layer_id=…) - removed:
constructor ReformerModel(config) - removed:
constructor ReformerModelWithLMHead(config) - removed:
constructor ReformerOnlyLMHead(config) - removed:
constructor ReformerSelfOutput(config) - removed:
field ReformerModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ReformerModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ReformerModelOutput.last_hidden_state: torch.FloatTensor - removed:
field ReformerModelOutput.past_buckets_states: Optional[List[Tuple[torch.LongTensor, torch.FloatTensor]]] - removed:
field ReformerModelWithLMHeadOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field ReformerModelWithLMHeadOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field ReformerModelWithLMHeadOutput.logits: torch.FloatTensor - removed:
field ReformerModelWithLMHeadOutput.loss: Optional[torch.FloatTensor] - removed:
field ReformerModelWithLMHeadOutput.past_buckets_states: Optional[List[Tuple[torch.LongTensor, torch.FloatTensor]]] - removed:
field ReformerPreTrainedModel.base_model_prefix - removed:
field ReformerPreTrainedModel.config_class - removed:
method AxialPositionEmbeddings.forward(position_ids) - removed:
method ChunkReformerFeedForward.forward(attention_output) - removed:
method ChunkReformerFeedForward.forward_chunk(hidden_states) - removed:
method LSHSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, buckets=…, past_buckets_states=…, use_cache=…, output_attentions=…, **kwargs) - removed:
method LocalSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, past_buckets_states=…, use_cache=…, output_attentions=…, **kwargs) - removed:
method PositionEmbeddings.forward(position_ids) - removed:
method ReformerAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…, buckets=…) - removed:
method ReformerClassificationHead.forward(hidden_states, **kwargs) - removed:
method ReformerEmbeddings.forward(input_ids=…, position_ids=…, inputs_embeds=…, start_idx_pos_encodings=…) - removed:
method ReformerEncoder.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_hidden_states=…, output_attentions=…) - removed:
method ReformerFeedForwardDense.forward(hidden_states) - removed:
method ReformerFeedForwardOutput.forward(hidden_states) - removed:
method ReformerForMaskedLM.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, labels=…, output_hidden_states=…, output_attentions=…, return_dict=…) - removed:
method ReformerForMaskedLM.get_output_embeddings() - removed:
method ReformerForQuestionAnswering.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, start_positions=…, end_positions=…, output_hidden_states=…, output_attentions=…, return_dict=…) - removed:
method ReformerForSequenceClassification.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, labels=…, output_hidden_states=…, output_attentions=…, return_dict=…) - removed:
method ReformerLayer.backward_pass(next_attn_output, hidden_states, grad_attn_output, grad_hidden_states, attention_mask=…, head_mask=…, buckets=…) - removed:
method ReformerLayer.forward(prev_attn_output, hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…) - removed:
method ReformerModel.forward(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, num_hashes=…, past_buckets_states=…, use_cache=…, output_hidden_states=…, output_attentions=…, return_dict=…) - removed:
method ReformerModel.get_input_embeddings() - removed:
method ReformerModel.set_input_embeddings(value) - removed:
method ReformerModelWithLMHead.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, past_buckets_states=…, use_cache=…, output_hidden_states=…, output_attentions=…, return_dict=…, labels=…) - removed:
method ReformerModelWithLMHead.get_output_embeddings() - removed:
method ReformerModelWithLMHead.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, num_hashes=…, **kwargs) - removed:
method ReformerOnlyLMHead.forward(hidden_states) - removed:
method ReformerOnlyLMHead.forward_chunk(hidden_states) - removed:
method ReformerSelfOutput.forward(hidden_states) - removed:
property ReformerPreTrainedModel.dummy_inputs - removed:
static method ReverseSort.backward(ctx, grad_out_vectors, grad_logits) - removed:
static method ReverseSort.forward(ctx, out_vectors, logits, sorted_bucket_idx, undo_sorted_bucket_idx) - removed:
variable AttentionOutput - removed:
variable LSHSelfAttentionOutput - removed:
variable LocalSelfAttentionOutput - removed:
variable REFORMER_INPUTS_DOCSTRING - removed:
variable REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable REFORMER_START_DOCSTRING - removed:
variable ReformerBackwardOutput - removed:
variable ReformerEncoderOutput - removed:
variable ReformerOutput - removed:
variable logger
- removed:
class RetriBertModel(RetriBertPreTrainedModel) - removed:
class RetriBertPreTrainedModel(PreTrainedModel) - removed:
constructor RetriBertModel(config) - removed:
field RetriBertPreTrainedModel.base_model_prefix - removed:
field RetriBertPreTrainedModel.config_class - removed:
field RetriBertPreTrainedModel.load_tf_weights - removed:
method RetriBertModel.embed_answers(input_ids, attention_mask=…, checkpoint_batch_size=…) - removed:
method RetriBertModel.embed_questions(input_ids, attention_mask=…, checkpoint_batch_size=…) - removed:
method RetriBertModel.embed_sentences_checkpointed(input_ids, attention_mask, sent_encoder, checkpoint_batch_size=…) - removed:
method RetriBertModel.forward(input_ids_query, attention_mask_query, input_ids_doc, attention_mask_doc, checkpoint_batch_size=…) - removed:
variable RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable RETRIBERT_START_DOCSTRING - removed:
variable logger
- removed:
class RobertaAttention(nn.Module) - removed:
class RobertaClassificationHead(nn.Module) - removed:
class RobertaEmbeddings(nn.Module) - removed:
class RobertaEncoder(nn.Module) - removed:
class RobertaForCausalLM(RobertaPreTrainedModel) - removed:
class RobertaForMaskedLM(RobertaPreTrainedModel) - removed:
class RobertaForMultipleChoice(RobertaPreTrainedModel) - removed:
class RobertaForQuestionAnswering(RobertaPreTrainedModel) - removed:
class RobertaForSequenceClassification(RobertaPreTrainedModel) - removed:
class RobertaForTokenClassification(RobertaPreTrainedModel) - removed:
class RobertaIntermediate(nn.Module) - removed:
class RobertaLMHead(nn.Module) - removed:
class RobertaLayer(nn.Module) - removed:
class RobertaModel(RobertaPreTrainedModel) - removed:
class RobertaOutput(nn.Module) - removed:
class RobertaPooler(nn.Module) - removed:
class RobertaPreTrainedModel(PreTrainedModel) - removed:
class RobertaSelfAttention(nn.Module) - removed:
class RobertaSelfOutput(nn.Module) - removed:
constructor RobertaAttention(config) - removed:
constructor RobertaClassificationHead(config) - removed:
constructor RobertaEmbeddings(config) - removed:
constructor RobertaEncoder(config) - removed:
constructor RobertaForCausalLM(config) - removed:
constructor RobertaForMaskedLM(config) - removed:
constructor RobertaForMultipleChoice(config) - removed:
constructor RobertaForQuestionAnswering(config) - removed:
constructor RobertaForSequenceClassification(config) - removed:
constructor RobertaForTokenClassification(config) - removed:
constructor RobertaIntermediate(config) - removed:
constructor RobertaLMHead(config) - removed:
constructor RobertaLayer(config) - removed:
constructor RobertaModel(config, add_pooling_layer=…) - removed:
constructor RobertaOutput(config) - removed:
constructor RobertaPooler(config) - removed:
constructor RobertaSelfAttention(config) - removed:
constructor RobertaSelfOutput(config) - removed:
field RobertaForCausalLM.authorized_missing_keys - removed:
field RobertaForCausalLM.authorized_unexpected_keys - removed:
field RobertaForMaskedLM.authorized_missing_keys - removed:
field RobertaForMaskedLM.authorized_unexpected_keys - removed:
field RobertaForMultipleChoice.authorized_missing_keys - removed:
field RobertaForQuestionAnswering.authorized_missing_keys - removed:
field RobertaForQuestionAnswering.authorized_unexpected_keys - removed:
field RobertaForSequenceClassification.authorized_missing_keys - removed:
field RobertaForTokenClassification.authorized_missing_keys - removed:
field RobertaForTokenClassification.authorized_unexpected_keys - removed:
field RobertaModel.authorized_missing_keys - removed:
field RobertaPreTrainedModel.base_model_prefix - removed:
field RobertaPreTrainedModel.config_class - removed:
function create_position_ids_from_input_ids(input_ids, padding_idx) - removed:
method RobertaAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method RobertaAttention.prune_heads(heads) - removed:
method RobertaClassificationHead.forward(features, **kwargs) - removed:
method RobertaEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - removed:
method RobertaEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method RobertaEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaForCausalLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaForCausalLM.get_output_embeddings() - removed:
method RobertaForCausalLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - removed:
method RobertaForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method RobertaForMaskedLM.get_output_embeddings() - removed:
method RobertaForMultipleChoice.forward(input_ids=…, token_type_ids=…, attention_mask=…, labels=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaIntermediate.forward(hidden_states) - removed:
method RobertaLMHead.forward(features, **kwargs) - removed:
method RobertaLayer.feed_forward_chunk(attention_output) - removed:
method RobertaLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method RobertaModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method RobertaModel.get_input_embeddings() - removed:
method RobertaModel.set_input_embeddings(value) - removed:
method RobertaOutput.forward(hidden_states, input_tensor) - removed:
method RobertaPooler.forward(hidden_states) - removed:
method RobertaSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - removed:
method RobertaSelfAttention.transpose_for_scores(x) - removed:
method RobertaSelfOutput.forward(hidden_states, input_tensor) - removed:
variable ROBERTA_INPUTS_DOCSTRING - removed:
variable ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable ROBERTA_START_DOCSTRING - removed:
variable logger
- added:
function dynamic_rope_update(rope_forward) - added:
function rope_config_validation(config: PretrainedConfig, ignore_keys: Optional[set]=…) - added:
variable ROPE_INIT_FUNCTIONS - added:
variable ROPE_VALIDATION_FUNCTIONS - added:
variable logger
- removed:
class ConvActivation(nn.Module) - removed:
class ConvDropoutLayerNorm(nn.Module) - removed:
class MatMulWrapper(torch.nn.Module) - removed:
class SqueezeBertEmbeddings(nn.Module) - removed:
class SqueezeBertEncoder(nn.Module) - removed:
class SqueezeBertForMaskedLM(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertForMultipleChoice(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertForQuestionAnswering(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertForSequenceClassification(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertForTokenClassification(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertLayerNorm(nn.LayerNorm) - removed:
class SqueezeBertModel(SqueezeBertPreTrainedModel) - removed:
class SqueezeBertModule(nn.Module) - removed:
class SqueezeBertPooler(nn.Module) - removed:
class SqueezeBertPreTrainedModel(PreTrainedModel) - removed:
class SqueezeBertSelfAttention(nn.Module) - removed:
constructor ConvActivation(cin, cout, groups, act) - removed:
constructor ConvDropoutLayerNorm(cin, cout, groups, dropout_prob) - removed:
constructor MatMulWrapper() - removed:
constructor SqueezeBertEmbeddings(config) - removed:
constructor SqueezeBertEncoder(config) - removed:
constructor SqueezeBertForMaskedLM(config) - removed:
constructor SqueezeBertForMultipleChoice(config) - removed:
constructor SqueezeBertForQuestionAnswering(config) - removed:
constructor SqueezeBertForSequenceClassification(config) - removed:
constructor SqueezeBertForTokenClassification(config) - removed:
constructor SqueezeBertLayerNorm(hidden_size, eps=…) - removed:
constructor SqueezeBertModel(config) - removed:
constructor SqueezeBertModule(config) - removed:
constructor SqueezeBertPooler(config) - removed:
constructor SqueezeBertSelfAttention(config, cin, q_groups=…, k_groups=…, v_groups=…) - removed:
field SqueezeBertPreTrainedModel.authorized_missing_keys - removed:
field SqueezeBertPreTrainedModel.base_model_prefix - removed:
field SqueezeBertPreTrainedModel.config_class - removed:
method ConvActivation.forward(x) - removed:
method ConvDropoutLayerNorm.forward(hidden_states, input_tensor) - removed:
method MatMulWrapper.forward(mat1, mat2) - removed:
method SqueezeBertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - removed:
method SqueezeBertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertForMaskedLM.get_output_embeddings() - removed:
method SqueezeBertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertLayerNorm.forward(x) - removed:
method SqueezeBertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method SqueezeBertModel.get_input_embeddings() - removed:
method SqueezeBertModel.set_input_embeddings(new_embeddings) - removed:
method SqueezeBertModule.forward(hidden_states, attention_mask, output_attentions) - removed:
method SqueezeBertPooler.forward(hidden_states) - removed:
method SqueezeBertSelfAttention.forward(hidden_states, attention_mask, output_attentions) - removed:
method SqueezeBertSelfAttention.transpose_for_scores(x) - removed:
method SqueezeBertSelfAttention.transpose_key_for_scores(x) - removed:
method SqueezeBertSelfAttention.transpose_output(x) - removed:
variable SQUEEZEBERT_INPUTS_DOCSTRING - removed:
variable SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable SQUEEZEBERT_START_DOCSTRING - removed:
variable logger
- removed:
class T5Attention(nn.Module) - removed:
class T5Block(nn.Module) - removed:
class T5DenseReluDense(nn.Module) - removed:
class T5ForConditionalGeneration(T5PreTrainedModel) - removed:
class T5LayerCrossAttention(nn.Module) - removed:
class T5LayerFF(nn.Module) - removed:
class T5LayerNorm(nn.Module) - removed:
class T5LayerSelfAttention(nn.Module) - removed:
class T5Model(T5PreTrainedModel) - removed:
class T5PreTrainedModel(PreTrainedModel) - removed:
class T5Stack(T5PreTrainedModel) - removed:
constructor T5Attention(config: T5Config, has_relative_attention_bias=…, is_bidirectional=…) - removed:
constructor T5Block(config, has_relative_attention_bias=…) - removed:
constructor T5DenseReluDense(config) - removed:
constructor T5ForConditionalGeneration(config) - removed:
constructor T5LayerCrossAttention(config, has_relative_attention_bias=…) - removed:
constructor T5LayerFF(config) - removed:
constructor T5LayerNorm(hidden_size, eps=…) - removed:
constructor T5LayerSelfAttention(config, has_relative_attention_bias=…) - removed:
constructor T5Model(config: T5Config) - removed:
constructor T5Stack(config, embed_tokens=…) - removed:
field T5ForConditionalGeneration.authorized_missing_keys - removed:
field T5PreTrainedModel.base_model_prefix - removed:
field T5PreTrainedModel.config_class - removed:
field T5PreTrainedModel.load_tf_weights - removed:
function load_tf_weights_in_t5(model, config, tf_checkpoint_path) - removed:
method T5Attention.compute_bias(qlen, klen) - removed:
method T5Attention.forward(input, mask=…, kv=…, position_bias=…, past_key_value=…, head_mask=…, query_length=…, use_cache=…, output_attentions=…) - removed:
method T5Attention.prune_heads(heads) - removed:
method T5Block.forward(hidden_states, attention_mask=…, position_bias=…, encoder_hidden_states=…, encoder_attention_mask=…, encoder_decoder_position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…, return_dict=…) - removed:
method T5DenseReluDense.forward(hidden_states) - removed:
method T5ForConditionalGeneration.forward(input_ids=…, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, head_mask=…, inputs_embeds=…, decoder_inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method T5ForConditionalGeneration.get_decoder() - removed:
method T5ForConditionalGeneration.get_encoder() - removed:
method T5ForConditionalGeneration.get_input_embeddings() - removed:
method T5ForConditionalGeneration.get_output_embeddings() - removed:
method T5ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - removed:
method T5ForConditionalGeneration.set_input_embeddings(new_embeddings) - removed:
method T5LayerCrossAttention.forward(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, query_length=…, output_attentions=…) - removed:
method T5LayerFF.forward(hidden_states) - removed:
method T5LayerNorm.forward(x) - removed:
method T5LayerSelfAttention.forward(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…) - removed:
method T5Model.forward(input_ids=…, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, head_mask=…, inputs_embeds=…, decoder_inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - removed:
method T5Model.get_decoder() - removed:
method T5Model.get_encoder() - removed:
method T5Model.get_input_embeddings() - removed:
method T5Model.set_input_embeddings(new_embeddings) - removed:
method T5Stack.forward(input_ids=…, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, inputs_embeds=…, head_mask=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method T5Stack.get_input_embeddings() - removed:
method T5Stack.get_output_embeddings() - removed:
method T5Stack.set_input_embeddings(new_embeddings) - removed:
property T5PreTrainedModel.dummy_inputs - removed:
variable T5_INPUTS_DOCSTRING - removed:
variable T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable T5_START_DOCSTRING - removed:
variable logger
- removed:
class TFAlbertAttention(tf.keras.layers.Layer) - removed:
class TFAlbertEmbeddings(tf.keras.layers.Layer) - removed:
class TFAlbertForMaskedLM(TFAlbertPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFAlbertForPreTraining(TFAlbertPreTrainedModel) - removed:
class TFAlbertForPreTrainingOutput(ModelOutput) - removed:
class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFAlbertLayer(tf.keras.layers.Layer) - removed:
class TFAlbertLayerGroup(tf.keras.layers.Layer) - removed:
class TFAlbertMLMHead(tf.keras.layers.Layer) - removed:
class TFAlbertMainLayer(tf.keras.layers.Layer) - removed:
class TFAlbertModel(TFAlbertPreTrainedModel) - removed:
class TFAlbertPreTrainedModel(TFPreTrainedModel) - removed:
class TFAlbertSOPHead(tf.keras.layers.Layer) - removed:
class TFAlbertSelfOutput(tf.keras.layers.Layer) - removed:
class TFAlbertTransformer(tf.keras.layers.Layer) - removed:
constructor TFAlbertAttention(config, **kwargs) - removed:
constructor TFAlbertEmbeddings(config, **kwargs) - removed:
constructor TFAlbertForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFAlbertForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFAlbertForPreTraining(config, *inputs, **kwargs) - removed:
constructor TFAlbertForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFAlbertForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFAlbertForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFAlbertLayer(config, **kwargs) - removed:
constructor TFAlbertLayerGroup(config, **kwargs) - removed:
constructor TFAlbertMLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFAlbertMainLayer(config, **kwargs) - removed:
constructor TFAlbertModel(config, *inputs, **kwargs) - removed:
constructor TFAlbertSOPHead(config, **kwargs) - removed:
constructor TFAlbertSelfOutput(config, **kwargs) - removed:
constructor TFAlbertTransformer(config, **kwargs) - removed:
field TFAlbertForMaskedLM.authorized_missing_keys - removed:
field TFAlbertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFAlbertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFAlbertForPreTrainingOutput.prediction_logits: tf.Tensor - removed:
field TFAlbertForPreTrainingOutput.sop_logits: tf.Tensor - removed:
field TFAlbertForQuestionAnswering.authorized_missing_keys - removed:
field TFAlbertForTokenClassification.authorized_missing_keys - removed:
field TFAlbertMainLayer.config_class - removed:
field TFAlbertPreTrainedModel.base_model_prefix - removed:
field TFAlbertPreTrainedModel.config_class - removed:
method TFAlbertAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFAlbertAttention.prune_heads(heads) - removed:
method TFAlbertAttention.transpose_for_scores(x, batch_size) - removed:
method TFAlbertEmbeddings.build(input_shape) - removed:
method TFAlbertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFAlbertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFAlbertForMaskedLM.get_output_embeddings() - removed:
method TFAlbertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFAlbertForPreTraining.call(inputs, **kwargs) - removed:
method TFAlbertForPreTraining.get_output_embeddings() - removed:
method TFAlbertForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFAlbertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFAlbertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFAlbertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFAlbertLayerGroup.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, training=…) - removed:
method TFAlbertMLMHead.build(input_shape) - removed:
method TFAlbertMLMHead.call(hidden_states) - removed:
method TFAlbertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFAlbertMainLayer.get_input_embeddings() - removed:
method TFAlbertMainLayer.set_input_embeddings(value) - removed:
method TFAlbertModel.call(inputs, **kwargs) - removed:
method TFAlbertSOPHead.call(pooled_output, training: bool) - removed:
method TFAlbertSelfOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFAlbertTransformer.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
property TFAlbertForMultipleChoice.dummy_inputs - removed:
variable ALBERT_INPUTS_DOCSTRING - removed:
variable ALBERT_START_DOCSTRING - removed:
variable TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFAutoModel(object) - removed:
class TFAutoModelForCausalLM - removed:
class TFAutoModelForMaskedLM - removed:
class TFAutoModelForMultipleChoice - removed:
class TFAutoModelForPreTraining(object) - removed:
class TFAutoModelForQuestionAnswering(object) - removed:
class TFAutoModelForSeq2SeqLM - removed:
class TFAutoModelForSequenceClassification(object) - removed:
class TFAutoModelForTokenClassification - removed:
class TFAutoModelWithLMHead(object) - removed:
constructor TFAutoModel() - removed:
constructor TFAutoModelForCausalLM() - removed:
constructor TFAutoModelForMaskedLM() - removed:
constructor TFAutoModelForMultipleChoice() - removed:
constructor TFAutoModelForPreTraining() - removed:
constructor TFAutoModelForQuestionAnswering() - removed:
constructor TFAutoModelForSeq2SeqLM() - removed:
constructor TFAutoModelForSequenceClassification() - removed:
constructor TFAutoModelForTokenClassification() - removed:
constructor TFAutoModelWithLMHead() - removed:
static method TFAutoModel.from_config(config) - removed:
static method TFAutoModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForCausalLM.from_config(config) - removed:
static method TFAutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForMaskedLM.from_config(config) - removed:
static method TFAutoModelForMaskedLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForMultipleChoice.from_config(config) - removed:
static method TFAutoModelForMultipleChoice.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForPreTraining.from_config(config) - removed:
static method TFAutoModelForPreTraining.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForQuestionAnswering.from_config(config) - removed:
static method TFAutoModelForQuestionAnswering.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForSeq2SeqLM.from_config(config) - removed:
static method TFAutoModelForSeq2SeqLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForSequenceClassification.from_config(config) - removed:
static method TFAutoModelForSequenceClassification.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelForTokenClassification.from_config(config) - removed:
static method TFAutoModelForTokenClassification.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
static method TFAutoModelWithLMHead.from_config(config) - removed:
static method TFAutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - removed:
variable TF_AUTO_MODEL_PRETRAINED_DOCSTRING - removed:
variable TF_MODEL_FOR_CAUSAL_LM_MAPPING - removed:
variable TF_MODEL_FOR_MASKED_LM_MAPPING - removed:
variable TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
variable TF_MODEL_FOR_PRETRAINING_MAPPING - removed:
variable TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
variable TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
variable TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
variable TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
variable TF_MODEL_MAPPING - removed:
variable TF_MODEL_WITH_LM_HEAD_MAPPING - removed:
variable logger
- removed:
class TFAttention(Layer) - removed:
class TFBartDecoder(Layer) - removed:
class TFBartEncoder(Layer) - removed:
class TFBartForConditionalGeneration(TFPretrainedBartModel) - removed:
class TFBartModel(TFPretrainedBartModel) - removed:
class TFDecoderLayer(Layer) - removed:
class TFEncoderLayer(Layer) - removed:
class TFLearnedPositionalEmbedding(TFSharedEmbeddings) - removed:
class TFPretrainedBartModel(TFPreTrainedModel) - removed:
class TFSinusoidalPositionalEmbedding(tf.keras.layers.Embedding) - removed:
constructor TFAttention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…, **kwargs) - removed:
constructor TFBartDecoder(config: BartConfig, embed_tokens, **kwargs) - removed:
constructor TFBartEncoder(config: BartConfig, embed_tokens: TFSharedEmbeddings, **kwargs) - removed:
constructor TFBartForConditionalGeneration(config: BartConfig, *args, **kwargs) - removed:
constructor TFBartModel(config: BartConfig, *inputs, **kwargs) - removed:
constructor TFDecoderLayer(config: BartConfig, **kwargs) - removed:
constructor TFEncoderLayer(config: BartConfig, **kwargs) - removed:
constructor TFLearnedPositionalEmbedding(num_embeddings: int, embedding_dim: int, padding_idx: int, offset, **kwargs) - removed:
constructor TFSinusoidalPositionalEmbedding(num_positions, embedding_dim, **kwargs) - removed:
field TFBartForConditionalGeneration.authorized_missing_keys - removed:
field TFBartForConditionalGeneration.authorized_unexpected_keys - removed:
field TFBartForConditionalGeneration.base_model_prefix - removed:
field TFPretrainedBartModel.base_model_prefix - removed:
field TFPretrainedBartModel.config_class - removed:
function causal_attention_mask(nd, ns, dtype) - removed:
function create_position_ids_from_input_ids(input_ids, padding_idx) - removed:
function invert_mask(attention_mask: tf.Tensor) - removed:
function make_padding_mask(input_ids, padding_idx=…) - removed:
method TFAttention.call(query: tf.Tensor, key: tf.Tensor, key_padding_mask: Optional[tf.Tensor]=…, layer_state: Optional[Dict[str, tf.Tensor]]=…, attn_mask: Optional[Tensor]=…, training=…) -> Tuple[Tensor, Optional[Tensor]] - removed:
method TFBartDecoder.call(input_ids, encoder_hidden_states, encoder_padding_mask, decoder_padding_mask, decoder_causal_mask, decoder_cached_states=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFBartEncoder.call(input_ids=…, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFBartForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
method TFBartForConditionalGeneration.call(inputs, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs: Optional[TFBaseModelOutput]=…, past_key_values=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFBartForConditionalGeneration.compute_loss(labels, logits) - removed:
method TFBartForConditionalGeneration.get_encoder() - removed:
method TFBartForConditionalGeneration.get_output_embeddings() - removed:
method TFBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past, attention_mask, use_cache=…, **kwargs) -> Dict - removed:
method TFBartModel.call(inputs, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs: Optional[TFBaseModelOutput]=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFBartModel.get_input_embeddings() - removed:
method TFBartModel.get_output_embeddings() - removed:
method TFBartModel.set_input_embeddings(value) - removed:
method TFDecoderLayer.call(x, encoder_hidden_states: tf.Tensor, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…, training=…) -> Tuple[tf.Tensor, tf.Tensor, Dict[str, tf.Tensor]] - removed:
method TFEncoderLayer.call(x, encoder_padding_mask, training=…) - removed:
method TFLearnedPositionalEmbedding.call(input_ids: tf.Tensor, use_cache=…) - removed:
method TFSinusoidalPositionalEmbedding.build(input_shape) - removed:
method TFSinusoidalPositionalEmbedding.call(input_ids, use_cache=…) - removed:
property TFPretrainedBartModel.dummy_inputs - removed:
variable BART_INPUTS_DOCSTRING - removed:
variable BART_START_DOCSTRING - removed:
variable LARGE_NEGATIVE - removed:
variable PAST_KV_DEPRECATION_WARNING - removed:
variable logger
- removed:
class TFBertAttention(tf.keras.layers.Layer) - removed:
class TFBertEmbeddings(tf.keras.layers.Layer) - removed:
class TFBertEncoder(tf.keras.layers.Layer) - removed:
class TFBertForMaskedLM(TFBertPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFBertForNextSentencePrediction(TFBertPreTrainedModel) - removed:
class TFBertForPreTraining(TFBertPreTrainedModel) - removed:
class TFBertForPreTrainingOutput(ModelOutput) - removed:
class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFBertIntermediate(tf.keras.layers.Layer) - removed:
class TFBertLMHeadModel(TFBertPreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFBertLMPredictionHead(tf.keras.layers.Layer) - removed:
class TFBertLayer(tf.keras.layers.Layer) - removed:
class TFBertMLMHead(tf.keras.layers.Layer) - removed:
class TFBertMainLayer(tf.keras.layers.Layer) - removed:
class TFBertModel(TFBertPreTrainedModel) - removed:
class TFBertNSPHead(tf.keras.layers.Layer) - removed:
class TFBertOutput(tf.keras.layers.Layer) - removed:
class TFBertPooler(tf.keras.layers.Layer) - removed:
class TFBertPreTrainedModel(TFPreTrainedModel) - removed:
class TFBertPredictionHeadTransform(tf.keras.layers.Layer) - removed:
class TFBertSelfAttention(tf.keras.layers.Layer) - removed:
class TFBertSelfOutput(tf.keras.layers.Layer) - removed:
constructor TFBertAttention(config, **kwargs) - removed:
constructor TFBertEmbeddings(config, **kwargs) - removed:
constructor TFBertEncoder(config, **kwargs) - removed:
constructor TFBertForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFBertForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFBertForNextSentencePrediction(config, *inputs, **kwargs) - removed:
constructor TFBertForPreTraining(config, *inputs, **kwargs) - removed:
constructor TFBertForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFBertForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFBertForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFBertIntermediate(config, **kwargs) - removed:
constructor TFBertLMHeadModel(config, *inputs, **kwargs) - removed:
constructor TFBertLMPredictionHead(config, input_embeddings, **kwargs) - removed:
constructor TFBertLayer(config, **kwargs) - removed:
constructor TFBertMLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFBertMainLayer(config, **kwargs) - removed:
constructor TFBertModel(config, *inputs, **kwargs) - removed:
constructor TFBertNSPHead(config, **kwargs) - removed:
constructor TFBertOutput(config, **kwargs) - removed:
constructor TFBertPooler(config, **kwargs) - removed:
constructor TFBertPredictionHeadTransform(config, **kwargs) - removed:
constructor TFBertSelfAttention(config, **kwargs) - removed:
constructor TFBertSelfOutput(config, **kwargs) - removed:
field TFBertForMaskedLM.authorized_missing_keys - removed:
field TFBertForMaskedLM.authorized_unexpected_keys - removed:
field TFBertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFBertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFBertForPreTrainingOutput.prediction_logits: tf.Tensor - removed:
field TFBertForPreTrainingOutput.seq_relationship_logits: tf.Tensor - removed:
field TFBertForQuestionAnswering.authorized_missing_keys - removed:
field TFBertForQuestionAnswering.authorized_unexpected_keys - removed:
field TFBertForTokenClassification.authorized_missing_keys - removed:
field TFBertForTokenClassification.authorized_unexpected_keys - removed:
field TFBertLMHeadModel.authorized_missing_keys - removed:
field TFBertLMHeadModel.authorized_unexpected_keys - removed:
field TFBertMainLayer.config_class - removed:
field TFBertPreTrainedModel.base_model_prefix - removed:
field TFBertPreTrainedModel.config_class - removed:
method TFBertAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFBertAttention.prune_heads(heads) - removed:
method TFBertEmbeddings.build(input_shape) - removed:
method TFBertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFBertEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
method TFBertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFBertForMaskedLM.get_output_embeddings() - removed:
method TFBertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFBertForNextSentencePrediction.call(inputs, **kwargs) - removed:
method TFBertForPreTraining.call(inputs, **kwargs) - removed:
method TFBertForPreTraining.get_output_embeddings() - removed:
method TFBertForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFBertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFBertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFBertIntermediate.call(hidden_states) - removed:
method TFBertLMHeadModel.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFBertLMHeadModel.get_output_embeddings() - removed:
method TFBertLMPredictionHead.build(input_shape) - removed:
method TFBertLMPredictionHead.call(hidden_states) - removed:
method TFBertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFBertMLMHead.call(sequence_output) - removed:
method TFBertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFBertMainLayer.get_input_embeddings() - removed:
method TFBertMainLayer.set_input_embeddings(value) - removed:
method TFBertModel.call(inputs, **kwargs) - removed:
method TFBertNSPHead.call(pooled_output) - removed:
method TFBertOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFBertPooler.call(hidden_states) - removed:
method TFBertPredictionHeadTransform.call(hidden_states) - removed:
method TFBertSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFBertSelfAttention.transpose_for_scores(x, batch_size) - removed:
method TFBertSelfOutput.call(hidden_states, input_tensor, training=…) - removed:
property TFBertForMultipleChoice.dummy_inputs - removed:
variable BERT_INPUTS_DOCSTRING - removed:
variable BERT_START_DOCSTRING - removed:
variable TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFBlenderbotForConditionalGeneration(TFBartForConditionalGeneration) - removed:
field TFBlenderbotForConditionalGeneration.config_class - removed:
method TFBlenderbotForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
variable START_DOCSTRING - removed:
variable logger
- removed:
class TFCamembertForMaskedLM(TFRobertaForMaskedLM) - removed:
class TFCamembertForMultipleChoice(TFRobertaForMultipleChoice) - removed:
class TFCamembertForQuestionAnswering(TFRobertaForQuestionAnswering) - removed:
class TFCamembertForSequenceClassification(TFRobertaForSequenceClassification) - removed:
class TFCamembertForTokenClassification(TFRobertaForTokenClassification) - removed:
class TFCamembertModel(TFRobertaModel) - removed:
field TFCamembertForMaskedLM.config_class - removed:
field TFCamembertForMultipleChoice.config_class - removed:
field TFCamembertForQuestionAnswering.config_class - removed:
field TFCamembertForSequenceClassification.config_class - removed:
field TFCamembertForTokenClassification.config_class - removed:
field TFCamembertModel.config_class - removed:
variable CAMEMBERT_START_DOCSTRING - removed:
variable TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFCTRLLMHead(tf.keras.layers.Layer) - removed:
class TFCTRLLMHeadModel(TFCTRLPreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFCTRLMainLayer(tf.keras.layers.Layer) - removed:
class TFCTRLModel(TFCTRLPreTrainedModel) - removed:
class TFCTRLPreTrainedModel(TFPreTrainedModel) - removed:
class TFEncoderLayer(tf.keras.layers.Layer) - removed:
class TFMultiHeadAttention(tf.keras.layers.Layer) - removed:
class TFPointWiseFeedForwardLayer(tf.keras.layers.Layer) - removed:
constructor TFCTRLLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFCTRLLMHeadModel(config, *inputs, **kwargs) - removed:
constructor TFCTRLMainLayer(config, **kwargs) - removed:
constructor TFCTRLModel(config, *inputs, **kwargs) - removed:
constructor TFEncoderLayer(d_model_size, num_heads, dff, rate=…, layer_norm_epsilon=…, output_attentions=…, **kwargs) - removed:
constructor TFMultiHeadAttention(d_model_size, num_heads, output_attentions=…, **kwargs) - removed:
constructor TFPointWiseFeedForwardLayer(d_model_size, dff, **kwargs) - removed:
field TFCTRLMainLayer.config_class - removed:
field TFCTRLPreTrainedModel.base_model_prefix - removed:
field TFCTRLPreTrainedModel.config_class - removed:
function angle_defn(pos, i, d_model_size) - removed:
function positional_encoding(position, d_model_size) - removed:
function scaled_dot_product_attention(q, k, v, mask, attention_mask=…, head_mask=…) - removed:
method TFCTRLLMHead.build(input_shape) - removed:
method TFCTRLLMHead.call(hidden_states) - removed:
method TFCTRLLMHeadModel.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFCTRLLMHeadModel.get_output_embeddings() - removed:
method TFCTRLLMHeadModel.prepare_inputs_for_generation(inputs, past, **kwargs) - removed:
method TFCTRLMainLayer.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFCTRLMainLayer.get_input_embeddings() - removed:
method TFCTRLMainLayer.set_input_embeddings(value) - removed:
method TFCTRLModel.call(inputs, **kwargs) - removed:
method TFEncoderLayer.call(x, mask, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - removed:
method TFMultiHeadAttention.call(v, k, q, mask, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - removed:
method TFMultiHeadAttention.split_into_heads(x, batch_size) - removed:
method TFPointWiseFeedForwardLayer.call(inputs, trainable=…) - removed:
variable CTRL_INPUTS_DOCSTRING - removed:
variable CTRL_START_DOCSTRING - removed:
variable TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFDistilBertLMHead(tf.keras.layers.Layer) - removed:
class TFDistilBertMainLayer(tf.keras.layers.Layer) - removed:
class TFDistilBertModel(TFDistilBertPreTrainedModel) - removed:
class TFDistilBertPreTrainedModel(TFPreTrainedModel) - removed:
class TFEmbeddings(tf.keras.layers.Layer) - removed:
class TFFFN(tf.keras.layers.Layer) - removed:
class TFMultiHeadSelfAttention(tf.keras.layers.Layer) - removed:
class TFTransformer(tf.keras.layers.Layer) - removed:
class TFTransformerBlock(tf.keras.layers.Layer) - removed:
constructor TFDistilBertForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFDistilBertForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFDistilBertForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFDistilBertForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFDistilBertForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFDistilBertLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFDistilBertMainLayer(config, **kwargs) - removed:
constructor TFDistilBertModel(config, *inputs, **kwargs) - removed:
constructor TFEmbeddings(config, **kwargs) - removed:
constructor TFFFN(config, **kwargs) - removed:
constructor TFMultiHeadSelfAttention(config, **kwargs) - removed:
constructor TFTransformer(config, **kwargs) - removed:
constructor TFTransformerBlock(config, **kwargs) - removed:
field TFDistilBertMainLayer.config_class - removed:
field TFDistilBertPreTrainedModel.base_model_prefix - removed:
field TFDistilBertPreTrainedModel.config_class - removed:
method TFDistilBertForMaskedLM.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFDistilBertForMaskedLM.get_output_embeddings() - removed:
method TFDistilBertForMultipleChoice.call(inputs, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFDistilBertForQuestionAnswering.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFDistilBertForSequenceClassification.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFDistilBertForTokenClassification.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFDistilBertLMHead.build(input_shape) - removed:
method TFDistilBertLMHead.call(hidden_states) - removed:
method TFDistilBertMainLayer.call(inputs, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFDistilBertMainLayer.get_input_embeddings() - removed:
method TFDistilBertMainLayer.set_input_embeddings(value) - removed:
method TFDistilBertModel.call(inputs, **kwargs) - removed:
method TFEmbeddings.build(input_shape) - removed:
method TFEmbeddings.call(input_ids=…, position_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFFFN.call(input, training=…) - removed:
method TFMultiHeadSelfAttention.call(query, key, value, mask, head_mask, output_attentions, training=…) - removed:
method TFMultiHeadSelfAttention.prune_heads(heads) - removed:
method TFTransformer.call(x, attn_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
method TFTransformerBlock.call(x, attn_mask, head_mask, output_attentions, training=…) - removed:
property TFDistilBertForMultipleChoice.dummy_inputs - removed:
variable DISTILBERT_INPUTS_DOCSTRING - removed:
variable DISTILBERT_START_DOCSTRING - removed:
variable TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFElectraAttention(tf.keras.layers.Layer) - removed:
class TFElectraClassificationHead(tf.keras.layers.Layer) - removed:
class TFElectraDiscriminatorPredictions(tf.keras.layers.Layer) - removed:
class TFElectraEmbeddings(tf.keras.layers.Layer) - removed:
class TFElectraEncoder(tf.keras.layers.Layer) - removed:
class TFElectraForMaskedLM(TFElectraPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFElectraForMultipleChoice(TFElectraPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFElectraForPreTraining(TFElectraPreTrainedModel) - removed:
class TFElectraForPreTrainingOutput(ModelOutput) - removed:
class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFElectraForSequenceClassification(TFElectraPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFElectraGeneratorPredictions(tf.keras.layers.Layer) - removed:
class TFElectraIntermediate(tf.keras.layers.Layer) - removed:
class TFElectraLayer(tf.keras.layers.Layer) - removed:
class TFElectraMainLayer(tf.keras.layers.Layer) - removed:
class TFElectraMaskedLMHead(tf.keras.layers.Layer) - removed:
class TFElectraModel(TFElectraPreTrainedModel) - removed:
class TFElectraOutput(tf.keras.layers.Layer) - removed:
class TFElectraPooler(tf.keras.layers.Layer) - removed:
class TFElectraPreTrainedModel(TFPreTrainedModel) - removed:
class TFElectraSelfAttention(tf.keras.layers.Layer) - removed:
class TFElectraSelfOutput(tf.keras.layers.Layer) - removed:
constructor TFElectraAttention(config, **kwargs) - removed:
constructor TFElectraClassificationHead(config, **kwargs) - removed:
constructor TFElectraDiscriminatorPredictions(config, **kwargs) - removed:
constructor TFElectraEmbeddings(config, **kwargs) - removed:
constructor TFElectraEncoder(config, **kwargs) - removed:
constructor TFElectraForMaskedLM(config, **kwargs) - removed:
constructor TFElectraForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFElectraForPreTraining(config, **kwargs) - removed:
constructor TFElectraForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFElectraForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFElectraForTokenClassification(config, **kwargs) - removed:
constructor TFElectraGeneratorPredictions(config, **kwargs) - removed:
constructor TFElectraIntermediate(config, **kwargs) - removed:
constructor TFElectraLayer(config, **kwargs) - removed:
constructor TFElectraMainLayer(config, **kwargs) - removed:
constructor TFElectraMaskedLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFElectraModel(config, *inputs, **kwargs) - removed:
constructor TFElectraOutput(config, **kwargs) - removed:
constructor TFElectraPooler(config, **kwargs) - removed:
constructor TFElectraSelfAttention(config, **kwargs) - removed:
constructor TFElectraSelfOutput(config, **kwargs) - removed:
field TFElectraForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFElectraForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFElectraForPreTrainingOutput.logits: tf.Tensor - removed:
field TFElectraMainLayer.config_class - removed:
field TFElectraPreTrainedModel.base_model_prefix - removed:
field TFElectraPreTrainedModel.config_class - removed:
method TFElectraAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFElectraAttention.prune_heads(heads) - removed:
method TFElectraClassificationHead.call(inputs, **kwargs) - removed:
method TFElectraDiscriminatorPredictions.call(discriminator_hidden_states, training=…) - removed:
method TFElectraEmbeddings.build(input_shape) - removed:
method TFElectraEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFElectraEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
method TFElectraForMaskedLM.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…, **kwargs) - removed:
method TFElectraForMaskedLM.get_output_embeddings() - removed:
method TFElectraForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFElectraForPreTraining.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFElectraForQuestionAnswering.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFElectraForSequenceClassification.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…, **kwargs) - removed:
method TFElectraForTokenClassification.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFElectraGeneratorPredictions.call(generator_hidden_states, training=…) - removed:
method TFElectraIntermediate.call(hidden_states) - removed:
method TFElectraLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFElectraMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFElectraMainLayer.get_extended_attention_mask(attention_mask, input_shape, dtype) - removed:
method TFElectraMainLayer.get_head_mask(head_mask) - removed:
method TFElectraMainLayer.get_input_embeddings() - removed:
method TFElectraMainLayer.set_input_embeddings(value) - removed:
method TFElectraMaskedLMHead.build(input_shape) - removed:
method TFElectraMaskedLMHead.call(hidden_states, training=…) - removed:
method TFElectraModel.call(inputs, **kwargs) - removed:
method TFElectraOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFElectraPooler.call(hidden_states) - removed:
method TFElectraSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFElectraSelfAttention.transpose_for_scores(x, batch_size) - removed:
method TFElectraSelfOutput.call(hidden_states, input_tensor, training=…) - removed:
property TFElectraForMultipleChoice.dummy_inputs - removed:
variable ELECTRA_INPUTS_DOCSTRING - removed:
variable ELECTRA_START_DOCSTRING - removed:
variable TF_ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFFlaubertForMultipleChoice(TFXLMForMultipleChoice) - removed:
class TFFlaubertForQuestionAnsweringSimple(TFXLMForQuestionAnsweringSimple) - removed:
class TFFlaubertForSequenceClassification(TFXLMForSequenceClassification) - removed:
class TFFlaubertForTokenClassification(TFXLMForTokenClassification) - removed:
class TFFlaubertMainLayer(tf.keras.layers.Layer) - removed:
class TFFlaubertModel(TFFlaubertPreTrainedModel) - removed:
class TFFlaubertMultiHeadAttention(tf.keras.layers.Layer) - removed:
class TFFlaubertPreTrainedModel(TFPreTrainedModel) - removed:
class TFFlaubertPredLayer(tf.keras.layers.Layer) - removed:
class TFFlaubertTransformerFFN(tf.keras.layers.Layer) - removed:
class TFFlaubertWithLMHeadModel(TFFlaubertPreTrainedModel) - removed:
class TFFlaubertWithLMHeadModelOutput(ModelOutput) - removed:
constructor TFFlaubertForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFFlaubertForQuestionAnsweringSimple(config, *inputs, **kwargs) - removed:
constructor TFFlaubertForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFFlaubertForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFFlaubertMainLayer(config, *inputs, **kwargs) - removed:
constructor TFFlaubertModel(config, *inputs, **kwargs) - removed:
constructor TFFlaubertMultiHeadAttention(n_heads, dim, config, **kwargs) - removed:
constructor TFFlaubertPredLayer(config, input_embeddings, **kwargs) - removed:
constructor TFFlaubertTransformerFFN(in_dim, dim_hidden, out_dim, config, **kwargs) - removed:
constructor TFFlaubertWithLMHeadModel(config, *inputs, **kwargs) - removed:
field TFFlaubertForQuestionAnsweringSimple.config_class - removed:
field TFFlaubertForSequenceClassification.config_class - removed:
field TFFlaubertMainLayer.config_class - removed:
field TFFlaubertMultiHeadAttention.NEW_ID - removed:
field TFFlaubertPreTrainedModel.base_model_prefix - removed:
field TFFlaubertPreTrainedModel.config_class - removed:
field TFFlaubertWithLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFFlaubertWithLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFFlaubertWithLMHeadModelOutput.logits: tf.Tensor - removed:
function get_masks(slen, lengths, causal, padding_mask=…, dtype=…) - removed:
method TFFlaubertMainLayer.call(inputs, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFFlaubertMainLayer.get_input_embeddings() - removed:
method TFFlaubertModel.call(inputs, **kwargs) - removed:
method TFFlaubertMultiHeadAttention.call(input, mask, kv, cache, head_mask, output_attentions, training=…) - removed:
method TFFlaubertMultiHeadAttention.prune_heads(heads) - removed:
method TFFlaubertPredLayer.build(input_shape) - removed:
method TFFlaubertPredLayer.call(hidden_states) - removed:
method TFFlaubertTransformerFFN.call(input, training=…) - removed:
method TFFlaubertWithLMHeadModel.call(inputs, **kwargs) - removed:
method TFFlaubertWithLMHeadModel.get_output_embeddings() - removed:
method TFFlaubertWithLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - removed:
property TFFlaubertPreTrainedModel.dummy_inputs - removed:
variable FLAUBERT_INPUTS_DOCSTRING - removed:
variable FLAUBERT_START_DOCSTRING - removed:
variable TF_FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFFunnelAttentionStructure - removed:
class TFFunnelBaseLayer(tf.keras.layers.Layer) - removed:
class TFFunnelBaseModel(TFFunnelPreTrainedModel) - removed:
class TFFunnelClassificationHead(tf.keras.layers.Layer) - removed:
class TFFunnelDecoder(tf.keras.layers.Layer) - removed:
class TFFunnelDiscriminatorPredictions(tf.keras.layers.Layer) - removed:
class TFFunnelEmbeddings(tf.keras.layers.Layer) - removed:
class TFFunnelEncoder(tf.keras.layers.Layer) - removed:
class TFFunnelForMaskedLM(TFFunnelPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFFunnelForMultipleChoice(TFFunnelPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFFunnelForPreTraining(TFFunnelPreTrainedModel) - removed:
class TFFunnelForPreTrainingOutput(ModelOutput) - removed:
class TFFunnelForQuestionAnswering(TFFunnelPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFFunnelForSequenceClassification(TFFunnelPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFFunnelForTokenClassification(TFFunnelPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFFunnelLayer(tf.keras.layers.Layer) - removed:
class TFFunnelMainLayer(tf.keras.layers.Layer) - removed:
class TFFunnelMaskedLMHead(tf.keras.layers.Layer) - removed:
class TFFunnelModel(TFFunnelPreTrainedModel) - removed:
class TFFunnelPositionwiseFFN(tf.keras.layers.Layer) - removed:
class TFFunnelPreTrainedModel(TFPreTrainedModel) - removed:
class TFFunnelRelMultiheadAttention(tf.keras.layers.Layer) - removed:
constructor TFFunnelAttentionStructure(config) - removed:
constructor TFFunnelBaseLayer(config, **kwargs) - removed:
constructor TFFunnelBaseModel(config, *inputs, **kwargs) - removed:
constructor TFFunnelClassificationHead(config, n_labels, **kwargs) - removed:
constructor TFFunnelDecoder(config, **kwargs) - removed:
constructor TFFunnelDiscriminatorPredictions(config, **kwargs) - removed:
constructor TFFunnelEmbeddings(config, **kwargs) - removed:
constructor TFFunnelEncoder(config, **kwargs) - removed:
constructor TFFunnelForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFFunnelForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFFunnelForPreTraining(config, **kwargs) - removed:
constructor TFFunnelForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFFunnelForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFFunnelForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFFunnelLayer(config, block_index, **kwargs) - removed:
constructor TFFunnelMainLayer(config, **kwargs) - removed:
constructor TFFunnelMaskedLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFFunnelModel(config, *inputs, **kwargs) - removed:
constructor TFFunnelPositionwiseFFN(config, **kwargs) - removed:
constructor TFFunnelRelMultiheadAttention(config, block_index, **kwargs) - removed:
field TFFunnelAttentionStructure.cls_token_type_id: int - removed:
field TFFunnelBaseLayer.config_class - removed:
field TFFunnelForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFFunnelForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFFunnelForPreTrainingOutput.logits: tf.Tensor - removed:
field TFFunnelMainLayer.config_class - removed:
field TFFunnelPreTrainedModel.base_model_prefix - removed:
field TFFunnelPreTrainedModel.config_class - removed:
function upsample(x, stride, target_len, separate_cls=…, truncate_seq=…) - removed:
method TFFunnelAttentionStructure.get_position_embeds(seq_len, dtype=…, training=…) - removed:
method TFFunnelAttentionStructure.init_attention_inputs(inputs_embeds, attention_mask=…, token_type_ids=…, training=…) - removed:
method TFFunnelAttentionStructure.pool_tensor(tensor, mode=…, stride=…) - removed:
method TFFunnelAttentionStructure.post_attention_pooling(attention_inputs) - removed:
method TFFunnelAttentionStructure.pre_attention_pooling(output, attention_inputs) - removed:
method TFFunnelAttentionStructure.relative_pos(pos, stride, pooled_pos=…, shift=…) - removed:
method TFFunnelAttentionStructure.stride_pool(tensor, axis) - removed:
method TFFunnelAttentionStructure.stride_pool_pos(pos_id, block_index) - removed:
method TFFunnelAttentionStructure.token_type_ids_to_mat(token_type_ids) - removed:
method TFFunnelBaseLayer.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFFunnelBaseLayer.get_input_embeddings() - removed:
method TFFunnelBaseLayer.set_input_embeddings(value) - removed:
method TFFunnelBaseModel.call(inputs, **kwargs) - removed:
method TFFunnelClassificationHead.call(hidden, training=…) - removed:
method TFFunnelDecoder.call(final_hidden, first_block_hidden, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFFunnelDiscriminatorPredictions.call(discriminator_hidden_states) - removed:
method TFFunnelEmbeddings.build(input_shape) - removed:
method TFFunnelEmbeddings.call(input_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFFunnelEncoder.call(inputs_embeds, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFFunnelForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFFunnelForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFFunnelForPreTraining.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFFunnelForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFFunnelForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFFunnelForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFFunnelLayer.call(query, key, value, attention_inputs, output_attentions=…, training=…) - removed:
method TFFunnelMainLayer.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFFunnelMainLayer.get_input_embeddings() - removed:
method TFFunnelMainLayer.set_input_embeddings(value) - removed:
method TFFunnelMaskedLMHead.build(input_shape) - removed:
method TFFunnelMaskedLMHead.call(hidden_states, training=…) - removed:
method TFFunnelModel.call(inputs, **kwargs) - removed:
method TFFunnelPositionwiseFFN.call(hidden, training=…) - removed:
method TFFunnelRelMultiheadAttention.build(input_shape) - removed:
method TFFunnelRelMultiheadAttention.call(query, key, value, attention_inputs, output_attentions=…, training=…) - removed:
method TFFunnelRelMultiheadAttention.relative_positional_attention(position_embeds, q_head, context_len, cls_mask=…) - removed:
method TFFunnelRelMultiheadAttention.relative_token_type_attention(token_type_mat, q_head, cls_mask=…) - removed:
property TFFunnelForMultipleChoice.dummy_inputs - removed:
variable FUNNEL_INPUTS_DOCSTRING - removed:
variable FUNNEL_START_DOCSTRING - removed:
variable INF - removed:
variable TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFAttention(tf.keras.layers.Layer) - removed:
class TFBlock(tf.keras.layers.Layer) - removed:
class TFGPT2DoubleHeadsModel(TFGPT2PreTrainedModel) - removed:
class TFGPT2DoubleHeadsModelOutput(ModelOutput) - removed:
class TFGPT2LMHeadModel(TFGPT2PreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFGPT2MainLayer(tf.keras.layers.Layer) - removed:
class TFGPT2Model(TFGPT2PreTrainedModel) - removed:
class TFGPT2PreTrainedModel(TFPreTrainedModel) - removed:
class TFMLP(tf.keras.layers.Layer) - removed:
constructor TFAttention(nx, n_ctx, config, scale=…, **kwargs) - removed:
constructor TFBlock(n_ctx, config, scale=…, **kwargs) - removed:
constructor TFGPT2DoubleHeadsModel(config, *inputs, **kwargs) - removed:
constructor TFGPT2LMHeadModel(config, *inputs, **kwargs) - removed:
constructor TFGPT2MainLayer(config, *inputs, **kwargs) - removed:
constructor TFGPT2Model(config, *inputs, **kwargs) - removed:
constructor TFMLP(n_state, config, **kwargs) - removed:
field TFGPT2DoubleHeadsModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFGPT2DoubleHeadsModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFGPT2DoubleHeadsModelOutput.logits: tf.Tensor - removed:
field TFGPT2DoubleHeadsModelOutput.mc_logits: tf.Tensor - removed:
field TFGPT2DoubleHeadsModelOutput.past_key_values: Optional[List[tf.Tensor]] - removed:
field TFGPT2MainLayer.config_class - removed:
field TFGPT2PreTrainedModel.base_model_prefix - removed:
field TFGPT2PreTrainedModel.config_class - removed:
method TFAttention.call(x, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - removed:
method TFAttention.merge_heads(x) - removed:
method TFAttention.prune_heads(heads) - removed:
method TFAttention.split_heads(x) - removed:
method TFBlock.call(x, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - removed:
method TFGPT2DoubleHeadsModel.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFGPT2DoubleHeadsModel.get_output_embeddings() - removed:
method TFGPT2LMHeadModel.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFGPT2LMHeadModel.get_output_embeddings() - removed:
method TFGPT2LMHeadModel.prepare_inputs_for_generation(inputs, past, **kwargs) - removed:
method TFGPT2MainLayer.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFGPT2MainLayer.get_input_embeddings() - removed:
method TFGPT2MainLayer.set_input_embeddings(value) - removed:
method TFGPT2Model.call(inputs, **kwargs) - removed:
method TFMLP.call(x, training=…) - removed:
static method TFAttention.causal_attention_mask(nd, ns, dtype) - removed:
variable GPT2_INPUTS_DOCSTRING - removed:
variable GPT2_START_DOCSTRING - removed:
variable TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFLongformerAttention(tf.keras.layers.Layer) - removed:
class TFLongformerBaseModelOutput(ModelOutput) - removed:
class TFLongformerBaseModelOutputWithPooling(ModelOutput) - removed:
class TFLongformerEmbeddings(tf.keras.layers.Layer) - removed:
class TFLongformerEncoder(tf.keras.layers.Layer) - removed:
class TFLongformerForMaskedLM(TFLongformerPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFLongformerForQuestionAnswering(TFLongformerPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFLongformerIntermediate(tf.keras.layers.Layer) - removed:
class TFLongformerLMHead(tf.keras.layers.Layer) - removed:
class TFLongformerLayer(tf.keras.layers.Layer) - removed:
class TFLongformerMainLayer(tf.keras.layers.Layer) - removed:
class TFLongformerModel(TFLongformerPreTrainedModel) - removed:
class TFLongformerOutput(tf.keras.layers.Layer) - removed:
class TFLongformerPooler(tf.keras.layers.Layer) - removed:
class TFLongformerPreTrainedModel(TFPreTrainedModel) - removed:
class TFLongformerQuestionAnsweringModelOutput(ModelOutput) - removed:
class TFLongformerSelfAttention(tf.keras.layers.Layer) - removed:
class TFLongformerSelfOutput(tf.keras.layers.Layer) - removed:
constructor TFLongformerAttention(config, layer_id=…, **kwargs) - removed:
constructor TFLongformerEmbeddings(config, **kwargs) - removed:
constructor TFLongformerEncoder(config, **kwargs) - removed:
constructor TFLongformerForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFLongformerForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFLongformerIntermediate(config, **kwargs) - removed:
constructor TFLongformerLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFLongformerLayer(config, layer_id=…, **kwargs) - removed:
constructor TFLongformerMainLayer(config, **kwargs) - removed:
constructor TFLongformerModel(config, *inputs, **kwargs) - removed:
constructor TFLongformerOutput(config, **kwargs) - removed:
constructor TFLongformerPooler(config, **kwargs) - removed:
constructor TFLongformerSelfAttention(config, layer_id, **kwargs) - removed:
constructor TFLongformerSelfOutput(config, **kwargs) - removed:
field TFLongformerBaseModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutput.global_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutput.last_hidden_state: tf.Tensor - removed:
field TFLongformerBaseModelOutputWithPooling.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutputWithPooling.global_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutputWithPooling.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerBaseModelOutputWithPooling.last_hidden_state: tf.Tensor - removed:
field TFLongformerBaseModelOutputWithPooling.pooler_output: tf.Tensor - removed:
field TFLongformerForMaskedLM.authorized_missing_keys - removed:
field TFLongformerForQuestionAnswering.authorized_missing_keys - removed:
field TFLongformerMainLayer.config_class - removed:
field TFLongformerPreTrainedModel.base_model_prefix - removed:
field TFLongformerPreTrainedModel.config_class - removed:
field TFLongformerQuestionAnsweringModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerQuestionAnsweringModelOutput.end_logits: tf.Tensor - removed:
field TFLongformerQuestionAnsweringModelOutput.global_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLongformerQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] - removed:
field TFLongformerQuestionAnsweringModelOutput.start_logits: tf.Tensor - removed:
method TFLongformerAttention.call(inputs, training=…) - removed:
method TFLongformerAttention.prune_heads(heads) - removed:
method TFLongformerEmbeddings.build(input_shape) - removed:
method TFLongformerEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFLongformerEmbeddings.create_position_ids_from_input_ids(x) - removed:
method TFLongformerEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - removed:
method TFLongformerEncoder.call(hidden_states, attention_mask=…, head_mask=…, padding_len=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFLongformerForMaskedLM.call(inputs=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFLongformerForMaskedLM.get_output_embeddings() - removed:
method TFLongformerForQuestionAnswering.call(inputs=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFLongformerIntermediate.call(hidden_states) - removed:
method TFLongformerLMHead.build(input_shape) - removed:
method TFLongformerLMHead.call(features) - removed:
method TFLongformerLayer.call(inputs, training=…) - removed:
method TFLongformerMainLayer.call(inputs, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFLongformerMainLayer.get_input_embeddings() - removed:
method TFLongformerMainLayer.set_input_embeddings(value) - removed:
method TFLongformerModel.call(inputs, **kwargs) - removed:
method TFLongformerOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFLongformerPooler.call(hidden_states) - removed:
method TFLongformerSelfAttention.call(inputs, training=…) - removed:
method TFLongformerSelfAttention.reshape_and_transpose(vector, batch_size) - removed:
method TFLongformerSelfOutput.call(hidden_states, input_tensor, training=…) - removed:
property TFLongformerPreTrainedModel.dummy_inputs - removed:
variable LONGFORMER_INPUTS_DOCSTRING - removed:
variable LONGFORMER_START_DOCSTRING - removed:
variable TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFLxmertAttention(tf.keras.layers.Layer) - removed:
class TFLxmertAttentionOutput(tf.keras.layers.Layer) - removed:
class TFLxmertCrossAttentionLayer(tf.keras.layers.Layer) - removed:
class TFLxmertEmbeddings(tf.keras.layers.Layer) - removed:
class TFLxmertEncoder(tf.keras.layers.Layer) - removed:
class TFLxmertForPreTraining(TFLxmertPreTrainedModel) - removed:
class TFLxmertForPreTrainingOutput(ModelOutput) - removed:
class TFLxmertIntermediate(tf.keras.layers.Layer) - removed:
class TFLxmertLMPredictionHead(tf.keras.layers.Layer) - removed:
class TFLxmertLayer(tf.keras.layers.Layer) - removed:
class TFLxmertMLMHead(tf.keras.layers.Layer) - removed:
class TFLxmertMainLayer(tf.keras.layers.Layer) - removed:
class TFLxmertModel(TFLxmertPreTrainedModel) - removed:
class TFLxmertModelOutput(ModelOutput) - removed:
class TFLxmertOutput(tf.keras.layers.Layer) - removed:
class TFLxmertPooler(tf.keras.layers.Layer) - removed:
class TFLxmertPreTrainedModel(TFPreTrainedModel) - removed:
class TFLxmertPreTrainingHeads(tf.keras.layers.Layer) - removed:
class TFLxmertPredictionHeadTransform(tf.keras.layers.Layer) - removed:
class TFLxmertSelfAttentionLayer(tf.keras.layers.Layer) - removed:
class TFLxmertVisualAnswerHead(tf.keras.layers.Layer) - removed:
class TFLxmertVisualFeatureEncoder(tf.keras.layers.Layer) - removed:
class TFLxmertVisualObjHead(tf.keras.layers.Layer) - removed:
class TFLxmertXLayer(tf.keras.layers.Layer) - removed:
constructor TFLxmertAttention(config, **kwargs) - removed:
constructor TFLxmertAttentionOutput(config, **kwargs) - removed:
constructor TFLxmertCrossAttentionLayer(config, **kwargs) - removed:
constructor TFLxmertEmbeddings(config, **kwargs) - removed:
constructor TFLxmertEncoder(config, **kwargs) - removed:
constructor TFLxmertForPreTraining(config, *inputs, **kwargs) - removed:
constructor TFLxmertIntermediate(config, **kwargs) - removed:
constructor TFLxmertLMPredictionHead(config, input_embeddings, **kwargs) - removed:
constructor TFLxmertLayer(config, **kwargs) - removed:
constructor TFLxmertMLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFLxmertMainLayer(config, **kwargs) - removed:
constructor TFLxmertModel(config, *inputs, **kwargs) - removed:
constructor TFLxmertOutput(config, **kwargs) - removed:
constructor TFLxmertPooler(config, **kwargs) - removed:
constructor TFLxmertPreTrainingHeads(config, input_embeddings, **kwargs) - removed:
constructor TFLxmertPredictionHeadTransform(config, **kwargs) - removed:
constructor TFLxmertSelfAttentionLayer(config, **kwargs) - removed:
constructor TFLxmertVisualAnswerHead(config, num_labels, **kwargs) - removed:
constructor TFLxmertVisualFeatureEncoder(config, **kwargs) - removed:
constructor TFLxmertVisualObjHead(config, **kwargs) - removed:
constructor TFLxmertXLayer(config, **kwargs) - removed:
field TFLxmertForPreTrainingOutput.cross_encoder_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertForPreTrainingOutput.cross_relationship_score: Optional[tf.Tensor] - removed:
field TFLxmertForPreTrainingOutput.language_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertForPreTrainingOutput.language_hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertForPreTrainingOutput.loss: [tf.Tensor] - removed:
field TFLxmertForPreTrainingOutput.prediction_logits: Optional[tf.Tensor] - removed:
field TFLxmertForPreTrainingOutput.question_answering_score: Optional[tf.Tensor] - removed:
field TFLxmertForPreTrainingOutput.vision_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertForPreTrainingOutput.vision_hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertMainLayer.config_class - removed:
field TFLxmertModelOutput.cross_encoder_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertModelOutput.language_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertModelOutput.language_hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertModelOutput.language_output: Optional[tf.Tensor] - removed:
field TFLxmertModelOutput.pooled_output: Optional[tf.Tensor] - removed:
field TFLxmertModelOutput.vision_attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertModelOutput.vision_hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFLxmertModelOutput.vision_output: Optional[tf.Tensor] - removed:
field TFLxmertPreTrainedModel.base_model_prefix - removed:
field TFLxmertPreTrainedModel.config_class - removed:
method TFLxmertAttention.call(hidden_states, context, attention_mask, output_attentions, training=…) - removed:
method TFLxmertAttention.transpose_for_scores(x, batch_size) - removed:
method TFLxmertAttentionOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFLxmertCrossAttentionLayer.call(input_tensor, ctx_tensor, ctx_att_mask, output_attentions=…, training=…) - removed:
method TFLxmertEmbeddings.build(input_shape) - removed:
method TFLxmertEmbeddings.call(inputs, mode=…, training=…) - removed:
method TFLxmertEncoder.call(lang_feats=…, lang_attention_mask=…, visual_feats=…, visual_pos=…, visual_attention_mask=…, output_attentions=…, training=…) - removed:
method TFLxmertForPreTraining.call(inputs=…, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, masked_lm_labels=…, obj_labels=…, matched_label=…, ans=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method TFLxmertIntermediate.call(hidden_states) - removed:
method TFLxmertLMPredictionHead.build(input_shape) - removed:
method TFLxmertLMPredictionHead.call(hidden_states) - removed:
method TFLxmertLayer.call(hidden_states, attention_mask, output_attentions, training=…) - removed:
method TFLxmertMLMHead.call(sequence_output) - removed:
method TFLxmertMainLayer.call(inputs, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFLxmertMainLayer.get_input_embeddings() - removed:
method TFLxmertMainLayer.set_input_embeddings(value) - removed:
method TFLxmertModel.call(inputs, *args, **kwargs) - removed:
method TFLxmertOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFLxmertPooler.call(hidden_states) - removed:
method TFLxmertPreTrainingHeads.call(sequence_output, pooled_output) - removed:
method TFLxmertPredictionHeadTransform.call(hidden_states) - removed:
method TFLxmertSelfAttentionLayer.call(input_tensor, attention_mask, output_attentions, training=…) - removed:
method TFLxmertVisualAnswerHead.call(hidden_states) - removed:
method TFLxmertVisualFeatureEncoder.call(visn_input, training=…) - removed:
method TFLxmertVisualObjHead.call(hidden_states) - removed:
method TFLxmertXLayer.call(lang_feats, lang_attention_mask, visn_feats, visn_attention_mask, output_attentions, training=…) - removed:
method TFLxmertXLayer.cross_att(lang_input, lang_attention_mask, visn_input, visn_attention_mask, output_attentions, training=…) - removed:
method TFLxmertXLayer.output_fc(lang_input, visn_input, training=…) - removed:
method TFLxmertXLayer.self_att(lang_input, lang_attention_mask, visn_input, visn_attention_mask, training=…) - removed:
property TFLxmertForPreTraining.dummy_inputs - removed:
property TFLxmertMainLayer.dummy_inputs - removed:
property TFLxmertPreTrainedModel.dummy_inputs -> Dict[str, tf.Tensor] - removed:
variable LXMERT_INPUTS_DOCSTRING - removed:
variable LXMERT_START_DOCSTRING - removed:
variable TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFMarianMTModel(TFBartForConditionalGeneration) - removed:
field TFMarianMTModel.authorized_missing_keys - removed:
field TFMarianMTModel.config_class - removed:
method TFMarianMTModel.adjust_logits_during_generation(logits, cur_len, max_length) - removed:
variable START_DOCSTRING - removed:
variable logger
- removed:
class TFMBartForConditionalGeneration(TFBartForConditionalGeneration) - removed:
field TFMBartForConditionalGeneration.config_class - removed:
variable START_DOCSTRING - removed:
variable logger
- removed:
class TFBottleneck(tf.keras.layers.Layer) - removed:
class TFBottleneckLayer(tf.keras.layers.Layer) - removed:
class TFFFNLayer(tf.keras.layers.Layer) - removed:
class TFFFNOutput(tf.keras.layers.Layer) - removed:
class TFLayerNorm(tf.keras.layers.LayerNormalization) - removed:
class TFMobileBertAttention(tf.keras.layers.Layer) - removed:
class TFMobileBertEmbeddings(tf.keras.layers.Layer) - removed:
class TFMobileBertEncoder(tf.keras.layers.Layer) - removed:
class TFMobileBertForMaskedLM(TFMobileBertPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFMobileBertForMultipleChoice(TFMobileBertPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFMobileBertForNextSentencePrediction(TFMobileBertPreTrainedModel) - removed:
class TFMobileBertForPreTraining(TFMobileBertPreTrainedModel) - removed:
class TFMobileBertForPreTrainingOutput(ModelOutput) - removed:
class TFMobileBertForQuestionAnswering(TFMobileBertPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFMobileBertForSequenceClassification(TFMobileBertPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFMobileBertForTokenClassification(TFMobileBertPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFMobileBertIntermediate(tf.keras.layers.Layer) - removed:
class TFMobileBertLMPredictionHead(tf.keras.layers.Layer) - removed:
class TFMobileBertLayer(tf.keras.layers.Layer) - removed:
class TFMobileBertMLMHead(tf.keras.layers.Layer) - removed:
class TFMobileBertMainLayer(tf.keras.layers.Layer) - removed:
class TFMobileBertModel(TFMobileBertPreTrainedModel) - removed:
class TFMobileBertOnlyNSPHead(tf.keras.layers.Layer) - removed:
class TFMobileBertOutput(tf.keras.layers.Layer) - removed:
class TFMobileBertPooler(tf.keras.layers.Layer) - removed:
class TFMobileBertPreTrainedModel(TFPreTrainedModel) - removed:
class TFMobileBertPredictionHeadTransform(tf.keras.layers.Layer) - removed:
class TFMobileBertSelfAttention(tf.keras.layers.Layer) - removed:
class TFMobileBertSelfOutput(tf.keras.layers.Layer) - removed:
class TFNoNorm(tf.keras.layers.Layer) - removed:
class TFOutputBottleneck(tf.keras.layers.Layer) - removed:
constructor TFBottleneck(config, **kwargs) - removed:
constructor TFBottleneckLayer(config, **kwargs) - removed:
constructor TFFFNLayer(config, **kwargs) - removed:
constructor TFFFNOutput(config, **kwargs) - removed:
constructor TFLayerNorm(feat_size, *args, **kwargs) - removed:
constructor TFMobileBertAttention(config, **kwargs) - removed:
constructor TFMobileBertEmbeddings(config, **kwargs) - removed:
constructor TFMobileBertEncoder(config, **kwargs) - removed:
constructor TFMobileBertForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForNextSentencePrediction(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForPreTraining(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFMobileBertForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFMobileBertIntermediate(config, **kwargs) - removed:
constructor TFMobileBertLMPredictionHead(config, **kwargs) - removed:
constructor TFMobileBertLayer(config, **kwargs) - removed:
constructor TFMobileBertMLMHead(config, **kwargs) - removed:
constructor TFMobileBertMainLayer(config, **kwargs) - removed:
constructor TFMobileBertModel(config, *inputs, **kwargs) - removed:
constructor TFMobileBertOnlyNSPHead(config, **kwargs) - removed:
constructor TFMobileBertOutput(config, **kwargs) - removed:
constructor TFMobileBertPooler(config, **kwargs) - removed:
constructor TFMobileBertPredictionHeadTransform(config, **kwargs) - removed:
constructor TFMobileBertSelfAttention(config, **kwargs) - removed:
constructor TFMobileBertSelfOutput(config, **kwargs) - removed:
constructor TFNoNorm(feat_size, epsilon=…, **kwargs) - removed:
constructor TFOutputBottleneck(config, **kwargs) - removed:
field TFMobileBertForMaskedLM.authorized_missing_keys - removed:
field TFMobileBertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFMobileBertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFMobileBertForPreTrainingOutput.loss: Optional[tf.Tensor] - removed:
field TFMobileBertForPreTrainingOutput.prediction_logits: tf.Tensor - removed:
field TFMobileBertForPreTrainingOutput.seq_relationship_logits: tf.Tensor - removed:
field TFMobileBertForQuestionAnswering.authorized_missing_keys - removed:
field TFMobileBertForTokenClassification.authorized_missing_keys - removed:
field TFMobileBertMainLayer.config_class - removed:
field TFMobileBertPreTrainedModel.base_model_prefix - removed:
field TFMobileBertPreTrainedModel.config_class - removed:
method TFBottleneck.call(hidden_states) - removed:
method TFBottleneckLayer.call(inputs) - removed:
method TFFFNLayer.call(hidden_states) - removed:
method TFFFNOutput.call(hidden_states, residual_tensor) - removed:
method TFMobileBertAttention.call(query_tensor, key_tensor, value_tensor, layer_input, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFMobileBertAttention.prune_heads(heads) - removed:
method TFMobileBertEmbeddings.build(input_shape) - removed:
method TFMobileBertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFMobileBertEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
method TFMobileBertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFMobileBertForMaskedLM.get_output_embeddings() - removed:
method TFMobileBertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFMobileBertForNextSentencePrediction.call(inputs, **kwargs) - removed:
method TFMobileBertForPreTraining.call(inputs, **kwargs) - removed:
method TFMobileBertForPreTraining.get_output_embeddings() - removed:
method TFMobileBertForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFMobileBertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFMobileBertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFMobileBertIntermediate.call(hidden_states) - removed:
method TFMobileBertLMPredictionHead.build(input_shape) - removed:
method TFMobileBertLMPredictionHead.call(hidden_states) - removed:
method TFMobileBertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFMobileBertMLMHead.call(sequence_output) - removed:
method TFMobileBertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFMobileBertMainLayer.get_input_embeddings() - removed:
method TFMobileBertModel.call(inputs, **kwargs) - removed:
method TFMobileBertOnlyNSPHead.call(pooled_output) - removed:
method TFMobileBertOutput.call(hidden_states, residual_tensor_1, residual_tensor_2, training=…) - removed:
method TFMobileBertPooler.call(hidden_states) - removed:
method TFMobileBertPredictionHeadTransform.call(hidden_states) - removed:
method TFMobileBertSelfAttention.call(query_tensor, key_tensor, value_tensor, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFMobileBertSelfAttention.transpose_for_scores(x, batch_size) - removed:
method TFMobileBertSelfOutput.call(hidden_states, residual_tensor, training=…) - removed:
method TFNoNorm.build(input_shape) - removed:
method TFNoNorm.call(inputs: tf.Tensor) - removed:
method TFOutputBottleneck.call(hidden_states, residual_tensor, training=…) - removed:
property TFMobileBertForMultipleChoice.dummy_inputs - removed:
variable MOBILEBERT_INPUTS_DOCSTRING - removed:
variable MOBILEBERT_START_DOCSTRING - removed:
variable NORM2FN - removed:
variable TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFAttention(tf.keras.layers.Layer) - removed:
class TFBlock(tf.keras.layers.Layer) - removed:
class TFMLP(tf.keras.layers.Layer) - removed:
class TFOpenAIGPTDoubleHeadsModel(TFOpenAIGPTPreTrainedModel) - removed:
class TFOpenAIGPTDoubleHeadsModelOutput(ModelOutput) - removed:
class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFOpenAIGPTMainLayer(tf.keras.layers.Layer) - removed:
class TFOpenAIGPTModel(TFOpenAIGPTPreTrainedModel) - removed:
class TFOpenAIGPTPreTrainedModel(TFPreTrainedModel) - removed:
constructor TFAttention(nx, n_ctx, config, scale=…, **kwargs) - removed:
constructor TFBlock(n_ctx, config, scale=…, **kwargs) - removed:
constructor TFMLP(n_state, config, **kwargs) - removed:
constructor TFOpenAIGPTDoubleHeadsModel(config, *inputs, **kwargs) - removed:
constructor TFOpenAIGPTLMHeadModel(config, *inputs, **kwargs) - removed:
constructor TFOpenAIGPTMainLayer(config, *inputs, **kwargs) - removed:
constructor TFOpenAIGPTModel(config, *inputs, **kwargs) - removed:
field TFOpenAIGPTDoubleHeadsModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFOpenAIGPTDoubleHeadsModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFOpenAIGPTDoubleHeadsModelOutput.logits: tf.Tensor - removed:
field TFOpenAIGPTDoubleHeadsModelOutput.mc_logits: tf.Tensor - removed:
field TFOpenAIGPTMainLayer.config_class - removed:
field TFOpenAIGPTPreTrainedModel.base_model_prefix - removed:
field TFOpenAIGPTPreTrainedModel.config_class - removed:
method TFAttention.call(x, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFAttention.merge_heads(x) - removed:
method TFAttention.prune_heads(heads) - removed:
method TFAttention.split_heads(x) - removed:
method TFBlock.call(x, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFMLP.call(x, training=…) - removed:
method TFOpenAIGPTDoubleHeadsModel.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFOpenAIGPTDoubleHeadsModel.get_output_embeddings() - removed:
method TFOpenAIGPTLMHeadModel.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFOpenAIGPTLMHeadModel.get_output_embeddings() - removed:
method TFOpenAIGPTMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFOpenAIGPTMainLayer.get_input_embeddings() - removed:
method TFOpenAIGPTMainLayer.set_input_embeddings(value) - removed:
method TFOpenAIGPTModel.call(inputs, **kwargs) - removed:
static method TFAttention.causal_attention_mask(nd, ns, dtype) - removed:
variable OPENAI_GPT_INPUTS_DOCSTRING - removed:
variable OPENAI_GPT_START_DOCSTRING - removed:
variable TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- changed:
field TFBaseModelOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutput.last_hidden_state: tf.Tensor → field TFBaseModelOutput.last_hidden_state: tf.Tensor | None - changed:
field TFBaseModelOutputWithPast.attentions: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutputWithPast.attentions: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutputWithPast.hidden_states: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutputWithPast.hidden_states: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutputWithPast.last_hidden_state: tf.Tensor → field TFBaseModelOutputWithPast.last_hidden_state: tf.Tensor | None - changed:
field TFBaseModelOutputWithPast.past_key_values: Optional[List[tf.Tensor]] → field TFBaseModelOutputWithPast.past_key_values: list[tf.Tensor] | None - changed:
field TFBaseModelOutputWithPooling.attentions: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutputWithPooling.attentions: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutputWithPooling.hidden_states: Optional[Tuple[tf.Tensor]] → field TFBaseModelOutputWithPooling.hidden_states: tuple[tf.Tensor] | None - changed:
field TFBaseModelOutputWithPooling.last_hidden_state: tf.Tensor → field TFBaseModelOutputWithPooling.last_hidden_state: tf.Tensor | None - changed:
field TFBaseModelOutputWithPooling.pooler_output: tf.Tensor → field TFBaseModelOutputWithPooling.pooler_output: tf.Tensor | None - changed:
field TFCausalLMOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFCausalLMOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFCausalLMOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFCausalLMOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFCausalLMOutput.logits: tf.Tensor → field TFCausalLMOutput.logits: tf.Tensor | None - changed:
field TFCausalLMOutput.loss: Optional[tf.Tensor] → field TFCausalLMOutput.loss: tf.Tensor | None - changed:
field TFCausalLMOutputWithPast.attentions: Optional[Tuple[tf.Tensor]] → field TFCausalLMOutputWithPast.attentions: tuple[tf.Tensor] | None - changed:
field TFCausalLMOutputWithPast.hidden_states: Optional[Tuple[tf.Tensor]] → field TFCausalLMOutputWithPast.hidden_states: tuple[tf.Tensor] | None - changed:
field TFCausalLMOutputWithPast.logits: tf.Tensor → field TFCausalLMOutputWithPast.logits: tf.Tensor | None - changed:
field TFCausalLMOutputWithPast.loss: Optional[tf.Tensor] → field TFCausalLMOutputWithPast.loss: tf.Tensor | None - changed:
field TFCausalLMOutputWithPast.past_key_values: Optional[List[tf.Tensor]] → field TFCausalLMOutputWithPast.past_key_values: list[tf.Tensor] | None - changed:
field TFMaskedLMOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFMaskedLMOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFMaskedLMOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFMaskedLMOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFMaskedLMOutput.logits: tf.Tensor → field TFMaskedLMOutput.logits: tf.Tensor | None - changed:
field TFMaskedLMOutput.loss: Optional[tf.Tensor] → field TFMaskedLMOutput.loss: tf.Tensor | None - changed:
field TFMultipleChoiceModelOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFMultipleChoiceModelOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFMultipleChoiceModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFMultipleChoiceModelOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFMultipleChoiceModelOutput.logits: tf.Tensor → field TFMultipleChoiceModelOutput.logits: tf.Tensor | None - changed:
field TFMultipleChoiceModelOutput.loss: Optional[tf.Tensor] → field TFMultipleChoiceModelOutput.loss: tf.Tensor | None - changed:
field TFNextSentencePredictorOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFNextSentencePredictorOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFNextSentencePredictorOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFNextSentencePredictorOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFNextSentencePredictorOutput.logits: tf.Tensor → field TFNextSentencePredictorOutput.logits: tf.Tensor | None - changed:
field TFQuestionAnsweringModelOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFQuestionAnsweringModelOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFQuestionAnsweringModelOutput.end_logits: tf.Tensor → field TFQuestionAnsweringModelOutput.end_logits: tf.Tensor | None - changed:
field TFQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFQuestionAnsweringModelOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] → field TFQuestionAnsweringModelOutput.loss: tf.Tensor | None - changed:
field TFQuestionAnsweringModelOutput.start_logits: tf.Tensor → field TFQuestionAnsweringModelOutput.start_logits: tf.Tensor | None - changed:
field TFSeq2SeqLMOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqLMOutput.decoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqLMOutput.decoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqLMOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqLMOutput.encoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqLMOutput.encoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqLMOutput.encoder_last_hidden_state: Optional[tf.Tensor] → field TFSeq2SeqLMOutput.encoder_last_hidden_state: tf.Tensor | None - changed:
field TFSeq2SeqLMOutput.logits: tf.Tensor → field TFSeq2SeqLMOutput.logits: tf.Tensor | None - changed:
field TFSeq2SeqLMOutput.loss: Optional[tf.Tensor] → field TFSeq2SeqLMOutput.loss: tf.Tensor | None - changed:
field TFSeq2SeqLMOutput.past_key_values: Optional[List[tf.Tensor]] → field TFSeq2SeqLMOutput.past_key_values: list[tf.Tensor] | None - changed:
field TFSeq2SeqModelOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqModelOutput.decoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqModelOutput.decoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqModelOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqModelOutput.encoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqModelOutput.encoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqModelOutput.encoder_last_hidden_state: Optional[tf.Tensor] → field TFSeq2SeqModelOutput.encoder_last_hidden_state: tf.Tensor | None - changed:
field TFSeq2SeqModelOutput.last_hidden_state: tf.Tensor → field TFSeq2SeqModelOutput.last_hidden_state: tf.Tensor | None - changed:
field TFSeq2SeqModelOutput.past_key_values: Optional[List[tf.Tensor]] → field TFSeq2SeqModelOutput.past_key_values: list[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqQuestionAnsweringModelOutput.decoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqQuestionAnsweringModelOutput.encoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_last_hidden_state: Optional[tf.Tensor] → field TFSeq2SeqQuestionAnsweringModelOutput.encoder_last_hidden_state: tf.Tensor | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.end_logits: tf.Tensor → field TFSeq2SeqQuestionAnsweringModelOutput.end_logits: tf.Tensor | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] → field TFSeq2SeqQuestionAnsweringModelOutput.loss: tf.Tensor | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[List[tf.Tensor]] → field TFSeq2SeqQuestionAnsweringModelOutput.past_key_values: list[tf.Tensor] | None - changed:
field TFSeq2SeqQuestionAnsweringModelOutput.start_logits: tf.Tensor → field TFSeq2SeqQuestionAnsweringModelOutput.start_logits: tf.Tensor | None - changed:
field TFSeq2SeqSequenceClassifierOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqSequenceClassifierOutput.decoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqSequenceClassifierOutput.decoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqSequenceClassifierOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqSequenceClassifierOutput.encoder_attentions: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] → field TFSeq2SeqSequenceClassifierOutput.encoder_hidden_states: tuple[tf.Tensor] | None - changed:
field TFSeq2SeqSequenceClassifierOutput.encoder_last_hidden_state: Optional[tf.Tensor] → field TFSeq2SeqSequenceClassifierOutput.encoder_last_hidden_state: tf.Tensor | None - changed:
field TFSeq2SeqSequenceClassifierOutput.logits: tf.Tensor → field TFSeq2SeqSequenceClassifierOutput.logits: tf.Tensor | None - changed:
field TFSeq2SeqSequenceClassifierOutput.loss: Optional[tf.Tensor] → field TFSeq2SeqSequenceClassifierOutput.loss: tf.Tensor | None - changed:
field TFSeq2SeqSequenceClassifierOutput.past_key_values: Optional[List[tf.Tensor]] → field TFSeq2SeqSequenceClassifierOutput.past_key_values: list[tf.Tensor] | None - changed:
field TFSequenceClassifierOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFSequenceClassifierOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFSequenceClassifierOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFSequenceClassifierOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFSequenceClassifierOutput.logits: tf.Tensor → field TFSequenceClassifierOutput.logits: tf.Tensor | None - changed:
field TFSequenceClassifierOutput.loss: Optional[tf.Tensor] → field TFSequenceClassifierOutput.loss: tf.Tensor | None - changed:
field TFTokenClassifierOutput.attentions: Optional[Tuple[tf.Tensor]] → field TFTokenClassifierOutput.attentions: tuple[tf.Tensor] | None - changed:
field TFTokenClassifierOutput.hidden_states: Optional[Tuple[tf.Tensor]] → field TFTokenClassifierOutput.hidden_states: tuple[tf.Tensor] | None - changed:
field TFTokenClassifierOutput.logits: tf.Tensor → field TFTokenClassifierOutput.logits: tf.Tensor | None - changed:
field TFTokenClassifierOutput.loss: Optional[tf.Tensor] → field TFTokenClassifierOutput.loss: tf.Tensor | None - added:
class TFBaseModelOutputWithCrossAttentions(ModelOutput) - added:
class TFBaseModelOutputWithNoAttention(ModelOutput) - added:
class TFBaseModelOutputWithPastAndCrossAttentions(ModelOutput) - added:
class TFBaseModelOutputWithPoolingAndCrossAttentions(ModelOutput) - added:
class TFBaseModelOutputWithPoolingAndNoAttention(ModelOutput) - added:
class TFCausalLMOutputWithCrossAttentions(ModelOutput) - added:
class TFImageClassifierOutput(ModelOutput) - added:
class TFImageClassifierOutputWithNoAttention(ModelOutput) - added:
class TFMaskedImageModelingOutput(ModelOutput) - added:
class TFSemanticSegmenterOutput(ModelOutput) - added:
class TFSemanticSegmenterOutputWithNoAttention(ModelOutput) - added:
class TFSequenceClassifierOutputWithPast(ModelOutput) - added:
field TFBaseModelOutputWithCrossAttentions.attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithCrossAttentions.cross_attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithCrossAttentions.hidden_states: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithCrossAttentions.last_hidden_state: tf.Tensor | None - added:
field TFBaseModelOutputWithNoAttention.hidden_states: tuple[tf.Tensor, ...] | None - added:
field TFBaseModelOutputWithNoAttention.last_hidden_state: tf.Tensor | None - added:
field TFBaseModelOutputWithPastAndCrossAttentions.attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPastAndCrossAttentions.cross_attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPastAndCrossAttentions.hidden_states: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPastAndCrossAttentions.last_hidden_state: tf.Tensor | None - added:
field TFBaseModelOutputWithPastAndCrossAttentions.past_key_values: list[tf.Tensor] | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.cross_attentions: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.hidden_states: tuple[tf.Tensor] | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.last_hidden_state: tf.Tensor | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.past_key_values: list[tf.Tensor] | None - added:
field TFBaseModelOutputWithPoolingAndCrossAttentions.pooler_output: tf.Tensor | None - added:
field TFBaseModelOutputWithPoolingAndNoAttention.hidden_states: tuple[tf.Tensor, ...] | None - added:
field TFBaseModelOutputWithPoolingAndNoAttention.last_hidden_state: tf.Tensor | None - added:
field TFBaseModelOutputWithPoolingAndNoAttention.pooler_output: tf.Tensor | None - added:
field TFCausalLMOutputWithCrossAttentions.attentions: tuple[tf.Tensor] | None - added:
field TFCausalLMOutputWithCrossAttentions.cross_attentions: tuple[tf.Tensor] | None - added:
field TFCausalLMOutputWithCrossAttentions.hidden_states: tuple[tf.Tensor] | None - added:
field TFCausalLMOutputWithCrossAttentions.logits: tf.Tensor | None - added:
field TFCausalLMOutputWithCrossAttentions.loss: tf.Tensor | None - added:
field TFCausalLMOutputWithCrossAttentions.past_key_values: list[tf.Tensor] | None - added:
field TFImageClassifierOutput.attentions: tuple[tf.Tensor] | None - added:
field TFImageClassifierOutput.hidden_states: tuple[tf.Tensor] | None - added:
field TFImageClassifierOutput.logits: tf.Tensor | None - added:
field TFImageClassifierOutput.loss: tf.Tensor | None - added:
field TFImageClassifierOutputWithNoAttention.hidden_states: tuple[tf.Tensor, ...] | None - added:
field TFImageClassifierOutputWithNoAttention.logits: tf.Tensor | None - added:
field TFImageClassifierOutputWithNoAttention.loss: tf.Tensor | None - added:
field TFMaskedImageModelingOutput.attentions: tuple[tf.Tensor] | None - added:
field TFMaskedImageModelingOutput.hidden_states: tuple[tf.Tensor] | None - added:
field TFMaskedImageModelingOutput.loss: tf.Tensor | None - added:
field TFMaskedImageModelingOutput.reconstruction: tf.Tensor | None - added:
field TFNextSentencePredictorOutput.loss: tf.Tensor | None - added:
field TFSemanticSegmenterOutput.attentions: tuple[tf.Tensor] | None - added:
field TFSemanticSegmenterOutput.hidden_states: tuple[tf.Tensor] | None - added:
field TFSemanticSegmenterOutput.logits: tf.Tensor | None - added:
field TFSemanticSegmenterOutput.loss: tf.Tensor | None - added:
field TFSemanticSegmenterOutputWithNoAttention.hidden_states: tuple[tf.Tensor] | None - added:
field TFSemanticSegmenterOutputWithNoAttention.logits: tf.Tensor | None - added:
field TFSemanticSegmenterOutputWithNoAttention.loss: tf.Tensor | None - added:
field TFSeq2SeqLMOutput.cross_attentions: tuple[tf.Tensor] | None - added:
field TFSeq2SeqModelOutput.cross_attentions: tuple[tf.Tensor] | None - added:
field TFSeq2SeqSequenceClassifierOutput.cross_attentions: tuple[tf.Tensor] | None - added:
field TFSequenceClassifierOutputWithPast.attentions: tuple[tf.Tensor] | None - added:
field TFSequenceClassifierOutputWithPast.hidden_states: tuple[tf.Tensor] | None - added:
field TFSequenceClassifierOutputWithPast.logits: tf.Tensor | None - added:
field TFSequenceClassifierOutputWithPast.loss: tf.Tensor | None - added:
field TFSequenceClassifierOutputWithPast.past_key_values: list[tf.Tensor] | None - added:
property TFMaskedImageModelingOutput.logits
- removed:
class TFPegasusForConditionalGeneration(TFBartForConditionalGeneration) - removed:
field TFPegasusForConditionalGeneration.authorized_missing_keys - removed:
field TFPegasusForConditionalGeneration.config_class - removed:
variable START_DOCSTRING - removed:
variable logger
- changed:
function convert_tf_weight_name_to_pt_weight_name(tf_name, start_prefix_to_remove=…) → function convert_tf_weight_name_to_pt_weight_name(tf_name, start_prefix_to_remove=…, tf_weight_shape=…, name_scope=…) - changed:
function load_pytorch_checkpoint_in_tf2_model(tf_model, pytorch_checkpoint_path, tf_inputs=…, allow_missing_keys=…) → function load_pytorch_checkpoint_in_tf2_model(tf_model, pytorch_checkpoint_path, tf_inputs=…, allow_missing_keys=…, output_loading_info=…, _prefix=…, tf_to_pt_weight_rename=…) - changed:
function load_pytorch_weights_in_tf2_model(tf_model, pt_state_dict, tf_inputs=…, allow_missing_keys=…) → function load_pytorch_weights_in_tf2_model(tf_model, pt_state_dict, tf_inputs=…, allow_missing_keys=…, output_loading_info=…, _prefix=…, tf_to_pt_weight_rename=…) - changed:
function load_tf2_checkpoint_in_pytorch_model(pt_model, tf_checkpoint_path, tf_inputs=…, allow_missing_keys=…) → function load_tf2_checkpoint_in_pytorch_model(pt_model, tf_checkpoint_path, tf_inputs=…, allow_missing_keys=…, output_loading_info=…) - changed:
function load_tf2_model_in_pytorch_model(pt_model, tf_model, allow_missing_keys=…) → function load_tf2_model_in_pytorch_model(pt_model, tf_model, allow_missing_keys=…, output_loading_info=…) - changed:
function load_tf2_weights_in_pytorch_model(pt_model, tf_weights, allow_missing_keys=…) → function load_tf2_weights_in_pytorch_model(pt_model, tf_weights, allow_missing_keys=…, output_loading_info=…) - added:
class TransposeType(ExplicitEnum) - added:
field TransposeType.CONV1D - added:
field TransposeType.CONV2D - added:
field TransposeType.NO - added:
field TransposeType.SIMPLE - added:
function apply_transpose(transpose: TransposeType, weight, match_shape=…, pt_to_tf=…) - added:
function load_pytorch_state_dict_in_tf2_model(tf_model, pt_state_dict, tf_inputs=…, allow_missing_keys=…, output_loading_info=…, _prefix=…, tf_to_pt_weight_rename=…, ignore_mismatched_sizes=…, skip_logger_warnings=…) - added:
function load_sharded_pytorch_safetensors_in_tf2_model(tf_model, safetensors_shards, tf_inputs=…, allow_missing_keys=…, output_loading_info=…, _prefix=…, tf_to_pt_weight_rename=…, ignore_mismatched_sizes=…) - added:
function load_tf2_state_dict_in_pytorch_model(pt_model, tf_state_dict, allow_missing_keys=…, output_loading_info=…)
- removed:
class TFRobertaAttention(tf.keras.layers.Layer) - removed:
class TFRobertaClassificationHead(tf.keras.layers.Layer) - removed:
class TFRobertaEmbeddings(tf.keras.layers.Layer) - removed:
class TFRobertaEncoder(tf.keras.layers.Layer) - removed:
class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLoss) - removed:
class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFRobertaIntermediate(tf.keras.layers.Layer) - removed:
class TFRobertaLMHead(tf.keras.layers.Layer) - removed:
class TFRobertaLayer(tf.keras.layers.Layer) - removed:
class TFRobertaMainLayer(tf.keras.layers.Layer) - removed:
class TFRobertaModel(TFRobertaPreTrainedModel) - removed:
class TFRobertaOutput(tf.keras.layers.Layer) - removed:
class TFRobertaPooler(tf.keras.layers.Layer) - removed:
class TFRobertaPreTrainedModel(TFPreTrainedModel) - removed:
class TFRobertaSelfAttention(tf.keras.layers.Layer) - removed:
class TFRobertaSelfOutput(tf.keras.layers.Layer) - removed:
constructor TFRobertaAttention(config, **kwargs) - removed:
constructor TFRobertaClassificationHead(config, **kwargs) - removed:
constructor TFRobertaEmbeddings(config, **kwargs) - removed:
constructor TFRobertaEncoder(config, **kwargs) - removed:
constructor TFRobertaForMaskedLM(config, *inputs, **kwargs) - removed:
constructor TFRobertaForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFRobertaForQuestionAnswering(config, *inputs, **kwargs) - removed:
constructor TFRobertaForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFRobertaForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFRobertaIntermediate(config, **kwargs) - removed:
constructor TFRobertaLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFRobertaLayer(config, **kwargs) - removed:
constructor TFRobertaMainLayer(config, **kwargs) - removed:
constructor TFRobertaModel(config, *inputs, **kwargs) - removed:
constructor TFRobertaOutput(config, **kwargs) - removed:
constructor TFRobertaPooler(config, **kwargs) - removed:
constructor TFRobertaSelfAttention(config, **kwargs) - removed:
constructor TFRobertaSelfOutput(config, **kwargs) - removed:
field TFRobertaForMaskedLM.authorized_missing_keys - removed:
field TFRobertaForQuestionAnswering.authorized_missing_keys - removed:
field TFRobertaForSequenceClassification.authorized_missing_keys - removed:
field TFRobertaForTokenClassification.authorized_missing_keys - removed:
field TFRobertaMainLayer.config_class - removed:
field TFRobertaPreTrainedModel.base_model_prefix - removed:
field TFRobertaPreTrainedModel.config_class - removed:
method TFRobertaAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFRobertaAttention.prune_heads(heads) - removed:
method TFRobertaClassificationHead.call(features, training=…) - removed:
method TFRobertaEmbeddings.build(input_shape) - removed:
method TFRobertaEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - removed:
method TFRobertaEmbeddings.create_position_ids_from_input_ids(x) - removed:
method TFRobertaEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - removed:
method TFRobertaEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - removed:
method TFRobertaForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFRobertaForMaskedLM.get_output_embeddings() - removed:
method TFRobertaForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFRobertaForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFRobertaForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFRobertaForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFRobertaIntermediate.call(hidden_states) - removed:
method TFRobertaLMHead.build(input_shape) - removed:
method TFRobertaLMHead.call(features) - removed:
method TFRobertaLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFRobertaMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFRobertaMainLayer.get_input_embeddings() - removed:
method TFRobertaMainLayer.set_input_embeddings(value) - removed:
method TFRobertaModel.call(inputs, **kwargs) - removed:
method TFRobertaOutput.call(hidden_states, input_tensor, training=…) - removed:
method TFRobertaPooler.call(hidden_states) - removed:
method TFRobertaSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - removed:
method TFRobertaSelfAttention.transpose_for_scores(x, batch_size) - removed:
method TFRobertaSelfOutput.call(hidden_states, input_tensor, training=…) - removed:
property TFRobertaForMultipleChoice.dummy_inputs - removed:
variable ROBERTA_INPUTS_DOCSTRING - removed:
variable ROBERTA_START_DOCSTRING - removed:
variable TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFT5Attention(tf.keras.layers.Layer) - removed:
class TFT5Block(tf.keras.layers.Layer) - removed:
class TFT5DenseReluDense(tf.keras.layers.Layer) - removed:
class TFT5ForConditionalGeneration(TFT5PreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFT5LayerCrossAttention(tf.keras.layers.Layer) - removed:
class TFT5LayerFF(tf.keras.layers.Layer) - removed:
class TFT5LayerNorm(tf.keras.layers.Layer) - removed:
class TFT5LayerSelfAttention(tf.keras.layers.Layer) - removed:
class TFT5MainLayer(tf.keras.layers.Layer) - removed:
class TFT5Model(TFT5PreTrainedModel) - removed:
class TFT5PreTrainedModel(TFPreTrainedModel) - removed:
constructor TFT5Attention(config, has_relative_attention_bias=…, is_bidirectional=…, **kwargs) - removed:
constructor TFT5Block(config, has_relative_attention_bias=…, **kwargs) - removed:
constructor TFT5DenseReluDense(config, **kwargs) - removed:
constructor TFT5ForConditionalGeneration(config, *inputs, **kwargs) - removed:
constructor TFT5LayerCrossAttention(config, has_relative_attention_bias=…, **kwargs) - removed:
constructor TFT5LayerFF(config, **kwargs) - removed:
constructor TFT5LayerNorm(epsilon=…, **kwargs) - removed:
constructor TFT5LayerSelfAttention(config, has_relative_attention_bias=…, **kwargs) - removed:
constructor TFT5MainLayer(config, embed_tokens=…, **kwargs) - removed:
constructor TFT5Model(config, *inputs, **kwargs) - removed:
field TFT5Attention.NEW_ID - removed:
field TFT5MainLayer.config_class - removed:
field TFT5PreTrainedModel.base_model_prefix - removed:
field TFT5PreTrainedModel.config_class - removed:
method TFT5Attention.call(input, mask=…, kv=…, position_bias=…, past_key_value=…, head_mask=…, query_length=…, use_cache=…, training=…, output_attentions=…) - removed:
method TFT5Attention.compute_bias(qlen, klen) - removed:
method TFT5Attention.prune_heads(heads) - removed:
method TFT5Block.call(hidden_states, attention_mask=…, position_bias=…, encoder_hidden_states=…, encoder_attention_mask=…, encoder_decoder_position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…, training=…) - removed:
method TFT5DenseReluDense.call(hidden_states, training=…) - removed:
method TFT5ForConditionalGeneration.call(inputs, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, head_mask=…, inputs_embeds=…, decoder_inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFT5ForConditionalGeneration.get_decoder() - removed:
method TFT5ForConditionalGeneration.get_encoder() - removed:
method TFT5ForConditionalGeneration.get_input_embeddings() - removed:
method TFT5ForConditionalGeneration.get_output_embeddings() - removed:
method TFT5ForConditionalGeneration.prepare_inputs_for_generation(inputs, past, attention_mask, use_cache, **kwargs) - removed:
method TFT5ForConditionalGeneration.set_input_embeddings(new_embeddings) - removed:
method TFT5LayerCrossAttention.call(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, query_length=…, use_cache=…, output_attentions=…, training=…) - removed:
method TFT5LayerFF.call(hidden_states, training=…) - removed:
method TFT5LayerNorm.build(input_shape) - removed:
method TFT5LayerNorm.call(x) - removed:
method TFT5LayerSelfAttention.call(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…, training=…) - removed:
method TFT5MainLayer.call(inputs, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, inputs_embeds=…, head_mask=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, training=…, **kwargs) -> Tuple - removed:
method TFT5MainLayer.get_input_embeddings() - removed:
method TFT5MainLayer.get_output_embeddings() - removed:
method TFT5MainLayer.set_embed_tokens(embed_tokens) - removed:
method TFT5Model.call(inputs, attention_mask=…, decoder_input_ids=…, decoder_attention_mask=…, encoder_outputs=…, past_key_values=…, head_mask=…, inputs_embeds=…, decoder_inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - removed:
method TFT5Model.get_decoder() - removed:
method TFT5Model.get_encoder() - removed:
method TFT5Model.get_input_embeddings() - removed:
method TFT5Model.get_output_embeddings() - removed:
method TFT5Model.set_input_embeddings(new_embeddings) - removed:
property TFT5PreTrainedModel.dummy_inputs - removed:
variable T5_INPUTS_DOCSTRING - removed:
variable T5_START_DOCSTRING - removed:
variable TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class TFAdaptiveEmbedding(tf.keras.layers.Layer) - removed:
class TFPositionalEmbedding(tf.keras.layers.Layer) - removed:
class TFPositionwiseFF(tf.keras.layers.Layer) - removed:
class TFRelPartialLearnableDecoderLayer(tf.keras.layers.Layer) - removed:
class TFRelPartialLearnableMultiHeadAttn(tf.keras.layers.Layer) - removed:
class TFTransfoXLLMHeadModel(TFTransfoXLPreTrainedModel) - removed:
class TFTransfoXLLMHeadModelOutput(ModelOutput) - removed:
class TFTransfoXLMHead(tf.keras.layers.Layer) - removed:
class TFTransfoXLMainLayer(tf.keras.layers.Layer) - removed:
class TFTransfoXLModel(TFTransfoXLPreTrainedModel) - removed:
class TFTransfoXLModelOutput(ModelOutput) - removed:
class TFTransfoXLPreTrainedModel(TFPreTrainedModel) - removed:
constructor TFAdaptiveEmbedding(n_token, d_embed, d_proj, cutoffs, div_val=…, init_std=…, sample_softmax=…, **kwargs) - removed:
constructor TFPositionalEmbedding(demb, **kwargs) - removed:
constructor TFPositionwiseFF(d_model, d_inner, dropout, pre_lnorm=…, layer_norm_epsilon=…, init_std=…, **kwargs) - removed:
constructor TFRelPartialLearnableDecoderLayer(n_head, d_model, d_head, d_inner, dropout, dropatt=…, pre_lnorm=…, r_w_bias=…, r_r_bias=…, layer_norm_epsilon=…, init_std=…, output_attentions=…, **kwargs) - removed:
constructor TFRelPartialLearnableMultiHeadAttn(n_head, d_model, d_head, dropout, dropatt=…, pre_lnorm=…, r_r_bias=…, r_w_bias=…, layer_norm_epsilon=…, init_std=…, output_attentions=…, **kwargs) - removed:
constructor TFTransfoXLLMHeadModel(config) - removed:
constructor TFTransfoXLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFTransfoXLMainLayer(config, **kwargs) - removed:
constructor TFTransfoXLModel(config, *inputs, **kwargs) - removed:
field TFTransfoXLLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFTransfoXLLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFTransfoXLLMHeadModelOutput.mems: List[tf.Tensor] - removed:
field TFTransfoXLLMHeadModelOutput.prediction_scores: tf.Tensor - removed:
field TFTransfoXLMainLayer.config_class - removed:
field TFTransfoXLModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFTransfoXLModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFTransfoXLModelOutput.last_hidden_state: tf.Tensor - removed:
field TFTransfoXLModelOutput.mems: List[tf.Tensor] - removed:
field TFTransfoXLPreTrainedModel.base_model_prefix - removed:
field TFTransfoXLPreTrainedModel.config_class - removed:
method TFAdaptiveEmbedding.build(input_shape) - removed:
method TFAdaptiveEmbedding.call(inp) - removed:
method TFPositionalEmbedding.call(pos_seq, bsz=…) - removed:
method TFPositionwiseFF.call(inp, training=…) - removed:
method TFRelPartialLearnableDecoderLayer.call(dec_inp, r, dec_attn_mask, mems, head_mask, output_attentions, training=…) - removed:
method TFRelPartialLearnableMultiHeadAttn.build(input_shape) - removed:
method TFRelPartialLearnableMultiHeadAttn.call(w, r, attn_mask, mems, head_mask, output_attentions, training=…) - removed:
method TFTransfoXLLMHeadModel.call(inputs, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFTransfoXLLMHeadModel.get_output_embeddings() - removed:
method TFTransfoXLLMHeadModel.init_mems(bsz) - removed:
method TFTransfoXLLMHeadModel.prepare_inputs_for_generation(inputs, past, **model_kwargs) - removed:
method TFTransfoXLLMHeadModel.reset_length(tgt_len, ext_len, mem_len) - removed:
method TFTransfoXLLMHeadModel.reset_memory_length(mem_len) - removed:
method TFTransfoXLMHead.build(input_shape) - removed:
method TFTransfoXLMHead.call(hidden_states) - removed:
method TFTransfoXLMainLayer.backward_compatible() - removed:
method TFTransfoXLMainLayer.build(input_shape) - removed:
method TFTransfoXLMainLayer.call(inputs, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFTransfoXLMainLayer.get_input_embeddings() - removed:
method TFTransfoXLMainLayer.init_mems(bsz) - removed:
method TFTransfoXLMainLayer.reset_memory_length(mem_len) - removed:
method TFTransfoXLMainLayer.set_input_embeddings(value) - removed:
method TFTransfoXLModel.call(inputs, **kwargs) - removed:
variable TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TRANSFO_XL_INPUTS_DOCSTRING - removed:
variable TRANSFO_XL_START_DOCSTRING - removed:
variable logger
- removed:
class TFAdaptiveSoftmaxMask(tf.keras.layers.Layer) - removed:
constructor TFAdaptiveSoftmaxMask(vocab_size, d_embed, d_proj, cutoffs, div_val=…, keep_order=…, **kwargs) - removed:
method TFAdaptiveSoftmaxMask.build(input_shape) - removed:
method TFAdaptiveSoftmaxMask.call(hidden, target, return_mean=…, training=…)
- removed:
class TFWrappedEmbeddings - removed:
constructor TFWrappedEmbeddings(layer, abs_scope_name=…) - removed:
field TFPreTrainedModel.authorized_missing_keys - removed:
field TFPreTrainedModel.authorized_unexpected_keys - removed:
function cast_bool_to_primitive(bool_variable: Union[tf.Tensor, bool], default_tensor_to_true=…) -> bool - removed:
function detect_tf_missing_unexpected_layers(model, resolved_archive_file) - removed:
function shape_list(x: tf.Tensor) -> List[int] - removed:
method TFCausalLanguageModelingLoss.compute_loss(labels, logits) - removed:
method TFQuestionAnsweringLoss.compute_loss(labels, logits) - removed:
method TFSequenceClassificationLoss.compute_loss(labels, logits) - removed:
method TFTokenClassificationLoss.compute_loss(labels, logits) - removed:
method TFWrappedEmbeddings.call(inputs, mode=…) - removed:
property TFPreTrainedModel.dummy_inputs -> Dict[str, tf.Tensor] - changed:
class TFConv1D(tf.keras.layers.Layer) → class TFConv1D(keras.layers.Layer) - changed:
class TFMultipleChoiceLoss(TFSequenceClassificationLoss) → class TFMultipleChoiceLoss - changed:
class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin) → class TFPreTrainedModel(keras.Model, TFModelUtilsMixin, TFGenerationMixin, PushToHubMixin) - changed:
class TFSequenceSummary(tf.keras.layers.Layer) → class TFSequenceSummary(keras.layers.Layer) - changed:
class TFSharedEmbeddings(tf.keras.layers.Layer) → class TFSharedEmbeddings(keras.layers.Layer) - changed:
constructor TFSharedEmbeddings(vocab_size: int, hidden_size: int, initializer_range: Optional[float]=…, **kwargs) → constructor TFSharedEmbeddings(vocab_size: int, hidden_size: int, initializer_range: float | None=…, **kwargs) - changed:
function get_initializer(initializer_range: float=…) -> tf.initializers.TruncatedNormal → function get_initializer(initializer_range: float=…) -> keras.initializers.TruncatedNormal - changed:
function load_tf_weights(model, resolved_archive_file) → function load_tf_weights(model, resolved_archive_file, ignore_mismatched_sizes=…, _prefix=…) - changed:
method TFPreTrainedModel.get_input_embeddings() -> tf.keras.layers.Layer → method TFPreTrainedModel.get_input_embeddings() -> keras.layers.Layer - changed:
method TFPreTrainedModel.get_output_embeddings() -> tf.keras.layers.Layer → method TFPreTrainedModel.get_output_embeddings() -> None | keras.layers.Layer - changed:
method TFPreTrainedModel.resize_token_embeddings(new_num_tokens=…) -> tf.Variable → method TFPreTrainedModel.resize_token_embeddings(new_num_tokens: int | None=…) -> keras.layers.Embedding | tf.Variable - changed:
method TFPreTrainedModel.save_pretrained(save_directory) → method TFPreTrainedModel.save_pretrained(save_directory, saved_model=…, version=…, push_to_hub=…, signatures=…, max_shard_size: int | str=…, create_pr: bool=…, safe_serialization: bool=…, token: str | bool | None=…, **kwargs) - changed:
static method TFPreTrainedModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) → static method TFPreTrainedModel.from_pretrained(pretrained_model_name_or_path: str | os.PathLike | None, *model_args, config: PretrainedConfig | str | os.PathLike | None=…, cache_dir: str | os.PathLike | None=…, ignore_mismatched_sizes: bool=…, force_download: bool=…, local_files_only: bool=…, token: str | bool | None=…, revision: str=…, use_safetensors: bool | None=…, **kwargs) - added:
class TFNextSentencePredictionLoss - added:
field TFPreTrainedModel.main_input_name - added:
function booleans_processing(config, **kwargs) - added:
function dummy_loss(y_true, y_pred) - added:
function init_copy_embeddings(old_embeddings, new_num_tokens) - added:
function input_processing(func, config, **kwargs) - added:
function load_tf_shard(model, model_layer_map, resolved_archive_file, ignore_mismatched_sizes=…, _prefix=…) - added:
function load_tf_sharded_weights(model, shard_files, ignore_mismatched_sizes=…, strict=…, _prefix=…) - added:
function load_tf_sharded_weights_from_safetensors(model, shard_files, ignore_mismatched_sizes=…, strict=…, _prefix=…) - added:
function load_tf_weights_from_h5(model, resolved_archive_file, ignore_mismatched_sizes=…, _prefix=…) - added:
function load_tf_weights_from_safetensors(model, resolved_archive_file, ignore_mismatched_sizes=…, _prefix=…) - added:
function strip_model_name_and_prefix(name, _prefix=…) - added:
function tf_shard_checkpoint(weights, max_shard_size=…, weights_name: str=…) - added:
function unpack_inputs(func) - added:
method TFCausalLanguageModelingLoss.hf_compute_loss(labels, logits) - added:
method TFMultipleChoiceLoss.hf_compute_loss(labels, logits) - added:
method TFNextSentencePredictionLoss.hf_compute_loss(labels, logits) - added:
method TFPreTrainedModel.build(input_shape=…) - added:
method TFPreTrainedModel.build_in_name_scope() - added:
method TFPreTrainedModel.compile(optimizer=…, loss=…, metrics=…, loss_weights=…, weighted_metrics=…, run_eagerly=…, steps_per_execution=…, **kwargs) - added:
method TFPreTrainedModel.compute_loss(*args, **kwargs) - added:
method TFPreTrainedModel.create_model_card(output_dir, model_name: str, language: str | None=…, license: str | None=…, tags: str | None=…, finetuned_from: str | None=…, tasks: str | None=…, dataset_tags: str | list[str] | None=…, dataset: str | list[str] | None=…, dataset_args: str | list[str] | None=…) - added:
method TFPreTrainedModel.evaluate(*args, **kwargs) - added:
method TFPreTrainedModel.fit(*args, **kwargs) - added:
method TFPreTrainedModel.get_bias() -> None | dict[str, tf.Variable] - added:
method TFPreTrainedModel.get_config() - added:
method TFPreTrainedModel.get_head_mask(head_mask: tf.Tensor | None, num_hidden_layers: int) -> tf.Tensor - added:
method TFPreTrainedModel.get_label_to_output_name_mapping() - added:
method TFPreTrainedModel.get_lm_head() -> keras.layers.Layer - added:
method TFPreTrainedModel.get_output_layer_with_bias() -> None | keras.layers.Layer - added:
method TFPreTrainedModel.get_prefix_bias_name() -> None | str - added:
method TFPreTrainedModel.predict(*args, **kwargs) - added:
method TFPreTrainedModel.predict_on_batch(*args, **kwargs) - added:
method TFPreTrainedModel.prepare_tf_dataset(dataset: datasets.Dataset, batch_size: int=…, shuffle: bool=…, tokenizer: PreTrainedTokenizerBase | None=…, collate_fn: Callable | None=…, collate_fn_args: dict[str, Any] | None=…, drop_remainder: bool | None=…, prefetch: bool=…) - added:
method TFPreTrainedModel.push_to_hub(repo_id: str, use_temp_dir: bool | None=…, commit_message: str | None=…, private: bool | None=…, max_shard_size: int | str | None=…, token: bool | str | None=…, use_auth_token: bool | str | None=…, create_pr: bool=…, **base_model_card_args) -> str - added:
method TFPreTrainedModel.serving(inputs) - added:
method TFPreTrainedModel.serving_output(output) - added:
method TFPreTrainedModel.set_bias(value) - added:
method TFPreTrainedModel.set_output_embeddings(value) - added:
method TFPreTrainedModel.test_on_batch(*args, **kwargs) - added:
method TFPreTrainedModel.test_step(data) - added:
method TFPreTrainedModel.train_on_batch(*args, **kwargs) - added:
method TFPreTrainedModel.train_step(data) - added:
method TFQuestionAnsweringLoss.hf_compute_loss(labels, logits) - added:
method TFSequenceClassificationLoss.hf_compute_loss(labels, logits) - added:
method TFSequenceSummary.build(input_shape) - added:
method TFTokenClassificationLoss.hf_compute_loss(labels, logits) - added:
property TFPreTrainedModel.dummy_inputs -> dict[str, tf.Tensor] - added:
property TFPreTrainedModel.framework -> str - added:
property TFPreTrainedModel.input_signature -> dict[str, tf.TensorSpec] - added:
static method TFPreTrainedModel.can_generate() -> bool - added:
static method TFPreTrainedModel.from_config(config, **kwargs) - added:
static method TFPreTrainedModel.register_for_auto_class(auto_class=…) - added:
variable TFModelInputType - added:
variable tf_logger
- removed:
class TFXLMForMultipleChoice(TFXLMPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFXLMForQuestionAnsweringSimple(TFXLMPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFXLMForSequenceClassification(TFXLMPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFXLMForTokenClassification(TFXLMPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFXLMMainLayer(tf.keras.layers.Layer) - removed:
class TFXLMModel(TFXLMPreTrainedModel) - removed:
class TFXLMMultiHeadAttention(tf.keras.layers.Layer) - removed:
class TFXLMPreTrainedModel(TFPreTrainedModel) - removed:
class TFXLMPredLayer(tf.keras.layers.Layer) - removed:
class TFXLMTransformerFFN(tf.keras.layers.Layer) - removed:
class TFXLMWithLMHeadModel(TFXLMPreTrainedModel) - removed:
class TFXLMWithLMHeadModelOutput(ModelOutput) - removed:
constructor TFXLMForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFXLMForQuestionAnsweringSimple(config, *inputs, **kwargs) - removed:
constructor TFXLMForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFXLMForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFXLMMainLayer(config, **kwargs) - removed:
constructor TFXLMModel(config, *inputs, **kwargs) - removed:
constructor TFXLMMultiHeadAttention(n_heads, dim, config, **kwargs) - removed:
constructor TFXLMPredLayer(config, input_embeddings, **kwargs) - removed:
constructor TFXLMTransformerFFN(in_dim, dim_hidden, out_dim, config, **kwargs) - removed:
constructor TFXLMWithLMHeadModel(config, *inputs, **kwargs) - removed:
field TFXLMMainLayer.config_class - removed:
field TFXLMMultiHeadAttention.NEW_ID - removed:
field TFXLMPreTrainedModel.base_model_prefix - removed:
field TFXLMPreTrainedModel.config_class - removed:
field TFXLMWithLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLMWithLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLMWithLMHeadModelOutput.logits: tf.Tensor - removed:
function create_sinusoidal_embeddings(n_pos, dim, out) - removed:
function get_masks(slen, lengths, causal, padding_mask=…, dtype=…) - removed:
method TFXLMForMultipleChoice.call(inputs, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLMForQuestionAnsweringSimple.call(inputs=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFXLMForSequenceClassification.call(inputs=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLMForTokenClassification.call(inputs=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLMMainLayer.call(inputs, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFXLMMainLayer.get_input_embeddings() - removed:
method TFXLMMainLayer.set_input_embeddings(value) - removed:
method TFXLMModel.call(inputs, **kwargs) - removed:
method TFXLMMultiHeadAttention.call(input, mask, kv, cache, head_mask, output_attentions, training=…) - removed:
method TFXLMMultiHeadAttention.prune_heads(heads) - removed:
method TFXLMPredLayer.build(input_shape) - removed:
method TFXLMPredLayer.call(hidden_states) - removed:
method TFXLMTransformerFFN.call(input, training=…) - removed:
method TFXLMWithLMHeadModel.call(inputs, **kwargs) - removed:
method TFXLMWithLMHeadModel.get_output_embeddings() - removed:
method TFXLMWithLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - removed:
property TFXLMForMultipleChoice.dummy_inputs - removed:
property TFXLMPreTrainedModel.dummy_inputs - removed:
variable TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_INPUTS_DOCSTRING - removed:
variable XLM_START_DOCSTRING - removed:
variable logger
- removed:
class TFXLMRobertaForMaskedLM(TFRobertaForMaskedLM) - removed:
class TFXLMRobertaForMultipleChoice(TFRobertaForMultipleChoice) - removed:
class TFXLMRobertaForQuestionAnswering(TFRobertaForQuestionAnswering) - removed:
class TFXLMRobertaForSequenceClassification(TFRobertaForSequenceClassification) - removed:
class TFXLMRobertaForTokenClassification(TFRobertaForTokenClassification) - removed:
class TFXLMRobertaModel(TFRobertaModel) - removed:
field TFXLMRobertaForMaskedLM.config_class - removed:
field TFXLMRobertaForMultipleChoice.config_class - removed:
field TFXLMRobertaForQuestionAnswering.config_class - removed:
field TFXLMRobertaForSequenceClassification.config_class - removed:
field TFXLMRobertaForTokenClassification.config_class - removed:
field TFXLMRobertaModel.config_class - removed:
variable TF_XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_ROBERTA_START_DOCSTRING - removed:
variable logger
- removed:
class TFXLNetFeedForward(tf.keras.layers.Layer) - removed:
class TFXLNetForMultipleChoice(TFXLNetPreTrainedModel, TFMultipleChoiceLoss) - removed:
class TFXLNetForMultipleChoiceOutput(ModelOutput) - removed:
class TFXLNetForQuestionAnsweringSimple(TFXLNetPreTrainedModel, TFQuestionAnsweringLoss) - removed:
class TFXLNetForQuestionAnsweringSimpleOutput(ModelOutput) - removed:
class TFXLNetForSequenceClassification(TFXLNetPreTrainedModel, TFSequenceClassificationLoss) - removed:
class TFXLNetForSequenceClassificationOutput(ModelOutput) - removed:
class TFXLNetForTokenClassification(TFXLNetPreTrainedModel, TFTokenClassificationLoss) - removed:
class TFXLNetForTokenClassificationOutput(ModelOutput) - removed:
class TFXLNetLMHead(tf.keras.layers.Layer) - removed:
class TFXLNetLMHeadModel(TFXLNetPreTrainedModel, TFCausalLanguageModelingLoss) - removed:
class TFXLNetLMHeadModelOutput(ModelOutput) - removed:
class TFXLNetLayer(tf.keras.layers.Layer) - removed:
class TFXLNetMainLayer(tf.keras.layers.Layer) - removed:
class TFXLNetModel(TFXLNetPreTrainedModel) - removed:
class TFXLNetModelOutput(ModelOutput) - removed:
class TFXLNetPreTrainedModel(TFPreTrainedModel) - removed:
class TFXLNetRelativeAttention(tf.keras.layers.Layer) - removed:
constructor TFXLNetFeedForward(config, **kwargs) - removed:
constructor TFXLNetForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor TFXLNetForQuestionAnsweringSimple(config, *inputs, **kwargs) - removed:
constructor TFXLNetForSequenceClassification(config, *inputs, **kwargs) - removed:
constructor TFXLNetForTokenClassification(config, *inputs, **kwargs) - removed:
constructor TFXLNetLMHead(config, input_embeddings, **kwargs) - removed:
constructor TFXLNetLMHeadModel(config, *inputs, **kwargs) - removed:
constructor TFXLNetLayer(config, **kwargs) - removed:
constructor TFXLNetMainLayer(config, **kwargs) - removed:
constructor TFXLNetModel(config, *inputs, **kwargs) - removed:
constructor TFXLNetRelativeAttention(config, **kwargs) - removed:
field TFXLNetForMultipleChoiceOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForMultipleChoiceOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForMultipleChoiceOutput.logits: tf.Tensor - removed:
field TFXLNetForMultipleChoiceOutput.loss: Optional[tf.Tensor] - removed:
field TFXLNetForMultipleChoiceOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.end_logits: tf.Tensor - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.loss: Optional[tf.Tensor] - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetForQuestionAnsweringSimpleOutput.start_logits: tf.Tensor - removed:
field TFXLNetForSequenceClassificationOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForSequenceClassificationOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForSequenceClassificationOutput.logits: tf.Tensor - removed:
field TFXLNetForSequenceClassificationOutput.loss: Optional[tf.Tensor] - removed:
field TFXLNetForSequenceClassificationOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetForTokenClassificationOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForTokenClassificationOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetForTokenClassificationOutput.logits: tf.Tensor - removed:
field TFXLNetForTokenClassificationOutput.loss: Optional[tf.Tensor] - removed:
field TFXLNetForTokenClassificationOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetLMHeadModelOutput.logits: tf.Tensor - removed:
field TFXLNetLMHeadModelOutput.loss: Optional[tf.Tensor] - removed:
field TFXLNetLMHeadModelOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetMainLayer.config_class - removed:
field TFXLNetModelOutput.attentions: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - removed:
field TFXLNetModelOutput.last_hidden_state: tf.Tensor - removed:
field TFXLNetModelOutput.mems: Optional[List[tf.Tensor]] - removed:
field TFXLNetPreTrainedModel.base_model_prefix - removed:
field TFXLNetPreTrainedModel.config_class - removed:
method TFXLNetFeedForward.call(inp, training=…) - removed:
method TFXLNetForMultipleChoice.call(inputs=…, token_type_ids=…, input_mask=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLNetForQuestionAnsweringSimple.call(inputs=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - removed:
method TFXLNetForSequenceClassification.call(inputs=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLNetForTokenClassification.call(inputs=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLNetLMHead.build(input_shape) - removed:
method TFXLNetLMHead.call(hidden_states) - removed:
method TFXLNetLMHeadModel.call(inputs, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - removed:
method TFXLNetLMHeadModel.get_output_embeddings() - removed:
method TFXLNetLMHeadModel.prepare_inputs_for_generation(inputs, past, **kwargs) - removed:
method TFXLNetLayer.call(output_h, output_g, non_tgt_mask, attn_mask, pos_emb, seg_mat, mems, target_mapping, head_mask, output_attentions, training=…) - removed:
method TFXLNetMainLayer.build(input_shape) - removed:
method TFXLNetMainLayer.cache_mem(curr_out, prev_mem) - removed:
method TFXLNetMainLayer.call(inputs, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - removed:
method TFXLNetMainLayer.create_mask(qlen, mlen, dtype=…) - removed:
method TFXLNetMainLayer.get_input_embeddings() - removed:
method TFXLNetMainLayer.relative_positional_encoding(qlen, klen, bsz=…, dtype=…) - removed:
method TFXLNetMainLayer.set_input_embeddings(value) - removed:
method TFXLNetModel.call(inputs, **kwargs) - removed:
method TFXLNetRelativeAttention.build(input_shape) - removed:
method TFXLNetRelativeAttention.call(h, g, attn_mask_h, attn_mask_g, r, seg_mat, mems, target_mapping, head_mask, output_attentions, training=…) - removed:
method TFXLNetRelativeAttention.post_attention(h, attn_vec, residual=…, training=…) - removed:
method TFXLNetRelativeAttention.prune_heads(heads) - removed:
method TFXLNetRelativeAttention.rel_attn_core(q_head, k_head_h, v_head_h, k_head_r, seg_mat, attn_mask, head_mask, output_attentions, training=…) - removed:
method TFXLNetRelativeAttention.rel_shift(x, klen=…) - removed:
property TFXLNetForMultipleChoice.dummy_inputs - removed:
static method TFXLNetMainLayer.positional_embedding(pos_seq, inv_freq, bsz=…) - removed:
variable TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLNET_INPUTS_DOCSTRING - removed:
variable XLNET_START_DOCSTRING - removed:
variable logger
- removed:
class AdaptiveEmbedding(nn.Module) - removed:
class PositionalEmbedding(nn.Module) - removed:
class PositionwiseFF(nn.Module) - removed:
class RelPartialLearnableDecoderLayer(nn.Module) - removed:
class RelPartialLearnableMultiHeadAttn(nn.Module) - removed:
class TransfoXLLMHeadModel(TransfoXLPreTrainedModel) - removed:
class TransfoXLLMHeadModelOutput(ModelOutput) - removed:
class TransfoXLModel(TransfoXLPreTrainedModel) - removed:
class TransfoXLModelOutput(ModelOutput) - removed:
class TransfoXLPreTrainedModel(PreTrainedModel) - removed:
constructor AdaptiveEmbedding(n_token, d_embed, d_proj, cutoffs, div_val=…, sample_softmax=…) - removed:
constructor PositionalEmbedding(demb) - removed:
constructor PositionwiseFF(d_model, d_inner, dropout, pre_lnorm=…, layer_norm_epsilon=…) - removed:
constructor RelPartialLearnableDecoderLayer(n_head, d_model, d_head, d_inner, dropout, layer_norm_epsilon=…, **kwargs) - removed:
constructor RelPartialLearnableMultiHeadAttn(n_head, d_model, d_head, dropout, dropatt=…, pre_lnorm=…, r_r_bias=…, r_w_bias=…, layer_norm_epsilon=…) - removed:
constructor TransfoXLLMHeadModel(config) - removed:
constructor TransfoXLModel(config) - removed:
field TransfoXLLMHeadModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field TransfoXLLMHeadModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field TransfoXLLMHeadModelOutput.losses: Optional[torch.FloatTensor] - removed:
field TransfoXLLMHeadModelOutput.mems: List[torch.FloatTensor] - removed:
field TransfoXLLMHeadModelOutput.prediction_scores: torch.FloatTensor - removed:
field TransfoXLModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field TransfoXLModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field TransfoXLModelOutput.last_hidden_state: torch.FloatTensor - removed:
field TransfoXLModelOutput.mems: List[torch.FloatTensor] - removed:
field TransfoXLPreTrainedModel.base_model_prefix - removed:
field TransfoXLPreTrainedModel.config_class - removed:
field TransfoXLPreTrainedModel.load_tf_weights - removed:
function build_tf_to_pytorch_map(model, config) - removed:
function load_tf_weights_in_transfo_xl(model, config, tf_path) - removed:
method AdaptiveEmbedding.forward(inp) - removed:
method PositionalEmbedding.forward(pos_seq, bsz=…) - removed:
method PositionwiseFF.forward(inp) - removed:
method RelPartialLearnableDecoderLayer.forward(dec_inp, r, dec_attn_mask=…, mems=…, head_mask=…, output_attentions=…) - removed:
method RelPartialLearnableMultiHeadAttn.forward(w, r, attn_mask=…, mems=…, head_mask=…, output_attentions=…) - removed:
method TransfoXLLMHeadModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method TransfoXLLMHeadModel.get_output_embeddings() - removed:
method TransfoXLLMHeadModel.init_mems(bsz) - removed:
method TransfoXLLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, **model_kwargs) - removed:
method TransfoXLLMHeadModel.reset_length(tgt_len, ext_len, mem_len) - removed:
method TransfoXLLMHeadModel.reset_memory_length(mem_len) - removed:
method TransfoXLLMHeadModel.tie_weights() - removed:
method TransfoXLModel.backward_compatible() - removed:
method TransfoXLModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method TransfoXLModel.get_input_embeddings() - removed:
method TransfoXLModel.init_mems(bsz) - removed:
method TransfoXLModel.reset_memory_length(mem_len) - removed:
method TransfoXLModel.set_input_embeddings(new_embeddings) - removed:
method TransfoXLPreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…, layer: Optional[int]=…) - removed:
property TransfoXLLMHeadModelOutput.logits - removed:
variable TRANSFO_XL_INPUTS_DOCSTRING - removed:
variable TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TRANSFO_XL_START_DOCSTRING - removed:
variable logger
- removed:
class ProjectedAdaptiveLogSoftmax(nn.Module) - removed:
constructor ProjectedAdaptiveLogSoftmax(n_token, d_embed, d_proj, cutoffs, div_val=…, keep_order=…) - removed:
method ProjectedAdaptiveLogSoftmax.forward(hidden, labels=…, keep_order=…) - removed:
method ProjectedAdaptiveLogSoftmax.log_prob(hidden)
- removed:
class Conv1D(nn.Module) - removed:
class Identity(nn.Module) - removed:
class PoolerAnswerClass(nn.Module) - removed:
class PoolerEndLogits(nn.Module) - removed:
class PoolerStartLogits(nn.Module) - removed:
class SQuADHead(nn.Module) - removed:
class SequenceSummary(nn.Module) - removed:
class SquadHeadOutput(ModelOutput) - removed:
constructor Conv1D(nf, nx) - removed:
constructor Identity(*args, **kwargs) - removed:
constructor PoolerAnswerClass(config) - removed:
constructor PoolerEndLogits(config: PretrainedConfig) - removed:
constructor PoolerStartLogits(config: PretrainedConfig) - removed:
constructor SQuADHead(config) - removed:
constructor SequenceSummary(config: PretrainedConfig) - removed:
field PreTrainedModel.authorized_missing_keys - removed:
field PreTrainedModel.authorized_unexpected_keys - removed:
field PreTrainedModel.keys_to_never_save - removed:
field SquadHeadOutput.cls_logits: Optional[torch.FloatTensor] - removed:
field SquadHeadOutput.end_top_index: Optional[torch.LongTensor] - removed:
field SquadHeadOutput.end_top_log_probs: Optional[torch.FloatTensor] - removed:
field SquadHeadOutput.loss: Optional[torch.FloatTensor] - removed:
field SquadHeadOutput.start_top_index: Optional[torch.LongTensor] - removed:
field SquadHeadOutput.start_top_log_probs: Optional[torch.FloatTensor] - removed:
function apply_chunking_to_forward(forward_fn: Callable[..., torch.Tensor], chunk_size: int, chunk_dim: int, *input_tensors) -> torch.Tensor - removed:
function find_pruneable_heads_and_indices(heads: List[int], n_heads: int, head_size: int, already_pruned_heads: Set[int]) -> Tuple[Set[int], torch.LongTensor] - removed:
function prune_conv1d_layer(layer: Conv1D, index: torch.LongTensor, dim: int=…) -> Conv1D - removed:
function prune_layer(layer: Union[torch.nn.Linear, Conv1D], index: torch.LongTensor, dim: Optional[int]=…) -> Union[torch.nn.Linear, Conv1D] - removed:
function prune_linear_layer(layer: torch.nn.Linear, index: torch.LongTensor, dim: int=…) -> torch.nn.Linear - removed:
method Conv1D.forward(x) - removed:
method Identity.forward(input) - removed:
method PoolerAnswerClass.forward(hidden_states: torch.FloatTensor, start_states: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, cls_index: Optional[torch.LongTensor]=…) -> torch.FloatTensor - removed:
method PoolerEndLogits.forward(hidden_states: torch.FloatTensor, start_states: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, p_mask: Optional[torch.FloatTensor]=…) -> torch.FloatTensor - removed:
method PoolerStartLogits.forward(hidden_states: torch.FloatTensor, p_mask: Optional[torch.FloatTensor]=…) -> torch.FloatTensor - removed:
method PreTrainedModel.get_input_embeddings() -> nn.Module - removed:
method PreTrainedModel.get_output_embeddings() -> nn.Module - removed:
method PreTrainedModel.set_input_embeddings(value: nn.Module) - removed:
method SQuADHead.forward(hidden_states: torch.FloatTensor, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, cls_index: Optional[torch.LongTensor]=…, is_impossible: Optional[torch.LongTensor]=…, p_mask: Optional[torch.FloatTensor]=…, return_dict: bool=…) -> Union[SquadHeadOutput, Tuple[torch.FloatTensor]] - removed:
method SequenceSummary.forward(hidden_states: torch.FloatTensor, cls_index: Optional[torch.LongTensor]=…) -> torch.FloatTensor - removed:
property ModuleUtilsMixin.device -> device - removed:
property ModuleUtilsMixin.dtype -> dtype - removed:
property PreTrainedModel.dummy_inputs -> Dict[str, torch.Tensor] - changed:
class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin) → class PreTrainedModel(nn.Module, EmbeddingAccessMixin, ModuleUtilsMixin, PushToHubMixin, PeftAdapterMixin) - changed:
method ModuleUtilsMixin.estimate_tokens(input_dict: Dict[str, Union[torch.Tensor, Any]]) -> int → method ModuleUtilsMixin.estimate_tokens(input_dict: dict[str, Union[torch.Tensor, Any]]) -> int - changed:
method ModuleUtilsMixin.floating_point_ops(input_dict: Dict[str, Union[torch.Tensor, Any]], exclude_embeddings: bool=…) -> int → method ModuleUtilsMixin.floating_point_ops(input_dict: dict[str, Union[torch.Tensor, Any]], exclude_embeddings: bool=…) -> int - changed:
method ModuleUtilsMixin.get_extended_attention_mask(attention_mask: Tensor, input_shape: Tuple[int], device: device) -> Tensor → method ModuleUtilsMixin.get_extended_attention_mask(attention_mask: Tensor, input_shape: tuple[int, ...], device: Optional[torch.device]=…, dtype: Optional[torch.dtype]=…) -> Tensor - changed:
method PreTrainedModel.prune_heads(heads_to_prune: Dict[int, List[int]]) → method PreTrainedModel.prune_heads(heads_to_prune: dict[int, list[int]]) - changed:
method PreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…) -> torch.nn.Embedding → method PreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - changed:
method PreTrainedModel.save_pretrained(save_directory) → method PreTrainedModel.save_pretrained(save_directory: Union[str, os.PathLike], is_main_process: bool=…, state_dict: Optional[dict]=…, save_function: Callable=…, push_to_hub: bool=…, max_shard_size: Union[int, str]=…, safe_serialization: bool=…, variant: Optional[str]=…, token: Optional[Union[str, bool]]=…, save_peft_format: bool=…, **kwargs) - changed:
static method PreTrainedModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) → static method PreTrainedModel.from_pretrained(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *model_args, config: Optional[Union[PretrainedConfig, str, os.PathLike]]=…, cache_dir: Optional[Union[str, os.PathLike]]=…, ignore_mismatched_sizes: bool=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, use_safetensors: Optional[… - added:
class AttentionInterface(GeneralInterface) - added:
class EmbeddingAccessMixin - added:
class PipelineParallel(Enum) - added:
class PreTrainedAudioTokenizerBase(PreTrainedModel) - added:
field PipelineParallel.inputs - added:
field PipelineParallel.outputs - added:
field PreTrainedModel.is_parallelizable - added:
field PreTrainedModel.main_input_name - added:
field PreTrainedModel.model_tags - added:
field PreTrainedModel.supports_gradient_checkpointing - added:
function caching_allocator_warmup(model: PreTrainedModel, expanded_device_map: dict, hf_quantizer: Optional[HfQuantizer]) - added:
function expand_device_map(device_map, param_names) - added:
function get_disk_only_shard_files(device_map, weight_map) - added:
function get_parameter_device(parameter: Union[nn.Module, ModuleUtilsMixin]) - added:
function get_parameter_dtype(parameter: Union[nn.Module, ModuleUtilsMixin]) - added:
function get_state_dict_dtype(state_dict) - added:
function get_torch_context_manager_or_global_device() - added:
function is_accelerator_device(device: Union[str, int, torch.device]) -> bool - added:
function is_local_dist_rank_0() - added:
function load_shard_file(args) - added:
function load_shard_files_with_threadpool(args_list) - added:
function load_sharded_checkpoint(model, folder, strict=…, prefer_safe=…) - added:
function load_state_dict(checkpoint_file: Union[str, os.PathLike], is_quantized: bool=…, map_location: Optional[Union[str, torch.device]]=…, weights_only: bool=…) - added:
function no_init_weights() - added:
function restore_default_dtype(func) - added:
function set_quantized_state() - added:
function set_zero3_state() - added:
function unwrap_model(model: nn.Module, recursive: bool=…) -> nn.Module - added:
method EmbeddingAccessMixin.get_input_embeddings() -> nn.Module - added:
method EmbeddingAccessMixin.get_output_embeddings() - added:
method EmbeddingAccessMixin.set_input_embeddings(value: nn.Module) - added:
method EmbeddingAccessMixin.set_output_embeddings(new_embeddings) - added:
method PreTrainedAudioTokenizerBase.decode(audio_codes: torch.Tensor, *args, **kwargs) - added:
method PreTrainedAudioTokenizerBase.encode(input_values: torch.Tensor, *args, **kwargs) - added:
method PreTrainedModel.add_model_tags(tags: Union[list[str], str]) -> None - added:
method PreTrainedModel.cuda(*args, **kwargs) - added:
method PreTrainedModel.dequantize() - added:
method PreTrainedModel.disable_input_require_grads() - added:
method PreTrainedModel.enable_input_require_grads() - added:
method PreTrainedModel.eval() - added:
method PreTrainedModel.float(*args) - added:
method PreTrainedModel.get_compiled_call(compile_config: Optional[CompileConfig]) -> Callable - added:
method PreTrainedModel.get_correct_attn_implementation(requested_attention: Optional[str], is_init_check: bool=…) -> str - added:
method PreTrainedModel.get_decoder() - added:
method PreTrainedModel.get_memory_footprint(return_buffers=…) - added:
method PreTrainedModel.get_parameter_or_buffer(target: str) - added:
method PreTrainedModel.get_position_embeddings() -> Union[nn.Embedding, tuple[nn.Embedding]] - added:
method PreTrainedModel.gradient_checkpointing_disable() - added:
method PreTrainedModel.gradient_checkpointing_enable(gradient_checkpointing_kwargs=…) - added:
method PreTrainedModel.half(*args) - added:
method PreTrainedModel.initialize_weights() - added:
method PreTrainedModel.kernelize() - added:
method PreTrainedModel.post_init() - added:
method PreTrainedModel.push_to_hub(*args, **kwargs) - added:
method PreTrainedModel.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PreTrainedModel.retrieve_modules_from_names(names, add_prefix=…, remove_prefix=…) - added:
method PreTrainedModel.reverse_bettertransformer() - added:
method PreTrainedModel.set_attn_implementation(attn_implementation: Union[str, dict]) - added:
method PreTrainedModel.set_decoder(decoder) - added:
method PreTrainedModel.tie_embeddings_and_encoder_decoder() - added:
method PreTrainedModel.to(*args, **kwargs) - added:
method PreTrainedModel.to_bettertransformer() -> PreTrainedModel - added:
method PreTrainedModel.train(mode: bool=…) - added:
method PreTrainedModel.warn_if_padding_and_no_attention_mask(input_ids, attention_mask) - added:
property ModuleUtilsMixin.device -> torch.device - added:
property ModuleUtilsMixin.dtype -> torch.dtype - added:
property PreTrainedModel.can_record_outputs -> dict[str, OutputRecorder] - added:
property PreTrainedModel.dummy_inputs -> dict[str, torch.Tensor] - added:
property PreTrainedModel.framework -> str - added:
property PreTrainedModel.is_gradient_checkpointing -> bool - added:
property PreTrainedModel.loss_function - added:
property PreTrainedModel.pp_plan -> dict[str, tuple[str, str]] - added:
property PreTrainedModel.supports_pp_plan - added:
property PreTrainedModel.supports_tp_plan - added:
property PreTrainedModel.tp_plan -> dict[str, str] - added:
property PreTrainedModel.tp_size - added:
property PreTrainedModel.use_kernels -> bool - added:
static method ModuleUtilsMixin.create_extended_attention_mask_for_decoder(input_shape, attention_mask, device=…) - added:
static method PreTrainedModel.can_generate() -> bool - added:
static method PreTrainedModel.get_init_context(is_quantized: bool, _is_ds_init_called: bool) - added:
static method PreTrainedModel.is_backend_compatible() - added:
static method PreTrainedModel.register_for_auto_class(auto_class=…) - added:
variable ALL_ATTENTION_FUNCTIONS: AttentionInterface - added:
variable IS_SAGEMAKER_MP_POST_1_10 - added:
variable SpecificPreTrainedModelType - added:
variable TORCH_INIT_FUNCTIONS - added:
variable VLMS - added:
variable XLA_DOWNCAST_BF16 - added:
variable XLA_USE_BF16 - added:
variable accelerate_version - added:
variable str_to_torch_dtype
- removed:
class MultiHeadAttention(nn.Module) - removed:
class TransformerFFN(nn.Module) - removed:
class XLMForMultipleChoice(XLMPreTrainedModel) - removed:
class XLMForQuestionAnswering(XLMPreTrainedModel) - removed:
class XLMForQuestionAnsweringOutput(ModelOutput) - removed:
class XLMForQuestionAnsweringSimple(XLMPreTrainedModel) - removed:
class XLMForSequenceClassification(XLMPreTrainedModel) - removed:
class XLMForTokenClassification(XLMPreTrainedModel) - removed:
class XLMModel(XLMPreTrainedModel) - removed:
class XLMPreTrainedModel(PreTrainedModel) - removed:
class XLMPredLayer(nn.Module) - removed:
class XLMWithLMHeadModel(XLMPreTrainedModel) - removed:
constructor MultiHeadAttention(n_heads, dim, config) - removed:
constructor TransformerFFN(in_dim, dim_hidden, out_dim, config) - removed:
constructor XLMForMultipleChoice(config, *inputs, **kwargs) - removed:
constructor XLMForQuestionAnswering(config) - removed:
constructor XLMForQuestionAnsweringSimple(config) - removed:
constructor XLMForSequenceClassification(config) - removed:
constructor XLMForTokenClassification(config) - removed:
constructor XLMModel(config) - removed:
constructor XLMPreTrainedModel(*inputs, **kwargs) - removed:
constructor XLMPredLayer(config) - removed:
constructor XLMWithLMHeadModel(config) - removed:
field MultiHeadAttention.NEW_ID - removed:
field XLMForQuestionAnsweringOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLMForQuestionAnsweringOutput.cls_logits: Optional[torch.FloatTensor] - removed:
field XLMForQuestionAnsweringOutput.end_top_index: Optional[torch.LongTensor] - removed:
field XLMForQuestionAnsweringOutput.end_top_log_probs: Optional[torch.FloatTensor] - removed:
field XLMForQuestionAnsweringOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLMForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - removed:
field XLMForQuestionAnsweringOutput.start_top_index: Optional[torch.LongTensor] - removed:
field XLMForQuestionAnsweringOutput.start_top_log_probs: Optional[torch.FloatTensor] - removed:
field XLMModel.authorized_missing_keys - removed:
field XLMPreTrainedModel.base_model_prefix - removed:
field XLMPreTrainedModel.config_class - removed:
field XLMPreTrainedModel.load_tf_weights - removed:
function create_sinusoidal_embeddings(n_pos, dim, out) - removed:
function get_masks(slen, lengths, causal, padding_mask=…) - removed:
method MultiHeadAttention.forward(input, mask, kv=…, cache=…, head_mask=…, output_attentions=…) - removed:
method MultiHeadAttention.prune_heads(heads) - removed:
method TransformerFFN.ff_chunk(input) - removed:
method TransformerFFN.forward(input) - removed:
method XLMForMultipleChoice.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMForQuestionAnswering.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, is_impossible=…, cls_index=…, p_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMForQuestionAnsweringSimple.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMForSequenceClassification.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMForTokenClassification.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMModel.get_input_embeddings() - removed:
method XLMModel.set_input_embeddings(new_embeddings) - removed:
method XLMPredLayer.forward(x, y=…) - removed:
method XLMWithLMHeadModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLMWithLMHeadModel.get_output_embeddings() - removed:
method XLMWithLMHeadModel.prepare_inputs_for_generation(input_ids, **kwargs) - removed:
property XLMPreTrainedModel.dummy_inputs - removed:
variable XLM_INPUTS_DOCSTRING - removed:
variable XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_START_DOCSTRING - removed:
variable logger
- removed:
class XLMProphetNetDecoder(ProphetNetDecoder) - removed:
class XLMProphetNetEncoder(ProphetNetEncoder) - removed:
class XLMProphetNetForCausalLM(ProphetNetForCausalLM) - removed:
class XLMProphetNetForConditionalGeneration(ProphetNetForConditionalGeneration) - removed:
class XLMProphetNetModel(ProphetNetModel) - removed:
field XLMProphetNetDecoder.config_class - removed:
field XLMProphetNetEncoder.config_class - removed:
field XLMProphetNetForCausalLM.config_class - removed:
field XLMProphetNetForConditionalGeneration.config_class - removed:
field XLMProphetNetModel.config_class - removed:
variable XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable logger
- removed:
class XLMRobertaForCausalLM(RobertaForCausalLM) - removed:
class XLMRobertaForMaskedLM(RobertaForMaskedLM) - removed:
class XLMRobertaForMultipleChoice(RobertaForMultipleChoice) - removed:
class XLMRobertaForQuestionAnswering(RobertaForQuestionAnswering) - removed:
class XLMRobertaForSequenceClassification(RobertaForSequenceClassification) - removed:
class XLMRobertaForTokenClassification(RobertaForTokenClassification) - removed:
class XLMRobertaModel(RobertaModel) - removed:
field XLMRobertaForCausalLM.config_class - removed:
field XLMRobertaForMaskedLM.config_class - removed:
field XLMRobertaForMultipleChoice.config_class - removed:
field XLMRobertaForQuestionAnswering.config_class - removed:
field XLMRobertaForSequenceClassification.config_class - removed:
field XLMRobertaForTokenClassification.config_class - removed:
field XLMRobertaModel.config_class - removed:
variable XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_ROBERTA_START_DOCSTRING - removed:
variable logger
- removed:
class XLNetFeedForward(nn.Module) - removed:
class XLNetForMultipleChoice(XLNetPreTrainedModel) - removed:
class XLNetForMultipleChoiceOutput(ModelOutput) - removed:
class XLNetForQuestionAnswering(XLNetPreTrainedModel) - removed:
class XLNetForQuestionAnsweringOutput(ModelOutput) - removed:
class XLNetForQuestionAnsweringSimple(XLNetPreTrainedModel) - removed:
class XLNetForQuestionAnsweringSimpleOutput(ModelOutput) - removed:
class XLNetForSequenceClassification(XLNetPreTrainedModel) - removed:
class XLNetForSequenceClassificationOutput(ModelOutput) - removed:
class XLNetForTokenClassification(XLNetPreTrainedModel) - removed:
class XLNetForTokenClassificationOutput(ModelOutput) - removed:
class XLNetLMHeadModel(XLNetPreTrainedModel) - removed:
class XLNetLMHeadModelOutput(ModelOutput) - removed:
class XLNetLayer(nn.Module) - removed:
class XLNetModel(XLNetPreTrainedModel) - removed:
class XLNetModelOutput(ModelOutput) - removed:
class XLNetPreTrainedModel(PreTrainedModel) - removed:
class XLNetRelativeAttention(nn.Module) - removed:
constructor XLNetFeedForward(config) - removed:
constructor XLNetForMultipleChoice(config) - removed:
constructor XLNetForQuestionAnswering(config) - removed:
constructor XLNetForQuestionAnsweringSimple(config) - removed:
constructor XLNetForSequenceClassification(config) - removed:
constructor XLNetForTokenClassification(config) - removed:
constructor XLNetLMHeadModel(config) - removed:
constructor XLNetLayer(config) - removed:
constructor XLNetModel(config) - removed:
constructor XLNetRelativeAttention(config) - removed:
field XLNetForMultipleChoiceOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForMultipleChoiceOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForMultipleChoiceOutput.logits: torch.FloatTensor - removed:
field XLNetForMultipleChoiceOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetForMultipleChoiceOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringOutput.cls_logits: Optional[torch.FloatTensor] - removed:
field XLNetForQuestionAnsweringOutput.end_top_index: Optional[torch.LongTensor] - removed:
field XLNetForQuestionAnsweringOutput.end_top_log_probs: Optional[torch.FloatTensor] - removed:
field XLNetForQuestionAnsweringOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetForQuestionAnsweringOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringOutput.start_top_index: Optional[torch.LongTensor] - removed:
field XLNetForQuestionAnsweringOutput.start_top_log_probs: Optional[torch.FloatTensor] - removed:
field XLNetForQuestionAnsweringSimpleOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringSimpleOutput.end_logits: torch.FloatTensor - removed:
field XLNetForQuestionAnsweringSimpleOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringSimpleOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetForQuestionAnsweringSimpleOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetForQuestionAnsweringSimpleOutput.start_logits: torch.FloatTensor - removed:
field XLNetForSequenceClassificationOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForSequenceClassificationOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForSequenceClassificationOutput.logits: torch.FloatTensor - removed:
field XLNetForSequenceClassificationOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetForSequenceClassificationOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetForTokenClassificationOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForTokenClassificationOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetForTokenClassificationOutput.logits: torch.FloatTensor - removed:
field XLNetForTokenClassificationOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetForTokenClassificationOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetLMHeadModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetLMHeadModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetLMHeadModelOutput.logits: torch.FloatTensor - removed:
field XLNetLMHeadModelOutput.loss: Optional[torch.FloatTensor] - removed:
field XLNetLMHeadModelOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - removed:
field XLNetModelOutput.last_hidden_state: torch.FloatTensor - removed:
field XLNetModelOutput.mems: Optional[List[torch.FloatTensor]] - removed:
field XLNetPreTrainedModel.base_model_prefix - removed:
field XLNetPreTrainedModel.config_class - removed:
field XLNetPreTrainedModel.load_tf_weights - removed:
function build_tf_xlnet_to_pytorch_map(model, config, tf_weights=…) - removed:
function load_tf_weights_in_xlnet(model, config, tf_path) - removed:
method XLNetFeedForward.forward(inp) - removed:
method XLNetForMultipleChoice.forward(input_ids=…, token_type_ids=…, input_mask=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetForQuestionAnswering.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, is_impossible=…, cls_index=…, p_mask=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetForQuestionAnsweringSimple.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetForSequenceClassification.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetForTokenClassification.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetLMHeadModel.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetLMHeadModel.get_output_embeddings() - removed:
method XLNetLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, **kwargs) - removed:
method XLNetLayer.ff_chunk(output_x) - removed:
method XLNetLayer.forward(output_h, output_g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…, output_attentions=…) - removed:
method XLNetModel.cache_mem(curr_out, prev_mem) - removed:
method XLNetModel.create_mask(qlen, mlen) - removed:
method XLNetModel.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…) - removed:
method XLNetModel.get_input_embeddings() - removed:
method XLNetModel.relative_positional_encoding(qlen, klen, bsz=…) - removed:
method XLNetModel.set_input_embeddings(new_embeddings) - removed:
method XLNetRelativeAttention.forward(h, g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…, output_attentions=…) - removed:
method XLNetRelativeAttention.post_attention(h, attn_vec, residual=…) - removed:
method XLNetRelativeAttention.prune_heads(heads) - removed:
method XLNetRelativeAttention.rel_attn_core(q_head, k_head_h, v_head_h, k_head_r, seg_mat=…, attn_mask=…, head_mask=…, output_attentions=…) - removed:
static method XLNetModel.positional_embedding(pos_seq, inv_freq, bsz=…) - removed:
static method XLNetRelativeAttention.rel_shift(x, klen=…) - removed:
static method XLNetRelativeAttention.rel_shift_bnij(x, klen=…) - removed:
variable XLNET_INPUTS_DOCSTRING - removed:
variable XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLNET_START_DOCSTRING - removed:
variable logger
- added:
export define_import_structure
- added:
export define_import_structure
- added:
class Aimv2Config(PretrainedConfig) - added:
class Aimv2TextConfig(PretrainedConfig) - added:
class Aimv2VisionConfig(PretrainedConfig) - added:
constructor Aimv2Config(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor Aimv2TextConfig(vocab_size: int=…, hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, rms_norm_eps: float=…, attention_dropout: float=…, qkv_bias: bool=…, mlp_bias: bool=…, hidden_act: str=…, pad_token_id: Optional[int]=…, bos_token_id: Optional[int]=…, eos_token_id: int=…, max_position_embeddings: int=…, initializer_range: bool=…, **kw… - added:
constructor Aimv2VisionConfig(hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_channels: int=…, image_size: int=…, patch_size: int=…, rms_norm_eps: float=…, attention_dropout: float=…, qkv_bias: bool=…, mlp_bias: bool=…, hidden_act: str=…, initializer_range: float=…, use_head: bool=…, is_native: bool=…, **kwargs) - added:
field Aimv2Config.model_type - added:
field Aimv2Config.sub_configs - added:
field Aimv2TextConfig.base_config_key - added:
field Aimv2TextConfig.model_type - added:
field Aimv2VisionConfig.base_config_key - added:
field Aimv2VisionConfig.model_type
- added:
class Aimv2Model(Aimv2PreTrainedModel) - added:
class Aimv2PreTrainedModel(PreTrainedModel) - added:
class Aimv2TextModel(Aimv2PreTrainedModel) - added:
class Aimv2VisionModel(Aimv2PreTrainedModel) - added:
constructor Aimv2Model(config: Aimv2Config) - added:
constructor Aimv2TextModel(config: Aimv2TextConfig) - added:
constructor Aimv2VisionModel(config: Aimv2VisionConfig) - added:
field Aimv2Model.config: Aimv2Config - added:
field Aimv2PreTrainedModel.base_model_prefix - added:
field Aimv2PreTrainedModel.config: Aimv2Config - added:
field Aimv2PreTrainedModel.supports_gradient_checkpointing - added:
field Aimv2TextModel.main_input_name - added:
field Aimv2VisionModel.config: Aimv2VisionConfig - added:
field Aimv2VisionModel.main_input_name - added:
method Aimv2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Aimv2Output - added:
method Aimv2Model.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method Aimv2Model.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method Aimv2TextModel.forward(input_ids, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Aimv2TextModel.get_input_embeddings() -> nn.Module - added:
method Aimv2TextModel.set_input_embeddings(value) - added:
method Aimv2VisionModel.forward(pixel_values, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Aimv2VisionModel.get_input_embeddings() -> nn.Module
- added:
class Aimv2Config(SiglipConfig) - added:
class Aimv2Model(CLIPModel) - added:
class Aimv2PreTrainedModel(PreTrainedModel) - added:
class Aimv2TextConfig(SiglipTextConfig) - added:
class Aimv2TextModel(Aimv2PreTrainedModel) - added:
class Aimv2VisionConfig(SiglipVisionConfig) - added:
class Aimv2VisionModel(Aimv2PreTrainedModel) - added:
constructor Aimv2Config(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor Aimv2Model(config: Aimv2Config) - added:
constructor Aimv2TextConfig(vocab_size: int=…, hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, rms_norm_eps: float=…, attention_dropout: float=…, qkv_bias: bool=…, mlp_bias: bool=…, hidden_act: str=…, pad_token_id: Optional[int]=…, bos_token_id: Optional[int]=…, eos_token_id: int=…, max_position_embeddings: int=…, initializer_range: bool=…, **kw… - added:
constructor Aimv2TextModel(config: Aimv2TextConfig) - added:
constructor Aimv2VisionConfig(hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_channels: int=…, image_size: int=…, patch_size: int=…, rms_norm_eps: float=…, attention_dropout: float=…, qkv_bias: bool=…, mlp_bias: bool=…, hidden_act: str=…, initializer_range: float=…, use_head: bool=…, is_native: bool=…, **kwargs) - added:
constructor Aimv2VisionModel(config: Aimv2VisionConfig) - added:
field Aimv2PreTrainedModel.base_model_prefix - added:
field Aimv2PreTrainedModel.config: Aimv2Config - added:
field Aimv2PreTrainedModel.supports_gradient_checkpointing - added:
field Aimv2TextModel.main_input_name - added:
field Aimv2VisionModel.config: Aimv2VisionConfig - added:
field Aimv2VisionModel.main_input_name - added:
method Aimv2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Aimv2Output - added:
method Aimv2TextModel.forward(input_ids, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Aimv2TextModel.get_input_embeddings() -> nn.Module - added:
method Aimv2TextModel.set_input_embeddings(value) - added:
method Aimv2VisionModel.forward(pixel_values, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Aimv2VisionModel.get_input_embeddings() -> nn.Module
- added:
export define_import_structure
- added:
class AlbertConfig(PretrainedConfig) - added:
class AlbertOnnxConfig(OnnxConfig) - added:
constructor AlbertConfig(vocab_size=…, embedding_size=…, hidden_size=…, num_hidden_layers=…, num_hidden_groups=…, num_attention_heads=…, intermediate_size=…, inner_group_num=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, classifier_dropout_prob=…, position_embedding_type=…, pad_token_id=…… - added:
field AlbertConfig.model_type - added:
property AlbertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class AlbertForMaskedLM(AlbertPreTrainedModel) - added:
class AlbertForMultipleChoice(AlbertPreTrainedModel) - added:
class AlbertForPreTraining(AlbertPreTrainedModel) - added:
class AlbertForQuestionAnswering(AlbertPreTrainedModel) - added:
class AlbertForSequenceClassification(AlbertPreTrainedModel) - added:
class AlbertForTokenClassification(AlbertPreTrainedModel) - added:
class AlbertModel(AlbertPreTrainedModel) - added:
class AlbertPreTrainedModel(PreTrainedModel) - added:
constructor AlbertForMaskedLM(config) - added:
constructor AlbertForMultipleChoice(config: AlbertConfig) - added:
constructor AlbertForPreTraining(config: AlbertConfig) - added:
constructor AlbertForQuestionAnswering(config: AlbertConfig) - added:
constructor AlbertForSequenceClassification(config: AlbertConfig) - added:
constructor AlbertForTokenClassification(config: AlbertConfig) - added:
constructor AlbertModel(config: AlbertConfig, add_pooling_layer: bool=…) - added:
field AlbertModel.base_model_prefix - added:
field AlbertModel.config: AlbertConfig - added:
field AlbertPreTrainedModel.base_model_prefix - added:
field AlbertPreTrainedModel.config: AlbertConfig - added:
field AlbertPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_albert(model, config, tf_checkpoint_path) - added:
method AlbertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Opt… - added:
method AlbertForMaskedLM.get_input_embeddings() -> nn.Embedding - added:
method AlbertForMaskedLM.get_output_embeddings() -> nn.Linear - added:
method AlbertForMaskedLM.set_output_embeddings(new_embeddings: nn.Linear) -> None - added:
method AlbertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_state… - added:
method AlbertForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, sentence_order_label: Optional[torch.LongTensor]=…, output_… - added:
method AlbertForPreTraining.get_input_embeddings() -> nn.Embedding - added:
method AlbertForPreTraining.get_output_embeddings() -> nn.Linear - added:
method AlbertForPreTraining.set_output_embeddings(new_embeddings: nn.Linear) -> None - added:
method AlbertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…,… - added:
method AlbertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method AlbertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_… - added:
method AlbertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…… - added:
method AlbertModel.get_input_embeddings() -> nn.Embedding - added:
method AlbertModel.set_input_embeddings(value: nn.Embedding) -> None
- added:
class FlaxAlbertForMaskedLM(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertForMultipleChoice(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertForPreTraining(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertForQuestionAnswering(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertForSequenceClassification(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertForTokenClassification(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertModel(FlaxAlbertPreTrainedModel) - added:
class FlaxAlbertPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxAlbertPreTrainedModel(config: AlbertConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxAlbertForMaskedLM.module_class - added:
field FlaxAlbertForMultipleChoice.module_class - added:
field FlaxAlbertForPreTraining.module_class - added:
field FlaxAlbertForQuestionAnswering.module_class - added:
field FlaxAlbertForSequenceClassification.module_class - added:
field FlaxAlbertForTokenClassification.module_class - added:
field FlaxAlbertModel.module_class - added:
field FlaxAlbertPreTrainedModel.base_model_prefix - added:
field FlaxAlbertPreTrainedModel.config_class - added:
field FlaxAlbertPreTrainedModel.module_class: nn.Module - added:
method FlaxAlbertPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFAlbertForMaskedLM(TFAlbertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFAlbertForPreTraining(TFAlbertPreTrainedModel, TFAlbertPreTrainingLoss) - added:
class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFAlbertMainLayer(keras.layers.Layer) - added:
class TFAlbertModel(TFAlbertPreTrainedModel) - added:
class TFAlbertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFAlbertForMaskedLM(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertForMultipleChoice(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertForPreTraining(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertForQuestionAnswering(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertForSequenceClassification(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertForTokenClassification(config: AlbertConfig, *inputs, **kwargs) - added:
constructor TFAlbertMainLayer(config: AlbertConfig, add_pooling_layer: bool=…, **kwargs) - added:
constructor TFAlbertModel(config: AlbertConfig, *inputs, **kwargs) - added:
field TFAlbertMainLayer.config_class - added:
field TFAlbertPreTrainedModel.base_model_prefix - added:
field TFAlbertPreTrainedModel.config_class - added:
method TFAlbertForMaskedLM.build(input_shape=…) - added:
method TFAlbertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | N… - added:
method TFAlbertForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFAlbertForMultipleChoice.build(input_shape=…) - added:
method TFAlbertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bo… - added:
method TFAlbertForPreTraining.build(input_shape=…) - added:
method TFAlbertForPreTraining.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFAlbertForPreTraining.get_lm_head() -> keras.layers.Layer - added:
method TFAlbertForQuestionAnswering.build(input_shape=…) - added:
method TFAlbertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict:… - added:
method TFAlbertForSequenceClassification.build(input_shape=…) - added:
method TFAlbertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_… - added:
method TFAlbertForTokenClassification.build(input_shape=…) - added:
method TFAlbertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dic… - added:
method TFAlbertMainLayer.build(input_shape=…) - added:
method TFAlbertMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | Non… - added:
method TFAlbertMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFAlbertMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFAlbertModel.build(input_shape=…) - added:
method TFAlbertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…,…
- added:
class AlbertTokenizer(PreTrainedTokenizer) - added:
constructor AlbertTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field AlbertTokenizer.vocab_files_names - added:
method AlbertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method AlbertTokenizer.convert_tokens_to_string(tokens) - added:
method AlbertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method AlbertTokenizer.get_vocab() -> dict[str, int] - added:
method AlbertTokenizer.preprocess_text(inputs) - added:
method AlbertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property AlbertTokenizer.vocab_size -> int
- added:
class AlbertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor AlbertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field AlbertTokenizerFast.slow_tokenizer_class - added:
field AlbertTokenizerFast.vocab_files_names - added:
method AlbertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method AlbertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class AlignConfig(PretrainedConfig) - added:
class AlignTextConfig(PretrainedConfig) - added:
class AlignVisionConfig(PretrainedConfig) - added:
constructor AlignConfig(text_config=…, vision_config=…, projection_dim=…, temperature_init_value=…, initializer_range=…, **kwargs) - added:
constructor AlignTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, **kwargs) - added:
constructor AlignVisionConfig(num_channels: int=…, image_size: int=…, width_coefficient: float=…, depth_coefficient: float=…, depth_divisor: int=…, kernel_sizes: list[int]=…, in_channels: list[int]=…, out_channels: list[int]=…, depthwise_padding: list[int]=…, strides: list[int]=…, num_block_repeats: list[int]=…, expand_ratios: list[int]=…, squeeze_expansion_ratio: float=…, hidden_act: str=…, hidd… - added:
field AlignConfig.model_type - added:
field AlignConfig.sub_configs - added:
field AlignTextConfig.base_config_key - added:
field AlignTextConfig.model_type - added:
field AlignVisionConfig.base_config_key - added:
field AlignVisionConfig.model_type
- added:
class AlignModel(AlignPreTrainedModel) - added:
class AlignPreTrainedModel(PreTrainedModel) - added:
class AlignTextModel(AlignPreTrainedModel) - added:
class AlignVisionModel(AlignPreTrainedModel) - added:
constructor AlignModel(config: AlignConfig) - added:
constructor AlignTextModel(config: AlignTextConfig, add_pooling_layer: bool=…) - added:
constructor AlignVisionModel(config: AlignVisionConfig) - added:
field AlignModel.config: AlignConfig - added:
field AlignPreTrainedModel.base_model_prefix - added:
field AlignPreTrainedModel.config: AlignConfig - added:
field AlignPreTrainedModel.supports_gradient_checkpointing - added:
field AlignTextModel.config: AlignTextConfig - added:
field AlignVisionModel.config: AlignVisionConfig - added:
field AlignVisionModel.main_input_name - added:
field AlignVisionModel.supports_gradient_checkpointing - added:
method AlignModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_sta… - added:
method AlignModel.get_image_features(pixel_values: torch.FloatTensor) -> torch.FloatTensor - added:
method AlignModel.get_text_features(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method AlignTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> Unio… - added:
method AlignTextModel.get_input_embeddings() - added:
method AlignTextModel.set_input_embeddings(value) - added:
method AlignVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention] - added:
method AlignVisionModel.get_input_embeddings() -> nn.Module
- added:
class AlignProcessor(ProcessorMixin) - added:
constructor AlignProcessor(image_processor, tokenizer) - added:
field AlignProcessor.attributes - added:
field AlignProcessor.image_processor_class - added:
field AlignProcessor.tokenizer_class - added:
field AlignProcessor.valid_processor_kwargs
- added:
export define_import_structure
- added:
class AltCLIPConfig(PretrainedConfig) - added:
class AltCLIPTextConfig(PretrainedConfig) - added:
class AltCLIPVisionConfig(PretrainedConfig) - added:
constructor AltCLIPConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor AltCLIPTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, initializer_factor=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, project_dim… - added:
constructor AltCLIPVisionConfig(hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field AltCLIPConfig.model_type - added:
field AltCLIPConfig.sub_configs - added:
field AltCLIPTextConfig.model_type - added:
field AltCLIPVisionConfig.base_config_key - added:
field AltCLIPVisionConfig.model_type
- added:
class AltCLIPModel(AltCLIPPreTrainedModel) - added:
class AltCLIPPreTrainedModel(PreTrainedModel) - added:
class AltCLIPTextModel(AltCLIPPreTrainedModel) - added:
class AltCLIPVisionModel(AltCLIPPreTrainedModel) - added:
constructor AltCLIPModel(config: AltCLIPConfig) - added:
constructor AltCLIPTextModel(config) - added:
constructor AltCLIPVisionModel(config: AltCLIPVisionConfig) - added:
field AltCLIPModel.config: AltCLIPConfig - added:
field AltCLIPPreTrainedModel.base_model_prefix - added:
field AltCLIPPreTrainedModel.config: AltCLIPConfig - added:
field AltCLIPPreTrainedModel.supports_gradient_checkpointing - added:
field AltCLIPTextModel.config: AltCLIPTextConfig - added:
field AltCLIPVisionModel.config: AltCLIPVisionConfig - added:
field AltCLIPVisionModel.main_input_name - added:
method AltCLIPModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.Tensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Op… - added:
method AltCLIPModel.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method AltCLIPModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method AltCLIPTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> Union[tuple, Base… - added:
method AltCLIPTextModel.get_input_embeddings() -> nn.Module - added:
method AltCLIPTextModel.resize_token_embeddings(new_num_tokens: Optional[int]=…) -> nn.Embedding - added:
method AltCLIPTextModel.set_input_embeddings(value: nn.Embedding) -> None - added:
method AltCLIPVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method AltCLIPVisionModel.get_input_embeddings() -> nn.Module
- added:
class AltCLIPProcessor(ProcessorMixin) - added:
constructor AltCLIPProcessor(image_processor=…, tokenizer=…) - added:
field AltCLIPProcessor.attributes - added:
field AltCLIPProcessor.image_processor_class - added:
field AltCLIPProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class ApertusConfig(PretrainedConfig) - added:
constructor ApertusConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
field ApertusConfig.base_model_pp_plan - added:
field ApertusConfig.base_model_tp_plan - added:
field ApertusConfig.keys_to_ignore_at_inference - added:
field ApertusConfig.model_type
- added:
class ApertusForCausalLM(ApertusPreTrainedModel, GenerationMixin) - added:
class ApertusForTokenClassification(GenericForTokenClassification, ApertusPreTrainedModel) - added:
class ApertusModel(ApertusPreTrainedModel) - added:
class ApertusPreTrainedModel(PreTrainedModel) - added:
constructor ApertusForCausalLM(config) - added:
constructor ApertusModel(config: ApertusConfig) - added:
field ApertusPreTrainedModel.base_model_prefix - added:
field ApertusPreTrainedModel.config: ApertusConfig - added:
field ApertusPreTrainedModel.supports_gradient_checkpointing - added:
method ApertusForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method ApertusModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class ApertusConfig(LlamaConfig) - added:
class ApertusForCausalLM(LlamaForCausalLM) - added:
class ApertusForTokenClassification(LlamaForTokenClassification) - added:
class ApertusModel(LlamaModel) - added:
class ApertusPreTrainedModel(LlamaPreTrainedModel) - added:
constructor ApertusConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
field ApertusConfig.base_model_tp_plan - added:
field ApertusConfig.model_type - added:
method ApertusForCausalLM.forward(**super_kwargs)
- added:
export define_import_structure
- added:
class ArceeConfig(PretrainedConfig) - added:
constructor ArceeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, head_dim=…, **k… - added:
field ArceeConfig.base_model_pp_plan - added:
field ArceeConfig.base_model_tp_plan - added:
field ArceeConfig.keys_to_ignore_at_inference - added:
field ArceeConfig.model_type
- added:
class ArceeForCausalLM(ArceePreTrainedModel, GenerationMixin) - added:
class ArceeForQuestionAnswering(GenericForQuestionAnswering, ArceePreTrainedModel) - added:
class ArceeForSequenceClassification(GenericForSequenceClassification, ArceePreTrainedModel) - added:
class ArceeForTokenClassification(GenericForTokenClassification, ArceePreTrainedModel) - added:
class ArceeModel(ArceePreTrainedModel) - added:
class ArceePreTrainedModel(PreTrainedModel) - added:
constructor ArceeForCausalLM(config) - added:
constructor ArceeModel(config: ArceeConfig) - added:
field ArceeForQuestionAnswering.base_model_prefix - added:
field ArceePreTrainedModel.base_model_prefix - added:
field ArceePreTrainedModel.config: ArceeConfig - added:
field ArceePreTrainedModel.supports_gradient_checkpointing - added:
method ArceeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method ArceeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class ArceeConfig(LlamaConfig) - added:
class ArceeForCausalLM(LlamaForCausalLM) - added:
class ArceeForQuestionAnswering(LlamaForQuestionAnswering) - added:
class ArceeForSequenceClassification(LlamaForSequenceClassification) - added:
class ArceeForTokenClassification(LlamaForTokenClassification) - added:
constructor ArceeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, head_dim=…, **k… - added:
field ArceeConfig.base_model_tp_plan - added:
field ArceeConfig.model_type
- added:
export define_import_structure
- added:
class AriaConfig(PretrainedConfig) - added:
class AriaTextConfig(PretrainedConfig) - added:
constructor AriaConfig(vision_config=…, vision_feature_layer: int=…, text_config: AriaTextConfig=…, projector_patch_to_query_dict: Optional[dict]=…, image_token_index: int=…, initializer_range: float=…, **kwargs) - added:
constructor AriaTextConfig(vocab_size=…, hidden_size=…, intermediate_size: int=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, pretraining_tp=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, m… - added:
field AriaConfig.attribute_map - added:
field AriaConfig.model_type - added:
field AriaConfig.sub_configs - added:
field AriaTextConfig.base_config_key - added:
field AriaTextConfig.base_model_pp_plan - added:
field AriaTextConfig.base_model_tp_plan - added:
field AriaTextConfig.keys_to_ignore_at_inference - added:
field AriaTextConfig.model_type
- added:
class AriaImageProcessor(BaseImageProcessor) - added:
constructor AriaImageProcessor(image_mean: Optional[list[float]]=…, image_std: Optional[list[float]]=…, max_image_size: int=…, min_image_size: int=…, split_resolutions: Optional[list[tuple[int, int]]]=…, split_image: Optional[bool]=…, do_convert_rgb: Optional[bool]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: Optional[bool]=…, resample: PILImageResampling=…, **kwargs) - added:
field AriaImageProcessor.model_input_names - added:
method AriaImageProcessor.get_image_patches(image: np.ndarray, grid_pinpoints: list[tuple[int, int]], patch_size: int, resample: PILImageResampling, data_format: ChannelDimension, input_data_format: ChannelDimension) -> list[np.ndarray] - added:
method AriaImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method AriaImageProcessor.pad(image: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method AriaImageProcessor.preprocess(images: Union[ImageInput, list[ImageInput]], image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, max_image_size: Optional[int]=…, min_image_size: Optional[int]=…, split_image: Optional[bool]=…, do_convert_rgb: Optional[bool]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Option…
- added:
class AriaForConditionalGeneration(AriaPreTrainedModel, GenerationMixin) - added:
class AriaModel(AriaPreTrainedModel) - added:
class AriaPreTrainedModel(PreTrainedModel) - added:
class AriaTextForCausalLM(AriaTextPreTrainedModel, GenerationMixin) - added:
class AriaTextModel(AriaTextPreTrainedModel) - added:
class AriaTextPreTrainedModel(PreTrainedModel) - added:
constructor AriaForConditionalGeneration(config: AriaConfig) - added:
constructor AriaModel(config: AriaConfig) - added:
constructor AriaTextForCausalLM(config: AriaTextConfig) - added:
constructor AriaTextModel(config: AriaTextConfig) - added:
field AriaPreTrainedModel.base_model_prefix - added:
field AriaPreTrainedModel.config: AriaConfig - added:
field AriaPreTrainedModel.supports_gradient_checkpointing - added:
field AriaTextPreTrainedModel.base_model_prefix - added:
field AriaTextPreTrainedModel.config: AriaTextConfig - added:
field AriaTextPreTrainedModel.supports_gradient_checkpointing - added:
method AriaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[b… - added:
method AriaForConditionalGeneration.get_decoder() - added:
method AriaForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, vision_feature_layer: int=…) - added:
method AriaForConditionalGeneration.get_input_embeddings() - added:
method AriaForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method AriaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_mask=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method AriaForConditionalGeneration.set_decoder(decoder) - added:
method AriaForConditionalGeneration.set_input_embeddings(value) - added:
method AriaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **k… - added:
method AriaModel.get_decoder() - added:
method AriaModel.get_image_features(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, vision_feature_layer: int=…) - added:
method AriaModel.get_input_embeddings() - added:
method AriaModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method AriaModel.set_decoder(decoder) - added:
method AriaModel.set_input_embeddings(value) - added:
method AriaTextForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…… - added:
method AriaTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
property AriaForConditionalGeneration.language_model - added:
property AriaForConditionalGeneration.multi_modal_projector - added:
property AriaForConditionalGeneration.vision_tower
- added:
class AriaConfig(PretrainedConfig) - added:
class AriaForConditionalGeneration(LlavaForConditionalGeneration) - added:
class AriaImageProcessor(BaseImageProcessor) - added:
class AriaModel(LlavaModel) - added:
class AriaPreTrainedModel(LlamaPreTrainedModel) - added:
class AriaProcessor(ProcessorMixin) - added:
class AriaTextConfig(LlamaConfig) - added:
class AriaTextForCausalLM(AriaTextPreTrainedModel, LlamaForCausalLM) - added:
class AriaTextModel(LlamaModel) - added:
class AriaTextPreTrainedModel(PreTrainedModel) - added:
constructor AriaConfig(vision_config=…, vision_feature_layer: int=…, text_config: AriaTextConfig=…, projector_patch_to_query_dict: Optional[dict]=…, image_token_index: int=…, initializer_range: float=…, **kwargs) - added:
constructor AriaImageProcessor(image_mean: Optional[list[float]]=…, image_std: Optional[list[float]]=…, max_image_size: int=…, min_image_size: int=…, split_resolutions: Optional[list[tuple[int, int]]]=…, split_image: Optional[bool]=…, do_convert_rgb: Optional[bool]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: Optional[bool]=…, resample: PILImageResampling=…, **kwargs) - added:
constructor AriaModel(config: AriaConfig) - added:
constructor AriaProcessor(image_processor=…, tokenizer: Union[AutoTokenizer, str]=…, chat_template: Optional[str]=…, size_conversion: Optional[dict[Union[float, int], int]]=…) - added:
constructor AriaTextConfig(intermediate_size: int=…, moe_num_experts: int=…, moe_topk: int=…, moe_num_shared_experts: int=…, pad_token_id=…, **super_kwargs) - added:
constructor AriaTextForCausalLM(config: AriaTextConfig) - added:
constructor AriaTextModel(config: AriaTextConfig) - added:
field AriaConfig.attribute_map - added:
field AriaConfig.model_type - added:
field AriaConfig.sub_configs - added:
field AriaImageProcessor.model_input_names - added:
field AriaPreTrainedModel.base_model_prefix - added:
field AriaPreTrainedModel.config: AriaConfig - added:
field AriaProcessor.attributes - added:
field AriaProcessor.image_processor_class - added:
field AriaProcessor.tokenizer_class - added:
field AriaTextConfig.base_config_key - added:
field AriaTextConfig.model_type - added:
field AriaTextPreTrainedModel.base_model_prefix - added:
field AriaTextPreTrainedModel.config: AriaTextConfig - added:
field AriaTextPreTrainedModel.supports_gradient_checkpointing - added:
method AriaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[b… - added:
method AriaForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, vision_feature_layer: int=…) - added:
method AriaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_mask=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method AriaImageProcessor.get_image_patches(image: np.ndarray, grid_pinpoints: list[tuple[int, int]], patch_size: int, resample: PILImageResampling, data_format: ChannelDimension, input_data_format: ChannelDimension) -> list[np.ndarray] - added:
method AriaImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method AriaImageProcessor.pad(image: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method AriaImageProcessor.preprocess(images: Union[ImageInput, list[ImageInput]], image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, max_image_size: Optional[int]=…, min_image_size: Optional[int]=…, split_image: Optional[bool]=…, do_convert_rgb: Optional[bool]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Option… - added:
method AriaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **k… - added:
method AriaModel.get_image_features(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, vision_feature_layer: int=…) - added:
method AriaTextForCausalLM.forward(**super_kwargs) - added:
property AriaProcessor.model_input_names
- added:
class AriaProcessor(ProcessorMixin) - added:
constructor AriaProcessor(image_processor=…, tokenizer: Union[AutoTokenizer, str]=…, chat_template: Optional[str]=…, size_conversion: Optional[dict[Union[float, int], int]]=…) - added:
field AriaProcessor.attributes - added:
field AriaProcessor.image_processor_class - added:
field AriaProcessor.tokenizer_class - added:
property AriaProcessor.model_input_names
- added:
export define_import_structure
- added:
class ASTConfig(PretrainedConfig) - added:
constructor ASTConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, patch_size=…, qkv_bias=…, frequency_stride=…, time_stride=…, max_length=…, num_mel_bins=…, **kwargs) - added:
field ASTConfig.model_type
- added:
class ASTFeatureExtractor(SequenceFeatureExtractor) - added:
constructor ASTFeatureExtractor(feature_size=…, sampling_rate=…, num_mel_bins=…, max_length=…, padding_value=…, do_normalize=…, mean=…, std=…, return_attention_mask=…, **kwargs) - added:
field ASTFeatureExtractor.model_input_names - added:
method ASTFeatureExtractor.normalize(input_values: np.ndarray) -> np.ndarray
- added:
class ASTForAudioClassification(ASTPreTrainedModel) - added:
class ASTModel(ASTPreTrainedModel) - added:
class ASTPreTrainedModel(PreTrainedModel) - added:
constructor ASTForAudioClassification(config: ASTConfig) -> None - added:
constructor ASTModel(config: ASTConfig) -> None - added:
field ASTPreTrainedModel.base_model_prefix - added:
field ASTPreTrainedModel.config: ASTConfig - added:
field ASTPreTrainedModel.main_input_name - added:
field ASTPreTrainedModel.supports_gradient_checkpointing - added:
method ASTForAudioClassification.forward(input_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> SequenceClassifierOutput - added:
method ASTModel.forward(input_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method ASTModel.get_input_embeddings() -> ASTPatchEmbeddings
- added:
export define_import_structure
- added:
function get_values(model_mapping)
- added:
class AutoConfig - added:
constructor AutoConfig() -> None - added:
static method AutoConfig.for_model(model_type: str, *args, **kwargs) -> PretrainedConfig - added:
static method AutoConfig.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike[str]], **kwargs) - added:
static method AutoConfig.register(model_type, config, exist_ok=…) -> None - added:
variable CONFIG_MAPPING - added:
variable MODEL_NAMES_MAPPING
- added:
class AutoFeatureExtractor - added:
constructor AutoFeatureExtractor() - added:
static method AutoFeatureExtractor.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
static method AutoFeatureExtractor.register(config_class, feature_extractor_class, exist_ok=…) - added:
variable FEATURE_EXTRACTOR_MAPPING
- added:
class AutoImageProcessor - added:
constructor AutoImageProcessor() - added:
static method AutoImageProcessor.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs) - added:
static method AutoImageProcessor.register(config_class, image_processor_class=…, slow_image_processor_class=…, fast_image_processor_class=…, exist_ok=…) - added:
variable IMAGE_PROCESSOR_MAPPING
- added:
class AutoBackbone(_BaseAutoBackboneClass) - added:
class AutoModel(_BaseAutoModelClass) - added:
class AutoModelForAudioClassification(_BaseAutoModelClass) - added:
class AutoModelForAudioFrameClassification(_BaseAutoModelClass) - added:
class AutoModelForAudioTokenization(_BaseAutoModelClass) - added:
class AutoModelForAudioXVector(_BaseAutoModelClass) - added:
class AutoModelForCTC(_BaseAutoModelClass) - added:
class AutoModelForCausalLM(_BaseAutoModelClass) - added:
class AutoModelForDepthEstimation(_BaseAutoModelClass) - added:
class AutoModelForDocumentQuestionAnswering(_BaseAutoModelClass) - added:
class AutoModelForImageClassification(_BaseAutoModelClass) - added:
class AutoModelForImageSegmentation(_BaseAutoModelClass) - added:
class AutoModelForImageTextToText(_BaseAutoModelClass) - added:
class AutoModelForImageToImage(_BaseAutoModelClass) - added:
class AutoModelForInstanceSegmentation(_BaseAutoModelClass) - added:
class AutoModelForKeypointDetection(_BaseAutoModelClass) - added:
class AutoModelForKeypointMatching(_BaseAutoModelClass) - added:
class AutoModelForMaskGeneration(_BaseAutoModelClass) - added:
class AutoModelForMaskedImageModeling(_BaseAutoModelClass) - added:
class AutoModelForMaskedLM(_BaseAutoModelClass) - added:
class AutoModelForMultipleChoice(_BaseAutoModelClass) - added:
class AutoModelForNextSentencePrediction(_BaseAutoModelClass) - added:
class AutoModelForObjectDetection(_BaseAutoModelClass) - added:
class AutoModelForPreTraining(_BaseAutoModelClass) - added:
class AutoModelForQuestionAnswering(_BaseAutoModelClass) - added:
class AutoModelForSemanticSegmentation(_BaseAutoModelClass) - added:
class AutoModelForSeq2SeqLM(_BaseAutoModelClass) - added:
class AutoModelForSequenceClassification(_BaseAutoModelClass) - added:
class AutoModelForSpeechSeq2Seq(_BaseAutoModelClass) - added:
class AutoModelForTableQuestionAnswering(_BaseAutoModelClass) - added:
class AutoModelForTextEncoding(_BaseAutoModelClass) - added:
class AutoModelForTextToSpectrogram(_BaseAutoModelClass) - added:
class AutoModelForTextToWaveform(_BaseAutoModelClass) - added:
class AutoModelForTimeSeriesPrediction(_BaseAutoModelClass) - added:
class AutoModelForTokenClassification(_BaseAutoModelClass) - added:
class AutoModelForUniversalSegmentation(_BaseAutoModelClass) - added:
class AutoModelForVideoClassification(_BaseAutoModelClass) - added:
class AutoModelForVision2Seq(_AutoModelForVision2Seq) - added:
class AutoModelForVisualQuestionAnswering(_BaseAutoModelClass) - added:
class AutoModelForZeroShotImageClassification(_BaseAutoModelClass) - added:
class AutoModelForZeroShotObjectDetection(_BaseAutoModelClass) - added:
class AutoModelWithLMHead(_AutoModelWithLMHead) - added:
static method AutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike[str]], *model_args, **kwargs) -> _BaseModelWithGenerate - added:
static method AutoModelForImageTextToText.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike[str]], *model_args, **kwargs) -> _BaseModelWithGenerate - added:
static method AutoModelForVision2Seq.from_config(config, **kwargs) - added:
static method AutoModelForVision2Seq.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method AutoModelWithLMHead.from_config(config, **kwargs) - added:
static method AutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable AutoModel - added:
variable AutoModelForAudioClassification - added:
variable AutoModelForAudioFrameClassification - added:
variable AutoModelForAudioTokenization - added:
variable AutoModelForAudioXVector - added:
variable AutoModelForCTC - added:
variable AutoModelForCausalLM - added:
variable AutoModelForDepthEstimation - added:
variable AutoModelForDocumentQuestionAnswering - added:
variable AutoModelForImageClassification - added:
variable AutoModelForImageSegmentation - added:
variable AutoModelForImageTextToText - added:
variable AutoModelForInstanceSegmentation - added:
variable AutoModelForMaskedImageModeling - added:
variable AutoModelForMaskedLM - added:
variable AutoModelForMultipleChoice - added:
variable AutoModelForNextSentencePrediction - added:
variable AutoModelForObjectDetection - added:
variable AutoModelForPreTraining - added:
variable AutoModelForQuestionAnswering - added:
variable AutoModelForSemanticSegmentation - added:
variable AutoModelForSeq2SeqLM - added:
variable AutoModelForSequenceClassification - added:
variable AutoModelForSpeechSeq2Seq - added:
variable AutoModelForTableQuestionAnswering - added:
variable AutoModelForTimeSeriesPrediction - added:
variable AutoModelForTokenClassification - added:
variable AutoModelForUniversalSegmentation - added:
variable AutoModelForVideoClassification - added:
variable AutoModelForVisualQuestionAnswering - added:
variable AutoModelForZeroShotImageClassification - added:
variable AutoModelForZeroShotObjectDetection - added:
variable MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_AUDIO_TOKENIZATION_MAPPING - added:
variable MODEL_FOR_AUDIO_XVECTOR_MAPPING - added:
variable MODEL_FOR_BACKBONE_MAPPING - added:
variable MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPING - added:
variable MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable MODEL_FOR_CTC_MAPPING - added:
variable MODEL_FOR_DEPTH_ESTIMATION_MAPPING - added:
variable MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING - added:
variable MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_IMAGE_MAPPING - added:
variable MODEL_FOR_IMAGE_SEGMENTATION_MAPPING - added:
variable MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING - added:
variable MODEL_FOR_IMAGE_TO_IMAGE_MAPPING - added:
variable MODEL_FOR_INSTANCE_SEGMENTATION_MAPPING - added:
variable MODEL_FOR_KEYPOINT_DETECTION_MAPPING - added:
variable MODEL_FOR_KEYPOINT_MATCHING_MAPPING - added:
variable MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING - added:
variable MODEL_FOR_MASKED_LM_MAPPING - added:
variable MODEL_FOR_MASK_GENERATION_MAPPING - added:
variable MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
variable MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
variable MODEL_FOR_OBJECT_DETECTION_MAPPING - added:
variable MODEL_FOR_PRETRAINING_MAPPING - added:
variable MODEL_FOR_QUESTION_ANSWERING_MAPPING - added:
variable MODEL_FOR_RETRIEVAL_MAPPING - added:
variable MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING - added:
variable MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
variable MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING - added:
variable MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING - added:
variable MODEL_FOR_TEXT_ENCODING_MAPPING - added:
variable MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING - added:
variable MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING - added:
variable MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_TIME_SERIES_PREDICTION_MAPPING - added:
variable MODEL_FOR_TIME_SERIES_REGRESSION_MAPPING - added:
variable MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING - added:
variable MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_VISION_2_SEQ_MAPPING - added:
variable MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING - added:
variable MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING - added:
variable MODEL_MAPPING - added:
variable MODEL_WITH_LM_HEAD_MAPPING
- added:
class FlaxAutoModel(_BaseAutoModelClass) - added:
class FlaxAutoModelForCausalLM(_BaseAutoModelClass) - added:
class FlaxAutoModelForImageClassification(_BaseAutoModelClass) - added:
class FlaxAutoModelForMaskedLM(_BaseAutoModelClass) - added:
class FlaxAutoModelForMultipleChoice(_BaseAutoModelClass) - added:
class FlaxAutoModelForNextSentencePrediction(_BaseAutoModelClass) - added:
class FlaxAutoModelForPreTraining(_BaseAutoModelClass) - added:
class FlaxAutoModelForQuestionAnswering(_BaseAutoModelClass) - added:
class FlaxAutoModelForSeq2SeqLM(_BaseAutoModelClass) - added:
class FlaxAutoModelForSequenceClassification(_BaseAutoModelClass) - added:
class FlaxAutoModelForSpeechSeq2Seq(_BaseAutoModelClass) - added:
class FlaxAutoModelForTokenClassification(_BaseAutoModelClass) - added:
class FlaxAutoModelForVision2Seq(_BaseAutoModelClass) - added:
variable FLAX_MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING - added:
variable FLAX_MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable FLAX_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING - added:
variable FLAX_MODEL_FOR_MASKED_LM_MAPPING - added:
variable FLAX_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
variable FLAX_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
variable FLAX_MODEL_FOR_PRETRAINING_MAPPING - added:
variable FLAX_MODEL_FOR_QUESTION_ANSWERING_MAPPING - added:
variable FLAX_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - added:
variable FLAX_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
variable FLAX_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING - added:
variable FLAX_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - added:
variable FLAX_MODEL_FOR_VISION_2_SEQ_MAPPING - added:
variable FLAX_MODEL_MAPPING - added:
variable FlaxAutoModel - added:
variable FlaxAutoModelForCausalLM - added:
variable FlaxAutoModelForImageClassification - added:
variable FlaxAutoModelForMaskedLM - added:
variable FlaxAutoModelForMultipleChoice - added:
variable FlaxAutoModelForNextSentencePrediction - added:
variable FlaxAutoModelForPreTraining - added:
variable FlaxAutoModelForQuestionAnswering - added:
variable FlaxAutoModelForSeq2SeqLM - added:
variable FlaxAutoModelForSequenceClassification - added:
variable FlaxAutoModelForSpeechSeq2Seq - added:
variable FlaxAutoModelForTokenClassification - added:
variable FlaxAutoModelForVision2Seq
- added:
class TFAutoModel(_BaseAutoModelClass) - added:
class TFAutoModelForAudioClassification(_BaseAutoModelClass) - added:
class TFAutoModelForCausalLM(_BaseAutoModelClass) - added:
class TFAutoModelForDocumentQuestionAnswering(_BaseAutoModelClass) - added:
class TFAutoModelForImageClassification(_BaseAutoModelClass) - added:
class TFAutoModelForMaskGeneration(_BaseAutoModelClass) - added:
class TFAutoModelForMaskedImageModeling(_BaseAutoModelClass) - added:
class TFAutoModelForMaskedLM(_BaseAutoModelClass) - added:
class TFAutoModelForMultipleChoice(_BaseAutoModelClass) - added:
class TFAutoModelForNextSentencePrediction(_BaseAutoModelClass) - added:
class TFAutoModelForPreTraining(_BaseAutoModelClass) - added:
class TFAutoModelForQuestionAnswering(_BaseAutoModelClass) - added:
class TFAutoModelForSemanticSegmentation(_BaseAutoModelClass) - added:
class TFAutoModelForSeq2SeqLM(_BaseAutoModelClass) - added:
class TFAutoModelForSequenceClassification(_BaseAutoModelClass) - added:
class TFAutoModelForSpeechSeq2Seq(_BaseAutoModelClass) - added:
class TFAutoModelForTableQuestionAnswering(_BaseAutoModelClass) - added:
class TFAutoModelForTextEncoding(_BaseAutoModelClass) - added:
class TFAutoModelForTokenClassification(_BaseAutoModelClass) - added:
class TFAutoModelForVision2Seq(_BaseAutoModelClass) - added:
class TFAutoModelForZeroShotImageClassification(_BaseAutoModelClass) - added:
class TFAutoModelWithLMHead(_TFAutoModelWithLMHead) - added:
static method TFAutoModelWithLMHead.from_config(config) - added:
static method TFAutoModelWithLMHead.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable TFAutoModel - added:
variable TFAutoModelForAudioClassification - added:
variable TFAutoModelForCausalLM - added:
variable TFAutoModelForDocumentQuestionAnswering - added:
variable TFAutoModelForImageClassification - added:
variable TFAutoModelForMaskedImageModeling - added:
variable TFAutoModelForMaskedLM - added:
variable TFAutoModelForMultipleChoice - added:
variable TFAutoModelForNextSentencePrediction - added:
variable TFAutoModelForPreTraining - added:
variable TFAutoModelForQuestionAnswering - added:
variable TFAutoModelForSemanticSegmentation - added:
variable TFAutoModelForSeq2SeqLM - added:
variable TFAutoModelForSequenceClassification - added:
variable TFAutoModelForSpeechSeq2Seq - added:
variable TFAutoModelForTableQuestionAnswering - added:
variable TFAutoModelForTokenClassification - added:
variable TFAutoModelForVision2Seq - added:
variable TFAutoModelForZeroShotImageClassification - added:
variable TF_MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable TF_MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING - added:
variable TF_MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING - added:
variable TF_MODEL_FOR_MASKED_LM_MAPPING - added:
variable TF_MODEL_FOR_MASK_GENERATION_MAPPING - added:
variable TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
variable TF_MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
variable TF_MODEL_FOR_PRETRAINING_MAPPING - added:
variable TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING - added:
variable TF_MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING - added:
variable TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
variable TF_MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING - added:
variable TF_MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING - added:
variable TF_MODEL_FOR_TEXT_ENCODING_MAPPING - added:
variable TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_FOR_VISION_2_SEQ_MAPPING - added:
variable TF_MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_MAPPING - added:
variable TF_MODEL_WITH_LM_HEAD_MAPPING
- added:
class AutoProcessor - added:
constructor AutoProcessor() - added:
static method AutoProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
static method AutoProcessor.register(config_class, processor_class, exist_ok=…) - added:
variable PROCESSOR_MAPPING
- added:
class AutoTokenizer - added:
constructor AutoTokenizer() - added:
static method AutoTokenizer.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs) - added:
static method AutoTokenizer.register(config_class, slow_tokenizer_class=…, fast_tokenizer_class=…, exist_ok=…) - added:
variable TOKENIZER_MAPPING
- added:
class AutoVideoProcessor - added:
constructor AutoVideoProcessor() - added:
static method AutoVideoProcessor.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs) - added:
static method AutoVideoProcessor.register(config_class, video_processor_class, exist_ok=…) - added:
variable VIDEO_PROCESSOR_MAPPING
- added:
export define_import_structure
- added:
class AutoformerConfig(PretrainedConfig) - added:
constructor AutoformerConfig(prediction_length: Optional[int]=…, context_length: Optional[int]=…, distribution_output: str=…, loss: str=…, input_size: int=…, lags_sequence: list[int]=…, scaling: bool=…, num_time_features: int=…, num_dynamic_real_features: int=…, num_static_categorical_features: int=…, num_static_real_features: int=…, cardinality: Optional[list[int]]=…, embedding_dimension: Option… - added:
field AutoformerConfig.attribute_map - added:
field AutoformerConfig.model_type
- added:
class AutoformerForPrediction(AutoformerPreTrainedModel) - added:
class AutoformerModel(AutoformerPreTrainedModel) - added:
class AutoformerPreTrainedModel(PreTrainedModel) - added:
constructor AutoformerForPrediction(config: AutoformerConfig) - added:
constructor AutoformerModel(config: AutoformerConfig) - added:
field AutoformerPreTrainedModel.base_model_prefix - added:
field AutoformerPreTrainedModel.config: AutoformerConfig - added:
field AutoformerPreTrainedModel.main_input_name - added:
field AutoformerPreTrainedModel.supports_gradient_checkpointing - added:
method AutoformerForPrediction.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, future_observed_mask: Optional[torch.Tensor]=…, decoder_attention_mask: O… - added:
method AutoformerForPrediction.generate(past_values: torch.Tensor, past_time_features: torch.Tensor, future_time_features: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> SampleTSPredictionOutput - added:
method AutoformerForPrediction.get_decoder() - added:
method AutoformerForPrediction.get_encoder() - added:
method AutoformerForPrediction.output_distribution(params, loc=…, scale=…, trailing_n=…) -> torch.distributions.Distribution - added:
method AutoformerForPrediction.output_params(decoder_output) - added:
method AutoformerModel.create_network_inputs(past_values: torch.Tensor, past_time_features: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, past_observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torc… - added:
method AutoformerModel.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.T… - added:
method AutoformerModel.get_encoder() - added:
method AutoformerModel.get_lagged_subsequences(sequence: torch.Tensor, subsequences_length: int, shift: int=…) -> torch.Tensor
- added:
export define_import_structure
- added:
class AyaVisionConfig(PretrainedConfig) - added:
constructor AyaVisionConfig(vision_config=…, text_config=…, vision_feature_select_strategy=…, vision_feature_layer=…, downsample_factor=…, adapter_layer_norm_eps=…, image_token_index=…, **kwargs) - added:
field AyaVisionConfig.attribute_map - added:
field AyaVisionConfig.model_type - added:
field AyaVisionConfig.sub_configs
- added:
class AyaVisionForConditionalGeneration(AyaVisionPreTrainedModel, GenerationMixin) - added:
class AyaVisionModel(AyaVisionPreTrainedModel) - added:
class AyaVisionPreTrainedModel(PreTrainedModel) - added:
constructor AyaVisionForConditionalGeneration(config: AyaVisionConfig) - added:
constructor AyaVisionModel(config: AyaVisionConfig) - added:
field AyaVisionPreTrainedModel.base_model_prefix - added:
field AyaVisionPreTrainedModel.config: AyaVisionConfig - added:
field AyaVisionPreTrainedModel.supports_gradient_checkpointing - added:
method AyaVisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optiona… - added:
method AyaVisionForConditionalGeneration.get_decoder() - added:
method AyaVisionForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method AyaVisionForConditionalGeneration.get_input_embeddings() - added:
method AyaVisionForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method AyaVisionForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method AyaVisionForConditionalGeneration.set_decoder(decoder) - added:
method AyaVisionForConditionalGeneration.set_input_embeddings(value) - added:
method AyaVisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, use_cache… - added:
method AyaVisionModel.get_decoder() - added:
method AyaVisionModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method AyaVisionModel.get_input_embeddings() - added:
method AyaVisionModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method AyaVisionModel.set_decoder(decoder) - added:
method AyaVisionModel.set_input_embeddings(value) - added:
property AyaVisionForConditionalGeneration.language_model - added:
property AyaVisionForConditionalGeneration.multi_modal_projector - added:
property AyaVisionForConditionalGeneration.vision_tower
- added:
class AyaVisionForConditionalGeneration(LlavaForConditionalGeneration) - added:
class AyaVisionModel(LlavaModel) - added:
class AyaVisionPreTrainedModel(LlavaPreTrainedModel) - added:
method AyaVisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optiona… - added:
method AyaVisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, use_cache… - added:
method AyaVisionModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs)
- added:
class AyaVisionProcessor(ProcessorMixin) - added:
constructor AyaVisionProcessor(image_processor=…, tokenizer=…, patch_size: int=…, img_size: int=…, image_token=…, downsample_factor: int=…, start_of_img_token=…, end_of_img_token=…, img_patch_token=…, img_line_break_token=…, tile_token=…, tile_global_token=…, chat_template=…, **kwargs) - added:
field AyaVisionProcessor.attributes - added:
field AyaVisionProcessor.image_processor_class - added:
field AyaVisionProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class BambaConfig(PretrainedConfig) - added:
constructor BambaConfig(vocab_size=…, tie_word_embeddings=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, num_logits_to_keep=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_position_embeddings=…, attention_dropout=…, attn_layer_indices=…, mamba_n_heads=…, mamba_d_head=…… - added:
field BambaConfig.keys_to_ignore_at_inference - added:
field BambaConfig.model_type - added:
property BambaConfig.layers_block_type
- added:
class BambaForCausalLM(BambaPreTrainedModel, GenerationMixin) - added:
class BambaModel(BambaPreTrainedModel) - added:
class BambaPreTrainedModel(PreTrainedModel) - added:
constructor BambaForCausalLM(config) - added:
constructor BambaModel(config: BambaConfig) - added:
field BambaPreTrainedModel.base_model_prefix - added:
field BambaPreTrainedModel.config: BambaConfig - added:
field BambaPreTrainedModel.supports_gradient_checkpointing - added:
method BambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optio… - added:
method BambaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method BambaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.…
- added:
class BambaForCausalLM(LlamaForCausalLM) - added:
class BambaModel(BambaPreTrainedModel) - added:
class BambaPreTrainedModel(PreTrainedModel) - added:
constructor BambaForCausalLM(config) - added:
constructor BambaModel(config: BambaConfig) - added:
field BambaPreTrainedModel.base_model_prefix - added:
field BambaPreTrainedModel.config: BambaConfig - added:
field BambaPreTrainedModel.supports_gradient_checkpointing - added:
method BambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optio… - added:
method BambaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method BambaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.…
- added:
export define_import_structure
- added:
class BarkCoarseConfig(BarkSubModelConfig) - added:
class BarkConfig(PretrainedConfig) - added:
class BarkFineConfig(BarkSubModelConfig) - added:
class BarkSemanticConfig(BarkSubModelConfig) - added:
constructor BarkConfig(semantic_config: Optional[dict]=…, coarse_acoustics_config: Optional[dict]=…, fine_acoustics_config: Optional[dict]=…, codec_config: Optional[dict]=…, initializer_range=…, **kwargs) - added:
constructor BarkFineConfig(tie_word_embeddings=…, n_codes_total=…, n_codes_given=…, **kwargs) - added:
field BarkCoarseConfig.base_config_key - added:
field BarkCoarseConfig.model_type - added:
field BarkConfig.model_type - added:
field BarkConfig.sub_configs - added:
field BarkFineConfig.base_config_key - added:
field BarkFineConfig.model_type - added:
field BarkSemanticConfig.base_config_key - added:
field BarkSemanticConfig.model_type - added:
static method BarkConfig.from_sub_model_configs(semantic_config: BarkSemanticConfig, coarse_acoustics_config: BarkCoarseConfig, fine_acoustics_config: BarkFineConfig, codec_config: PretrainedConfig, **kwargs)
- added:
class BarkCoarseGenerationConfig(GenerationConfig) - added:
class BarkFineGenerationConfig(GenerationConfig) - added:
class BarkGenerationConfig(GenerationConfig) - added:
class BarkSemanticGenerationConfig(GenerationConfig) - added:
constructor BarkCoarseGenerationConfig(renormalize_logits=…, output_scores=…, return_dict_in_generate=…, output_hidden_states=…, output_attentions=…, temperature=…, do_sample=…, coarse_semantic_pad_token=…, coarse_rate_hz=…, n_coarse_codebooks=…, coarse_infer_token=…, max_coarse_input_length=…, max_coarse_history: int=…, sliding_window_len: int=…, **kwargs) - added:
constructor BarkFineGenerationConfig(temperature=…, max_fine_history_length=…, max_fine_input_length=…, n_fine_codebooks=…, **kwargs) - added:
constructor BarkGenerationConfig(semantic_config: Optional[dict]=…, coarse_acoustics_config: Optional[dict]=…, fine_acoustics_config: Optional[dict]=…, sample_rate=…, codebook_size=…, **kwargs) - added:
constructor BarkSemanticGenerationConfig(eos_token_id=…, renormalize_logits=…, max_new_tokens=…, output_scores=…, return_dict_in_generate=…, output_hidden_states=…, output_attentions=…, temperature=…, do_sample=…, text_encoding_offset=…, text_pad_token=…, semantic_infer_token=…, semantic_vocab_size=…, max_input_semantic_length=…, semantic_rate_hz=…, min_eos_p=…, **kwargs) - added:
field BarkCoarseGenerationConfig.model_type - added:
field BarkFineGenerationConfig.model_type - added:
field BarkGenerationConfig.model_type - added:
field BarkSemanticGenerationConfig.model_type - added:
method BarkFineGenerationConfig.validate(**kwargs) - added:
method BarkGenerationConfig.to_dict() - added:
static method BarkGenerationConfig.from_sub_model_configs(semantic_config: BarkSemanticGenerationConfig, coarse_acoustics_config: BarkCoarseGenerationConfig, fine_acoustics_config: BarkFineGenerationConfig, **kwargs) - added:
variable logger
- added:
class BarkCausalModel(BarkPreTrainedModel, GenerationMixin) - added:
class BarkCoarseModel(BarkCausalModel) - added:
class BarkFineModel(BarkPreTrainedModel) - added:
class BarkModel(BarkPreTrainedModel) - added:
class BarkPreTrainedModel(PreTrainedModel) - added:
class BarkSemanticModel(BarkCausalModel) - added:
constructor BarkCausalModel(config) - added:
constructor BarkFineModel(config) - added:
constructor BarkModel(config) - added:
constructor BarkPreTrainedModel(*inputs, **kwargs) - added:
field BarkCausalModel.config: BarkSubModelConfig - added:
field BarkCoarseModel.base_model_prefix - added:
field BarkCoarseModel.config: BarkCoarseConfig - added:
field BarkFineModel.base_model_prefix - added:
field BarkFineModel.config: BarkFineConfig - added:
field BarkFineModel.main_input_name - added:
field BarkModel.config: BarkConfig - added:
field BarkPreTrainedModel.config: BarkConfig - added:
field BarkPreTrainedModel.supports_gradient_checkpointing - added:
field BarkSemanticModel.base_model_prefix - added:
field BarkSemanticModel.config: BarkSemanticConfig - added:
method BarkCausalModel.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, input_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[b… - added:
method BarkCausalModel.get_input_embeddings() - added:
method BarkCausalModel.get_output_embeddings() - added:
method BarkCausalModel.prepare_inputs_for_generation(input_ids, attention_mask=…, input_embeds=…, past_key_values=…, position_ids=…, use_cache=…, cache_position=…, **kwargs) - added:
method BarkCausalModel.set_input_embeddings(new_embeddings) - added:
method BarkCoarseModel.generate(semantic_output: torch.Tensor, semantic_generation_config: Optional[BarkSemanticGenerationConfig]=…, coarse_generation_config: Optional[BarkCoarseGenerationConfig]=…, codebook_size: int=…, history_prompt: Optional[dict[str, torch.Tensor]]=…, return_output_lengths: Optional[bool]=…, **kwargs) -> Union[torch.LongTensor, tuple[torch.LongTensor, torch.LongTensor]] - added:
method BarkCoarseModel.preprocess_histories(max_coarse_history: int, semantic_to_coarse_ratio: int, batch_size: int, semantic_generation_config: int, codebook_size: int, history_prompt: Optional[dict[str, torch.Tensor]]=…) - added:
method BarkFineModel.forward(codebook_idx: int, input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, input_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[… - added:
method BarkFineModel.generate(coarse_output: torch.Tensor, semantic_generation_config: Optional[BarkSemanticGenerationConfig]=…, coarse_generation_config: Optional[BarkCoarseGenerationConfig]=…, fine_generation_config: BarkFineGenerationConfig=…, codebook_size: int=…, history_prompt: Optional[dict[str, torch.Tensor]]=…, **kwargs) -> torch.LongTensor - added:
method BarkFineModel.get_input_embeddings() - added:
method BarkFineModel.get_output_embeddings() - added:
method BarkFineModel.resize_token_embeddings(new_num_tokens: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method BarkFineModel.set_input_embeddings(new_embeddings) - added:
method BarkFineModel.set_output_embeddings(new_output_embeddings) - added:
method BarkFineModel.tie_weights() - added:
method BarkModel.codec_decode(fine_output, output_lengths=…) - added:
method BarkModel.enable_cpu_offload(accelerator_id: Optional[int]=…, **kwargs) - added:
method BarkModel.generate(input_ids: Optional[torch.Tensor]=…, history_prompt: Optional[dict[str, torch.Tensor]]=…, return_output_lengths: Optional[bool]=…, **kwargs) -> torch.LongTensor - added:
method BarkModel.tie_weights() - added:
method BarkSemanticModel.generate(input_ids: torch.Tensor, semantic_generation_config: Optional[BarkSemanticGenerationConfig]=…, history_prompt: Optional[dict[str, torch.Tensor]]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) -> torch.LongTensor - added:
property BarkModel.device -> torch.device - added:
property BarkPreTrainedModel.device -> torch.device - added:
static method BarkModel.can_generate() -> bool
- added:
class BarkProcessor(ProcessorMixin) - added:
constructor BarkProcessor(tokenizer, speaker_embeddings=…) - added:
field BarkProcessor.attributes - added:
field BarkProcessor.preset_shape - added:
field BarkProcessor.tokenizer_class - added:
method BarkProcessor.save_pretrained(save_directory, speaker_embeddings_dict_path=…, speaker_embeddings_directory=…, push_to_hub: bool=…, **kwargs) - added:
property BarkProcessor.available_voice_presets -> list - added:
static method BarkProcessor.from_pretrained(pretrained_processor_name_or_path, speaker_embeddings_dict_path=…, **kwargs)
- added:
export define_import_structure
- added:
class BartConfig(PretrainedConfig) - added:
class BartOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor BartConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, classifier_dropout=…, scale_embedding=…, use_cache=…, num_labels=… - added:
field BartConfig.attribute_map - added:
field BartConfig.keys_to_ignore_at_inference - added:
field BartConfig.model_type - added:
method BartOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property BartOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property BartOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class BartForCausalLM(BartPreTrainedModel, GenerationMixin) - added:
class BartForConditionalGeneration(BartPreTrainedModel, GenerationMixin) - added:
class BartForQuestionAnswering(BartPreTrainedModel) - added:
class BartForSequenceClassification(BartPreTrainedModel) - added:
class BartModel(BartPreTrainedModel) - added:
class BartPreTrainedModel(PreTrainedModel) - added:
class BartPretrainedModel(BartPreTrainedModel) - added:
class PretrainedBartModel(BartPreTrainedModel) - added:
constructor BartForCausalLM(config) - added:
constructor BartForConditionalGeneration(config: BartConfig) - added:
constructor BartForQuestionAnswering(config) - added:
constructor BartForSequenceClassification(config: BartConfig, **kwargs) - added:
constructor BartModel(config: BartConfig) - added:
field BartForConditionalGeneration.base_model_prefix - added:
field BartPreTrainedModel.base_model_prefix - added:
field BartPreTrainedModel.config: BartConfig - added:
field BartPreTrainedModel.supports_gradient_checkpointing - added:
method BartForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: O… - added:
method BartForCausalLM.get_decoder() - added:
method BartForCausalLM.get_input_embeddings() - added:
method BartForCausalLM.set_decoder(decoder) - added:
method BartForCausalLM.set_input_embeddings(value) - added:
method BartForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.Fl… - added:
method BartForConditionalGeneration.get_decoder() - added:
method BartForConditionalGeneration.get_encoder() - added:
method BartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method BartForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method BartForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTenso… - added:
method BartForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.F… - added:
method BartModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor]]=…, past… - added:
method BartModel.get_encoder() - added:
method BartModel.get_input_embeddings() - added:
method BartModel.set_input_embeddings(value) - added:
property BartPreTrainedModel.dummy_inputs
- added:
class FlaxBartDecoderPreTrainedModel(FlaxPreTrainedModel) - added:
class FlaxBartForCausalLM(FlaxBartDecoderPreTrainedModel) - added:
class FlaxBartForConditionalGeneration(FlaxBartPreTrainedModel) - added:
class FlaxBartForQuestionAnswering(FlaxBartPreTrainedModel) - added:
class FlaxBartForSequenceClassification(FlaxBartPreTrainedModel) - added:
class FlaxBartModel(FlaxBartPreTrainedModel) - added:
class FlaxBartPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBartDecoderPreTrainedModel(config: BartConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
constructor FlaxBartPreTrainedModel(config: BartConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxBartDecoderPreTrainedModel.base_model_prefix: str - added:
field FlaxBartDecoderPreTrainedModel.config_class - added:
field FlaxBartDecoderPreTrainedModel.module_class: nn.Module - added:
field FlaxBartForCausalLM.module_class - added:
field FlaxBartForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxBartForConditionalGeneration.module_class - added:
field FlaxBartForQuestionAnswering.dtype - added:
field FlaxBartForQuestionAnswering.module_class - added:
field FlaxBartForSequenceClassification.dtype - added:
field FlaxBartForSequenceClassification.module_class - added:
field FlaxBartModel.config: BartConfig - added:
field FlaxBartModel.dtype: jnp.dtype - added:
field FlaxBartModel.module_class - added:
field FlaxBartPreTrainedModel.base_model_prefix: str - added:
field FlaxBartPreTrainedModel.config_class - added:
field FlaxBartPreTrainedModel.module_class: nn.Module - added:
method FlaxBartDecoderPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxBartDecoderPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxBartForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxBartForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBartForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optio… - added:
method FlaxBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxBartForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBartPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]… - added:
method FlaxBartPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxBartPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxBartPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFBartForConditionalGeneration(TFBartPretrainedModel, TFCausalLanguageModelingLoss) - added:
class TFBartForSequenceClassification(TFBartPretrainedModel, TFSequenceClassificationLoss) - added:
class TFBartModel(TFBartPretrainedModel) - added:
class TFBartPretrainedModel(TFPreTrainedModel) - added:
constructor TFBartForConditionalGeneration(config, load_weight_prefix=…, *inputs, **kwargs) - added:
constructor TFBartForSequenceClassification(config: BartConfig, load_weight_prefix=…, *inputs, **kwargs) - added:
constructor TFBartModel(config: BartConfig, load_weight_prefix=…, *inputs, **kwargs) - added:
field TFBartPretrainedModel.base_model_prefix - added:
field TFBartPretrainedModel.config_class - added:
method TFBartForConditionalGeneration.build(input_shape=…) - added:
method TFBartForConditionalGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_att… - added:
method TFBartForConditionalGeneration.get_bias() - added:
method TFBartForConditionalGeneration.get_decoder() - added:
method TFBartForConditionalGeneration.get_encoder() - added:
method TFBartForConditionalGeneration.get_output_embeddings() - added:
method TFBartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFBartForConditionalGeneration.serving_output(output) - added:
method TFBartForConditionalGeneration.set_bias(value) - added:
method TFBartForConditionalGeneration.set_output_embeddings(value) - added:
method TFBartForSequenceClassification.build(input_shape=…) - added:
method TFBartForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_at… - added:
method TFBartForSequenceClassification.serving_output(output) - added:
method TFBartModel.build(input_shape=…) - added:
method TFBartModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.nda… - added:
method TFBartModel.get_decoder() - added:
method TFBartModel.get_encoder() - added:
method TFBartModel.serving_output(output) - added:
method TFBartPretrainedModel.tf_to_pt_weight_rename(tf_weight) - added:
property TFBartPretrainedModel.dummy_inputs
- added:
class BartTokenizer(PreTrainedTokenizer) - added:
constructor BartTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field BartTokenizer.model_input_names - added:
field BartTokenizer.vocab_files_names - added:
method BartTokenizer.bpe(token) - added:
method BartTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BartTokenizer.convert_tokens_to_string(tokens) - added:
method BartTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BartTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BartTokenizer.get_vocab() - added:
method BartTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method BartTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BartTokenizer.vocab_size
- added:
class BartTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BartTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field BartTokenizerFast.model_input_names - added:
field BartTokenizerFast.slow_tokenizer_class - added:
field BartTokenizerFast.vocab_files_names - added:
method BartTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method BartTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BartTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BartTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class BarthezTokenizer(PreTrainedTokenizer) - added:
constructor BarthezTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field BarthezTokenizer.model_input_names - added:
field BarthezTokenizer.vocab_files_names - added:
method BarthezTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BarthezTokenizer.convert_tokens_to_string(tokens) - added:
method BarthezTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BarthezTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BarthezTokenizer.get_vocab() - added:
method BarthezTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BarthezTokenizer.vocab_size
- added:
class BarthezTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BarthezTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
field BarthezTokenizerFast.model_input_names - added:
field BarthezTokenizerFast.slow_tokenizer_class - added:
field BarthezTokenizerFast.vocab_files_names - added:
method BarthezTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BarthezTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BarthezTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class BartphoTokenizer(PreTrainedTokenizer) - added:
constructor BartphoTokenizer(vocab_file, monolingual_vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field BartphoTokenizer.model_input_names - added:
field BartphoTokenizer.vocab_files_names - added:
method BartphoTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BartphoTokenizer.convert_tokens_to_string(tokens) - added:
method BartphoTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BartphoTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BartphoTokenizer.get_vocab() - added:
method BartphoTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BartphoTokenizer.vocab_size
- added:
export define_import_structure
- added:
class BeitConfig(BackboneConfigMixin, PretrainedConfig) - added:
class BeitOnnxConfig(OnnxConfig) - added:
constructor BeitConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, use_mask_token=…, use_absolute_position_embeddings=…, use_relative_position_bias=…, use_shared_relative_position_bias=…, layer_sca… - added:
field BeitConfig.model_type - added:
field BeitOnnxConfig.torch_onnx_minimum_version - added:
property BeitOnnxConfig.atol_for_validation -> float - added:
property BeitOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class BeitFeatureExtractor(BeitImageProcessor) - added:
constructor BeitFeatureExtractor(*args, **kwargs) -> None
- added:
class BeitImageProcessor(BaseImageProcessor) - added:
constructor BeitImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, rescale_factor: Union[int, float]=…, do_rescale: bool=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_reduce_labels: bool=…, **kwargs) -> N… - added:
field BeitImageProcessor.model_input_names - added:
method BeitImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method BeitImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Uni… - added:
method BeitImageProcessor.reduce_label(label: ImageInput) -> np.ndarray - added:
method BeitImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class BeitImageProcessorFast(BaseImageProcessorFast) - added:
constructor BeitImageProcessorFast(**kwargs: Unpack[BeitFastImageProcessorKwargs]) - added:
field BeitImageProcessorFast.crop_size - added:
field BeitImageProcessorFast.default_to_square - added:
field BeitImageProcessorFast.do_center_crop - added:
field BeitImageProcessorFast.do_normalize - added:
field BeitImageProcessorFast.do_reduce_labels - added:
field BeitImageProcessorFast.do_rescale - added:
field BeitImageProcessorFast.do_resize - added:
field BeitImageProcessorFast.image_mean - added:
field BeitImageProcessorFast.image_std - added:
field BeitImageProcessorFast.resample - added:
field BeitImageProcessorFast.size - added:
field BeitImageProcessorFast.valid_kwargs - added:
method BeitImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method BeitImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[BeitFastImageProcessorKwargs]) -> BatchFeature - added:
method BeitImageProcessorFast.reduce_label(labels: list[torch.Tensor])
- added:
class BeitBackbone(BeitPreTrainedModel, BackboneMixin) - added:
class BeitForImageClassification(BeitPreTrainedModel) - added:
class BeitForMaskedImageModeling(BeitPreTrainedModel) - added:
class BeitForSemanticSegmentation(BeitPreTrainedModel) - added:
class BeitModel(BeitPreTrainedModel) - added:
class BeitPreTrainedModel(PreTrainedModel) - added:
constructor BeitBackbone(config) - added:
constructor BeitForImageClassification(config: BeitConfig) -> None - added:
constructor BeitForMaskedImageModeling(config: BeitConfig) -> None - added:
constructor BeitForSemanticSegmentation(config: BeitConfig) -> None - added:
constructor BeitModel(config: BeitConfig, add_pooling_layer: bool=…) -> None - added:
field BeitPreTrainedModel.base_model_prefix - added:
field BeitPreTrainedModel.config: BeitConfig - added:
field BeitPreTrainedModel.main_input_name - added:
field BeitPreTrainedModel.supports_gradient_checkpointing - added:
method BeitBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method BeitBackbone.get_input_embeddings() - added:
method BeitForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method BeitForMaskedImageModeling.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, MaskedLMOutput] - added:
method BeitForMaskedImageModeling.get_output_embeddings() - added:
method BeitForSemanticSegmentation.compute_loss(logits, auxiliary_logits, labels) - added:
method BeitForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method BeitModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BeitModelOutputWithPooling] - added:
method BeitModel.get_input_embeddings()
- added:
class FlaxBeitForImageClassification(FlaxBeitPreTrainedModel) - added:
class FlaxBeitForMaskedImageModeling(FlaxBeitPreTrainedModel) - added:
class FlaxBeitModel(FlaxBeitPreTrainedModel) - added:
class FlaxBeitPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBeitPreTrainedModel(config: BeitConfig, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxBeitForImageClassification.module_class - added:
field FlaxBeitForMaskedImageModeling.module_class - added:
field FlaxBeitModel.module_class - added:
field FlaxBeitPreTrainedModel.base_model_prefix - added:
field FlaxBeitPreTrainedModel.config_class - added:
field FlaxBeitPreTrainedModel.main_input_name - added:
field FlaxBeitPreTrainedModel.module_class: nn.Module - added:
method FlaxBeitPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
export define_import_structure
- added:
class BertConfig(PretrainedConfig) - added:
class BertOnnxConfig(OnnxConfig) - added:
constructor BertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field BertConfig.model_type - added:
property BertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class BertForMaskedLM(BertPreTrainedModel) - added:
class BertForMultipleChoice(BertPreTrainedModel) - added:
class BertForNextSentencePrediction(BertPreTrainedModel) - added:
class BertForPreTraining(BertPreTrainedModel) - added:
class BertForQuestionAnswering(BertPreTrainedModel) - added:
class BertForSequenceClassification(BertPreTrainedModel) - added:
class BertForTokenClassification(BertPreTrainedModel) - added:
class BertLMHeadModel(BertPreTrainedModel, GenerationMixin) - added:
class BertLayer(GradientCheckpointingLayer) - added:
class BertModel(BertPreTrainedModel) - added:
class BertPreTrainedModel(PreTrainedModel) - added:
constructor BertForMaskedLM(config) - added:
constructor BertForMultipleChoice(config) - added:
constructor BertForNextSentencePrediction(config) - added:
constructor BertForPreTraining(config) - added:
constructor BertForQuestionAnswering(config) - added:
constructor BertForSequenceClassification(config) - added:
constructor BertForTokenClassification(config) - added:
constructor BertLMHeadModel(config) - added:
constructor BertLayer(config, layer_idx=…) - added:
constructor BertModel(config, add_pooling_layer=…) - added:
field BertPreTrainedModel.base_model_prefix - added:
field BertPreTrainedModel.config: BertConfig - added:
field BertPreTrainedModel.load_tf_weights - added:
field BertPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_bert(model, config, tf_checkpoint_path) - added:
method BertForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]… - added:
method BertForMaskedLM.get_output_embeddings() - added:
method BertForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method BertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method BertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
method BertForNextSentencePrediction.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retu… - added:
method BertForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, next_sentence_label: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method BertForPreTraining.get_output_embeddings() - added:
method BertForPreTraining.set_output_embeddings(new_embeddings) - added:
method BertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…,… - added:
method BertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retu… - added:
method BertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_… - added:
method BertLMHeadModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]… - added:
method BertLMHeadModel.get_output_embeddings() - added:
method BertLMHeadModel.set_output_embeddings(new_embeddings) - added:
method BertLayer.feed_forward_chunk(attention_output) - added:
method BertLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…) -> tuple[torch.Tensor] - added:
method BertModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…,… - added:
method BertModel.get_input_embeddings() - added:
method BertModel.set_input_embeddings(value) - added:
static method BertForMaskedLM.can_generate() -> bool
- added:
class FlaxBertForCausalLM(FlaxBertPreTrainedModel) - added:
class FlaxBertForMaskedLM(FlaxBertPreTrainedModel) - added:
class FlaxBertForMultipleChoice(FlaxBertPreTrainedModel) - added:
class FlaxBertForNextSentencePrediction(FlaxBertPreTrainedModel) - added:
class FlaxBertForPreTraining(FlaxBertPreTrainedModel) - added:
class FlaxBertForQuestionAnswering(FlaxBertPreTrainedModel) - added:
class FlaxBertForSequenceClassification(FlaxBertPreTrainedModel) - added:
class FlaxBertForTokenClassification(FlaxBertPreTrainedModel) - added:
class FlaxBertModel(FlaxBertPreTrainedModel) - added:
class FlaxBertPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBertPreTrainedModel(config: BertConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxBertForCausalLM.module_class - added:
field FlaxBertForMaskedLM.module_class - added:
field FlaxBertForMultipleChoice.module_class - added:
field FlaxBertForNextSentencePrediction.module_class - added:
field FlaxBertForPreTraining.module_class - added:
field FlaxBertForQuestionAnswering.module_class - added:
field FlaxBertForSequenceClassification.module_class - added:
field FlaxBertForTokenClassification.module_class - added:
field FlaxBertModel.module_class - added:
field FlaxBertPreTrainedModel.base_model_prefix - added:
field FlaxBertPreTrainedModel.config_class - added:
field FlaxBertPreTrainedModel.module_class: nn.Module - added:
method FlaxBertForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxBertForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBertPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxBertPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxBertPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFBertEmbeddings(keras.layers.Layer) - added:
class TFBertForMaskedLM(TFBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFBertForNextSentencePrediction(TFBertPreTrainedModel, TFNextSentencePredictionLoss) - added:
class TFBertForPreTraining(TFBertPreTrainedModel, TFBertPreTrainingLoss) - added:
class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFBertLMHeadModel(TFBertPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFBertMainLayer(keras.layers.Layer) - added:
class TFBertModel(TFBertPreTrainedModel) - added:
class TFBertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFBertEmbeddings(config: BertConfig, **kwargs) - added:
constructor TFBertForMaskedLM(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForMultipleChoice(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForNextSentencePrediction(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForPreTraining(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForQuestionAnswering(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForSequenceClassification(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertForTokenClassification(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertLMHeadModel(config: BertConfig, *inputs, **kwargs) - added:
constructor TFBertMainLayer(config: BertConfig, add_pooling_layer: bool=…, **kwargs) - added:
constructor TFBertModel(config: BertConfig, add_pooling_layer: bool=…, *inputs, **kwargs) - added:
field TFBertMainLayer.config_class - added:
field TFBertPreTrainedModel.base_model_prefix - added:
field TFBertPreTrainedModel.config_class - added:
method TFBertEmbeddings.build(input_shape=…) - added:
method TFBertEmbeddings.call(input_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, past_key_values_length=…, training: bool=…) -> tf.Tensor - added:
method TFBertForMaskedLM.build(input_shape=…) - added:
method TFBertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | Non… - added:
method TFBertForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFBertForMaskedLM.get_prefix_bias_name() -> str - added:
method TFBertForMultipleChoice.build(input_shape=…) - added:
method TFBertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFBertForNextSentencePrediction.build(input_shape=…) - added:
method TFBertForNextSentencePrediction.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFBertForPreTraining.build(input_shape=…) - added:
method TFBertForPreTraining.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool |… - added:
method TFBertForPreTraining.get_lm_head() -> keras.layers.Layer - added:
method TFBertForPreTraining.get_prefix_bias_name() -> str - added:
method TFBertForQuestionAnswering.build(input_shape=…) - added:
method TFBertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: b… - added:
method TFBertForSequenceClassification.build(input_shape=…) - added:
method TFBertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFBertForTokenClassification.build(input_shape=…) - added:
method TFBertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict:… - added:
method TFBertLMHeadModel.build(input_shape=…) - added:
method TFBertLMHeadModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | t… - added:
method TFBertLMHeadModel.get_lm_head() -> keras.layers.Layer - added:
method TFBertLMHeadModel.get_prefix_bias_name() -> str - added:
method TFBertLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFBertMainLayer.build(input_shape=…) - added:
method TFBertMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.… - added:
method TFBertMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFBertMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFBertModel.build(input_shape=…) - added:
method TFBertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.Tens…
- added:
class BasicTokenizer - added:
class BertTokenizer(PreTrainedTokenizer) - added:
class WordpieceTokenizer - added:
constructor BasicTokenizer(do_lower_case=…, never_split=…, tokenize_chinese_chars=…, strip_accents=…, do_split_on_punc=…) - added:
constructor BertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
constructor WordpieceTokenizer(vocab, unk_token, max_input_chars_per_word=…) - added:
field BertTokenizer.vocab_files_names - added:
method BasicTokenizer.tokenize(text, never_split=…) - added:
method BertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BertTokenizer.convert_tokens_to_string(tokens) - added:
method BertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BertTokenizer.get_vocab() - added:
method BertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method WordpieceTokenizer.tokenize(text) - added:
property BertTokenizer.do_lower_case - added:
property BertTokenizer.vocab_size
- added:
class BertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field BertTokenizerFast.slow_tokenizer_class - added:
field BertTokenizerFast.vocab_files_names - added:
method BertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method BertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
class TFBertTokenizer(keras.layers.Layer) - added:
constructor TFBertTokenizer(vocab_list: list, do_lower_case: bool, cls_token_id: Optional[int]=…, sep_token_id: Optional[int]=…, pad_token_id: Optional[int]=…, padding: str=…, truncation: bool=…, max_length: int=…, pad_to_multiple_of: Optional[int]=…, return_token_type_ids: bool=…, return_attention_mask: bool=…, use_fast_bert_tokenizer: bool=…, **tokenizer_kwargs) - added:
method TFBertTokenizer.call(text, text_pair=…, padding=…, truncation=…, max_length=…, pad_to_multiple_of=…, return_token_type_ids=…, return_attention_mask=…) - added:
method TFBertTokenizer.get_config() - added:
method TFBertTokenizer.unpaired_tokenize(texts) - added:
static method TFBertTokenizer.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], *init_inputs, **kwargs) - added:
static method TFBertTokenizer.from_tokenizer(tokenizer: PreTrainedTokenizerBase, **kwargs)
- added:
export define_import_structure
- added:
class BertGenerationConfig(PretrainedConfig) - added:
constructor BertGenerationConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, **kwargs) - added:
field BertGenerationConfig.model_type
- added:
class BertGenerationDecoder(BertGenerationPreTrainedModel, GenerationMixin) - added:
class BertGenerationEncoder(BertGenerationPreTrainedModel) - added:
class BertGenerationPreTrainedModel(PreTrainedModel) - added:
constructor BertGenerationDecoder(config) - added:
constructor BertGenerationEncoder(config) - added:
field BertGenerationPreTrainedModel.base_model_prefix - added:
field BertGenerationPreTrainedModel.config: BertGenerationConfig - added:
field BertGenerationPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_bert_generation(model, tf_hub_path, model_class, is_encoder_named_decoder=…, is_encoder=…) - added:
method BertGenerationDecoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]… - added:
method BertGenerationDecoder.get_output_embeddings() - added:
method BertGenerationDecoder.set_output_embeddings(new_embeddings) - added:
method BertGenerationEncoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, o… - added:
method BertGenerationEncoder.get_input_embeddings() - added:
method BertGenerationEncoder.set_input_embeddings(value)
- added:
class BertGenerationTokenizer(PreTrainedTokenizer) - added:
constructor BertGenerationTokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, sep_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field BertGenerationTokenizer.model_input_names - added:
field BertGenerationTokenizer.prefix_tokens: list[int] - added:
field BertGenerationTokenizer.vocab_files_names - added:
method BertGenerationTokenizer.convert_tokens_to_string(tokens) - added:
method BertGenerationTokenizer.get_vocab() - added:
method BertGenerationTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BertGenerationTokenizer.vocab_size
- added:
export define_import_structure
- added:
class BertJapaneseTokenizer(PreTrainedTokenizer) - added:
class CharacterTokenizer - added:
class MecabTokenizer - added:
constructor BertJapaneseTokenizer(vocab_file, spm_file=…, do_lower_case=…, do_word_tokenize=…, do_subword_tokenize=…, word_tokenizer_type=…, subword_tokenizer_type=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, mecab_kwargs=…, sudachi_kwargs=…, jumanpp_kwargs=…, **kwargs) - added:
constructor CharacterTokenizer(vocab, unk_token, normalize_text=…) - added:
constructor MecabTokenizer(do_lower_case=…, never_split=…, normalize_text=…, mecab_dic: Optional[str]=…, mecab_option: Optional[str]=…) - added:
field BertJapaneseTokenizer.vocab_files_names - added:
method BertJapaneseTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BertJapaneseTokenizer.convert_tokens_to_string(tokens) - added:
method BertJapaneseTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BertJapaneseTokenizer.get_vocab() - added:
method BertJapaneseTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method CharacterTokenizer.tokenize(text) - added:
method MecabTokenizer.tokenize(text, never_split=…, **kwargs) - added:
property BertJapaneseTokenizer.do_lower_case - added:
property BertJapaneseTokenizer.vocab_size
- added:
export define_import_structure
- added:
class BertweetTokenizer(PreTrainedTokenizer) - added:
constructor BertweetTokenizer(vocab_file, merges_file, normalization=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
field BertweetTokenizer.vocab_files_names - added:
method BertweetTokenizer.add_from_file(f) - added:
method BertweetTokenizer.bpe(token) - added:
method BertweetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BertweetTokenizer.convert_tokens_to_string(tokens) - added:
method BertweetTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BertweetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BertweetTokenizer.get_vocab() - added:
method BertweetTokenizer.normalizeToken(token) - added:
method BertweetTokenizer.normalizeTweet(tweet) - added:
method BertweetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BertweetTokenizer.vocab_size
- added:
export define_import_structure
- added:
class BigBirdConfig(PretrainedConfig) - added:
class BigBirdOnnxConfig(OnnxConfig) - added:
constructor BigBirdConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, sep_token_id=…, attention_type=…, use_bias=…, rescale_embeddings… - added:
field BigBirdConfig.model_type - added:
property BigBirdOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class BigBirdForCausalLM(BigBirdPreTrainedModel, GenerationMixin) - added:
class BigBirdForMaskedLM(BigBirdPreTrainedModel) - added:
class BigBirdForMultipleChoice(BigBirdPreTrainedModel) - added:
class BigBirdForPreTraining(BigBirdPreTrainedModel) - added:
class BigBirdForQuestionAnswering(BigBirdPreTrainedModel) - added:
class BigBirdForSequenceClassification(BigBirdPreTrainedModel) - added:
class BigBirdForTokenClassification(BigBirdPreTrainedModel) - added:
class BigBirdLayer(GradientCheckpointingLayer) - added:
class BigBirdModel(BigBirdPreTrainedModel) - added:
class BigBirdPreTrainedModel(PreTrainedModel) - added:
constructor BigBirdForCausalLM(config) - added:
constructor BigBirdForMaskedLM(config) - added:
constructor BigBirdForMultipleChoice(config) - added:
constructor BigBirdForPreTraining(config) - added:
constructor BigBirdForQuestionAnswering(config, add_pooling_layer=…) - added:
constructor BigBirdForSequenceClassification(config) - added:
constructor BigBirdForTokenClassification(config) - added:
constructor BigBirdLayer(config, seed=…) - added:
constructor BigBirdModel(config, add_pooling_layer=…) - added:
field BigBirdPreTrainedModel.base_model_prefix - added:
field BigBirdPreTrainedModel.config: BigBirdConfig - added:
field BigBirdPreTrainedModel.load_tf_weights - added:
field BigBirdPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_big_bird(model, tf_checkpoint_path, is_trivia_qa=…) - added:
method BigBirdForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method BigBirdForCausalLM.get_output_embeddings() - added:
method BigBirdForCausalLM.set_output_embeddings(new_embeddings) - added:
method BigBirdForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method BigBirdForMaskedLM.get_output_embeddings() - added:
method BigBirdForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method BigBirdForMaskedLM.set_output_embeddings(new_embeddings) - added:
method BigBirdForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method BigBirdForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.FloatTensor]=…, next_sentence_label: Optional[torch.LongTensor]=…, output… - added:
method BigBirdForPreTraining.get_output_embeddings() - added:
method BigBirdForPreTraining.set_output_embeddings(new_embeddings) - added:
method BigBirdForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, question_lengths: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor… - added:
method BigBirdForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method BigBirdForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method BigBirdLayer.feed_forward_chunk(attention_output) - added:
method BigBirdLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, band_mask=…, from_mask=…, to_mask=…, blocked_encoder_mask=…, past_key_values=…, output_attentions=…, cache_position=…) - added:
method BigBirdLayer.set_attention_type(value: str, layer_idx=…) - added:
method BigBirdModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]… - added:
method BigBirdModel.get_input_embeddings() - added:
method BigBirdModel.set_attention_type(value: str) - added:
method BigBirdModel.set_input_embeddings(value) - added:
static method BigBirdForQuestionAnswering.prepare_question_mask(q_lengths: torch.Tensor, maxlen: int) - added:
static method BigBirdModel.create_masks_for_block_sparse_attn(attention_mask: torch.Tensor, block_size: int)
- added:
class FlaxBigBirdForCausalLM(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForMaskedLM(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForMultipleChoice(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForPreTraining(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForQuestionAnswering(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForSequenceClassification(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdForTokenClassification(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdModel(FlaxBigBirdPreTrainedModel) - added:
class FlaxBigBirdPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBigBirdForMultipleChoice(config: BigBirdConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
constructor FlaxBigBirdPreTrainedModel(config: BigBirdConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxBigBirdForCausalLM.module_class - added:
field FlaxBigBirdForMaskedLM.module_class - added:
field FlaxBigBirdForMultipleChoice.module_class - added:
field FlaxBigBirdForPreTraining.module_class - added:
field FlaxBigBirdForQuestionAnswering.module_class - added:
field FlaxBigBirdForSequenceClassification.module_class - added:
field FlaxBigBirdForTokenClassification.module_class - added:
field FlaxBigBirdModel.module_class - added:
field FlaxBigBirdPreTrainedModel.base_model_prefix - added:
field FlaxBigBirdPreTrainedModel.config_class - added:
field FlaxBigBirdPreTrainedModel.module_class: nn.Module - added:
method FlaxBigBirdForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxBigBirdForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBigBirdPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxBigBirdPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxBigBirdPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
static method FlaxBigBirdForQuestionAnswering.prepare_question_mask(q_lengths, maxlen: int)
- added:
class BigBirdTokenizer(PreTrainedTokenizer) - added:
constructor BigBirdTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, pad_token=…, sep_token=…, mask_token=…, cls_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field BigBirdTokenizer.model_input_names - added:
field BigBirdTokenizer.prefix_tokens: list[int] - added:
field BigBirdTokenizer.vocab_files_names - added:
method BigBirdTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BigBirdTokenizer.convert_tokens_to_string(tokens) - added:
method BigBirdTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BigBirdTokenizer.get_vocab() - added:
method BigBirdTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BigBirdTokenizer.vocab_size
- added:
class BigBirdTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BigBirdTokenizerFast(vocab_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, sep_token=…, mask_token=…, cls_token=…, **kwargs) - added:
field BigBirdTokenizerFast.model_input_names - added:
field BigBirdTokenizerFast.prefix_tokens: list[int] - added:
field BigBirdTokenizerFast.slow_tokenizer_class - added:
field BigBirdTokenizerFast.vocab_files_names - added:
method BigBirdTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BigBirdTokenizerFast.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BigBirdTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class BigBirdPegasusConfig(PretrainedConfig) - added:
class BigBirdPegasusOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor BigBirdPegasusConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_… - added:
field BigBirdPegasusConfig.attribute_map - added:
field BigBirdPegasusConfig.keys_to_ignore_at_inference - added:
field BigBirdPegasusConfig.model_type - added:
method BigBirdPegasusOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property BigBirdPegasusOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property BigBirdPegasusOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class BigBirdPegasusForCausalLM(BigBirdPegasusPreTrainedModel, GenerationMixin) - added:
class BigBirdPegasusForConditionalGeneration(BigBirdPegasusPreTrainedModel, GenerationMixin) - added:
class BigBirdPegasusForQuestionAnswering(BigBirdPegasusPreTrainedModel) - added:
class BigBirdPegasusForSequenceClassification(BigBirdPegasusPreTrainedModel) - added:
class BigBirdPegasusModel(BigBirdPegasusPreTrainedModel) - added:
class BigBirdPegasusPreTrainedModel(PreTrainedModel) - added:
constructor BigBirdPegasusForCausalLM(config) - added:
constructor BigBirdPegasusForConditionalGeneration(config: BigBirdPegasusConfig) - added:
constructor BigBirdPegasusForQuestionAnswering(config) - added:
constructor BigBirdPegasusForSequenceClassification(config: BigBirdPegasusConfig, **kwargs) - added:
constructor BigBirdPegasusModel(config: BigBirdPegasusConfig) - added:
field BigBirdPegasusForConditionalGeneration.base_model_prefix - added:
field BigBirdPegasusPreTrainedModel.base_model_prefix - added:
field BigBirdPegasusPreTrainedModel.config: BigBirdPegasusConfig - added:
field BigBirdPegasusPreTrainedModel.supports_gradient_checkpointing - added:
method BigBirdPegasusForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…,… - added:
method BigBirdPegasusForCausalLM.get_decoder() - added:
method BigBirdPegasusForCausalLM.get_input_embeddings() - added:
method BigBirdPegasusForCausalLM.set_decoder(decoder) - added:
method BigBirdPegasusForCausalLM.set_input_embeddings(value) - added:
method BigBirdPegasusForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[lis… - added:
method BigBirdPegasusForConditionalGeneration.get_decoder() - added:
method BigBirdPegasusForConditionalGeneration.get_encoder() - added:
method BigBirdPegasusForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method BigBirdPegasusForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method BigBirdPegasusForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.… - added:
method BigBirdPegasusForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[li… - added:
method BigBirdPegasusModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor… - added:
method BigBirdPegasusModel.get_encoder() - added:
method BigBirdPegasusModel.get_input_embeddings() - added:
method BigBirdPegasusModel.set_input_embeddings(value) - added:
property BigBirdPegasusPreTrainedModel.dummy_inputs
- added:
export define_import_structure
- added:
class BioGptConfig(PretrainedConfig) - added:
constructor BioGptConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, scale_embedding=…, use_cache=…, layerdrop=…, activation_dropout=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field BioGptConfig.model_type
- added:
class BioGptForCausalLM(BioGptPreTrainedModel, GenerationMixin) - added:
class BioGptForSequenceClassification(BioGptPreTrainedModel) - added:
class BioGptForTokenClassification(BioGptPreTrainedModel) - added:
class BioGptModel(BioGptPreTrainedModel) - added:
class BioGptPreTrainedModel(PreTrainedModel) - added:
constructor BioGptForCausalLM(config) - added:
constructor BioGptForSequenceClassification(config: BioGptConfig) - added:
constructor BioGptForTokenClassification(config) - added:
constructor BioGptModel(config: BioGptConfig) - added:
field BioGptPreTrainedModel.base_model_prefix - added:
field BioGptPreTrainedModel.config: BioGptConfig - added:
field BioGptPreTrainedModel.supports_gradient_checkpointing - added:
method BioGptForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output… - added:
method BioGptForCausalLM.get_output_embeddings() - added:
method BioGptForCausalLM.set_output_embeddings(new_embeddings) - added:
method BioGptForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[b… - added:
method BioGptForSequenceClassification.get_input_embeddings() - added:
method BioGptForSequenceClassification.set_input_embeddings(value) - added:
method BioGptForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.L… - added:
method BioGptModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic…
- added:
class BioGptForCausalLM(BioGptPreTrainedModel, GenerationMixin) - added:
class BioGptForSequenceClassification(BioGptPreTrainedModel) - added:
class BioGptForTokenClassification(BioGptPreTrainedModel) - added:
class BioGptModel(BioGptPreTrainedModel) - added:
class BioGptPreTrainedModel(PreTrainedModel) - added:
constructor BioGptForCausalLM(config) - added:
constructor BioGptForSequenceClassification(config: BioGptConfig) - added:
constructor BioGptForTokenClassification(config) - added:
constructor BioGptModel(config: BioGptConfig) - added:
field BioGptPreTrainedModel.base_model_prefix - added:
field BioGptPreTrainedModel.config: BioGptConfig - added:
field BioGptPreTrainedModel.supports_gradient_checkpointing - added:
method BioGptForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output… - added:
method BioGptForCausalLM.get_output_embeddings() - added:
method BioGptForCausalLM.set_output_embeddings(new_embeddings) - added:
method BioGptForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[b… - added:
method BioGptForSequenceClassification.get_input_embeddings() - added:
method BioGptForSequenceClassification.set_input_embeddings(value) - added:
method BioGptForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.L… - added:
method BioGptModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic…
- added:
class BioGptTokenizer(PreTrainedTokenizer) - added:
constructor BioGptTokenizer(vocab_file, merges_file, unk_token=…, bos_token=…, eos_token=…, sep_token=…, pad_token=…, **kwargs) - added:
field BioGptTokenizer.model_input_names - added:
field BioGptTokenizer.vocab_files_names - added:
method BioGptTokenizer.bpe(token) - added:
method BioGptTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BioGptTokenizer.convert_tokens_to_string(tokens) - added:
method BioGptTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BioGptTokenizer.get_vocab() - added:
method BioGptTokenizer.moses_detokenize(tokens, lang) - added:
method BioGptTokenizer.moses_tokenize(text, lang) - added:
method BioGptTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BioGptTokenizer.vocab_size
- added:
export define_import_structure
- added:
class BitConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor BitConfig(num_channels=…, embedding_size=…, hidden_sizes=…, depths=…, layer_type=…, hidden_act=…, global_padding=…, num_groups=…, drop_path_rate=…, embedding_dynamic_padding=…, output_stride=…, width_factor=…, out_features=…, out_indices=…, **kwargs) - added:
field BitConfig.layer_types - added:
field BitConfig.model_type - added:
field BitConfig.supported_padding
- added:
class BitImageProcessor(BaseImageProcessor) - added:
constructor BitImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field BitImageProcessor.model_input_names - added:
method BitImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[fl… - added:
method BitImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class BitImageProcessorFast(BaseImageProcessorFast) - added:
field BitImageProcessorFast.crop_size - added:
field BitImageProcessorFast.default_to_square - added:
field BitImageProcessorFast.do_center_crop - added:
field BitImageProcessorFast.do_convert_rgb - added:
field BitImageProcessorFast.do_normalize - added:
field BitImageProcessorFast.do_rescale - added:
field BitImageProcessorFast.do_resize - added:
field BitImageProcessorFast.image_mean - added:
field BitImageProcessorFast.image_std - added:
field BitImageProcessorFast.resample - added:
field BitImageProcessorFast.rescale_factor - added:
field BitImageProcessorFast.size
- added:
class BitBackbone(BitPreTrainedModel, BackboneMixin) - added:
class BitForImageClassification(BitPreTrainedModel) - added:
class BitModel(BitPreTrainedModel) - added:
class BitPreTrainedModel(PreTrainedModel) - added:
constructor BitBackbone(config) - added:
constructor BitForImageClassification(config) - added:
constructor BitModel(config) - added:
field BitBackbone.has_attentions - added:
field BitPreTrainedModel.base_model_prefix - added:
field BitPreTrainedModel.config: BitConfig - added:
field BitPreTrainedModel.main_input_name - added:
method BitBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method BitForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention - added:
method BitModel.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
export define_import_structure
- added:
class BitNetConfig(PretrainedConfig) - added:
constructor BitNetConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
field BitNetConfig.keys_to_ignore_at_inference - added:
field BitNetConfig.model_type
- added:
class BitNetForCausalLM(BitNetPreTrainedModel, GenerationMixin) - added:
class BitNetModel(BitNetPreTrainedModel) - added:
class BitNetPreTrainedModel(PreTrainedModel) - added:
constructor BitNetForCausalLM(config) - added:
constructor BitNetModel(config: BitNetConfig) - added:
field BitNetPreTrainedModel.base_model_prefix - added:
field BitNetPreTrainedModel.config: BitNetConfig - added:
field BitNetPreTrainedModel.supports_gradient_checkpointing - added:
method BitNetForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method BitNetModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class BitNetForCausalLM(LlamaForCausalLM) - added:
class BitNetModel(LlamaModel) - added:
method BitNetForCausalLM.forward(**super_kwargs) -> CausalLMOutputWithPast
- added:
export define_import_structure
- added:
class BlenderbotConfig(PretrainedConfig) - added:
class BlenderbotOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor BlenderbotConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…… - added:
field BlenderbotConfig.attribute_map - added:
field BlenderbotConfig.keys_to_ignore_at_inference - added:
field BlenderbotConfig.model_type - added:
method BlenderbotOnnxConfig.fill_with_past_key_values_(inputs_or_outputs: Mapping[str, Mapping[int, str]], direction: str) - added:
method BlenderbotOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property BlenderbotOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property BlenderbotOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class BlenderbotForCausalLM(BlenderbotPreTrainedModel, GenerationMixin) - added:
class BlenderbotForConditionalGeneration(BlenderbotPreTrainedModel, GenerationMixin) - added:
class BlenderbotModel(BlenderbotPreTrainedModel) - added:
class BlenderbotPreTrainedModel(PreTrainedModel) - added:
constructor BlenderbotForCausalLM(config) - added:
constructor BlenderbotForConditionalGeneration(config: BlenderbotConfig) - added:
constructor BlenderbotModel(config: BlenderbotConfig) - added:
field BlenderbotForConditionalGeneration.base_model_prefix - added:
field BlenderbotPreTrainedModel.base_model_prefix - added:
field BlenderbotPreTrainedModel.config: BlenderbotConfig - added:
field BlenderbotPreTrainedModel.supports_gradient_checkpointing - added:
method BlenderbotForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, lab… - added:
method BlenderbotForCausalLM.get_decoder() - added:
method BlenderbotForCausalLM.get_input_embeddings() - added:
method BlenderbotForCausalLM.set_decoder(decoder) - added:
method BlenderbotForCausalLM.set_input_embeddings(value) - added:
method BlenderbotForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Union[t… - added:
method BlenderbotForConditionalGeneration.get_decoder() - added:
method BlenderbotForConditionalGeneration.get_encoder() - added:
method BlenderbotForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method BlenderbotModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Union[tuple, BaseModelOutp… - added:
method BlenderbotModel.get_encoder() - added:
method BlenderbotModel.get_input_embeddings() - added:
method BlenderbotModel.set_input_embeddings(value) - added:
property BlenderbotPreTrainedModel.dummy_inputs - added:
static method BlenderbotForConditionalGeneration.from_pretrained(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *model_args, **kwargs) - added:
static method BlenderbotModel.from_pretrained(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *model_args, **kwargs)
- added:
class FlaxBlenderbotForConditionalGeneration(FlaxBlenderbotPreTrainedModel) - added:
class FlaxBlenderbotModel(FlaxBlenderbotPreTrainedModel) - added:
class FlaxBlenderbotPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBlenderbotPreTrainedModel(config: BlenderbotConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxBlenderbotForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxBlenderbotForConditionalGeneration.module_class - added:
field FlaxBlenderbotModel.config: BlenderbotConfig - added:
field FlaxBlenderbotModel.dtype: jnp.dtype - added:
field FlaxBlenderbotModel.module_class - added:
field FlaxBlenderbotPreTrainedModel.base_model_prefix: str - added:
field FlaxBlenderbotPreTrainedModel.config_class - added:
field FlaxBlenderbotPreTrainedModel.module_class: nn.Module - added:
method FlaxBlenderbotForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params:… - added:
method FlaxBlenderbotForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxBlenderbotForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBlenderbotPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional… - added:
method FlaxBlenderbotPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxBlenderbotPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxBlenderbotPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFBlenderbotForConditionalGeneration(TFBlenderbotPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFBlenderbotModel(TFBlenderbotPreTrainedModel) - added:
class TFBlenderbotPreTrainedModel(TFPreTrainedModel) - added:
constructor TFBlenderbotForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFBlenderbotModel(config: BlenderbotConfig, *inputs, **kwargs) - added:
field TFBlenderbotPreTrainedModel.base_model_prefix - added:
field TFBlenderbotPreTrainedModel.config_class - added:
method TFBlenderbotForConditionalGeneration.build(input_shape=…) - added:
method TFBlenderbotForConditionalGeneration.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: tuple | TFBaseModelOutput | N… - added:
method TFBlenderbotForConditionalGeneration.get_bias() - added:
method TFBlenderbotForConditionalGeneration.get_decoder() - added:
method TFBlenderbotForConditionalGeneration.get_encoder() - added:
method TFBlenderbotForConditionalGeneration.get_output_embeddings() - added:
method TFBlenderbotForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFBlenderbotForConditionalGeneration.serving_output(output) - added:
method TFBlenderbotForConditionalGeneration.set_bias(value) - added:
method TFBlenderbotForConditionalGeneration.set_output_embeddings(value) - added:
method TFBlenderbotModel.build(input_shape=…) - added:
method TFBlenderbotModel.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: tuple | TFBaseModelOutput | None=…, past_key_val… - added:
method TFBlenderbotModel.get_decoder() - added:
method TFBlenderbotModel.get_encoder() - added:
method TFBlenderbotModel.serving_output(output) - added:
static method TFBlenderbotForConditionalGeneration.from_pretrained(pretrained_model_name_or_path: str | os.PathLike | None, *model_args, **kwargs) - added:
static method TFBlenderbotModel.from_pretrained(pretrained_model_name_or_path: str | os.PathLike | None, *model_args, **kwargs)
- added:
class BlenderbotTokenizer(PreTrainedTokenizer) - added:
constructor BlenderbotTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field BlenderbotTokenizer.model_input_names - added:
field BlenderbotTokenizer.vocab_files_names - added:
method BlenderbotTokenizer.bpe(token) - added:
method BlenderbotTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) - added:
method BlenderbotTokenizer.convert_tokens_to_string(tokens) - added:
method BlenderbotTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BlenderbotTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method BlenderbotTokenizer.get_vocab() - added:
method BlenderbotTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method BlenderbotTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BlenderbotTokenizer.vocab_size
- added:
class BlenderbotTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BlenderbotTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field BlenderbotTokenizerFast.model_input_names - added:
field BlenderbotTokenizerFast.slow_tokenizer_class - added:
field BlenderbotTokenizerFast.vocab_files_names - added:
method BlenderbotTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) - added:
method BlenderbotTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method BlenderbotTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BlenderbotTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class BlenderbotSmallConfig(PretrainedConfig) - added:
class BlenderbotSmallOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor BlenderbotSmallConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token… - added:
field BlenderbotSmallConfig.attribute_map - added:
field BlenderbotSmallConfig.keys_to_ignore_at_inference - added:
field BlenderbotSmallConfig.model_type - added:
method BlenderbotSmallOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property BlenderbotSmallOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property BlenderbotSmallOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class BlenderbotSmallForCausalLM(BlenderbotSmallPreTrainedModel, GenerationMixin) - added:
class BlenderbotSmallForConditionalGeneration(BlenderbotSmallPreTrainedModel, GenerationMixin) - added:
class BlenderbotSmallModel(BlenderbotSmallPreTrainedModel) - added:
class BlenderbotSmallPreTrainedModel(PreTrainedModel) - added:
constructor BlenderbotSmallForCausalLM(config) - added:
constructor BlenderbotSmallForConditionalGeneration(config: BlenderbotSmallConfig) - added:
constructor BlenderbotSmallModel(config: BlenderbotSmallConfig) - added:
field BlenderbotSmallForConditionalGeneration.base_model_prefix - added:
field BlenderbotSmallPreTrainedModel.base_model_prefix - added:
field BlenderbotSmallPreTrainedModel.config: BlenderbotSmallConfig - added:
field BlenderbotSmallPreTrainedModel.supports_gradient_checkpointing - added:
method BlenderbotSmallForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…… - added:
method BlenderbotSmallForCausalLM.get_decoder() - added:
method BlenderbotSmallForCausalLM.get_input_embeddings() - added:
method BlenderbotSmallForCausalLM.set_decoder(decoder) - added:
method BlenderbotSmallForCausalLM.set_input_embeddings(value) - added:
method BlenderbotSmallForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Un… - added:
method BlenderbotSmallForConditionalGeneration.get_decoder() - added:
method BlenderbotSmallForConditionalGeneration.get_encoder() - added:
method BlenderbotSmallForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method BlenderbotSmallModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Union[tuple, BaseMode… - added:
method BlenderbotSmallModel.get_encoder() - added:
method BlenderbotSmallModel.get_input_embeddings() - added:
method BlenderbotSmallModel.set_input_embeddings(value) - added:
property BlenderbotSmallPreTrainedModel.dummy_inputs
- added:
class FlaxBlenderbotSmallForConditionalGeneration(FlaxBlenderbotSmallPreTrainedModel) - added:
class FlaxBlenderbotSmallModel(FlaxBlenderbotSmallPreTrainedModel) - added:
class FlaxBlenderbotSmallPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBlenderbotSmallPreTrainedModel(config: BlenderbotSmallConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxBlenderbotSmallForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxBlenderbotSmallForConditionalGeneration.module_class - added:
field FlaxBlenderbotSmallModel.config: BlenderbotSmallConfig - added:
field FlaxBlenderbotSmallModel.dtype: jnp.dtype - added:
field FlaxBlenderbotSmallModel.module_class - added:
field FlaxBlenderbotSmallPreTrainedModel.base_model_prefix: str - added:
field FlaxBlenderbotSmallPreTrainedModel.config_class - added:
field FlaxBlenderbotSmallPreTrainedModel.module_class: nn.Module - added:
method FlaxBlenderbotSmallForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, deterministic: bo… - added:
method FlaxBlenderbotSmallForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxBlenderbotSmallForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBlenderbotSmallPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Opt… - added:
method FlaxBlenderbotSmallPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxBlenderbotSmallPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxBlenderbotSmallPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFBlenderbotSmallForConditionalGeneration(TFBlenderbotSmallPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFBlenderbotSmallModel(TFBlenderbotSmallPreTrainedModel) - added:
class TFBlenderbotSmallPreTrainedModel(TFPreTrainedModel) - added:
constructor TFBlenderbotSmallForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFBlenderbotSmallModel(config: BlenderbotSmallConfig, *inputs, **kwargs) - added:
field TFBlenderbotSmallPreTrainedModel.base_model_prefix - added:
field TFBlenderbotSmallPreTrainedModel.config_class - added:
method TFBlenderbotSmallForConditionalGeneration.build(input_shape=…) - added:
method TFBlenderbotSmallForConditionalGeneration.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: TFBaseModelOutput | None… - added:
method TFBlenderbotSmallForConditionalGeneration.get_bias() - added:
method TFBlenderbotSmallForConditionalGeneration.get_decoder() - added:
method TFBlenderbotSmallForConditionalGeneration.get_encoder() - added:
method TFBlenderbotSmallForConditionalGeneration.get_output_embeddings() - added:
method TFBlenderbotSmallForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFBlenderbotSmallForConditionalGeneration.serving_output(output) - added:
method TFBlenderbotSmallForConditionalGeneration.set_bias(value) - added:
method TFBlenderbotSmallForConditionalGeneration.set_output_embeddings(value) - added:
method TFBlenderbotSmallModel.build(input_shape=…) - added:
method TFBlenderbotSmallModel.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: tuple | TFBaseModelOutput | None=…, past_ke… - added:
method TFBlenderbotSmallModel.get_decoder() - added:
method TFBlenderbotSmallModel.get_encoder() - added:
method TFBlenderbotSmallModel.serving_output(output)
- added:
class BlenderbotSmallTokenizer(PreTrainedTokenizer) - added:
constructor BlenderbotSmallTokenizer(vocab_file, merges_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, **kwargs) - added:
field BlenderbotSmallTokenizer.model_input_names - added:
field BlenderbotSmallTokenizer.vocab_files_names - added:
method BlenderbotSmallTokenizer.bpe(token: str) -> str - added:
method BlenderbotSmallTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method BlenderbotSmallTokenizer.get_vocab() -> dict - added:
method BlenderbotSmallTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property BlenderbotSmallTokenizer.vocab_size -> int
- added:
class BlenderbotSmallTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BlenderbotSmallTokenizerFast(vocab_file=…, merges_file=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field BlenderbotSmallTokenizerFast.slow_tokenizer_class - added:
field BlenderbotSmallTokenizerFast.vocab_files_names - added:
method BlenderbotSmallTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method BlenderbotSmallTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int]
- added:
export define_import_structure
- added:
class BlipConfig(PretrainedConfig) - added:
class BlipTextConfig(PretrainedConfig) - added:
class BlipVisionConfig(PretrainedConfig) - added:
constructor BlipConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, image_text_hidden_size=…, label_smoothing=…, **kwargs) - added:
constructor BlipTextConfig(vocab_size=…, hidden_size=…, encoder_hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, bos_token_id=…, eos_token_id=…, pad_token_id=…, sep_token_id=…, is_decoder=…, use_cache=…, label_smoot… - added:
constructor BlipVisionConfig(hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, **kwargs) - added:
field BlipConfig.model_type - added:
field BlipConfig.sub_configs - added:
field BlipTextConfig.base_config_key - added:
field BlipTextConfig.model_type - added:
field BlipVisionConfig.base_config_key - added:
field BlipVisionConfig.model_type
- added:
class BlipImageProcessor(BaseImageProcessor) - added:
constructor BlipImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field BlipImageProcessor.model_input_names - added:
method BlipImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, Ten… - added:
method BlipImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class BlipImageProcessorFast(BaseImageProcessorFast) - added:
field BlipImageProcessorFast.do_convert_rgb - added:
field BlipImageProcessorFast.do_normalize - added:
field BlipImageProcessorFast.do_rescale - added:
field BlipImageProcessorFast.do_resize - added:
field BlipImageProcessorFast.image_mean - added:
field BlipImageProcessorFast.image_std - added:
field BlipImageProcessorFast.resample - added:
field BlipImageProcessorFast.size
- added:
class BlipForConditionalGeneration(BlipPreTrainedModel, GenerationMixin) - added:
class BlipForImageTextRetrieval(BlipPreTrainedModel) - added:
class BlipForQuestionAnswering(BlipPreTrainedModel, GenerationMixin) - added:
class BlipModel(BlipPreTrainedModel) - added:
class BlipPreTrainedModel(PreTrainedModel) - added:
class BlipVisionModel(BlipPreTrainedModel) - added:
constructor BlipForConditionalGeneration(config: BlipConfig) - added:
constructor BlipForImageTextRetrieval(config: BlipConfig) - added:
constructor BlipForQuestionAnswering(config: BlipConfig) - added:
constructor BlipModel(config: BlipConfig) - added:
constructor BlipVisionModel(config: BlipVisionConfig) - added:
export BlipTextModel - added:
field BlipForConditionalGeneration.config: BlipConfig - added:
field BlipForConditionalGeneration.main_input_name - added:
field BlipForImageTextRetrieval.config: BlipConfig - added:
field BlipForQuestionAnswering.config: BlipConfig - added:
field BlipModel.config: BlipConfig - added:
field BlipPreTrainedModel.base_model_prefix - added:
field BlipPreTrainedModel.config: BlipConfig - added:
field BlipPreTrainedModel.supports_gradient_checkpointing - added:
field BlipVisionModel.config: BlipVisionConfig - added:
field BlipVisionModel.main_input_name - added:
method BlipForConditionalGeneration.forward(pixel_values: torch.FloatTensor, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BlipForConditionalGenerationModelOutput] - added:
method BlipForConditionalGeneration.generate(pixel_values: torch.FloatTensor, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method BlipForConditionalGeneration.get_input_embeddings() - added:
method BlipForConditionalGeneration.set_input_embeddings(value) - added:
method BlipForImageTextRetrieval.forward(input_ids: torch.LongTensor, pixel_values: torch.FloatTensor, use_itm_head: Optional[bool]=…, attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BlipTextVisionModelOutput] - added:
method BlipForImageTextRetrieval.get_input_embeddings() - added:
method BlipForImageTextRetrieval.set_input_embeddings(value) - added:
method BlipForQuestionAnswering.forward(input_ids: torch.LongTensor, pixel_values: torch.FloatTensor, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BlipTextVisionModelOut… - added:
method BlipForQuestionAnswering.generate(input_ids: torch.LongTensor, pixel_values: torch.FloatTensor, attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method BlipForQuestionAnswering.get_input_embeddings() - added:
method BlipForQuestionAnswering.set_input_embeddings(value) - added:
method BlipModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BlipOutput] - added:
method BlipModel.get_image_features(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method BlipModel.get_input_embeddings() - added:
method BlipModel.get_multimodal_features(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method BlipModel.get_text_features(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method BlipModel.set_input_embeddings(value) - added:
method BlipVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPooling] - added:
method BlipVisionModel.get_input_embeddings()
- added:
class BlipTextLMHeadModel(BlipTextPreTrainedModel, GenerationMixin) - added:
class BlipTextModel(BlipTextPreTrainedModel) - added:
class BlipTextPreTrainedModel(PreTrainedModel) - added:
constructor BlipTextLMHeadModel(config) - added:
constructor BlipTextModel(config, add_pooling_layer=…) - added:
field BlipTextPreTrainedModel.base_model_prefix - added:
field BlipTextPreTrainedModel.config: BlipTextConfig - added:
method BlipTextLMHeadModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…… - added:
method BlipTextLMHeadModel.get_input_embeddings() - added:
method BlipTextLMHeadModel.get_output_embeddings() - added:
method BlipTextLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method BlipTextLMHeadModel.set_input_embeddings(new_embeddings) - added:
method BlipTextLMHeadModel.set_output_embeddings(new_embeddings) - added:
method BlipTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]… - added:
method BlipTextModel.get_extended_attention_mask(attention_mask: Tensor, input_shape: tuple[int], device: device, is_decoder: bool) -> Tensor - added:
method BlipTextModel.get_input_embeddings() - added:
method BlipTextModel.set_input_embeddings(value)
- added:
class TFBlipForConditionalGeneration(TFBlipPreTrainedModel) - added:
class TFBlipForImageTextRetrieval(TFBlipPreTrainedModel) - added:
class TFBlipForQuestionAnswering(TFBlipPreTrainedModel) - added:
class TFBlipModel(TFBlipPreTrainedModel) - added:
class TFBlipPreTrainedModel(TFPreTrainedModel) - added:
class TFBlipVisionModel(TFBlipPreTrainedModel) - added:
constructor TFBlipForConditionalGeneration(config: BlipConfig, *args, **kwargs) - added:
constructor TFBlipForImageTextRetrieval(config: BlipConfig, *args, **kwargs) - added:
constructor TFBlipForQuestionAnswering(config: BlipConfig, *args, **kwargs) - added:
constructor TFBlipModel(config: BlipConfig, *inputs, **kwargs) - added:
constructor TFBlipVisionModel(config: BlipVisionConfig, *args, **kwargs) - added:
export TFBlipTextModel - added:
field TFBlipForConditionalGeneration.config_class - added:
field TFBlipForConditionalGeneration.main_input_name - added:
field TFBlipForImageTextRetrieval.config_class - added:
field TFBlipForQuestionAnswering.config_class - added:
field TFBlipModel.config_class - added:
field TFBlipModel.main_input_name - added:
field TFBlipPreTrainedModel.base_model_prefix - added:
field TFBlipPreTrainedModel.config_class - added:
field TFBlipVisionModel.config_class - added:
field TFBlipVisionModel.main_input_name - added:
method TFBlipForConditionalGeneration.build(input_shape=…) - added:
method TFBlipForConditionalGeneration.call(pixel_values: tf.Tensor, input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, labels: tf.Tensor | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBlipForConditionalGenerationModelOutput - added:
method TFBlipForConditionalGeneration.generate(pixel_values: tf.Tensor, input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, **generate_kwargs) -> tf.Tensor - added:
method TFBlipForConditionalGeneration.get_input_embeddings() -> keras.layers.Layer - added:
method TFBlipForImageTextRetrieval.build(input_shape=…) - added:
method TFBlipForImageTextRetrieval.call(input_ids: tf.Tensor, pixel_values: tf.Tensor | None=…, use_itm_head: bool | None=…, attention_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBlipImageTextMatchingModelOutput - added:
method TFBlipForImageTextRetrieval.get_input_embeddings() -> keras.layers.Layer - added:
method TFBlipForQuestionAnswering.build(input_shape=…) - added:
method TFBlipForQuestionAnswering.call(input_ids: tf.Tensor, pixel_values: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, labels: tf.Tensor | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBlipTextVisionModelO… - added:
method TFBlipForQuestionAnswering.generate(input_ids: tf.Tensor, pixel_values: tf.Tensor, attention_mask: tf.Tensor | None=…, **generate_kwargs) -> tf.Tensor - added:
method TFBlipForQuestionAnswering.get_input_embeddings() -> keras.layers.Layer - added:
method TFBlipModel.build(input_shape=…) - added:
method TFBlipModel.call(input_ids: tf.Tensor | None=…, pixel_values: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, return_loss: bool | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBlipOutput - added:
method TFBlipModel.get_image_features(pixel_values: tf.Tensor | None=…, return_dict: bool | None=…) -> tf.Tensor - added:
method TFBlipModel.get_text_features(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, return_dict: bool | None=…) -> tf.Tensor - added:
method TFBlipModel.serving_output(output: TFBlipOutput) -> TFBlipOutput - added:
method TFBlipVisionModel.build(input_shape=…) - added:
method TFBlipVisionModel.call(pixel_values: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBaseModelOutputWithPooling - added:
method TFBlipVisionModel.get_input_embeddings() - added:
method TFBlipVisionModel.serving_output(output: TFBaseModelOutputWithPooling) -> TFBaseModelOutputWithPooling
- added:
class TFBlipTextLMHeadModel(TFBlipTextPreTrainedModel) - added:
class TFBlipTextModel(TFBlipTextPreTrainedModel) - added:
class TFBlipTextPreTrainedModel(TFPreTrainedModel) - added:
constructor TFBlipTextLMHeadModel(config, **kwargs) - added:
constructor TFBlipTextModel(config, add_pooling_layer=…, name=…, **kwargs) - added:
field TFBlipTextPreTrainedModel.base_model_prefix - added:
field TFBlipTextPreTrainedModel.config_class - added:
method TFBlipTextLMHeadModel.build(input_shape=…) - added:
method TFBlipTextLMHeadModel.call(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, past_key_values=…, use_cache=…, output_attentions=…, output_hidden_states=…, return_dict=…, return_logits=…, is_decoder=…, training=…) - added:
method TFBlipTextLMHeadModel.get_output_embeddings() - added:
method TFBlipTextLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFBlipTextLMHeadModel.set_output_embeddings(new_embeddings) - added:
method TFBlipTextModel.build(input_shape=…) - added:
method TFBlipTextModel.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, encoder_embeds: tf.Tensor | None=…, encoder_hidden_states: tf.Tensor | None=…, encoder_attention_mask: tf.Tensor | None=…, past_key_values: tuple[tuple[tf.Tensor]] | None=…, use_cache: bool | None=… - added:
method TFBlipTextModel.get_extended_attention_mask(attention_mask: tf.Tensor, input_shape: tuple[int], is_decoder: bool) -> tf.Tensor - added:
method TFBlipTextModel.get_input_embeddings() - added:
method TFBlipTextModel.set_input_embeddings(value)
- added:
class BlipProcessor(ProcessorMixin) - added:
constructor BlipProcessor(image_processor, tokenizer, **kwargs) - added:
field BlipProcessor.attributes - added:
field BlipProcessor.image_processor_class - added:
field BlipProcessor.tokenizer_class - added:
property BlipProcessor.model_input_names
- added:
export define_import_structure
- added:
class Blip2Config(PretrainedConfig) - added:
class Blip2QFormerConfig(PretrainedConfig) - added:
class Blip2VisionConfig(PretrainedConfig) - added:
constructor Blip2Config(vision_config=…, qformer_config=…, text_config=…, num_query_tokens=…, image_text_hidden_size=…, image_token_index=…, **kwargs) - added:
constructor Blip2QFormerConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, cross_attention_frequency=…, encoder_hidden_size=…, use_qformer_text_input=…, **kwargs) - added:
constructor Blip2VisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, **kwargs) - added:
field Blip2Config.attribute_map - added:
field Blip2Config.model_type - added:
field Blip2Config.sub_configs - added:
field Blip2QFormerConfig.base_config_key - added:
field Blip2QFormerConfig.model_type - added:
field Blip2VisionConfig.base_config_key - added:
field Blip2VisionConfig.model_type - added:
static method Blip2Config.from_vision_qformer_text_configs(vision_config: Blip2VisionConfig, qformer_config: Blip2QFormerConfig, text_config: Optional[PretrainedConfig]=…, **kwargs)
- added:
class Blip2ForConditionalGeneration(Blip2PreTrainedModel, GenerationMixin) - added:
class Blip2ForImageTextRetrieval(Blip2PreTrainedModel) - added:
class Blip2Model(Blip2PreTrainedModel) - added:
class Blip2PreTrainedModel(PreTrainedModel) - added:
class Blip2QFormerModel(Blip2PreTrainedModel) - added:
class Blip2TextModelWithProjection(Blip2PreTrainedModel) - added:
class Blip2VisionModel(Blip2PreTrainedModel) - added:
class Blip2VisionModelWithProjection(Blip2PreTrainedModel) - added:
constructor Blip2ForConditionalGeneration(config: Blip2Config) - added:
constructor Blip2ForImageTextRetrieval(config: Blip2Config) - added:
constructor Blip2Model(config: Blip2Config) - added:
constructor Blip2QFormerModel(config: Blip2QFormerConfig) - added:
constructor Blip2TextModelWithProjection(config: Blip2Config) - added:
constructor Blip2VisionModel(config: Blip2VisionConfig) - added:
constructor Blip2VisionModelWithProjection(config: Blip2Config) - added:
field Blip2ForConditionalGeneration.config: Blip2Config - added:
field Blip2ForConditionalGeneration.main_input_name - added:
field Blip2ForImageTextRetrieval.main_input_name - added:
field Blip2Model.config: Blip2Config - added:
field Blip2Model.main_input_name - added:
field Blip2PreTrainedModel.base_model_prefix - added:
field Blip2PreTrainedModel.config: Blip2Config - added:
field Blip2PreTrainedModel.supports_gradient_checkpointing - added:
field Blip2TextModelWithProjection.supports_gradient_checkpointing - added:
field Blip2VisionModel.config: Blip2VisionConfig - added:
field Blip2VisionModel.main_input_name - added:
field Blip2VisionModelWithProjection.main_input_name - added:
method Blip2ForConditionalGeneration.forward(pixel_values: torch.FloatTensor, input_ids: torch.LongTensor, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[Transfor… - added:
method Blip2ForConditionalGeneration.generate(pixel_values: torch.FloatTensor, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method Blip2ForConditionalGeneration.get_decoder() - added:
method Blip2ForConditionalGeneration.get_encoder() - added:
method Blip2ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method Blip2ForConditionalGeneration.get_input_embeddings() - added:
method Blip2ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Blip2ForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method Blip2ForConditionalGeneration.set_input_embeddings(value) - added:
method Blip2ForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method Blip2ForImageTextRetrieval.forward(pixel_values: torch.FloatTensor, input_ids: torch.LongTensor, attention_mask: Optional[torch.LongTensor]=…, use_image_text_matching_head: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Blip2ImageTextMatchingModelOutput] - added:
method Blip2ForImageTextRetrieval.get_input_embeddings() - added:
method Blip2ForImageTextRetrieval.set_input_embeddings(value) - added:
method Blip2Model.forward(pixel_values: torch.FloatTensor, input_ids: torch.FloatTensor, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Blip2ForConditionalGenerationModelO… - added:
method Blip2Model.get_decoder() - added:
method Blip2Model.get_encoder() - added:
method Blip2Model.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…, legacy_output: bool=…) -> Union[torch.FloatTensor, CausalLMOutputWithPast] - added:
method Blip2Model.get_input_embeddings() - added:
method Blip2Model.get_output_embeddings() -> nn.Module - added:
method Blip2Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method Blip2Model.get_qformer_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…, legacy_output: bool=…) -> Union[torch.FloatTensor, BaseModelOutputWithPooling] - added:
method Blip2Model.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, legacy_output: bool=…) -> Union[torch.FloatTensor, CausalLMOutputWithPast] - added:
method Blip2Model.set_input_embeddings(value) - added:
method Blip2Model.set_output_embeddings(new_embeddings) - added:
method Blip2QFormerModel.forward(query_embeds: torch.FloatTensor, query_length: Optional[int]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple[torch.Tensor], BaseModelOutputWithPoolingAndCrossAt… - added:
method Blip2QFormerModel.get_extended_attention_mask(attention_mask: torch.Tensor, input_shape: tuple[int], device: torch.device, has_query: bool=…) -> torch.Tensor - added:
method Blip2QFormerModel.get_input_embeddings() - added:
method Blip2QFormerModel.set_input_embeddings(value) - added:
method Blip2TextModelWithProjection.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Blip2TextModelOutput] - added:
method Blip2TextModelWithProjection.get_input_embeddings() - added:
method Blip2TextModelWithProjection.set_input_embeddings(value) - added:
method Blip2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPooling] - added:
method Blip2VisionModel.get_input_embeddings() - added:
method Blip2VisionModelWithProjection.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Blip2VisionModelOutput] - added:
method Blip2VisionModelWithProjection.get_input_embeddings() -> nn.Module
- added:
class Blip2Processor(ProcessorMixin) - added:
constructor Blip2Processor(image_processor, tokenizer, num_query_tokens=…, **kwargs) - added:
field Blip2Processor.attributes - added:
field Blip2Processor.image_processor_class - added:
field Blip2Processor.tokenizer_class
- added:
export define_import_structure
- added:
class BloomConfig(PretrainedConfig) - added:
class BloomOnnxConfig(OnnxConfigWithPast) - added:
constructor BloomConfig(vocab_size=…, hidden_size=…, n_layer=…, n_head=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, bos_token_id=…, eos_token_id=…, apply_residual_connection_post_layernorm=…, hidden_dropout=…, attention_dropout=…, pretraining_tp=…, slow_but_exact=…, **kwargs) - added:
constructor BloomOnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…, use_past: bool=…) - added:
field BloomConfig.attribute_map - added:
field BloomConfig.keys_to_ignore_at_inference - added:
field BloomConfig.model_type - added:
field BloomOnnxConfig.torch_onnx_minimum_version - added:
method BloomOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property BloomOnnxConfig.atol_for_validation -> float - added:
property BloomOnnxConfig.default_onnx_opset -> int - added:
property BloomOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property BloomOnnxConfig.num_attention_heads -> int - added:
property BloomOnnxConfig.num_layers -> int
- added:
class BloomForCausalLM(BloomPreTrainedModel, GenerationMixin) - added:
class BloomForQuestionAnswering(BloomPreTrainedModel) - added:
class BloomForSequenceClassification(BloomPreTrainedModel) - added:
class BloomForTokenClassification(BloomPreTrainedModel) - added:
class BloomModel(BloomPreTrainedModel) - added:
class BloomPreTrainedModel(PreTrainedModel) - added:
constructor BloomForCausalLM(config: BloomConfig) - added:
constructor BloomForQuestionAnswering(config) - added:
constructor BloomForSequenceClassification(config: BloomConfig) - added:
constructor BloomForTokenClassification(config: BloomConfig) - added:
constructor BloomModel(config: BloomConfig) - added:
constructor BloomPreTrainedModel(*inputs, **kwargs) - added:
field BloomPreTrainedModel.base_model_prefix - added:
field BloomPreTrainedModel.config: BloomConfig - added:
field BloomPreTrainedModel.supports_gradient_checkpointing - added:
method BloomForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_st… - added:
method BloomForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, use_cache=…, **kwargs) - added:
method BloomForCausalLM.set_output_embeddings(new_embeddings: torch.Tensor) - added:
method BloomForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hi… - added:
method BloomForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, ou… - added:
method BloomForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, outpu… - added:
method BloomModel.build_alibi_tensor(attention_mask: torch.Tensor, num_heads: int, dtype: torch.dtype) -> torch.Tensor - added:
method BloomModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_d… - added:
method BloomModel.get_input_embeddings() - added:
method BloomModel.set_input_embeddings(new_embeddings: torch.Tensor)
- added:
class FlaxBloomForCausalLM(FlaxBloomPreTrainedModel) - added:
class FlaxBloomModel(FlaxBloomPreTrainedModel) - added:
class FlaxBloomPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxBloomPreTrainedModel(config: BloomConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxBloomForCausalLM.module_class - added:
field FlaxBloomModel.module_class - added:
field FlaxBloomPreTrainedModel.base_model_prefix - added:
field FlaxBloomPreTrainedModel.config_class - added:
field FlaxBloomPreTrainedModel.module_class: nn.Module - added:
method FlaxBloomForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxBloomForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxBloomPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxBloomPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class BloomTokenizerFast(PreTrainedTokenizerFast) - added:
constructor BloomTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_prefix_space=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field BloomTokenizerFast.model_input_names - added:
field BloomTokenizerFast.slow_tokenizer_class - added:
field BloomTokenizerFast.vocab_files_names - added:
method BloomTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class BltConfig(PretrainedConfig) - added:
class BltGlobalTransformerConfig(PretrainedConfig) - added:
class BltLocalDecoderConfig(PretrainedConfig) - added:
class BltLocalEncoderConfig(PretrainedConfig) - added:
class BltPatcherConfig(PretrainedConfig) - added:
constructor BltConfig(vocab_size=…, max_position_embeddings=…, patch_in_forward=…, patch_size=…, patching_mode=…, patching_threshold=…, patching_batch_size=…, max_patch_length=…, cross_attn_k=…, encoder_hash_byte_group_size=…, encoder_hash_byte_group_vocab=…, encoder_hash_byte_group_nb_functions=…, patcher_config=…, encoder_config=…, decoder_config=…, global_config=…, tie_word_embeddings=…, initi… - added:
constructor BltGlobalTransformerConfig(hidden_size=…, num_attention_heads=…, num_key_value_heads=…, num_hidden_layers=…, rms_norm_eps=…, dropout=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, hidden_act=…, intermediate_size=…, initializer_range=…, **kwargs) - added:
constructor BltLocalDecoderConfig(vocab_size=…, cross_attn_all_layers=…, cross_attn_k=…, hidden_size_global=…, hidden_size=…, num_attention_heads=…, num_key_value_heads=…, num_hidden_layers=…, rms_norm_eps=…, dropout=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, hidden_act=…, intermediate_size=…, initializer_range=…, **kwargs) - added:
constructor BltLocalEncoderConfig(vocab_size=…, cross_attn_all_layers=…, cross_attn_k=…, hidden_size_global=…, hidden_size=…, num_attention_heads=…, num_key_value_heads=…, num_hidden_layers=…, rms_norm_eps=…, dropout=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, hidden_act=…, intermediate_size=…, initializer_range=…, **kwargs) - added:
constructor BltPatcherConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, max_position_embeddings=…, rms_norm_eps=…, dropout=…, rope_theta=…, intermediate_size=…, rope_scaling=…, initializer_range=…, **kwargs) - added:
field BltConfig.keys_to_ignore_at_inference - added:
field BltConfig.model_type - added:
field BltConfig.sub_configs - added:
field BltGlobalTransformerConfig.model_type - added:
field BltLocalDecoderConfig.model_type - added:
field BltLocalEncoderConfig.model_type - added:
field BltPatcherConfig.model_type
- added:
class BltForCausalLM(BltPreTrainedModel, GenerationMixin) - added:
class BltModel(BltPreTrainedModel) - added:
class BltPatcher(BltPreTrainedModel) - added:
class BltPreTrainedModel(PreTrainedModel) - added:
constructor BltForCausalLM(config: BltConfig) - added:
constructor BltModel(config: BltConfig) - added:
constructor BltPatcher(config: BltPatcherConfig) - added:
field BltForCausalLM.base_model_prefix - added:
field BltForCausalLM.config: BltConfig - added:
field BltPatcher.config: BltPatcherConfig - added:
field BltPreTrainedModel.base_model_prefix - added:
field BltPreTrainedModel.config: BltConfig - added:
field BltPreTrainedModel.supports_gradient_checkpointing - added:
method BltForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, cross_attention_states: Optional[torch.LongTensor]=…, cross_attention_mask: Optional[torch.LongTensor]=…, full_text_row_masked_out_mask: Optional[tuple[torch.Tensor, torch.Tensor]]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTen… - added:
method BltModel.forward(input_ids: Optional[torch.LongTensor]=…, patch_lengths: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModel… - added:
method BltModel.get_input_embeddings() - added:
method BltModel.set_input_embeddings(value) - added:
method BltPatcher.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, patch_size: Optional[int]=…, threshold: Optional[float]=…, max_patch_length: Optional[int… - added:
static method BltPatcher.patch_lengths_from_entropies(entropies, sequence_length, patch_size=…, threshold=…)
- added:
class BltForCausalLM(MllamaForCausalLM) - added:
class BltModel(BltPreTrainedModel) - added:
class BltPatcher(BltPreTrainedModel) - added:
class BltPreTrainedModel(MllamaPreTrainedModel) - added:
constructor BltForCausalLM(config: BltConfig) - added:
constructor BltModel(config: BltConfig) - added:
constructor BltPatcher(config: BltPatcherConfig) - added:
field BltForCausalLM.base_model_prefix - added:
field BltForCausalLM.config: BltConfig - added:
field BltPatcher.config: BltPatcherConfig - added:
field BltPreTrainedModel.config: BltConfig - added:
method BltForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, cross_attention_states: Optional[torch.LongTensor]=…, cross_attention_mask: Optional[torch.LongTensor]=…, full_text_row_masked_out_mask: Optional[tuple[torch.Tensor, torch.Tensor]]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTen… - added:
method BltModel.forward(input_ids: Optional[torch.LongTensor]=…, patch_lengths: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModel… - added:
method BltModel.get_input_embeddings() - added:
method BltModel.set_input_embeddings(value) - added:
method BltPatcher.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, patch_size: Optional[int]=…, threshold: Optional[float]=…, max_patch_length: Optional[int… - added:
static method BltPatcher.patch_lengths_from_entropies(entropies, sequence_length, patch_size=…, threshold=…)
- added:
export define_import_structure
- added:
class BridgeTowerConfig(PretrainedConfig) - added:
class BridgeTowerTextConfig(PretrainedConfig) - added:
class BridgeTowerVisionConfig(PretrainedConfig) - added:
constructor BridgeTowerConfig(share_cross_modal_transformer_layers=…, hidden_act=…, hidden_size=…, initializer_factor=…, layer_norm_eps=…, share_link_tower_layers=…, link_tower_type=…, num_attention_heads=…, num_hidden_layers=…, tie_word_embeddings=…, init_layernorm_from_vision_encoder=…, text_config=…, vision_config=…, **kwargs) - added:
constructor BridgeTowerTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, initializer_factor=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, **kwargs) - added:
constructor BridgeTowerVisionConfig(hidden_size=…, num_hidden_layers=…, num_channels=…, patch_size=…, image_size=…, initializer_factor=…, layer_norm_eps=…, stop_gradient=…, share_layernorm=…, remove_last_layer=…, **kwargs) - added:
field BridgeTowerConfig.model_type - added:
field BridgeTowerConfig.sub_configs - added:
field BridgeTowerTextConfig.base_config_key - added:
field BridgeTowerTextConfig.model_type - added:
field BridgeTowerVisionConfig.base_config_key - added:
field BridgeTowerVisionConfig.model_type
- added:
class BridgeTowerImageProcessor(BaseImageProcessor) - added:
constructor BridgeTowerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, size_divisor: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_pad: bool… - added:
field BridgeTowerImageProcessor.model_input_names - added:
method BridgeTowerImageProcessor.center_crop(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method BridgeTowerImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method BridgeTowerImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, size_divisor: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, d… - added:
method BridgeTowerImageProcessor.resize(image: np.ndarray, size: dict[str, int], size_divisor: int=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class BridgeTowerImageProcessorFast(BaseImageProcessorFast) - added:
constructor BridgeTowerImageProcessorFast(**kwargs: Unpack[BridgeTowerFastImageProcessorKwargs]) - added:
field BridgeTowerImageProcessorFast.crop_size - added:
field BridgeTowerImageProcessorFast.default_to_square - added:
field BridgeTowerImageProcessorFast.do_center_crop - added:
field BridgeTowerImageProcessorFast.do_normalize - added:
field BridgeTowerImageProcessorFast.do_pad - added:
field BridgeTowerImageProcessorFast.do_rescale - added:
field BridgeTowerImageProcessorFast.do_resize - added:
field BridgeTowerImageProcessorFast.image_mean - added:
field BridgeTowerImageProcessorFast.image_std - added:
field BridgeTowerImageProcessorFast.model_input_names - added:
field BridgeTowerImageProcessorFast.resample - added:
field BridgeTowerImageProcessorFast.size - added:
field BridgeTowerImageProcessorFast.size_divisor - added:
field BridgeTowerImageProcessorFast.valid_kwargs - added:
method BridgeTowerImageProcessorFast.center_crop(image: torch.Tensor, size: dict[str, int], **kwargs) -> torch.Tensor - added:
method BridgeTowerImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[BridgeTowerFastImageProcessorKwargs]) -> BatchFeature - added:
method BridgeTowerImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, size_divisor: int=…, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method BridgeTowerImageProcessorFast.to_dict()
- added:
class BridgeTowerForContrastiveLearning(BridgeTowerPreTrainedModel) - added:
class BridgeTowerForImageAndTextRetrieval(BridgeTowerPreTrainedModel) - added:
class BridgeTowerForMaskedLM(BridgeTowerPreTrainedModel) - added:
class BridgeTowerModel(BridgeTowerPreTrainedModel) - added:
class BridgeTowerPreTrainedModel(PreTrainedModel) - added:
constructor BridgeTowerForContrastiveLearning(config) - added:
constructor BridgeTowerForImageAndTextRetrieval(config) - added:
constructor BridgeTowerForMaskedLM(config) - added:
constructor BridgeTowerModel(config) - added:
field BridgeTowerPreTrainedModel.base_model_prefix - added:
field BridgeTowerPreTrainedModel.config: BridgeTowerConfig - added:
field BridgeTowerPreTrainedModel.supports_gradient_checkpointing - added:
method BridgeTowerForContrastiveLearning.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]… - added:
method BridgeTowerForImageAndTextRetrieval.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTenso… - added:
method BridgeTowerForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, output_… - added:
method BridgeTowerForMaskedLM.get_output_embeddings() - added:
method BridgeTowerForMaskedLM.set_output_embeddings(new_embeddings) - added:
method BridgeTowerModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, image_token_t… - added:
method BridgeTowerModel.get_cls_features(text_features, image_features) - added:
method BridgeTowerModel.get_input_embeddings() - added:
method BridgeTowerModel.set_input_embeddings(value)
- added:
class BridgeTowerProcessor(ProcessorMixin) - added:
constructor BridgeTowerProcessor(image_processor, tokenizer) - added:
field BridgeTowerProcessor.attributes - added:
field BridgeTowerProcessor.image_processor_class - added:
field BridgeTowerProcessor.tokenizer_class - added:
field BridgeTowerProcessor.valid_processor_kwargs
- added:
export define_import_structure
- added:
class BrosConfig(PretrainedConfig) - added:
constructor BrosConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, dim_bbox=…, bbox_scale=…, n_relations=…, classifier_dropout_prob=…, **kwargs) - added:
field BrosConfig.model_type
- added:
class BrosForTokenClassification(BrosPreTrainedModel) - added:
class BrosModel(BrosPreTrainedModel) - added:
class BrosPreTrainedModel(PreTrainedModel) - added:
class BrosSpadeEEForTokenClassification(BrosPreTrainedModel) - added:
class BrosSpadeELForTokenClassification(BrosPreTrainedModel) - added:
constructor BrosForTokenClassification(config) - added:
constructor BrosModel(config, add_pooling_layer=…) - added:
constructor BrosSpadeEEForTokenClassification(config) - added:
constructor BrosSpadeELForTokenClassification(config) - added:
field BrosPreTrainedModel.base_model_prefix - added:
field BrosPreTrainedModel.config: BrosConfig - added:
method BrosForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, bbox_first_token_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, out… - added:
method BrosModel.forward(input_ids: Optional[torch.Tensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, outp… - added:
method BrosModel.get_input_embeddings() - added:
method BrosModel.set_input_embeddings(value) - added:
method BrosSpadeEEForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, bbox_first_token_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, initial_token_labels: Optional… - added:
method BrosSpadeELForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, bbox_first_token_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]…
- added:
class BrosProcessor(ProcessorMixin) - added:
constructor BrosProcessor(tokenizer=…, **kwargs) - added:
field BrosProcessor.attributes - added:
field BrosProcessor.tokenizer_class - added:
field BrosProcessor.valid_processor_kwargs
- added:
export define_import_structure
- added:
class ByT5Tokenizer(PreTrainedTokenizer) - added:
constructor ByT5Tokenizer(eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, **kwargs) -> None - added:
field ByT5Tokenizer.model_input_names - added:
method ByT5Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ByT5Tokenizer.convert_tokens_to_string(tokens) - added:
method ByT5Tokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ByT5Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method ByT5Tokenizer.get_vocab() - added:
method ByT5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ByT5Tokenizer.vocab_size
- added:
export define_import_structure
- added:
class CamembertConfig(PretrainedConfig) - added:
class CamembertOnnxConfig(OnnxConfig) - added:
constructor CamembertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field CamembertConfig.model_type - added:
property CamembertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class CamembertForCausalLM(CamembertPreTrainedModel, GenerationMixin) - added:
class CamembertForMaskedLM(CamembertPreTrainedModel) - added:
class CamembertForMultipleChoice(CamembertPreTrainedModel) - added:
class CamembertForQuestionAnswering(CamembertPreTrainedModel) - added:
class CamembertForSequenceClassification(CamembertPreTrainedModel) - added:
class CamembertForTokenClassification(CamembertPreTrainedModel) - added:
class CamembertModel(CamembertPreTrainedModel) - added:
class CamembertPreTrainedModel(PreTrainedModel) - added:
constructor CamembertForCausalLM(config) - added:
constructor CamembertForMaskedLM(config) - added:
constructor CamembertForMultipleChoice(config) - added:
constructor CamembertForQuestionAnswering(config) - added:
constructor CamembertForSequenceClassification(config) - added:
constructor CamembertForTokenClassification(config) - added:
constructor CamembertModel(config, add_pooling_layer=…) - added:
field CamembertPreTrainedModel.base_model_prefix - added:
field CamembertPreTrainedModel.config: CamembertConfig - added:
field CamembertPreTrainedModel.supports_gradient_checkpointing - added:
method CamembertForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.Floa… - added:
method CamembertForCausalLM.get_output_embeddings() - added:
method CamembertForCausalLM.set_output_embeddings(new_embeddings) - added:
method CamembertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.Floa… - added:
method CamembertForMaskedLM.get_output_embeddings() - added:
method CamembertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method CamembertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_st… - added:
method CamembertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]… - added:
method CamembertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method CamembertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method CamembertModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache… - added:
method CamembertModel.get_input_embeddings() - added:
method CamembertModel.set_input_embeddings(value)
- added:
class TFCamembertForCausalLM(TFCamembertPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFCamembertForMaskedLM(TFCamembertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFCamembertForMultipleChoice(TFCamembertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFCamembertForQuestionAnswering(TFCamembertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFCamembertForSequenceClassification(TFCamembertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFCamembertForTokenClassification(TFCamembertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFCamembertModel(TFCamembertPreTrainedModel) - added:
class TFCamembertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFCamembertForCausalLM(config: CamembertConfig, *inputs, **kwargs) - added:
constructor TFCamembertForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFCamembertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFCamembertForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFCamembertForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFCamembertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFCamembertModel(config, *inputs, **kwargs) - added:
field TFCamembertPreTrainedModel.base_model_prefix - added:
field TFCamembertPreTrainedModel.config_class - added:
method TFCamembertForCausalLM.build(input_shape=…) - added:
method TFCamembertForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarra… - added:
method TFCamembertForCausalLM.get_lm_head() - added:
method TFCamembertForCausalLM.get_prefix_bias_name() - added:
method TFCamembertForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFCamembertForMaskedLM.build(input_shape=…) - added:
method TFCamembertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFCamembertForMaskedLM.get_lm_head() - added:
method TFCamembertForMaskedLM.get_prefix_bias_name() - added:
method TFCamembertForMultipleChoice.build(input_shape=…) - added:
method TFCamembertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict:… - added:
method TFCamembertForQuestionAnswering.build(input_shape=…) - added:
method TFCamembertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFCamembertForSequenceClassification.build(input_shape=…) - added:
method TFCamembertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, retu… - added:
method TFCamembertForTokenClassification.build(input_shape=…) - added:
method TFCamembertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_… - added:
method TFCamembertModel.build(input_shape=…) - added:
method TFCamembertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf…
- added:
class CamembertTokenizer(PreTrainedTokenizer) - added:
constructor CamembertTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field CamembertTokenizer.model_input_names - added:
field CamembertTokenizer.vocab_files_names - added:
method CamembertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CamembertTokenizer.convert_tokens_to_string(tokens) - added:
method CamembertTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CamembertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CamembertTokenizer.get_vocab() - added:
method CamembertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property CamembertTokenizer.vocab_size
- added:
class CamembertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CamembertTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - added:
field CamembertTokenizerFast.model_input_names - added:
field CamembertTokenizerFast.slow_tokenizer_class - added:
field CamembertTokenizerFast.vocab_files_names - added:
method CamembertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CamembertTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CamembertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class CanineConfig(PretrainedConfig) - added:
constructor CanineConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, downsampling_rate=…, upsampling_kernel_size=…, num_hash_functions=…, num_hash_buckets=…, loc… - added:
field CanineConfig.model_type
- added:
class CanineForMultipleChoice(CaninePreTrainedModel) - added:
class CanineForQuestionAnswering(CaninePreTrainedModel) - added:
class CanineForSequenceClassification(CaninePreTrainedModel) - added:
class CanineForTokenClassification(CaninePreTrainedModel) - added:
class CanineLayer(GradientCheckpointingLayer) - added:
class CanineModel(CaninePreTrainedModel) - added:
class CaninePreTrainedModel(PreTrainedModel) - added:
constructor CanineForMultipleChoice(config) - added:
constructor CanineForQuestionAnswering(config) - added:
constructor CanineForSequenceClassification(config) - added:
constructor CanineForTokenClassification(config) - added:
constructor CanineLayer(config, local, always_attend_to_first_position, first_position_attends_to_all, attend_from_chunk_width, attend_from_chunk_stride, attend_to_chunk_width, attend_to_chunk_stride) - added:
constructor CanineModel(config, add_pooling_layer=…) - added:
field CaninePreTrainedModel.base_model_prefix - added:
field CaninePreTrainedModel.config: CanineConfig - added:
field CaninePreTrainedModel.load_tf_weights - added:
field CaninePreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_canine(model, config, tf_checkpoint_path) - added:
method CanineForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_state… - added:
method CanineForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…,… - added:
method CanineForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method CanineForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_… - added:
method CanineLayer.feed_forward_chunk(attention_output) - added:
method CanineLayer.forward(hidden_states: tuple[torch.FloatTensor], attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…) -> tuple[torch.FloatTensor, Optional[torch.FloatTensor]] - added:
method CanineModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=……
- added:
class CanineTokenizer(PreTrainedTokenizer) - added:
constructor CanineTokenizer(bos_token=…, eos_token=…, sep_token=…, cls_token=…, pad_token=…, mask_token=…, add_prefix_space=…, model_max_length=…, **kwargs) - added:
method CanineTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CanineTokenizer.convert_tokens_to_string(tokens) - added:
method CanineTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CanineTokenizer.get_vocab() - added:
method CanineTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) - added:
property CanineTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class ChameleonConfig(PretrainedConfig) - added:
class ChameleonVQVAEConfig(PretrainedConfig) - added:
constructor ChameleonConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, model_parallel_size=…,… - added:
constructor ChameleonVQVAEConfig(embed_dim: int=…, num_embeddings: int=…, double_latent: bool=…, latent_channels: int=…, resolution: int=…, in_channels: int=…, base_channels: int=…, channel_multiplier: list[int]=…, num_res_blocks: int=…, attn_resolutions: Optional[list[int]]=…, dropout: float=…, attn_type: str=…, initializer_range=…, **kwargs) - added:
field ChameleonConfig.keys_to_ignore_at_inference - added:
field ChameleonConfig.model_type - added:
field ChameleonConfig.sub_configs - added:
field ChameleonVQVAEConfig.base_config_key - added:
field ChameleonVQVAEConfig.model_type
- added:
class ChameleonImageProcessor(BaseImageProcessor) - added:
constructor ChameleonImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -… - added:
field ChameleonImageProcessor.model_input_names - added:
method ChameleonImageProcessor.blend_rgba(image: ImageInput) -> ImageInput - added:
method ChameleonImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Un… - added:
method ChameleonImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class ChameleonImageProcessorFast(BaseImageProcessorFast) - added:
field ChameleonImageProcessorFast.crop_size - added:
field ChameleonImageProcessorFast.default_to_square - added:
field ChameleonImageProcessorFast.do_center_crop - added:
field ChameleonImageProcessorFast.do_convert_rgb - added:
field ChameleonImageProcessorFast.do_normalize - added:
field ChameleonImageProcessorFast.do_rescale - added:
field ChameleonImageProcessorFast.do_resize - added:
field ChameleonImageProcessorFast.image_mean - added:
field ChameleonImageProcessorFast.image_std - added:
field ChameleonImageProcessorFast.resample - added:
field ChameleonImageProcessorFast.rescale_factor - added:
field ChameleonImageProcessorFast.size - added:
method ChameleonImageProcessorFast.convert_to_rgb(image: ImageInput) -> ImageInput - added:
method ChameleonImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor
- added:
class ChameleonForConditionalGeneration(ChameleonPreTrainedModel, GenerationMixin) - added:
class ChameleonModel(ChameleonPreTrainedModel) - added:
class ChameleonPreTrainedModel(PreTrainedModel) - added:
class ChameleonVQVAE(ChameleonPreTrainedModel) - added:
constructor ChameleonForConditionalGeneration(config) - added:
constructor ChameleonModel(config: ChameleonConfig) - added:
constructor ChameleonVQVAE(config: ChameleonVQVAEConfig) - added:
field ChameleonPreTrainedModel.base_model_prefix - added:
field ChameleonPreTrainedModel.config: ChameleonConfig - added:
field ChameleonPreTrainedModel.supports_gradient_checkpointing - added:
field ChameleonVQVAE.config: ChameleonVQVAEConfig - added:
method ChameleonForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[b… - added:
method ChameleonForConditionalGeneration.get_image_features(pixel_values) - added:
method ChameleonForConditionalGeneration.get_image_tokens(pixel_values) - added:
method ChameleonForConditionalGeneration.prepare_inputs_for_generation(input_ids, pixel_values=…, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method ChameleonModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_di… - added:
method ChameleonModel.get_image_features(pixel_values: torch.FloatTensor) - added:
method ChameleonModel.get_image_tokens(pixel_values: torch.FloatTensor) - added:
method ChameleonModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method ChameleonVQVAE.encode(pixel_values: torch.LongTensor)
- added:
class ChameleonProcessor(ProcessorMixin) - added:
constructor ChameleonProcessor(image_processor, tokenizer, image_seq_length: int=…, image_token: str=…) - added:
field ChameleonProcessor.attributes - added:
field ChameleonProcessor.image_processor_class - added:
field ChameleonProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class ChineseCLIPConfig(PretrainedConfig) - added:
class ChineseCLIPOnnxConfig(OnnxConfig) - added:
class ChineseCLIPTextConfig(PretrainedConfig) - added:
class ChineseCLIPVisionConfig(PretrainedConfig) - added:
constructor ChineseCLIPConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor ChineseCLIPTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, initializer_factor=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, **kwargs) - added:
constructor ChineseCLIPVisionConfig(hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field ChineseCLIPConfig.model_type - added:
field ChineseCLIPConfig.sub_configs - added:
field ChineseCLIPTextConfig.base_config_key - added:
field ChineseCLIPTextConfig.model_type - added:
field ChineseCLIPVisionConfig.base_config_key - added:
field ChineseCLIPVisionConfig.model_type - added:
method ChineseCLIPOnnxConfig.generate_dummy_inputs(processor: ProcessorMixin, batch_size: int=…, seq_length: int=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property ChineseCLIPOnnxConfig.atol_for_validation -> float - added:
property ChineseCLIPOnnxConfig.default_onnx_opset -> int - added:
property ChineseCLIPOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property ChineseCLIPOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class ChineseCLIPFeatureExtractor(ChineseCLIPImageProcessor) - added:
constructor ChineseCLIPFeatureExtractor(*args, **kwargs) -> None
- added:
class ChineseCLIPImageProcessor(BaseImageProcessor) - added:
constructor ChineseCLIPImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs)… - added:
field ChineseCLIPImageProcessor.model_input_names - added:
method ChineseCLIPImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[… - added:
method ChineseCLIPImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class ChineseCLIPImageProcessorFast(BaseImageProcessorFast) - added:
field ChineseCLIPImageProcessorFast.crop_size - added:
field ChineseCLIPImageProcessorFast.default_to_square - added:
field ChineseCLIPImageProcessorFast.do_center_crop - added:
field ChineseCLIPImageProcessorFast.do_convert_rgb - added:
field ChineseCLIPImageProcessorFast.do_normalize - added:
field ChineseCLIPImageProcessorFast.do_rescale - added:
field ChineseCLIPImageProcessorFast.do_resize - added:
field ChineseCLIPImageProcessorFast.image_mean - added:
field ChineseCLIPImageProcessorFast.image_std - added:
field ChineseCLIPImageProcessorFast.resample - added:
field ChineseCLIPImageProcessorFast.size
- added:
class ChineseCLIPModel(ChineseCLIPPreTrainedModel) - added:
class ChineseCLIPPreTrainedModel(PreTrainedModel) - added:
class ChineseCLIPTextModel(ChineseCLIPPreTrainedModel) - added:
class ChineseCLIPVisionModel(ChineseCLIPPreTrainedModel) - added:
constructor ChineseCLIPModel(config: ChineseCLIPConfig) - added:
constructor ChineseCLIPTextModel(config, add_pooling_layer=…) - added:
constructor ChineseCLIPVisionModel(config: ChineseCLIPVisionConfig) - added:
field ChineseCLIPModel.config: ChineseCLIPConfig - added:
field ChineseCLIPPreTrainedModel.base_model_prefix - added:
field ChineseCLIPPreTrainedModel.config: ChineseCLIPConfig - added:
field ChineseCLIPPreTrainedModel.supports_gradient_checkpointing - added:
field ChineseCLIPTextModel.config: ChineseCLIPTextConfig - added:
field ChineseCLIPVisionModel.config: ChineseCLIPVisionConfig - added:
field ChineseCLIPVisionModel.main_input_name - added:
method ChineseCLIPModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict… - added:
method ChineseCLIPModel.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method ChineseCLIPModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method ChineseCLIPTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional… - added:
method ChineseCLIPTextModel.get_input_embeddings() - added:
method ChineseCLIPTextModel.set_input_embeddings(value) - added:
method ChineseCLIPVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method ChineseCLIPVisionModel.get_input_embeddings() -> nn.Module
- added:
class ChineseCLIPProcessor(ProcessorMixin) - added:
constructor ChineseCLIPProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field ChineseCLIPProcessor.attributes - added:
field ChineseCLIPProcessor.image_processor_class - added:
field ChineseCLIPProcessor.tokenizer_class - added:
property ChineseCLIPProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class ClapAudioConfig(PretrainedConfig) - added:
class ClapConfig(PretrainedConfig) - added:
class ClapTextConfig(PretrainedConfig) - added:
constructor ClapAudioConfig(window_size=…, num_mel_bins=…, spec_size=…, hidden_act=…, patch_size=…, patch_stride=…, num_classes=…, hidden_size=…, projection_dim=…, depths=…, num_attention_heads=…, enable_fusion=…, hidden_dropout_prob=…, fusion_type=…, patch_embed_input_channels=…, flatten_patch_embeds=…, patch_embeds_hidden_size=…, enable_patch_layer_norm=…, drop_path_rate=…, attention_probs_drop… - added:
constructor ClapConfig(text_config=…, audio_config=…, logit_scale_init_value=…, projection_dim=…, projection_hidden_act=…, initializer_factor=…, **kwargs) - added:
constructor ClapTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_factor=…, layer_norm_eps=…, projection_dim=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, projection_hidden… - added:
field ClapAudioConfig.base_config_key - added:
field ClapAudioConfig.model_type - added:
field ClapConfig.model_type - added:
field ClapConfig.sub_configs - added:
field ClapTextConfig.base_config_key - added:
field ClapTextConfig.model_type
- added:
class ClapFeatureExtractor(SequenceFeatureExtractor) - added:
constructor ClapFeatureExtractor(feature_size=…, sampling_rate=…, hop_length=…, max_length_s=…, fft_window_size=…, padding_value=…, return_attention_mask=…, frequency_min: float=…, frequency_max: float=…, top_db: Optional[int]=…, truncation: str=…, padding: str=…, **kwargs) - added:
field ClapFeatureExtractor.model_input_names - added:
method ClapFeatureExtractor.to_dict() -> dict[str, Any]
- added:
class ClapAudioModel(ClapPreTrainedModel) - added:
class ClapAudioModelWithProjection(ClapPreTrainedModel) - added:
class ClapModel(ClapPreTrainedModel) - added:
class ClapPreTrainedModel(PreTrainedModel) - added:
class ClapTextModel(ClapPreTrainedModel) - added:
class ClapTextModelWithProjection(ClapPreTrainedModel) - added:
constructor ClapAudioModel(config: ClapAudioConfig) - added:
constructor ClapAudioModelWithProjection(config: ClapAudioConfig) - added:
constructor ClapModel(config: ClapConfig) - added:
constructor ClapTextModel(config, add_pooling_layer=…) - added:
constructor ClapTextModelWithProjection(config: ClapTextConfig) - added:
field ClapAudioModel.config: ClapAudioConfig - added:
field ClapAudioModel.main_input_name - added:
field ClapAudioModelWithProjection.config: ClapAudioConfig - added:
field ClapAudioModelWithProjection.main_input_name - added:
field ClapModel.config: ClapConfig - added:
field ClapPreTrainedModel.base_model_prefix - added:
field ClapPreTrainedModel.config: ClapConfig - added:
field ClapPreTrainedModel.supports_gradient_checkpointing - added:
field ClapTextModel.config: ClapTextConfig - added:
field ClapTextModelWithProjection.config: ClapTextConfig - added:
method ClapAudioModel.forward(input_features: Optional[torch.FloatTensor]=…, is_longer: Optional[torch.BoolTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method ClapAudioModel.get_input_embeddings() -> nn.Module - added:
method ClapAudioModelWithProjection.forward(input_features: Optional[torch.FloatTensor]=…, is_longer: Optional[torch.BoolTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ClapAudioModelOutput] - added:
method ClapAudioModelWithProjection.get_input_embeddings() -> nn.Module - added:
method ClapModel.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, is_longer: Optional[torch.BoolTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Clap… - added:
method ClapModel.get_audio_features(input_features: torch.Tensor, is_longer: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method ClapModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method ClapTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Te… - added:
method ClapTextModel.get_input_embeddings() - added:
method ClapTextModel.set_input_embeddings(value) - added:
method ClapTextModelWithProjection.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ClapTextModelOutput] - added:
method ClapTextModelWithProjection.get_input_embeddings() -> nn.Module - added:
method ClapTextModelWithProjection.set_input_embeddings(value)
- added:
class ClapProcessor(ProcessorMixin) - added:
constructor ClapProcessor(feature_extractor, tokenizer) - added:
field ClapProcessor.feature_extractor_class - added:
field ClapProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class CLIPConfig(PretrainedConfig) - added:
class CLIPOnnxConfig(OnnxConfig) - added:
class CLIPTextConfig(PretrainedConfig) - added:
class CLIPVisionConfig(PretrainedConfig) - added:
constructor CLIPConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor CLIPTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor CLIPVisionConfig(hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field CLIPConfig.model_type - added:
field CLIPConfig.sub_configs - added:
field CLIPTextConfig.base_config_key - added:
field CLIPTextConfig.model_type - added:
field CLIPVisionConfig.base_config_key - added:
field CLIPVisionConfig.model_type - added:
method CLIPOnnxConfig.generate_dummy_inputs(processor: ProcessorMixin, batch_size: int=…, seq_length: int=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property CLIPOnnxConfig.atol_for_validation -> float - added:
property CLIPOnnxConfig.default_onnx_opset -> int - added:
property CLIPOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property CLIPOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class CLIPFeatureExtractor(CLIPImageProcessor) - added:
constructor CLIPFeatureExtractor(*args, **kwargs) -> None
- added:
class CLIPImageProcessor(BaseImageProcessor) - added:
constructor CLIPImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field CLIPImageProcessor.model_input_names - added:
method CLIPImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[f… - added:
method CLIPImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class CLIPImageProcessorFast(BaseImageProcessorFast) - added:
field CLIPImageProcessorFast.crop_size - added:
field CLIPImageProcessorFast.default_to_square - added:
field CLIPImageProcessorFast.do_center_crop - added:
field CLIPImageProcessorFast.do_convert_rgb - added:
field CLIPImageProcessorFast.do_normalize - added:
field CLIPImageProcessorFast.do_rescale - added:
field CLIPImageProcessorFast.do_resize - added:
field CLIPImageProcessorFast.image_mean - added:
field CLIPImageProcessorFast.image_std - added:
field CLIPImageProcessorFast.resample - added:
field CLIPImageProcessorFast.size
- added:
class CLIPForImageClassification(CLIPPreTrainedModel) - added:
class CLIPModel(CLIPPreTrainedModel) - added:
class CLIPPreTrainedModel(PreTrainedModel) - added:
class CLIPTextModel(CLIPPreTrainedModel) - added:
class CLIPTextModelWithProjection(CLIPPreTrainedModel) - added:
class CLIPVisionModel(CLIPPreTrainedModel) - added:
class CLIPVisionModelWithProjection(CLIPPreTrainedModel) - added:
constructor CLIPForImageClassification(config: CLIPConfig) -> None - added:
constructor CLIPModel(config: CLIPConfig) - added:
constructor CLIPTextModel(config: CLIPTextConfig) - added:
constructor CLIPTextModelWithProjection(config: CLIPTextConfig) - added:
constructor CLIPVisionModel(config: CLIPVisionConfig) - added:
constructor CLIPVisionModelWithProjection(config: CLIPVisionConfig) - added:
field CLIPForImageClassification.main_input_name - added:
field CLIPModel.config: CLIPConfig - added:
field CLIPPreTrainedModel.base_model_prefix - added:
field CLIPPreTrainedModel.config: CLIPConfig - added:
field CLIPPreTrainedModel.supports_gradient_checkpointing - added:
field CLIPTextModel.config: CLIPTextConfig - added:
field CLIPTextModelWithProjection.config: CLIPTextConfig - added:
field CLIPVisionModel.config: CLIPVisionConfig - added:
field CLIPVisionModel.main_input_name - added:
field CLIPVisionModelWithProjection.config: CLIPVisionConfig - added:
field CLIPVisionModelWithProjection.main_input_name - added:
method CLIPForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> ImageClassifierOutput - added:
method CLIPModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> CLIPOutput - added:
method CLIPModel.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method CLIPModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method CLIPTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPooling - added:
method CLIPTextModel.get_input_embeddings() -> nn.Module - added:
method CLIPTextModel.set_input_embeddings(value) - added:
method CLIPTextModelWithProjection.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> CLIPTextModelOutput - added:
method CLIPTextModelWithProjection.get_input_embeddings() -> nn.Module - added:
method CLIPTextModelWithProjection.set_input_embeddings(value) - added:
method CLIPVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> BaseModelOutputWithPooling - added:
method CLIPVisionModel.get_input_embeddings() -> nn.Module - added:
method CLIPVisionModelWithProjection.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> CLIPVisionModelOutput - added:
method CLIPVisionModelWithProjection.get_input_embeddings() -> nn.Module
- added:
class FlaxCLIPModel(FlaxCLIPPreTrainedModel) - added:
class FlaxCLIPPreTrainedModel(FlaxPreTrainedModel) - added:
class FlaxCLIPTextModel(FlaxCLIPTextPreTrainedModel) - added:
class FlaxCLIPTextModelWithProjection(FlaxCLIPTextPreTrainedModel) - added:
class FlaxCLIPTextPreTrainedModel(FlaxPreTrainedModel) - added:
class FlaxCLIPVisionModel(FlaxCLIPVisionPreTrainedModel) - added:
class FlaxCLIPVisionPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxCLIPPreTrainedModel(config: CLIPConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
constructor FlaxCLIPTextPreTrainedModel(config: CLIPTextConfig, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
constructor FlaxCLIPVisionPreTrainedModel(config: CLIPVisionConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxCLIPModel.module_class - added:
field FlaxCLIPPreTrainedModel.config_class - added:
field FlaxCLIPPreTrainedModel.module_class: nn.Module - added:
field FlaxCLIPTextModel.module_class - added:
field FlaxCLIPTextModelWithProjection.module_class - added:
field FlaxCLIPTextPreTrainedModel.config_class - added:
field FlaxCLIPTextPreTrainedModel.module_class: nn.Module - added:
field FlaxCLIPVisionModel.module_class - added:
field FlaxCLIPVisionPreTrainedModel.config_class - added:
field FlaxCLIPVisionPreTrainedModel.main_input_name - added:
field FlaxCLIPVisionPreTrainedModel.module_class: nn.Module - added:
method FlaxCLIPPreTrainedModel.get_image_features(pixel_values, params: Optional[dict]=…, dropout_rng: jax.random.PRNGKey=…, train=…) - added:
method FlaxCLIPPreTrainedModel.get_text_features(input_ids, attention_mask=…, position_ids=…, params: Optional[dict]=…, dropout_rng: jax.random.PRNGKey=…, train=…) - added:
method FlaxCLIPPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxCLIPTextPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxCLIPVisionPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFCLIPModel(TFCLIPPreTrainedModel) - added:
class TFCLIPPreTrainedModel(TFPreTrainedModel) - added:
class TFCLIPTextModel(TFCLIPPreTrainedModel) - added:
class TFCLIPVisionModel(TFCLIPPreTrainedModel) - added:
constructor TFCLIPModel(config: CLIPConfig, *inputs, **kwargs) - added:
constructor TFCLIPTextModel(config: CLIPTextConfig, *inputs, **kwargs) - added:
constructor TFCLIPVisionModel(config: CLIPVisionConfig, *inputs, **kwargs) - added:
field TFCLIPModel.config_class - added:
field TFCLIPPreTrainedModel.base_model_prefix - added:
field TFCLIPPreTrainedModel.config_class - added:
field TFCLIPTextModel.config_class - added:
field TFCLIPVisionModel.config_class - added:
field TFCLIPVisionModel.main_input_name - added:
method TFCLIPModel.build(input_shape=…) - added:
method TFCLIPModel.call(input_ids: TFModelInputType | None=…, pixel_values: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, return_loss: bool | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFCLIPOutput | tuple[tf.Tensor] - added:
method TFCLIPModel.get_image_features(pixel_values: TFModelInputType | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tf.Tensor - added:
method TFCLIPModel.get_text_features(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tf.Tensor - added:
method TFCLIPModel.serving_output(output: TFCLIPOutput) -> TFCLIPOutput - added:
method TFCLIPTextModel.build(input_shape=…) - added:
method TFCLIPTextModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor] - added:
method TFCLIPVisionModel.build(input_shape=…) - added:
method TFCLIPVisionModel.call(pixel_values: TFModelInputType | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor]
- added:
class CLIPProcessor(ProcessorMixin) - added:
constructor CLIPProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field CLIPProcessor.attributes - added:
field CLIPProcessor.image_processor_class - added:
field CLIPProcessor.tokenizer_class - added:
property CLIPProcessor.feature_extractor - added:
property CLIPProcessor.feature_extractor_class
- added:
class CLIPTokenizer(PreTrainedTokenizer) - added:
constructor CLIPTokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, **kwargs) - added:
field CLIPTokenizer.model_input_names - added:
field CLIPTokenizer.vocab_files_names - added:
method CLIPTokenizer.bpe(token) - added:
method CLIPTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CLIPTokenizer.convert_tokens_to_string(tokens) - added:
method CLIPTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CLIPTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CLIPTokenizer.get_vocab() - added:
method CLIPTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property CLIPTokenizer.vocab_size
- added:
class CLIPTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CLIPTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, **kwargs) - added:
field CLIPTokenizerFast.model_input_names - added:
field CLIPTokenizerFast.slow_tokenizer_class - added:
field CLIPTokenizerFast.vocab_files_names - added:
method CLIPTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CLIPTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CLIPTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class CLIPSegConfig(PretrainedConfig) - added:
class CLIPSegTextConfig(PretrainedConfig) - added:
class CLIPSegVisionConfig(PretrainedConfig) - added:
constructor CLIPSegConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, extract_layers=…, reduce_dim=…, decoder_num_attention_heads=…, decoder_attention_dropout=…, decoder_hidden_act=…, decoder_intermediate_size=…, conditional_layer=…, use_complex_transposed_convolution=…, **kwargs) - added:
constructor CLIPSegTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor CLIPSegVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field CLIPSegConfig.model_type - added:
field CLIPSegConfig.sub_configs - added:
field CLIPSegTextConfig.base_config_key - added:
field CLIPSegTextConfig.model_type - added:
field CLIPSegVisionConfig.base_config_key - added:
field CLIPSegVisionConfig.model_type
- added:
class CLIPSegForImageSegmentation(CLIPSegPreTrainedModel) - added:
class CLIPSegModel(CLIPSegPreTrainedModel) - added:
class CLIPSegPreTrainedModel(PreTrainedModel) - added:
class CLIPSegTextModel(CLIPSegPreTrainedModel) - added:
class CLIPSegVisionModel(CLIPSegPreTrainedModel) - added:
constructor CLIPSegForImageSegmentation(config: CLIPSegConfig) - added:
constructor CLIPSegModel(config: CLIPSegConfig) - added:
constructor CLIPSegTextModel(config: CLIPSegTextConfig) - added:
constructor CLIPSegVisionModel(config: CLIPSegVisionConfig) - added:
field CLIPSegForImageSegmentation.config: CLIPSegConfig - added:
field CLIPSegModel.config: CLIPSegConfig - added:
field CLIPSegPreTrainedModel.base_model_prefix - added:
field CLIPSegPreTrainedModel.config: CLIPSegConfig - added:
field CLIPSegPreTrainedModel.supports_gradient_checkpointing - added:
field CLIPSegTextModel.config: CLIPSegTextConfig - added:
field CLIPSegVisionModel.config: CLIPSegVisionConfig - added:
field CLIPSegVisionModel.main_input_name - added:
method CLIPSegForImageSegmentation.forward(input_ids: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, conditional_pixel_values: Optional[torch.FloatTensor]=…, conditional_embeddings: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]… - added:
method CLIPSegForImageSegmentation.get_conditional_embeddings(batch_size: Optional[int]=…, input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, conditional_pixel_values: Optional[torch.Tensor]=…) - added:
method CLIPSegModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, CLIPSegOut… - added:
method CLIPSegModel.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method CLIPSegModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method CLIPSegTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method CLIPSegTextModel.get_input_embeddings() -> nn.Module - added:
method CLIPSegTextModel.set_input_embeddings(value) - added:
method CLIPSegVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method CLIPSegVisionModel.get_input_embeddings() -> nn.Module
- added:
class CLIPSegProcessor(ProcessorMixin) - added:
constructor CLIPSegProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field CLIPSegProcessor.attributes - added:
field CLIPSegProcessor.image_processor_class - added:
field CLIPSegProcessor.tokenizer_class - added:
property CLIPSegProcessor.feature_extractor - added:
property CLIPSegProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class ClvpConfig(PretrainedConfig) - added:
class ClvpDecoderConfig(PretrainedConfig) - added:
class ClvpEncoderConfig(PretrainedConfig) - added:
constructor ClvpConfig(text_config=…, speech_config=…, decoder_config=…, projection_dim=…, logit_scale_init_value=…, initializer_factor=…, **kwargs) - added:
constructor ClvpDecoderConfig(vocab_size=…, max_position_embeddings=…, max_text_tokens=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, n_inner=…, num_mel_attn_blocks=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attention_dropout=…, layer_norm_epsilon=…, initializer_range=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dr… - added:
constructor ClvpEncoderConfig(vocab_size=…, hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, dropout=…, use_rotary_embedding=…, use_attention_bias=…, summary_type=…, initializer_factor=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field ClvpConfig.model_type - added:
field ClvpConfig.sub_configs - added:
field ClvpDecoderConfig.base_config_key - added:
field ClvpDecoderConfig.model_type - added:
field ClvpEncoderConfig.base_config_key - added:
field ClvpEncoderConfig.model_type - added:
static method ClvpConfig.from_sub_model_configs(text_config: ClvpEncoderConfig, speech_config: ClvpEncoderConfig, decoder_config: ClvpDecoderConfig, **kwargs) - added:
static method ClvpEncoderConfig.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], config_type: str=…, **kwargs)
- added:
class ClvpFeatureExtractor(SequenceFeatureExtractor) - added:
constructor ClvpFeatureExtractor(feature_size=…, sampling_rate=…, default_audio_length=…, hop_length=…, chunk_length=…, n_fft=…, padding_value=…, mel_norms=…, return_attention_mask=…, **kwargs) - added:
field ClvpFeatureExtractor.model_input_names
- added:
class ClvpDecoder(ClvpPreTrainedModel) - added:
class ClvpEncoder(ClvpPreTrainedModel) - added:
class ClvpForCausalLM(ClvpPreTrainedModel, GenerationMixin) - added:
class ClvpModel(ClvpPreTrainedModel) - added:
class ClvpModelForConditionalGeneration(ClvpPreTrainedModel, GenerationMixin) - added:
class ClvpPreTrainedModel(PreTrainedModel) - added:
constructor ClvpDecoder(config) - added:
constructor ClvpEncoder(config: ClvpConfig) - added:
constructor ClvpForCausalLM(config) - added:
constructor ClvpModel(config: ClvpDecoderConfig) - added:
constructor ClvpModelForConditionalGeneration(config: ClvpConfig) - added:
field ClvpModelForConditionalGeneration.config: ClvpConfig - added:
field ClvpPreTrainedModel.base_model_prefix - added:
field ClvpPreTrainedModel.config: ClvpConfig - added:
field ClvpPreTrainedModel.supports_gradient_checkpointing - added:
method ClvpDecoder.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, outp… - added:
method ClvpDecoder.get_input_embeddings() - added:
method ClvpDecoder.set_input_embeddings(new_embeddings) - added:
method ClvpEncoder.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method ClvpEncoder.get_input_embeddings() - added:
method ClvpEncoder.set_input_embeddings(value) - added:
method ClvpForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…,… - added:
method ClvpForCausalLM.get_input_embeddings() - added:
method ClvpForCausalLM.get_output_embeddings() - added:
method ClvpForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, conditioning_embeds=…, cache_position=…, **kwargs) - added:
method ClvpForCausalLM.set_input_embeddings(new_embeddings) - added:
method ClvpModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output… - added:
method ClvpModel.get_input_embeddings() - added:
method ClvpModel.set_input_embeddings(value) - added:
method ClvpModelForConditionalGeneration.fix_speech_decoder_output(speech_ids: torch.LongTensor) -> torch.LongTensor - added:
method ClvpModelForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, conditioning_encoder_inputs_embeds: Optional[torch.FloatTensor]=…, text_encoder_inputs_embeds: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions:… - added:
method ClvpModelForConditionalGeneration.generate(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, generation_config: Optional[GenerationConfig]=…, pad_to_max_mel_tokens: Optional[int]=…, output_hidden_states: Optional[bool]=…, **kwargs) - added:
method ClvpModelForConditionalGeneration.get_speech_features(speech_ids: Optional[torch.LongTensor]=…, input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, conditioning_encoder_inputs_embeds: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, **kwargs) -> torch.FloatTensor - added:
method ClvpModelForConditionalGeneration.get_text_features(input_ids: Optional[torch.LongTensor]=…, text_encoder_inputs_embeds: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…) -> torch.FloatTensor
- added:
class EnglishNormalizer - added:
constructor EnglishNormalizer() - added:
method EnglishNormalizer.collapse_whitespace(text: str) -> str - added:
method EnglishNormalizer.convert_to_ascii(text: str) -> str - added:
method EnglishNormalizer.expand_abbreviations(text: str) -> str - added:
method EnglishNormalizer.normalize_numbers(text: str) -> str - added:
method EnglishNormalizer.number_to_words(num: int) -> str
- added:
class ClvpProcessor(ProcessorMixin) - added:
constructor ClvpProcessor(feature_extractor, tokenizer) - added:
field ClvpProcessor.feature_extractor_class - added:
field ClvpProcessor.tokenizer_class
- added:
class ClvpTokenizer(PreTrainedTokenizer) - added:
constructor ClvpTokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_prefix_space=…, add_bos_token=…, add_eos_token=…, **kwargs) - added:
field ClvpTokenizer.model_input_names - added:
field ClvpTokenizer.vocab_files_names - added:
method ClvpTokenizer.bpe(token) - added:
method ClvpTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method ClvpTokenizer.clean_up_tokenization(text) - added:
method ClvpTokenizer.convert_tokens_to_string(tokens) - added:
method ClvpTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method ClvpTokenizer.get_vocab() - added:
method ClvpTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ClvpTokenizer.normalizer - added:
property ClvpTokenizer.vocab_size
- added:
export define_import_structure
- added:
class CodeLlamaTokenizer(PreTrainedTokenizer) - added:
constructor CodeLlamaTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, prefix_token=…, middle_token=…, suffix_token=…, eot_token=…, fill_token=…, suffix_first=…, sp_model_kwargs: Optional[dict[str, Any]]=…, add_bos_token=…, add_eos_token=…, clean_up_tokenization_spaces=…, additional_special_tokens=…, use_default_system_prompt=…, **kwargs) - added:
field CodeLlamaTokenizer.model_input_names - added:
field CodeLlamaTokenizer.vocab_files_names - added:
method CodeLlamaTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method CodeLlamaTokenizer.convert_tokens_to_string(tokens) - added:
method CodeLlamaTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CodeLlamaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CodeLlamaTokenizer.get_spm_processor() - added:
method CodeLlamaTokenizer.get_vocab() - added:
method CodeLlamaTokenizer.save_vocabulary(save_directory, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method CodeLlamaTokenizer.tokenize(prefix, suffix=…, suffix_first=…, **kwargs) -> list[int] - added:
property CodeLlamaTokenizer.eot_id - added:
property CodeLlamaTokenizer.eot_token - added:
property CodeLlamaTokenizer.middle_id - added:
property CodeLlamaTokenizer.middle_token - added:
property CodeLlamaTokenizer.prefix_id - added:
property CodeLlamaTokenizer.prefix_token - added:
property CodeLlamaTokenizer.suffix_id - added:
property CodeLlamaTokenizer.suffix_token - added:
property CodeLlamaTokenizer.unk_token_length - added:
property CodeLlamaTokenizer.vocab_size
- added:
class CodeLlamaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CodeLlamaTokenizerFast(vocab_file=…, tokenizer_file=…, clean_up_tokenization_spaces=…, unk_token=…, bos_token=…, eos_token=…, prefix_token=…, middle_token=…, suffix_token=…, eot_token=…, fill_token=…, additional_special_tokens=…, add_bos_token=…, add_eos_token=…, use_default_system_prompt=…, **kwargs) - added:
field CodeLlamaTokenizerFast.model_input_names - added:
field CodeLlamaTokenizerFast.padding_side - added:
field CodeLlamaTokenizerFast.slow_tokenizer_class - added:
field CodeLlamaTokenizerFast.vocab_files_names - added:
method CodeLlamaTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CodeLlamaTokenizerFast.encode_plus(text, text_pair=…, suffix_first=…, add_special_tokens=…, **kwargs) - added:
method CodeLlamaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method CodeLlamaTokenizerFast.set_infilling_processor(reset, suffix_first=…, add_special_tokens=…) - added:
method CodeLlamaTokenizerFast.update_post_processor() - added:
property CodeLlamaTokenizerFast.add_bos_token - added:
property CodeLlamaTokenizerFast.add_eos_token - added:
property CodeLlamaTokenizerFast.eot_id - added:
property CodeLlamaTokenizerFast.eot_token - added:
property CodeLlamaTokenizerFast.middle_id - added:
property CodeLlamaTokenizerFast.middle_token - added:
property CodeLlamaTokenizerFast.prefix_id - added:
property CodeLlamaTokenizerFast.prefix_token - added:
property CodeLlamaTokenizerFast.suffix_id - added:
property CodeLlamaTokenizerFast.suffix_token
- added:
export define_import_structure
- added:
class CodeGenConfig(PretrainedConfig) - added:
class CodeGenOnnxConfig(OnnxConfigWithPast) - added:
constructor CodeGenConfig(vocab_size=…, n_positions=…, n_ctx=…, n_embd=…, n_layer=…, n_head=…, rotary_dim=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, **kwargs) - added:
constructor CodeGenOnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…, use_past: bool=…) - added:
field CodeGenConfig.attribute_map - added:
field CodeGenConfig.model_type - added:
method CodeGenOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property CodeGenOnnxConfig.default_onnx_opset -> int - added:
property CodeGenOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property CodeGenOnnxConfig.num_attention_heads -> int - added:
property CodeGenOnnxConfig.num_layers -> int
- added:
class CodeGenForCausalLM(CodeGenPreTrainedModel, GenerationMixin) - added:
class CodeGenModel(CodeGenPreTrainedModel) - added:
class CodeGenPreTrainedModel(PreTrainedModel) - added:
constructor CodeGenForCausalLM(config) - added:
constructor CodeGenModel(config) - added:
constructor CodeGenPreTrainedModel(*inputs, **kwargs) - added:
field CodeGenPreTrainedModel.base_model_prefix - added:
field CodeGenPreTrainedModel.config: CodeGenConfig - added:
field CodeGenPreTrainedModel.supports_gradient_checkpointing - added:
method CodeGenForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongT… - added:
method CodeGenModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, outpu… - added:
method CodeGenModel.get_input_embeddings() - added:
method CodeGenModel.set_input_embeddings(new_embeddings)
- added:
class CodeGenTokenizer(PreTrainedTokenizer) - added:
constructor CodeGenTokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_prefix_space=…, add_bos_token=…, return_token_type_ids=…, **kwargs) - added:
field CodeGenTokenizer.model_input_names - added:
field CodeGenTokenizer.vocab_files_names - added:
method CodeGenTokenizer.bpe(token) - added:
method CodeGenTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method CodeGenTokenizer.convert_tokens_to_string(tokens) - added:
method CodeGenTokenizer.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, truncate_before_pattern: Optional[list[str]]=…, **kwargs) -> str - added:
method CodeGenTokenizer.get_vocab() - added:
method CodeGenTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method CodeGenTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method CodeGenTokenizer.truncate(completion, truncate_before_pattern) - added:
property CodeGenTokenizer.vocab_size
- added:
class CodeGenTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CodeGenTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, return_token_type_ids=…, **kwargs) - added:
field CodeGenTokenizerFast.model_input_names - added:
field CodeGenTokenizerFast.slow_tokenizer_class - added:
field CodeGenTokenizerFast.vocab_files_names - added:
method CodeGenTokenizerFast.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, truncate_before_pattern: Optional[list[str]]=…, **kwargs) -> str - added:
method CodeGenTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method CodeGenTokenizerFast.truncate(completion, truncate_before_pattern)
- added:
export define_import_structure
- added:
class CohereConfig(PretrainedConfig) - added:
constructor CohereConfig(vocab_size=…, hidden_size=…, intermediate_size=…, logit_scale=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, use_qk_no… - added:
field CohereConfig.base_model_pp_plan - added:
field CohereConfig.base_model_tp_plan - added:
field CohereConfig.keys_to_ignore_at_inference - added:
field CohereConfig.model_type
- added:
class CohereForCausalLM(CoherePreTrainedModel, GenerationMixin) - added:
class CohereModel(CoherePreTrainedModel) - added:
class CoherePreTrainedModel(PreTrainedModel) - added:
constructor CohereForCausalLM(config) - added:
constructor CohereModel(config: CohereConfig) - added:
field CoherePreTrainedModel.base_model_prefix - added:
field CoherePreTrainedModel.config: CohereConfig - added:
field CoherePreTrainedModel.supports_gradient_checkpointing - added:
method CohereForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states:… - added:
method CohereModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class CohereForCausalLM(LlamaForCausalLM) - added:
class CohereModel(LlamaModel) - added:
constructor CohereForCausalLM(config) - added:
constructor CohereModel(config: CohereConfig) - added:
method CohereForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states:…
- added:
class CohereTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CohereTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, clean_up_tokenization_spaces=…, unk_token=…, bos_token=…, eos_token=…, add_bos_token=…, add_eos_token=…, use_default_system_prompt=…, add_prefix_space=…, **kwargs) - added:
field CohereTokenizerFast.model_input_names - added:
field CohereTokenizerFast.padding_side - added:
field CohereTokenizerFast.pretrained_vocab_files_map - added:
field CohereTokenizerFast.slow_tokenizer_class - added:
field CohereTokenizerFast.vocab_files_names - added:
method CohereTokenizerFast.apply_grounded_generation_template(conversation: list[dict[str, str]], documents: list[dict], citation_mode: Literal['fast', 'accurate']=…, **kwargs) -> Union[str, list[int]] - added:
method CohereTokenizerFast.apply_tool_use_template(conversation: list[dict[str, str]], tools: list[dict], **kwargs) -> Union[str, list[int]] - added:
method CohereTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method CohereTokenizerFast.update_post_processor() - added:
property CohereTokenizerFast.add_bos_token - added:
property CohereTokenizerFast.add_eos_token
- added:
export define_import_structure
- added:
class Cohere2Config(PretrainedConfig) - added:
constructor Cohere2Config(vocab_size=…, hidden_size=…, intermediate_size=…, logit_scale=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, sliding_… - added:
field Cohere2Config.base_model_pp_plan - added:
field Cohere2Config.base_model_tp_plan - added:
field Cohere2Config.keys_to_ignore_at_inference - added:
field Cohere2Config.model_type
- added:
class Cohere2ForCausalLM(Cohere2PreTrainedModel, GenerationMixin) - added:
class Cohere2Model(Cohere2PreTrainedModel) - added:
class Cohere2PreTrainedModel(PreTrainedModel) - added:
constructor Cohere2ForCausalLM(config) - added:
constructor Cohere2Model(config: Cohere2Config) - added:
field Cohere2PreTrainedModel.base_model_prefix - added:
field Cohere2PreTrainedModel.config: Cohere2Config - added:
field Cohere2PreTrainedModel.supports_gradient_checkpointing - added:
method Cohere2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states… - added:
method Cohere2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Cohere2Config(PretrainedConfig) - added:
class Cohere2ForCausalLM(CohereForCausalLM) - added:
class Cohere2Model(Gemma2Model) - added:
class Cohere2PreTrainedModel(CoherePreTrainedModel) - added:
constructor Cohere2Config(vocab_size=…, hidden_size=…, intermediate_size=…, logit_scale=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, sliding_… - added:
constructor Cohere2Model(config: Cohere2Config) - added:
field Cohere2Config.base_model_pp_plan - added:
field Cohere2Config.base_model_tp_plan - added:
field Cohere2Config.keys_to_ignore_at_inference - added:
field Cohere2Config.model_type - added:
field Cohere2PreTrainedModel.config: Cohere2Config - added:
method Cohere2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
export define_import_structure
- added:
class Cohere2VisionConfig(PretrainedConfig) - added:
constructor Cohere2VisionConfig(vision_config=…, text_config=…, downsample_factor=…, image_token_id=…, alignment_intermediate_size=…, **kwargs) - added:
field Cohere2VisionConfig.model_type - added:
field Cohere2VisionConfig.sub_configs
- added:
class Cohere2VisionImageProcessorFast(BaseImageProcessorFast) - added:
constructor Cohere2VisionImageProcessorFast(**kwargs: Unpack[Cohere2VisionFastImageProcessorKwargs]) - added:
field Cohere2VisionImageProcessorFast.crop_to_patches - added:
field Cohere2VisionImageProcessorFast.do_convert_rgb - added:
field Cohere2VisionImageProcessorFast.do_normalize - added:
field Cohere2VisionImageProcessorFast.do_rescale - added:
field Cohere2VisionImageProcessorFast.do_resize - added:
field Cohere2VisionImageProcessorFast.image_mean - added:
field Cohere2VisionImageProcessorFast.image_std - added:
field Cohere2VisionImageProcessorFast.max_patches - added:
field Cohere2VisionImageProcessorFast.min_patches - added:
field Cohere2VisionImageProcessorFast.patch_size - added:
field Cohere2VisionImageProcessorFast.resample - added:
field Cohere2VisionImageProcessorFast.size - added:
field Cohere2VisionImageProcessorFast.valid_kwargs - added:
method Cohere2VisionImageProcessorFast.crop_image_to_patches(images: torch.Tensor, min_patches: int, max_patches: int, use_thumbnail: bool=…, patch_size: Optional[Union[tuple, int, dict]]=…, interpolation: Optional[F.InterpolationMode]=…) - added:
method Cohere2VisionImageProcessorFast.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Cohere2VisionImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Cohere2VisionFastImageProcessorKwargs]) -> BatchFeature
- added:
class Cohere2VisionForConditionalGeneration(Cohere2VisionPreTrainedModel, GenerationMixin) - added:
class Cohere2VisionModel(Cohere2VisionPreTrainedModel) - added:
class Cohere2VisionPreTrainedModel(PreTrainedModel) - added:
constructor Cohere2VisionForConditionalGeneration(config: Cohere2VisionConfig) - added:
constructor Cohere2VisionModel(config: Cohere2VisionConfig) - added:
field Cohere2VisionPreTrainedModel.base_model_prefix - added:
field Cohere2VisionPreTrainedModel.config: Cohere2VisionConfig - added:
field Cohere2VisionPreTrainedModel.supports_gradient_checkpointing - added:
method Cohere2VisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[… - added:
method Cohere2VisionForConditionalGeneration.get_decoder() - added:
method Cohere2VisionForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor) - added:
method Cohere2VisionForConditionalGeneration.get_input_embeddings() - added:
method Cohere2VisionForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Cohere2VisionForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Cohere2VisionForConditionalGeneration.set_decoder(decoder) - added:
method Cohere2VisionForConditionalGeneration.set_input_embeddings(value) - added:
method Cohere2VisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[FlashAttentionKwarg… - added:
method Cohere2VisionModel.get_decoder() - added:
method Cohere2VisionModel.get_image_features(pixel_values: torch.FloatTensor) - added:
method Cohere2VisionModel.get_input_embeddings() - added:
method Cohere2VisionModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Cohere2VisionModel.set_decoder(decoder) - added:
method Cohere2VisionModel.set_input_embeddings(value) - added:
property Cohere2VisionForConditionalGeneration.language_model - added:
property Cohere2VisionForConditionalGeneration.multi_modal_projector - added:
property Cohere2VisionForConditionalGeneration.vision_tower
- added:
class Cohere2VisionForConditionalGeneration(AyaVisionForConditionalGeneration) - added:
class Cohere2VisionImageProcessorFast(GotOcr2ImageProcessorFast) - added:
class Cohere2VisionModel(AyaVisionModel) - added:
field Cohere2VisionImageProcessorFast.crop_to_patches - added:
field Cohere2VisionImageProcessorFast.max_patches - added:
field Cohere2VisionImageProcessorFast.min_patches - added:
field Cohere2VisionImageProcessorFast.patch_size - added:
field Cohere2VisionImageProcessorFast.size - added:
method Cohere2VisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[… - added:
method Cohere2VisionForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor) - added:
method Cohere2VisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[FlashAttentionKwarg… - added:
method Cohere2VisionModel.get_image_features(pixel_values: torch.FloatTensor)
- added:
class Cohere2VisionProcessor(ProcessorMixin) - added:
constructor Cohere2VisionProcessor(image_processor=…, tokenizer=…, chat_template=…, **kwargs) - added:
field Cohere2VisionProcessor.attributes - added:
field Cohere2VisionProcessor.image_processor_class - added:
field Cohere2VisionProcessor.tokenizer_class - added:
method Cohere2VisionProcessor.batch_decode(*args, **kwargs) - added:
method Cohere2VisionProcessor.decode(*args, **kwargs) - added:
property Cohere2VisionProcessor.model_input_names
- added:
export define_import_structure
- added:
class ColPaliConfig(PretrainedConfig) - added:
constructor ColPaliConfig(vlm_config=…, text_config=…, embedding_dim: int=…, **kwargs) - added:
field ColPaliConfig.model_type - added:
field ColPaliConfig.sub_configs
- added:
class ColPaliForRetrieval(ColPaliPreTrainedModel) - added:
class ColPaliPreTrainedModel(PreTrainedModel) - added:
constructor ColPaliForRetrieval(config: ColPaliConfig) - added:
field ColPaliPreTrainedModel.base_model_prefix - added:
field ColPaliPreTrainedModel.config: ColPaliConfig - added:
method ColPaliForRetrieval.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> ColPaliForRetrievalOutput - added:
method ColPaliForRetrieval.get_input_embeddings() - added:
method ColPaliForRetrieval.get_output_embeddings() - added:
method ColPaliForRetrieval.resize_token_embeddings(new_num_tokens: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method ColPaliForRetrieval.set_input_embeddings(value) - added:
method ColPaliForRetrieval.set_output_embeddings(new_embeddings) - added:
method ColPaliForRetrieval.tie_weights()
- added:
class ColPaliProcessor(PaliGemmaProcessor) - added:
constructor ColPaliProcessor(image_processor=…, tokenizer=…, chat_template=…, visual_prompt_prefix: str=…, query_prefix: str=…) - added:
method ColPaliProcessor.process_images(images: Optional[ImageInput]=…, **kwargs: Unpack[ColPaliProcessorKwargs]) -> BatchFeature - added:
method ColPaliProcessor.process_queries(text: Union[TextInput, list[TextInput]], **kwargs: Unpack[ColPaliProcessorKwargs]) -> BatchFeature - added:
method ColPaliProcessor.score_retrieval(query_embeddings: Union[torch.Tensor, list[torch.Tensor]], passage_embeddings: Union[torch.Tensor, list[torch.Tensor]], batch_size: int=…, output_dtype: Optional[torch.dtype]=…, output_device: Union[torch.device, str]=…) -> torch.Tensor - added:
property ColPaliProcessor.query_augmentation_token -> str
- added:
class ColPaliProcessor(ProcessorMixin) - added:
constructor ColPaliProcessor(image_processor=…, tokenizer=…, chat_template=…, visual_prompt_prefix: str=…, query_prefix: str=…) - added:
field ColPaliProcessor.attributes - added:
field ColPaliProcessor.image_processor_class - added:
field ColPaliProcessor.tokenizer_class - added:
method ColPaliProcessor.process_images(images: Optional[ImageInput]=…, **kwargs: Unpack[ColPaliProcessorKwargs]) -> BatchFeature - added:
method ColPaliProcessor.process_queries(text: Union[TextInput, list[TextInput]], **kwargs: Unpack[ColPaliProcessorKwargs]) -> BatchFeature - added:
method ColPaliProcessor.score_retrieval(query_embeddings: Union[torch.Tensor, list[torch.Tensor]], passage_embeddings: Union[torch.Tensor, list[torch.Tensor]], batch_size: int=…, output_dtype: Optional[torch.dtype]=…, output_device: Union[torch.device, str]=…) -> torch.Tensor - added:
property ColPaliProcessor.query_augmentation_token -> str
- added:
export define_import_structure
- added:
class ColQwen2Config(PretrainedConfig) - added:
constructor ColQwen2Config(vlm_config=…, embedding_dim: int=…, initializer_range: float=…, **kwargs) - added:
field ColQwen2Config.model_type - added:
field ColQwen2Config.sub_configs: dict[str, Any] - added:
method ColQwen2Config.get_text_config(*args, **kwargs) -> PretrainedConfig
- added:
class ColQwen2ForRetrieval(ColQwen2PreTrainedModel) - added:
class ColQwen2PreTrainedModel(PreTrainedModel) - added:
constructor ColQwen2ForRetrieval(config: ColQwen2Config) - added:
field ColQwen2PreTrainedModel.base_model_prefix - added:
field ColQwen2PreTrainedModel.config: ColQwen2Config - added:
method ColQwen2ForRetrieval.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, labels: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic… - added:
method ColQwen2ForRetrieval.get_input_embeddings() - added:
method ColQwen2ForRetrieval.get_output_embeddings() - added:
method ColQwen2ForRetrieval.resize_token_embeddings(new_num_tokens: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method ColQwen2ForRetrieval.set_input_embeddings(value) - added:
method ColQwen2ForRetrieval.set_output_embeddings(new_embeddings) - added:
method ColQwen2ForRetrieval.tie_weights()
- added:
class ColQwen2ForRetrieval(ColPaliForRetrieval) - added:
class ColQwen2PreTrainedModel(ColPaliPreTrainedModel) - added:
class ColQwen2Processor(ColPaliProcessor) - added:
constructor ColQwen2ForRetrieval(config: ColQwen2Config) - added:
constructor ColQwen2Processor(image_processor=…, tokenizer=…, chat_template=…, visual_prompt_prefix: Optional[str]=…, query_prefix: Optional[str]=…, **kwargs) - added:
field ColQwen2Processor.image_processor_class - added:
field ColQwen2Processor.tokenizer_class - added:
method ColQwen2ForRetrieval.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, labels: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic… - added:
property ColQwen2Processor.model_input_names
- added:
class ColQwen2Processor(ProcessorMixin) - added:
constructor ColQwen2Processor(image_processor=…, tokenizer=…, chat_template=…, visual_prompt_prefix: Optional[str]=…, query_prefix: Optional[str]=…, **kwargs) - added:
field ColQwen2Processor.attributes - added:
field ColQwen2Processor.image_processor_class - added:
field ColQwen2Processor.tokenizer_class - added:
method ColQwen2Processor.process_images(images: Optional[ImageInput]=…, **kwargs: Unpack[ColQwen2ProcessorKwargs]) -> BatchFeature - added:
method ColQwen2Processor.process_queries(text: Union[TextInput, list[TextInput]], **kwargs: Unpack[ColQwen2ProcessorKwargs]) -> BatchFeature - added:
method ColQwen2Processor.score_retrieval(query_embeddings: Union[torch.Tensor, list[torch.Tensor]], passage_embeddings: Union[torch.Tensor, list[torch.Tensor]], batch_size: int=…, output_dtype: Optional[torch.dtype]=…, output_device: Union[torch.device, str]=…) -> torch.Tensor - added:
property ColQwen2Processor.model_input_names - added:
property ColQwen2Processor.query_augmentation_token -> str
- added:
export define_import_structure
- added:
class ConditionalDetrConfig(PretrainedConfig) - added:
class ConditionalDetrOnnxConfig(OnnxConfig) - added:
constructor ConditionalDetrConfig(use_timm_backbone=…, backbone_config=…, num_channels=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, in… - added:
field ConditionalDetrConfig.attribute_map - added:
field ConditionalDetrConfig.keys_to_ignore_at_inference - added:
field ConditionalDetrConfig.model_type - added:
field ConditionalDetrOnnxConfig.torch_onnx_minimum_version - added:
property ConditionalDetrConfig.hidden_size -> int - added:
property ConditionalDetrConfig.num_attention_heads -> int - added:
property ConditionalDetrConfig.sub_configs - added:
property ConditionalDetrOnnxConfig.atol_for_validation -> float - added:
property ConditionalDetrOnnxConfig.default_onnx_opset -> int - added:
property ConditionalDetrOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ConditionalDetrFeatureExtractor(ConditionalDetrImageProcessor) - added:
constructor ConditionalDetrFeatureExtractor(*args, **kwargs) -> None
- added:
class ConditionalDetrImageProcessor(BaseImageProcessor) - added:
constructor ConditionalDetrImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: Optional[bool]=…, do_pad: bo… - added:
field ConditionalDetrImageProcessor.model_input_names - added:
method ConditionalDetrImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method ConditionalDetrImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_si… - added:
method ConditionalDetrImageProcessor.post_process(outputs, target_sizes) - added:
method ConditionalDetrImageProcessor.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) -> list[dict] - added:
method ConditionalDetrImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method ConditionalDetrImageProcessor.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method ConditionalDetrImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) - added:
method ConditionalDetrImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method ConditionalDetrImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize:… - added:
method ConditionalDetrImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method ConditionalDetrImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method ConditionalDetrImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict - added:
static method ConditionalDetrImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class ConditionalDetrImageProcessorFast(BaseImageProcessorFast) - added:
constructor ConditionalDetrImageProcessorFast(**kwargs: Unpack[ConditionalDetrFastImageProcessorKwargs]) -> None - added:
field ConditionalDetrImageProcessorFast.default_to_square - added:
field ConditionalDetrImageProcessorFast.do_normalize - added:
field ConditionalDetrImageProcessorFast.do_pad - added:
field ConditionalDetrImageProcessorFast.do_rescale - added:
field ConditionalDetrImageProcessorFast.do_resize - added:
field ConditionalDetrImageProcessorFast.format - added:
field ConditionalDetrImageProcessorFast.image_mean - added:
field ConditionalDetrImageProcessorFast.image_std - added:
field ConditionalDetrImageProcessorFast.model_input_names - added:
field ConditionalDetrImageProcessorFast.resample - added:
field ConditionalDetrImageProcessorFast.size - added:
field ConditionalDetrImageProcessorFast.valid_kwargs - added:
method ConditionalDetrImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method ConditionalDetrImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method ConditionalDetrImageProcessorFast.post_process(outputs, target_sizes) - added:
method ConditionalDetrImageProcessorFast.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) -> list[dict] - added:
method ConditionalDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method ConditionalDetrImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method ConditionalDetrImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) - added:
method ConditionalDetrImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method ConditionalDetrImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[ConditionalDetrFastImageProcessorKwargs]) -> BatchFeature - added:
method ConditionalDetrImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method ConditionalDetrImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…) - added:
static method ConditionalDetrImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class ConditionalDetrForObjectDetection(ConditionalDetrPreTrainedModel) - added:
class ConditionalDetrForSegmentation(ConditionalDetrPreTrainedModel) - added:
class ConditionalDetrModel(ConditionalDetrPreTrainedModel) - added:
class ConditionalDetrPreTrainedModel(PreTrainedModel) - added:
constructor ConditionalDetrForObjectDetection(config: ConditionalDetrConfig) - added:
constructor ConditionalDetrForSegmentation(config: ConditionalDetrConfig) - added:
constructor ConditionalDetrModel(config: ConditionalDetrConfig) - added:
field ConditionalDetrPreTrainedModel.base_model_prefix - added:
field ConditionalDetrPreTrainedModel.config: ConditionalDetrConfig - added:
field ConditionalDetrPreTrainedModel.main_input_name - added:
method ConditionalDetrForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, outp… - added:
method ConditionalDetrForSegmentation.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output… - added:
method ConditionalDetrModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_di… - added:
method ConditionalDetrModel.freeze_backbone() - added:
method ConditionalDetrModel.get_encoder() - added:
method ConditionalDetrModel.unfreeze_backbone()
- added:
class ConditionalDetrImageProcessorFast(DetrImageProcessorFast) - added:
method ConditionalDetrImageProcessorFast.post_process(outputs, target_sizes) - added:
method ConditionalDetrImageProcessorFast.post_process_instance() - added:
method ConditionalDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method ConditionalDetrImageProcessorFast.post_process_panoptic() - added:
method ConditionalDetrImageProcessorFast.post_process_segmentation()
- added:
export define_import_structure
- added:
class ConvBertConfig(PretrainedConfig) - added:
class ConvBertOnnxConfig(OnnxConfig) - added:
constructor ConvBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, embedding_size=…, head_ratio=…, conv_kernel_size=…, num_groups=…, classifier… - added:
field ConvBertConfig.model_type - added:
property ConvBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ConvBertForMaskedLM(ConvBertPreTrainedModel) - added:
class ConvBertForMultipleChoice(ConvBertPreTrainedModel) - added:
class ConvBertForQuestionAnswering(ConvBertPreTrainedModel) - added:
class ConvBertForSequenceClassification(ConvBertPreTrainedModel) - added:
class ConvBertForTokenClassification(ConvBertPreTrainedModel) - added:
class ConvBertLayer(GradientCheckpointingLayer) - added:
class ConvBertModel(ConvBertPreTrainedModel) - added:
class ConvBertPreTrainedModel(PreTrainedModel) - added:
constructor ConvBertForMaskedLM(config) - added:
constructor ConvBertForMultipleChoice(config) - added:
constructor ConvBertForQuestionAnswering(config) - added:
constructor ConvBertForSequenceClassification(config) - added:
constructor ConvBertForTokenClassification(config) - added:
constructor ConvBertLayer(config) - added:
constructor ConvBertModel(config) - added:
field ConvBertPreTrainedModel.base_model_prefix - added:
field ConvBertPreTrainedModel.config: ConvBertConfig - added:
field ConvBertPreTrainedModel.load_tf_weights - added:
field ConvBertPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_convbert(model, config, tf_checkpoint_path) - added:
method ConvBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: O… - added:
method ConvBertForMaskedLM.get_output_embeddings() - added:
method ConvBertForMaskedLM.set_output_embeddings(word_embeddings) - added:
method ConvBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_sta… - added:
method ConvBertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=… - added:
method ConvBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hi… - added:
method ConvBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidde… - added:
method ConvBertLayer.feed_forward_chunk(attention_output) - added:
method ConvBertLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…) -> tuple[torch.Tensor, Optional[torch.FloatTensor]] - added:
method ConvBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]… - added:
method ConvBertModel.get_input_embeddings() - added:
method ConvBertModel.set_input_embeddings(value)
- added:
class TFConvBertForMaskedLM(TFConvBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFConvBertForMultipleChoice(TFConvBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFConvBertForQuestionAnswering(TFConvBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFConvBertForSequenceClassification(TFConvBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFConvBertForTokenClassification(TFConvBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFConvBertLayer(keras.layers.Layer) - added:
class TFConvBertModel(TFConvBertPreTrainedModel) - added:
class TFConvBertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFConvBertForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFConvBertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFConvBertForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFConvBertForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFConvBertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFConvBertLayer(config, **kwargs) - added:
constructor TFConvBertModel(config, *inputs, **kwargs) - added:
field TFConvBertPreTrainedModel.base_model_prefix - added:
field TFConvBertPreTrainedModel.config_class - added:
method TFConvBertForMaskedLM.build(input_shape=…) - added:
method TFConvBertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labe… - added:
method TFConvBertForMaskedLM.get_lm_head() - added:
method TFConvBertForMaskedLM.get_prefix_bias_name() - added:
method TFConvBertForMultipleChoice.build(input_shape=…) - added:
method TFConvBertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…… - added:
method TFConvBertForQuestionAnswering.build(input_shape=…) - added:
method TFConvBertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | Non… - added:
method TFConvBertForSequenceClassification.build(input_shape=…) - added:
method TFConvBertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFConvBertForTokenClassification.build(input_shape=…) - added:
method TFConvBertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | N… - added:
method TFConvBertLayer.build(input_shape=…) - added:
method TFConvBertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFConvBertModel.build(input_shape=…) - added:
method TFConvBertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.array | tf.Tensor | None=…, token_type_ids: np.array | tf.Tensor | None=…, position_ids: np.array | tf.Tensor | None=…, head_mask: np.array | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…)…
- added:
class ConvBertTokenizer(PreTrainedTokenizer) - added:
constructor ConvBertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field ConvBertTokenizer.vocab_files_names - added:
method ConvBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ConvBertTokenizer.convert_tokens_to_string(tokens) - added:
method ConvBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method ConvBertTokenizer.get_vocab() - added:
method ConvBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ConvBertTokenizer.do_lower_case - added:
property ConvBertTokenizer.vocab_size
- added:
class ConvBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor ConvBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field ConvBertTokenizerFast.slow_tokenizer_class - added:
field ConvBertTokenizerFast.vocab_files_names - added:
method ConvBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method ConvBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class ConvNextConfig(BackboneConfigMixin, PretrainedConfig) - added:
class ConvNextOnnxConfig(OnnxConfig) - added:
constructor ConvNextConfig(num_channels=…, patch_size=…, num_stages=…, hidden_sizes=…, depths=…, hidden_act=…, initializer_range=…, layer_norm_eps=…, layer_scale_init_value=…, drop_path_rate=…, image_size=…, out_features=…, out_indices=…, **kwargs) - added:
field ConvNextConfig.model_type - added:
field ConvNextOnnxConfig.torch_onnx_minimum_version - added:
property ConvNextOnnxConfig.atol_for_validation -> float - added:
property ConvNextOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ConvNextFeatureExtractor(ConvNextImageProcessor) - added:
constructor ConvNextFeatureExtractor(*args, **kwargs) -> None
- added:
class ConvNextImageProcessor(BaseImageProcessor) - added:
constructor ConvNextImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, crop_pct: Optional[float]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field ConvNextImageProcessor.model_input_names - added:
method ConvNextImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, crop_pct: Optional[float]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return… - added:
method ConvNextImageProcessor.resize(image: np.ndarray, size: dict[str, int], crop_pct: float, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class ConvNextImageProcessorFast(BaseImageProcessorFast) - added:
constructor ConvNextImageProcessorFast(**kwargs: Unpack[ConvNextFastImageProcessorKwargs]) - added:
field ConvNextImageProcessorFast.crop_pct - added:
field ConvNextImageProcessorFast.default_to_square - added:
field ConvNextImageProcessorFast.do_normalize - added:
field ConvNextImageProcessorFast.do_rescale - added:
field ConvNextImageProcessorFast.do_resize - added:
field ConvNextImageProcessorFast.image_mean - added:
field ConvNextImageProcessorFast.image_std - added:
field ConvNextImageProcessorFast.resample - added:
field ConvNextImageProcessorFast.size - added:
field ConvNextImageProcessorFast.valid_kwargs - added:
method ConvNextImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[ConvNextFastImageProcessorKwargs]) -> BatchFeature - added:
method ConvNextImageProcessorFast.resize(image: torch.Tensor, size: dict[str, int], crop_pct: float, interpolation: PILImageResampling=…, **kwargs) -> torch.Tensor
- added:
class ConvNextBackbone(ConvNextPreTrainedModel, BackboneMixin) - added:
class ConvNextForImageClassification(ConvNextPreTrainedModel) - added:
class ConvNextModel(ConvNextPreTrainedModel) - added:
class ConvNextPreTrainedModel(PreTrainedModel) - added:
constructor ConvNextBackbone(config) - added:
constructor ConvNextForImageClassification(config) - added:
constructor ConvNextModel(config) - added:
field ConvNextBackbone.has_attentions - added:
field ConvNextForImageClassification.accepts_loss_kwargs - added:
field ConvNextPreTrainedModel.base_model_prefix - added:
field ConvNextPreTrainedModel.config: ConvNextConfig - added:
field ConvNextPreTrainedModel.main_input_name - added:
method ConvNextBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…) -> BackboneOutput - added:
method ConvNextForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, **kwargs) -> ImageClassifierOutputWithNoAttention - added:
method ConvNextModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
class TFConvNextForImageClassification(TFConvNextPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFConvNextModel(TFConvNextPreTrainedModel) - added:
class TFConvNextPreTrainedModel(TFPreTrainedModel) - added:
constructor TFConvNextForImageClassification(config: ConvNextConfig, *inputs, **kwargs) - added:
constructor TFConvNextModel(config, *inputs, add_pooling_layer=…, **kwargs) - added:
field TFConvNextPreTrainedModel.base_model_prefix - added:
field TFConvNextPreTrainedModel.config_class - added:
field TFConvNextPreTrainedModel.main_input_name - added:
method TFConvNextForImageClassification.build(input_shape=…) - added:
method TFConvNextForImageClassification.call(pixel_values: TFModelInputType | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFSequenceClassifierOutput | tuple[tf.Tensor] - added:
method TFConvNextModel.build(input_shape=…) - added:
method TFConvNextModel.call(pixel_values: TFModelInputType | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor]
- added:
export define_import_structure
- added:
class ConvNextV2Config(BackboneConfigMixin, PretrainedConfig) - added:
constructor ConvNextV2Config(num_channels=…, patch_size=…, num_stages=…, hidden_sizes=…, depths=…, hidden_act=…, initializer_range=…, layer_norm_eps=…, drop_path_rate=…, image_size=…, out_features=…, out_indices=…, **kwargs) - added:
field ConvNextV2Config.model_type
- added:
class ConvNextV2Backbone(ConvNextV2PreTrainedModel, BackboneMixin) - added:
class ConvNextV2ForImageClassification(ConvNextV2PreTrainedModel) - added:
class ConvNextV2Model(ConvNextV2PreTrainedModel) - added:
class ConvNextV2PreTrainedModel(PreTrainedModel) - added:
constructor ConvNextV2Backbone(config) - added:
constructor ConvNextV2ForImageClassification(config) - added:
constructor ConvNextV2Model(config) - added:
field ConvNextV2Backbone.has_attentions - added:
field ConvNextV2ForImageClassification.accepts_loss_kwargs - added:
field ConvNextV2PreTrainedModel.base_model_prefix - added:
field ConvNextV2PreTrainedModel.config: ConvNextV2Config - added:
field ConvNextV2PreTrainedModel.main_input_name - added:
method ConvNextV2Backbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…) -> BackboneOutput - added:
method ConvNextV2ForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, **kwargs) -> ImageClassifierOutputWithNoAttention - added:
method ConvNextV2Model.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
class TFConvNextV2ForImageClassification(TFConvNextV2PreTrainedModel, TFSequenceClassificationLoss) - added:
class TFConvNextV2Model(TFConvNextV2PreTrainedModel) - added:
class TFConvNextV2PreTrainedModel(TFPreTrainedModel) - added:
constructor TFConvNextV2ForImageClassification(config: ConvNextV2Config, *inputs, **kwargs) - added:
constructor TFConvNextV2Model(config: ConvNextV2Config, *inputs, **kwargs) - added:
field TFConvNextV2PreTrainedModel.base_model_prefix - added:
field TFConvNextV2PreTrainedModel.config_class - added:
field TFConvNextV2PreTrainedModel.main_input_name - added:
method TFConvNextV2ForImageClassification.build(input_shape=…) - added:
method TFConvNextV2ForImageClassification.call(pixel_values: TFModelInputType | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFImageClassifierOutputWithNoAttention | tuple[tf.Tensor] - added:
method TFConvNextV2Model.build(input_shape=…) - added:
method TFConvNextV2Model.call(pixel_values: TFModelInputType | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutputWithPoolingAndNoAttention | tuple[tf.Tensor]
- added:
export define_import_structure
- added:
class CpmTokenizer(PreTrainedTokenizer) - added:
constructor CpmTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field CpmTokenizer.vocab_files_names - added:
method CpmTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CpmTokenizer.convert_tokens_to_string(tokens) - added:
method CpmTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CpmTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CpmTokenizer.get_vocab() - added:
method CpmTokenizer.preprocess_text(inputs) - added:
method CpmTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property CpmTokenizer.vocab_size
- added:
class CpmTokenizerFast(PreTrainedTokenizerFast) - added:
constructor CpmTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - added:
method CpmTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CpmTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CpmTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class CpmAntConfig(PretrainedConfig) - added:
constructor CpmAntConfig(vocab_size: int=…, hidden_size: int=…, num_attention_heads: int=…, dim_head: int=…, dim_ff: int=…, num_hidden_layers: int=…, dropout_p: int=…, position_bias_num_buckets: int=…, position_bias_max_distance: int=…, eps: int=…, init_std: float=…, prompt_types: int=…, prompt_length: int=…, segment_types: int=…, use_cache: bool=…, **kwargs) - added:
field CpmAntConfig.model_type
- added:
class CpmAntForCausalLM(CpmAntPreTrainedModel, GenerationMixin) - added:
class CpmAntModel(CpmAntPreTrainedModel) - added:
class CpmAntPreTrainedModel(PreTrainedModel) - added:
constructor CpmAntForCausalLM(config: CpmAntConfig) - added:
constructor CpmAntModel(config: CpmAntConfig) - added:
field CpmAntPreTrainedModel.base_model_prefix - added:
field CpmAntPreTrainedModel.config: CpmAntConfig - added:
method CpmAntForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, attention_mask: Optional[torch.Tensor]=…, cache_position: Optional[torch.Tensor]=…, **kwargs) -> Union[tuple, CausalLMOutp… - added:
method CpmAntForCausalLM.get_input_embeddings() - added:
method CpmAntForCausalLM.set_input_embeddings(embeddings) - added:
method CpmAntModel.forward(input_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…, **kwargs) -> Union[tuple[torch.Tensor], BaseModelOutputWithPast] - added:
method CpmAntModel.get_input_embeddings() - added:
method CpmAntModel.set_input_embeddings(embeddings, **kwargs)
- added:
class CpmAntTokenizer(PreTrainedTokenizer) - added:
constructor CpmAntTokenizer(vocab_file, bod_token=…, eod_token=…, bos_token=…, eos_token=…, pad_token=…, unk_token=…, line_token=…, space_token=…, padding_side=…, **kwargs) - added:
field CpmAntTokenizer.add_prefix_space - added:
field CpmAntTokenizer.model_input_names - added:
field CpmAntTokenizer.vocab_files_names - added:
method CpmAntTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method CpmAntTokenizer.check(token) - added:
method CpmAntTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method CpmAntTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method CpmAntTokenizer.get_vocab() - added:
method CpmAntTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property CpmAntTokenizer.bod_token_id - added:
property CpmAntTokenizer.eod_token_id - added:
property CpmAntTokenizer.newline_id - added:
property CpmAntTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class CsmConfig(PretrainedConfig) - added:
class CsmDepthDecoderConfig(PretrainedConfig) - added:
constructor CsmConfig(num_codebooks=…, vocab_size=…, text_vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, codebook_pad_token_id=…, codebook_eos_token_id=…, bos_token_id=…, eos_token_id=…, audio_token_id=…, audio_eos_token_i… - added:
constructor CsmDepthDecoderConfig(num_codebooks=…, backbone_hidden_size=…, vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…,… - added:
field CsmConfig.base_config_key - added:
field CsmConfig.keys_to_ignore_at_inference - added:
field CsmConfig.model_type - added:
field CsmConfig.sub_configs - added:
field CsmDepthDecoderConfig.base_config_key - added:
field CsmDepthDecoderConfig.keys_to_ignore_at_inference - added:
field CsmDepthDecoderConfig.model_type
- added:
class CsmGenerateOutput(GenerateDecoderOnlyOutput) - added:
class CsmGenerationMixin(GenerationMixin) - added:
field CsmGenerateOutput.audio: Optional[list[torch.Tensor]] - added:
method CsmGenerationMixin.generate(input_ids: Optional[torch.Tensor]=…, input_values: Optional[torch.Tensor]=…, input_values_cutoffs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, synced_gpus: Optional[bool]=…, streamer: Optional[BaseStreamer]=…, output_audio: Optio… - added:
variable logger
- added:
class CsmBackboneModel(CsmPreTrainedModel) - added:
class CsmDepthDecoderForCausalLM(CsmPreTrainedModel, GenerationMixin) - added:
class CsmDepthDecoderModel(CsmPreTrainedModel) - added:
class CsmForConditionalGeneration(CsmPreTrainedModel, CsmGenerationMixin) - added:
class CsmPreTrainedModel(PreTrainedModel) - added:
constructor CsmBackboneModel(config) - added:
constructor CsmDepthDecoderForCausalLM(config) - added:
constructor CsmDepthDecoderModel(config) - added:
constructor CsmForConditionalGeneration(config) - added:
field CsmDepthDecoderModel.config: CsmDepthDecoderConfig - added:
field CsmPreTrainedModel.base_model_prefix - added:
field CsmPreTrainedModel.config: CsmConfig - added:
field CsmPreTrainedModel.supports_gradient_checkpointing - added:
method CsmBackboneModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method CsmDepthDecoderForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, backbone_last_hidden_state: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional… - added:
method CsmDepthDecoderForCausalLM.prepare_inputs_for_generation(input_ids: torch.LongTensor, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) - added:
method CsmDepthDecoderModel.forward(input_ids: Optional[torch.LongTensor]=…, backbone_last_hidden_state: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[Tra… - added:
method CsmForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_values: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, input_values_cutoffs: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.Long… - added:
method CsmForConditionalGeneration.get_input_embeddings() - added:
method CsmForConditionalGeneration.prepare_inputs_for_generation(input_ids: torch.LongTensor, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) - added:
method CsmForConditionalGeneration.save_pretrained(*args, **kwargs) - added:
method CsmForConditionalGeneration.set_input_embeddings(value) - added:
static method CsmForConditionalGeneration.from_pretrained(*args, **kwargs)
- added:
class CsmBackboneModel(LlamaModel) - added:
class CsmDepthDecoderForCausalLM(LlamaForCausalLM, GenerationMixin) - added:
class CsmDepthDecoderModel(LlamaModel, CsmPreTrainedModel) - added:
class CsmForConditionalGeneration(CsmPreTrainedModel, CsmGenerationMixin) - added:
class CsmPreTrainedModel(PreTrainedModel) - added:
constructor CsmBackboneModel(config) - added:
constructor CsmDepthDecoderForCausalLM(config) - added:
constructor CsmDepthDecoderModel(config) - added:
constructor CsmForConditionalGeneration(config) - added:
field CsmDepthDecoderModel.config: CsmDepthDecoderConfig - added:
field CsmPreTrainedModel.base_model_prefix - added:
field CsmPreTrainedModel.config: CsmConfig - added:
field CsmPreTrainedModel.supports_gradient_checkpointing - added:
method CsmBackboneModel.forward(**super_kwargs) - added:
method CsmDepthDecoderForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, backbone_last_hidden_state: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional… - added:
method CsmDepthDecoderForCausalLM.prepare_inputs_for_generation(input_ids: torch.LongTensor, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) - added:
method CsmDepthDecoderModel.forward(input_ids: Optional[torch.LongTensor]=…, backbone_last_hidden_state: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[Tra… - added:
method CsmForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_values: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, input_values_cutoffs: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.Long… - added:
method CsmForConditionalGeneration.get_input_embeddings() - added:
method CsmForConditionalGeneration.prepare_inputs_for_generation(input_ids: torch.LongTensor, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) - added:
method CsmForConditionalGeneration.save_pretrained(*args, **kwargs) - added:
method CsmForConditionalGeneration.set_input_embeddings(value) - added:
static method CsmForConditionalGeneration.from_pretrained(*args, **kwargs)
- added:
class CsmProcessor(ProcessorMixin) - added:
constructor CsmProcessor(feature_extractor, tokenizer, chat_template=…) - added:
field CsmProcessor.attributes - added:
field CsmProcessor.feature_extractor_class - added:
field CsmProcessor.tokenizer_class - added:
method CsmProcessor.save_audio(audio: AudioInput, saving_path: Union[str, Path, list[Union[str, Path]]], **kwargs: Unpack[CsmProcessorKwargs]) - added:
property CsmProcessor.model_input_names
- added:
export define_import_structure
- added:
class CTRLConfig(PretrainedConfig) - added:
constructor CTRLConfig(vocab_size=…, n_positions=…, n_embd=…, dff=…, n_layer=…, n_head=…, resid_pdrop=…, embd_pdrop=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, **kwargs) - added:
field CTRLConfig.attribute_map - added:
field CTRLConfig.keys_to_ignore_at_inference - added:
field CTRLConfig.model_type
- added:
class CTRLForSequenceClassification(CTRLPreTrainedModel) - added:
class CTRLLMHeadModel(CTRLPreTrainedModel, GenerationMixin) - added:
class CTRLModel(CTRLPreTrainedModel) - added:
class CTRLPreTrainedModel(PreTrainedModel) - added:
constructor CTRLForSequenceClassification(config) - added:
constructor CTRLLMHeadModel(config) - added:
constructor CTRLModel(config) - added:
field CTRLPreTrainedModel.base_model_prefix - added:
field CTRLPreTrainedModel.config: CTRLConfig - added:
method CTRLForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Opt… - added:
method CTRLLMHeadModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…,… - added:
method CTRLLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, use_cache=…, **kwargs) - added:
method CTRLModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output… - added:
method CTRLModel.get_input_embeddings() - added:
method CTRLModel.set_input_embeddings(new_embeddings)
- added:
class TFCTRLForSequenceClassification(TFCTRLPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFCTRLLMHeadModel(TFCTRLPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFCTRLModel(TFCTRLPreTrainedModel) - added:
class TFCTRLPreTrainedModel(TFPreTrainedModel) - added:
constructor TFCTRLForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFCTRLLMHeadModel(config, *inputs, **kwargs) - added:
constructor TFCTRLModel(config, *inputs, **kwargs) - added:
field TFCTRLPreTrainedModel.base_model_prefix - added:
field TFCTRLPreTrainedModel.config_class - added:
method TFCTRLForSequenceClassification.build(input_shape=…) - added:
method TFCTRLForSequenceClassification.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool… - added:
method TFCTRLForSequenceClassification.get_output_embeddings() - added:
method TFCTRLLMHeadModel.build(input_shape=…) - added:
method TFCTRLLMHeadModel.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool | None=…, outp… - added:
method TFCTRLLMHeadModel.get_bias() - added:
method TFCTRLLMHeadModel.get_output_embeddings() - added:
method TFCTRLLMHeadModel.prepare_inputs_for_generation(inputs, past_key_values=…, use_cache=…, **kwargs) - added:
method TFCTRLLMHeadModel.set_bias(value) - added:
method TFCTRLLMHeadModel.set_output_embeddings(value) - added:
method TFCTRLModel.build(input_shape=…) - added:
method TFCTRLModel.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool | None=…, output_att…
- added:
class CTRLTokenizer(PreTrainedTokenizer) - added:
constructor CTRLTokenizer(vocab_file, merges_file, unk_token=…, **kwargs) - added:
field CTRLTokenizer.control_codes - added:
field CTRLTokenizer.vocab_files_names - added:
method CTRLTokenizer.bpe(token) - added:
method CTRLTokenizer.convert_tokens_to_string(tokens) - added:
method CTRLTokenizer.get_vocab() - added:
method CTRLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property CTRLTokenizer.vocab_size
- added:
export define_import_structure
- added:
class CvtConfig(PretrainedConfig) - added:
constructor CvtConfig(num_channels=…, patch_sizes=…, patch_stride=…, patch_padding=…, embed_dim=…, num_heads=…, depth=…, mlp_ratio=…, attention_drop_rate=…, drop_rate=…, drop_path_rate=…, qkv_bias=…, cls_token=…, qkv_projection_method=…, kernel_qkv=…, padding_kv=…, stride_kv=…, padding_q=…, stride_q=…, initializer_range=…, layer_norm_eps=…, **kwargs) - added:
field CvtConfig.model_type
- added:
class CvtForImageClassification(CvtPreTrainedModel) - added:
class CvtModel(CvtPreTrainedModel) - added:
class CvtPreTrainedModel(PreTrainedModel) - added:
constructor CvtForImageClassification(config) - added:
constructor CvtModel(config, add_pooling_layer=…) - added:
field CvtPreTrainedModel.base_model_prefix - added:
field CvtPreTrainedModel.config: CvtConfig - added:
field CvtPreTrainedModel.main_input_name - added:
method CvtForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method CvtModel.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithCLSToken]
- added:
class TFCvtForImageClassification(TFCvtPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFCvtModel(TFCvtPreTrainedModel) - added:
class TFCvtPreTrainedModel(TFPreTrainedModel) - added:
constructor TFCvtForImageClassification(config: CvtConfig, *inputs, **kwargs) - added:
constructor TFCvtModel(config: CvtConfig, *inputs, **kwargs) - added:
field TFCvtPreTrainedModel.base_model_prefix - added:
field TFCvtPreTrainedModel.config_class - added:
field TFCvtPreTrainedModel.main_input_name - added:
method TFCvtForImageClassification.build(input_shape=…) - added:
method TFCvtForImageClassification.call(pixel_values: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFImageClassifierOutputWithNoAttention | tuple[tf.Tensor] - added:
method TFCvtModel.build(input_shape=…) - added:
method TFCvtModel.call(pixel_values: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOutputWithCLSToken | tuple[tf.Tensor]
- added:
export define_import_structure
- added:
class DFineConfig(PretrainedConfig) - added:
constructor DFineConfig(initializer_range=…, initializer_bias_prior_prob=…, layer_norm_eps=…, batch_norm_eps=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, freeze_backbone_batch_norms=…, backbone_kwargs=…, encoder_hidden_dim=…, encoder_in_channels=…, feat_strides=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, dropout=…, activation_dropout=…,… - added:
field DFineConfig.attribute_map - added:
field DFineConfig.layer_types - added:
field DFineConfig.model_type - added:
property DFineConfig.hidden_size -> int - added:
property DFineConfig.num_attention_heads -> int - added:
property DFineConfig.sub_configs - added:
static method DFineConfig.from_backbone_configs(backbone_config: PretrainedConfig, **kwargs)
- added:
class DFineForObjectDetection(DFinePreTrainedModel) - added:
class DFineModel(DFinePreTrainedModel) - added:
class DFinePreTrainedModel(PreTrainedModel) - added:
constructor DFineForObjectDetection(config: DFineConfig) - added:
constructor DFineModel(config: DFineConfig) - added:
field DFinePreTrainedModel.base_model_prefix - added:
field DFinePreTrainedModel.config: DFineConfig - added:
field DFinePreTrainedModel.main_input_name - added:
method DFineForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=… - added:
method DFineModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[t… - added:
method DFineModel.freeze_backbone() - added:
method DFineModel.generate_anchors(spatial_shapes=…, grid_size=…, device=…, dtype=…) - added:
method DFineModel.get_encoder() - added:
method DFineModel.unfreeze_backbone()
- added:
class DFineConfig(PretrainedConfig) - added:
class DFineForObjectDetection(RTDetrForObjectDetection, DFinePreTrainedModel) - added:
class DFineModel(RTDetrModel) - added:
class DFinePreTrainedModel(RTDetrPreTrainedModel) - added:
constructor DFineConfig(initializer_range=…, initializer_bias_prior_prob=…, layer_norm_eps=…, batch_norm_eps=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, freeze_backbone_batch_norms=…, backbone_kwargs=…, encoder_hidden_dim=…, encoder_in_channels=…, feat_strides=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, dropout=…, activation_dropout=…,… - added:
constructor DFineForObjectDetection(config: DFineConfig) - added:
constructor DFineModel(config: DFineConfig) - added:
field DFineConfig.attribute_map - added:
field DFineConfig.layer_types - added:
field DFineConfig.model_type - added:
method DFineForObjectDetection.forward(**super_kwargs) - added:
property DFineConfig.hidden_size -> int - added:
property DFineConfig.num_attention_heads -> int - added:
property DFineConfig.sub_configs - added:
static method DFineConfig.from_backbone_configs(backbone_config: PretrainedConfig, **kwargs)
- added:
export define_import_structure
- added:
class DabDetrConfig(PretrainedConfig) - added:
constructor DabDetrConfig(use_timm_backbone=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, backbone_kwargs=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, is_encoder_decoder=…, activation_function=…, hidden_size=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, init_x… - added:
field DabDetrConfig.attribute_map - added:
field DabDetrConfig.keys_to_ignore_at_inference - added:
field DabDetrConfig.model_type - added:
property DabDetrConfig.sub_configs
- added:
class DabDetrForObjectDetection(DabDetrPreTrainedModel) - added:
class DabDetrModel(DabDetrPreTrainedModel) - added:
class DabDetrPreTrainedModel(PreTrainedModel) - added:
constructor DabDetrForObjectDetection(config: DabDetrConfig) - added:
constructor DabDetrModel(config: DabDetrConfig) - added:
field DabDetrPreTrainedModel.base_model_prefix - added:
field DabDetrPreTrainedModel.config: DabDetrConfig - added:
field DabDetrPreTrainedModel.main_input_name - added:
method DabDetrForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidde… - added:
method DabDetrModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Opti… - added:
method DabDetrModel.freeze_backbone() - added:
method DabDetrModel.get_encoder() - added:
method DabDetrModel.unfreeze_backbone()
- added:
export define_import_structure
- added:
class DacConfig(PretrainedConfig) - added:
constructor DacConfig(encoder_hidden_size=…, downsampling_ratios=…, decoder_hidden_size=…, n_codebooks=…, codebook_size=…, codebook_dim=…, quantizer_dropout=…, commitment_loss_weight=…, codebook_loss_weight=…, sampling_rate=…, **kwargs) - added:
field DacConfig.model_type - added:
property DacConfig.frame_rate -> int
- added:
class DacFeatureExtractor(SequenceFeatureExtractor) - added:
constructor DacFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, hop_length: int=…, **kwargs) - added:
field DacFeatureExtractor.model_input_names
- added:
class DacModel(DacPreTrainedModel) - added:
class DacPreTrainedModel(PreTrainedAudioTokenizerBase) - added:
constructor DacModel(config: DacConfig) - added:
field DacPreTrainedModel.base_model_prefix - added:
field DacPreTrainedModel.config: DacConfig - added:
field DacPreTrainedModel.main_input_name - added:
method DacModel.decode(quantized_representation: Optional[torch.Tensor]=…, audio_codes: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) - added:
method DacModel.encode(input_values: torch.Tensor, n_quantizers: Optional[int]=…, return_dict: Optional[bool]=…) - added:
method DacModel.forward(input_values: torch.Tensor, n_quantizers: Optional[int]=…, return_dict: Optional[bool]=…) - added:
method DacPreTrainedModel.apply_weight_norm() - added:
method DacPreTrainedModel.remove_weight_norm()
- added:
export define_import_structure
- added:
class Data2VecAudioConfig(PretrainedConfig) - added:
constructor Data2VecAudioConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel=…, conv_bias=…, num_conv_pos_embedding_gro… - added:
field Data2VecAudioConfig.model_type - added:
property Data2VecAudioConfig.inputs_to_logits_ratio
- added:
class Data2VecTextConfig(PretrainedConfig) - added:
class Data2VecTextOnnxConfig(OnnxConfig) - added:
constructor Data2VecTextConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field Data2VecTextConfig.model_type - added:
property Data2VecTextOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class Data2VecVisionConfig(PretrainedConfig) - added:
class Data2VecVisionOnnxConfig(OnnxConfig) - added:
constructor Data2VecVisionConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, use_mask_token=…, use_absolute_position_embeddings=…, use_relative_position_bias=…, use_shared_relative_position_bias=…, layer_scale_i… - added:
field Data2VecVisionConfig.model_type - added:
field Data2VecVisionOnnxConfig.torch_onnx_minimum_version - added:
property Data2VecVisionOnnxConfig.atol_for_validation -> float - added:
property Data2VecVisionOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class Data2VecAudioForAudioFrameClassification(Data2VecAudioPreTrainedModel) - added:
class Data2VecAudioForCTC(Data2VecAudioPreTrainedModel) - added:
class Data2VecAudioForSequenceClassification(Data2VecAudioPreTrainedModel) - added:
class Data2VecAudioForXVector(Data2VecAudioPreTrainedModel) - added:
class Data2VecAudioModel(Data2VecAudioPreTrainedModel) - added:
class Data2VecAudioPreTrainedModel(PreTrainedModel) - added:
constructor Data2VecAudioForAudioFrameClassification(config) - added:
constructor Data2VecAudioForCTC(config) - added:
constructor Data2VecAudioForSequenceClassification(config) - added:
constructor Data2VecAudioForXVector(config) - added:
constructor Data2VecAudioModel(config: Data2VecAudioConfig) - added:
field Data2VecAudioPreTrainedModel.base_model_prefix - added:
field Data2VecAudioPreTrainedModel.config: Data2VecAudioConfig - added:
field Data2VecAudioPreTrainedModel.main_input_name - added:
field Data2VecAudioPreTrainedModel.supports_gradient_checkpointing - added:
method Data2VecAudioForAudioFrameClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method Data2VecAudioForAudioFrameClassification.freeze_base_model() - added:
method Data2VecAudioForAudioFrameClassification.freeze_feature_encoder() - added:
method Data2VecAudioForAudioFrameClassification.freeze_feature_extractor() - added:
method Data2VecAudioForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method Data2VecAudioForCTC.freeze_feature_encoder() - added:
method Data2VecAudioForCTC.freeze_feature_extractor() - added:
method Data2VecAudioForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method Data2VecAudioForSequenceClassification.freeze_base_model() - added:
method Data2VecAudioForSequenceClassification.freeze_feature_encoder() - added:
method Data2VecAudioForSequenceClassification.freeze_feature_extractor() - added:
method Data2VecAudioForXVector.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method Data2VecAudioForXVector.freeze_base_model() - added:
method Data2VecAudioForXVector.freeze_feature_encoder() - added:
method Data2VecAudioForXVector.freeze_feature_extractor() - added:
method Data2VecAudioModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Data2VecAudioBaseModelOutput] - added:
method Data2VecAudioModel.freeze_feature_encoder()
- added:
class Data2VecTextForCausalLM(Data2VecTextPreTrainedModel, GenerationMixin) - added:
class Data2VecTextForMaskedLM(Data2VecTextPreTrainedModel) - added:
class Data2VecTextForMultipleChoice(Data2VecTextPreTrainedModel) - added:
class Data2VecTextForQuestionAnswering(Data2VecTextPreTrainedModel) - added:
class Data2VecTextForSequenceClassification(Data2VecTextPreTrainedModel) - added:
class Data2VecTextForTokenClassification(Data2VecTextPreTrainedModel) - added:
class Data2VecTextModel(Data2VecTextPreTrainedModel) - added:
class Data2VecTextPreTrainedModel(PreTrainedModel) - added:
constructor Data2VecTextForCausalLM(config) - added:
constructor Data2VecTextForMaskedLM(config) - added:
constructor Data2VecTextForMultipleChoice(config) - added:
constructor Data2VecTextForQuestionAnswering(config) - added:
constructor Data2VecTextForSequenceClassification(config) - added:
constructor Data2VecTextForTokenClassification(config) - added:
constructor Data2VecTextModel(config, add_pooling_layer=…) - added:
field Data2VecTextPreTrainedModel.base_model_prefix - added:
field Data2VecTextPreTrainedModel.config: Data2VecTextConfig - added:
field Data2VecTextPreTrainedModel.supports_gradient_checkpointing - added:
method Data2VecTextForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.F… - added:
method Data2VecTextForCausalLM.get_output_embeddings() - added:
method Data2VecTextForCausalLM.set_output_embeddings(new_embeddings) - added:
method Data2VecTextForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.F… - added:
method Data2VecTextForMaskedLM.get_output_embeddings() - added:
method Data2VecTextForMaskedLM.set_output_embeddings(new_embeddings) - added:
method Data2VecTextForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method Data2VecTextForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTens… - added:
method Data2VecTextForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outpu… - added:
method Data2VecTextForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method Data2VecTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Ca… - added:
method Data2VecTextModel.get_input_embeddings() - added:
method Data2VecTextModel.set_input_embeddings(value)
- added:
class Data2VecVisionForImageClassification(Data2VecVisionPreTrainedModel) - added:
class Data2VecVisionForSemanticSegmentation(Data2VecVisionPreTrainedModel) - added:
class Data2VecVisionModel(Data2VecVisionPreTrainedModel) - added:
class Data2VecVisionPreTrainedModel(PreTrainedModel) - added:
constructor Data2VecVisionForImageClassification(config: Data2VecVisionConfig) -> None - added:
constructor Data2VecVisionForSemanticSegmentation(config: Data2VecVisionConfig) -> None - added:
constructor Data2VecVisionModel(config: Data2VecVisionConfig, add_pooling_layer: bool=…) -> None - added:
field Data2VecVisionPreTrainedModel.base_model_prefix - added:
field Data2VecVisionPreTrainedModel.config: Data2VecVisionConfig - added:
field Data2VecVisionPreTrainedModel.main_input_name - added:
field Data2VecVisionPreTrainedModel.supports_gradient_checkpointing - added:
method Data2VecVisionForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method Data2VecVisionForSemanticSegmentation.compute_loss(logits, auxiliary_logits, labels) - added:
method Data2VecVisionForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method Data2VecVisionModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, Data2VecVisionModelOutputWithPooling] - added:
method Data2VecVisionModel.get_input_embeddings()
- added:
class TFData2VecVisionForImageClassification(TFData2VecVisionPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFData2VecVisionForSemanticSegmentation(TFData2VecVisionPreTrainedModel) - added:
class TFData2VecVisionModel(TFData2VecVisionPreTrainedModel) - added:
class TFData2VecVisionPreTrainedModel(TFPreTrainedModel) - added:
constructor TFData2VecVisionForImageClassification(config: Data2VecVisionConfig, *inputs, **kwargs) - added:
constructor TFData2VecVisionForSemanticSegmentation(config: Data2VecVisionConfig, *inputs, **kwargs) -> None - added:
constructor TFData2VecVisionModel(config: Data2VecVisionConfig, add_pooling_layer: bool=…, *inputs, **kwargs) - added:
field TFData2VecVisionPreTrainedModel.base_model_prefix - added:
field TFData2VecVisionPreTrainedModel.config_class - added:
field TFData2VecVisionPreTrainedModel.main_input_name - added:
method TFData2VecVisionForImageClassification.build(input_shape=…) - added:
method TFData2VecVisionForImageClassification.call(pixel_values: TFModelInputType | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFSequenceClassifierOutput | tuple - added:
method TFData2VecVisionForSemanticSegmentation.build(input_shape=…) - added:
method TFData2VecVisionForSemanticSegmentation.call(pixel_values: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…) -> tuple | TFSemanticSegmenterOutput - added:
method TFData2VecVisionForSemanticSegmentation.compute_loss(logits, auxiliary_logits, labels) - added:
method TFData2VecVisionModel.build(input_shape=…) - added:
method TFData2VecVisionModel.call(pixel_values: TFModelInputType | None=…, bool_masked_pos: tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple | TFData2VecVisionModelOutputWithPooling - added:
method TFData2VecVisionModel.get_input_embeddings()
- added:
class Data2VecAudioForAudioFrameClassification(Wav2Vec2ForAudioFrameClassification) - added:
class Data2VecAudioForCTC(Data2VecAudioPreTrainedModel, Wav2Vec2ForCTC) - added:
class Data2VecAudioForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class Data2VecAudioForXVector(Wav2Vec2ForXVector) - added:
class Data2VecAudioModel(Data2VecAudioPreTrainedModel, Wav2Vec2Model) - added:
class Data2VecAudioPreTrainedModel(PreTrainedModel, Wav2Vec2PreTrainedModel) - added:
constructor Data2VecAudioForCTC(config) - added:
constructor Data2VecAudioModel(config: Data2VecAudioConfig) - added:
field Data2VecAudioPreTrainedModel.base_model_prefix - added:
field Data2VecAudioPreTrainedModel.config: Data2VecAudioConfig - added:
field Data2VecAudioPreTrainedModel.main_input_name - added:
field Data2VecAudioPreTrainedModel.supports_gradient_checkpointing - added:
method Data2VecAudioForCTC.forward(**super_kwargs) - added:
method Data2VecAudioForCTC.freeze_base_model() - added:
method Data2VecAudioForCTC.tie_weights() - added:
method Data2VecAudioModel.forward(**super_kwargs) - added:
method Data2VecAudioModel.freeze_feature_encoder() - added:
method Data2VecAudioModel.freeze_feature_extractor() - added:
method Data2VecAudioPreTrainedModel.init_adapter_layers() - added:
method Data2VecAudioPreTrainedModel.load_adapter()
- added:
export define_import_structure
- added:
class DbrxConfig(PretrainedConfig) - added:
constructor DbrxConfig(d_model: int=…, n_heads: int=…, n_layers: int=…, max_seq_len: int=…, vocab_size: int=…, resid_pdrop: float=…, emb_pdrop: float=…, attn_config: Optional[DbrxAttentionConfig]=…, ffn_config: Optional[DbrxFFNConfig]=…, use_cache: bool=…, initializer_range: float=…, output_router_logits: bool=…, **kwargs: Any) - added:
field DbrxConfig.attribute_map - added:
field DbrxConfig.model_type - added:
field DbrxConfig.sub_configs
- added:
class DbrxForCausalLM(DbrxPreTrainedModel, GenerationMixin) - added:
class DbrxModel(DbrxPreTrainedModel) - added:
class DbrxPreTrainedModel(PreTrainedModel) - added:
constructor DbrxForCausalLM(config: DbrxConfig) - added:
constructor DbrxModel(config: DbrxConfig) - added:
field DbrxPreTrainedModel.base_model_prefix - added:
field DbrxPreTrainedModel.config: DbrxConfig - added:
field DbrxPreTrainedModel.supports_gradient_checkpointing - added:
method DbrxForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits… - added:
method DbrxForCausalLM.get_decoder() -> DbrxModel - added:
method DbrxForCausalLM.get_input_embeddings() -> nn.Embedding - added:
method DbrxForCausalLM.get_output_embeddings() -> nn.Linear - added:
method DbrxForCausalLM.set_decoder(decoder: DbrxModel) - added:
method DbrxForCausalLM.set_input_embeddings(value: nn.Embedding) - added:
method DbrxForCausalLM.set_output_embeddings(new_embeddings: nn.Linear) - added:
method DbrxModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Optional[bool]=…, return_dict: Optional[bo… - added:
method DbrxModel.get_input_embeddings() -> nn.Embedding - added:
method DbrxModel.set_input_embeddings(value: nn.Embedding)
- added:
export define_import_structure
- added:
class DebertaConfig(PretrainedConfig) - added:
class DebertaOnnxConfig(OnnxConfig) - added:
constructor DebertaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, relative_attention=…, max_relative_positions=…, pad_token_id=…, position_biased_input=…, pos_att_type=…, pooler_dropout=…, po… - added:
field DebertaConfig.model_type - added:
method DebertaOnnxConfig.generate_dummy_inputs(preprocessor: Union[PreTrainedTokenizerBase, FeatureExtractionMixin], batch_size: int=…, seq_length: int=…, num_choices: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…, tokenizer: PreTrainedTokenizerBase=…) -> Mapping[str, Any] - added:
property DebertaOnnxConfig.default_onnx_opset -> int - added:
property DebertaOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class DebertaForMaskedLM(DebertaPreTrainedModel) - added:
class DebertaForQuestionAnswering(DebertaPreTrainedModel) - added:
class DebertaForSequenceClassification(DebertaPreTrainedModel) - added:
class DebertaForTokenClassification(DebertaPreTrainedModel) - added:
class DebertaModel(DebertaPreTrainedModel) - added:
class DebertaPreTrainedModel(PreTrainedModel) - added:
constructor DebertaForMaskedLM(config) - added:
constructor DebertaForQuestionAnswering(config) - added:
constructor DebertaForSequenceClassification(config) - added:
constructor DebertaForTokenClassification(config) - added:
constructor DebertaModel(config) - added:
field DebertaPreTrainedModel.base_model_prefix - added:
field DebertaPreTrainedModel.config: DebertaConfig - added:
field DebertaPreTrainedModel.supports_gradient_checkpointing - added:
method DebertaForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Maske… - added:
method DebertaForMaskedLM.get_output_embeddings() - added:
method DebertaForMaskedLM.set_output_embeddings(new_embeddings) - added:
method DebertaForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bo… - added:
method DebertaForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Unio… - added:
method DebertaForSequenceClassification.get_input_embeddings() - added:
method DebertaForSequenceClassification.set_input_embeddings(new_embeddings) - added:
method DebertaForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[t… - added:
method DebertaModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method DebertaModel.get_input_embeddings() - added:
method DebertaModel.set_input_embeddings(new_embeddings)
- added:
class TFDebertaForMaskedLM(TFDebertaPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFDebertaForQuestionAnswering(TFDebertaPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFDebertaForSequenceClassification(TFDebertaPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFDebertaForTokenClassification(TFDebertaPreTrainedModel, TFTokenClassificationLoss) - added:
class TFDebertaModel(TFDebertaPreTrainedModel) - added:
class TFDebertaPreTrainedModel(TFPreTrainedModel) - added:
constructor TFDebertaForMaskedLM(config: DebertaConfig, *inputs, **kwargs) - added:
constructor TFDebertaForQuestionAnswering(config: DebertaConfig, *inputs, **kwargs) - added:
constructor TFDebertaForSequenceClassification(config: DebertaConfig, *inputs, **kwargs) - added:
constructor TFDebertaForTokenClassification(config: DebertaConfig, *inputs, **kwargs) - added:
constructor TFDebertaModel(config: DebertaConfig, *inputs, **kwargs) - added:
field TFDebertaPreTrainedModel.base_model_prefix - added:
field TFDebertaPreTrainedModel.config_class - added:
method TFDebertaForMaskedLM.build(input_shape=…) - added:
method TFDebertaForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | Non… - added:
method TFDebertaForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFDebertaForQuestionAnswering.build(input_shape=…) - added:
method TFDebertaForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: np.ndarray… - added:
method TFDebertaForSequenceClassification.build(input_shape=…) - added:
method TFDebertaForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | t… - added:
method TFDebertaForTokenClassification.build(input_shape=…) - added:
method TFDebertaForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.T… - added:
method TFDebertaModel.build(input_shape=…) - added:
method TFDebertaModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOut…
- added:
class DebertaTokenizer(PreTrainedTokenizer) - added:
constructor DebertaTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, add_bos_token=…, **kwargs) - added:
field DebertaTokenizer.model_input_names - added:
field DebertaTokenizer.vocab_files_names - added:
method DebertaTokenizer.bpe(token) - added:
method DebertaTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method DebertaTokenizer.convert_tokens_to_string(tokens) - added:
method DebertaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method DebertaTokenizer.get_vocab() - added:
method DebertaTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method DebertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property DebertaTokenizer.vocab_size
- added:
class DebertaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor DebertaTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field DebertaTokenizerFast.model_input_names - added:
field DebertaTokenizerFast.slow_tokenizer_class - added:
field DebertaTokenizerFast.vocab_files_names - added:
method DebertaTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method DebertaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property DebertaTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class DebertaV2Config(PretrainedConfig) - added:
class DebertaV2OnnxConfig(OnnxConfig) - added:
constructor DebertaV2Config(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, relative_attention=…, max_relative_positions=…, pad_token_id=…, position_biased_input=…, pos_att_type=…, pooler_dropout=…,… - added:
field DebertaV2Config.model_type - added:
method DebertaV2OnnxConfig.generate_dummy_inputs(preprocessor: Union[PreTrainedTokenizerBase, FeatureExtractionMixin], batch_size: int=…, seq_length: int=…, num_choices: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…, tokenizer: PreTrainedTokenizerBase=…) -> Mapping[str, Any] - added:
property DebertaV2OnnxConfig.default_onnx_opset -> int - added:
property DebertaV2OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class DebertaV2ForMaskedLM(DebertaV2PreTrainedModel) - added:
class DebertaV2ForMultipleChoice(DebertaV2PreTrainedModel) - added:
class DebertaV2ForQuestionAnswering(DebertaV2PreTrainedModel) - added:
class DebertaV2ForSequenceClassification(DebertaV2PreTrainedModel) - added:
class DebertaV2ForTokenClassification(DebertaV2PreTrainedModel) - added:
class DebertaV2Model(DebertaV2PreTrainedModel) - added:
class DebertaV2PreTrainedModel(PreTrainedModel) - added:
constructor DebertaV2ForMaskedLM(config) - added:
constructor DebertaV2ForMultipleChoice(config) - added:
constructor DebertaV2ForQuestionAnswering(config) - added:
constructor DebertaV2ForSequenceClassification(config) - added:
constructor DebertaV2ForTokenClassification(config) - added:
constructor DebertaV2Model(config) - added:
field DebertaV2PreTrainedModel.base_model_prefix - added:
field DebertaV2PreTrainedModel.config: DebertaV2Config - added:
field DebertaV2PreTrainedModel.supports_gradient_checkpointing - added:
method DebertaV2ForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Mas… - added:
method DebertaV2ForMaskedLM.get_output_embeddings() - added:
method DebertaV2ForMaskedLM.set_output_embeddings(new_embeddings) - added:
method DebertaV2ForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tupl… - added:
method DebertaV2ForMultipleChoice.get_input_embeddings() - added:
method DebertaV2ForMultipleChoice.set_input_embeddings(new_embeddings) - added:
method DebertaV2ForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[… - added:
method DebertaV2ForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Un… - added:
method DebertaV2ForSequenceClassification.get_input_embeddings() - added:
method DebertaV2ForSequenceClassification.set_input_embeddings(new_embeddings) - added:
method DebertaV2ForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union… - added:
method DebertaV2Model.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method DebertaV2Model.get_input_embeddings() - added:
method DebertaV2Model.set_input_embeddings(new_embeddings)
- added:
class TFDebertaV2ForMaskedLM(TFDebertaV2PreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFDebertaV2ForMultipleChoice(TFDebertaV2PreTrainedModel, TFMultipleChoiceLoss) - added:
class TFDebertaV2ForQuestionAnswering(TFDebertaV2PreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFDebertaV2ForSequenceClassification(TFDebertaV2PreTrainedModel, TFSequenceClassificationLoss) - added:
class TFDebertaV2ForTokenClassification(TFDebertaV2PreTrainedModel, TFTokenClassificationLoss) - added:
class TFDebertaV2Model(TFDebertaV2PreTrainedModel) - added:
class TFDebertaV2PreTrainedModel(TFPreTrainedModel) - added:
constructor TFDebertaV2ForMaskedLM(config: DebertaV2Config, *inputs, **kwargs) - added:
constructor TFDebertaV2ForMultipleChoice(config: DebertaV2Config, *inputs, **kwargs) - added:
constructor TFDebertaV2ForQuestionAnswering(config: DebertaV2Config, *inputs, **kwargs) - added:
constructor TFDebertaV2ForSequenceClassification(config: DebertaV2Config, *inputs, **kwargs) - added:
constructor TFDebertaV2ForTokenClassification(config: DebertaV2Config, *inputs, **kwargs) - added:
constructor TFDebertaV2Model(config: DebertaV2Config, *inputs, **kwargs) - added:
field TFDebertaV2PreTrainedModel.base_model_prefix - added:
field TFDebertaV2PreTrainedModel.config_class - added:
method TFDebertaV2ForMaskedLM.build(input_shape=…) - added:
method TFDebertaV2ForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | N… - added:
method TFDebertaV2ForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFDebertaV2ForMultipleChoice.build(input_shape=…) - added:
method TFDebertaV2ForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tens… - added:
method TFDebertaV2ForQuestionAnswering.build(input_shape=…) - added:
method TFDebertaV2ForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: np.ndarr… - added:
method TFDebertaV2ForSequenceClassification.build(input_shape=…) - added:
method TFDebertaV2ForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray |… - added:
method TFDebertaV2ForTokenClassification.build(input_shape=…) - added:
method TFDebertaV2ForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf… - added:
method TFDebertaV2Model.build(input_shape=…) - added:
method TFDebertaV2Model.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelO…
- added:
class DebertaV2Tokenizer(PreTrainedTokenizer) - added:
constructor DebertaV2Tokenizer(vocab_file, do_lower_case=…, split_by_punct=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field DebertaV2Tokenizer.vocab_files_names - added:
method DebertaV2Tokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method DebertaV2Tokenizer.convert_tokens_to_string(tokens) - added:
method DebertaV2Tokenizer.get_special_tokens_mask(token_ids_0, token_ids_1=…, already_has_special_tokens=…) - added:
method DebertaV2Tokenizer.get_vocab() - added:
method DebertaV2Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method DebertaV2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property DebertaV2Tokenizer.vocab - added:
property DebertaV2Tokenizer.vocab_size
- added:
class DebertaV2TokenizerFast(PreTrainedTokenizerFast) - added:
constructor DebertaV2TokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, split_by_punct=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) -> None - added:
field DebertaV2TokenizerFast.slow_tokenizer_class - added:
field DebertaV2TokenizerFast.vocab_files_names - added:
method DebertaV2TokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method DebertaV2TokenizerFast.get_special_tokens_mask(token_ids_0, token_ids_1=…, already_has_special_tokens=…) - added:
method DebertaV2TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class DecisionTransformerConfig(PretrainedConfig) - added:
constructor DecisionTransformerConfig(state_dim=…, act_dim=…, hidden_size=…, max_ep_len=…, action_tanh=…, vocab_size=…, n_positions=…, n_layer=…, n_head=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, scale_attn_weights=…, use_cache=…, bos_token_id=…, eos_token_id=…, scale_attn_by_inverse_layer_idx=…, reorder_and_upcast_at… - added:
field DecisionTransformerConfig.attribute_map - added:
field DecisionTransformerConfig.keys_to_ignore_at_inference - added:
field DecisionTransformerConfig.model_type
- added:
class DecisionTransformerGPT2Model(DecisionTransformerGPT2PreTrainedModel) - added:
class DecisionTransformerGPT2PreTrainedModel(PreTrainedModel) - added:
class DecisionTransformerModel(DecisionTransformerPreTrainedModel) - added:
class DecisionTransformerPreTrainedModel(PreTrainedModel) - added:
constructor DecisionTransformerGPT2Model(config) - added:
constructor DecisionTransformerGPT2PreTrainedModel(*inputs, **kwargs) - added:
constructor DecisionTransformerModel(config) - added:
field DecisionTransformerGPT2PreTrainedModel.base_model_prefix - added:
field DecisionTransformerGPT2PreTrainedModel.config: DecisionTransformerConfig - added:
field DecisionTransformerGPT2PreTrainedModel.is_parallelizable - added:
field DecisionTransformerGPT2PreTrainedModel.load_tf_weights - added:
field DecisionTransformerGPT2PreTrainedModel.supports_gradient_checkpointing - added:
field DecisionTransformerPreTrainedModel.base_model_prefix - added:
field DecisionTransformerPreTrainedModel.config: DecisionTransformerConfig - added:
field DecisionTransformerPreTrainedModel.main_input_name - added:
field DecisionTransformerPreTrainedModel.supports_gradient_checkpointing - added:
method DecisionTransformerGPT2Model.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder… - added:
method DecisionTransformerGPT2Model.get_input_embeddings() - added:
method DecisionTransformerGPT2Model.set_input_embeddings(new_embeddings) - added:
method DecisionTransformerModel.forward(states: Optional[torch.FloatTensor]=…, actions: Optional[torch.FloatTensor]=…, rewards: Optional[torch.FloatTensor]=…, returns_to_go: Optional[torch.FloatTensor]=…, timesteps: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=……
- added:
export define_import_structure
- added:
class DeepseekV2Config(PretrainedConfig) - added:
constructor DeepseekV2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, aux_loss_a… - added:
field DeepseekV2Config.base_model_pp_plan - added:
field DeepseekV2Config.base_model_tp_plan - added:
field DeepseekV2Config.keys_to_ignore_at_inference - added:
field DeepseekV2Config.model_type
- added:
class DeepseekV2ForCausalLM(DeepseekV2PreTrainedModel, GenerationMixin) - added:
class DeepseekV2ForSequenceClassification(GenericForSequenceClassification, DeepseekV2PreTrainedModel) - added:
class DeepseekV2Model(DeepseekV2PreTrainedModel) - added:
class DeepseekV2PreTrainedModel(PreTrainedModel) - added:
constructor DeepseekV2ForCausalLM(config) - added:
constructor DeepseekV2Model(config: DeepseekV2Config) - added:
field DeepseekV2PreTrainedModel.base_model_prefix - added:
field DeepseekV2PreTrainedModel.config: DeepseekV2Config - added:
field DeepseekV2PreTrainedModel.supports_gradient_checkpointing - added:
method DeepseekV2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]… - added:
method DeepseekV2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class DeepseekV2Config(LlamaConfig) - added:
class DeepseekV2ForCausalLM(LlamaForCausalLM) - added:
class DeepseekV2ForSequenceClassification(LlamaForSequenceClassification) - added:
class DeepseekV2Model(LlamaModel) - added:
class DeepseekV2PreTrainedModel(LlamaPreTrainedModel) - added:
constructor DeepseekV2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, aux_loss_a… - added:
field DeepseekV2Config.base_model_tp_plan - added:
field DeepseekV2Config.keys_to_ignore_at_inference - added:
field DeepseekV2Config.model_type
- added:
export define_import_structure
- added:
class DeepseekV3Config(PretrainedConfig) - added:
constructor DeepseekV3Config(vocab_size=…, hidden_size=…, intermediate_size=…, moe_intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, n_shared_experts=…, n_routed_experts=…, routed_scaling_factor=…, kv_lora_rank=…, q_lora_rank=…, qk_rope_head_dim=…, v_head_dim=…, qk_nope_head_dim=…, n_group=…, topk_group=…, num_experts_per_tok=…, first_k_dense_replace=…, norm_… - added:
field DeepseekV3Config.base_model_pp_plan - added:
field DeepseekV3Config.base_model_tp_plan - added:
field DeepseekV3Config.keys_to_ignore_at_inference - added:
field DeepseekV3Config.model_type
- added:
class DeepseekV3ForCausalLM(DeepseekV3PreTrainedModel, GenerationMixin) - added:
class DeepseekV3ForSequenceClassification(GenericForSequenceClassification, DeepseekV3PreTrainedModel) - added:
class DeepseekV3ForTokenClassification(GenericForTokenClassification, DeepseekV3PreTrainedModel) - added:
class DeepseekV3Model(DeepseekV3PreTrainedModel) - added:
class DeepseekV3PreTrainedModel(PreTrainedModel) - added:
constructor DeepseekV3ForCausalLM(config) - added:
constructor DeepseekV3Model(config: DeepseekV3Config) - added:
field DeepseekV3PreTrainedModel.base_model_prefix - added:
field DeepseekV3PreTrainedModel.config: DeepseekV3Config - added:
field DeepseekV3PreTrainedModel.supports_gradient_checkpointing - added:
method DeepseekV3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]… - added:
method DeepseekV3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class DeepseekV3ForCausalLM(LlamaForCausalLM) - added:
class DeepseekV3ForSequenceClassification(GenericForSequenceClassification, DeepseekV3PreTrainedModel) - added:
class DeepseekV3ForTokenClassification(GenericForTokenClassification, DeepseekV3PreTrainedModel) - added:
class DeepseekV3Model(LlamaModel) - added:
class DeepseekV3PreTrainedModel(LlamaPreTrainedModel)
- added:
export define_import_structure
- added:
class DeepseekVLConfig(PretrainedConfig) - added:
constructor DeepseekVLConfig(text_config: Optional[AutoConfig]=…, vision_config: Optional[AutoConfig]=…, image_token_id: int=…, **kwargs) - added:
field DeepseekVLConfig.model_type - added:
field DeepseekVLConfig.sub_configs
- added:
class DeepseekVLImageProcessor(BaseImageProcessor) - added:
constructor DeepseekVLImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, min_size: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, do_pad: Optional[bool]=…, **kwargs) -> None - added:
field DeepseekVLImageProcessor.model_input_names - added:
method DeepseekVLImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DeepseekVLImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[st… - added:
method DeepseekVLImageProcessor.resize(image: np.ndarray, size: Union[dict[str, int], int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DeepseekVLImageProcessorFast(BaseImageProcessorFast) - added:
constructor DeepseekVLImageProcessorFast(**kwargs: Unpack[DeepseekVLFastImageProcessorKwargs]) - added:
field DeepseekVLImageProcessorFast.do_normalize - added:
field DeepseekVLImageProcessorFast.do_pad - added:
field DeepseekVLImageProcessorFast.do_rescale - added:
field DeepseekVLImageProcessorFast.do_resize - added:
field DeepseekVLImageProcessorFast.image_mean - added:
field DeepseekVLImageProcessorFast.image_std - added:
field DeepseekVLImageProcessorFast.min_size - added:
field DeepseekVLImageProcessorFast.resample - added:
field DeepseekVLImageProcessorFast.size - added:
field DeepseekVLImageProcessorFast.valid_kwargs - added:
method DeepseekVLImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method DeepseekVLImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, min_size: int, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor
- added:
class DeepseekVLForConditionalGeneration(DeepseekVLPreTrainedModel, GenerationMixin) - added:
class DeepseekVLModel(DeepseekVLPreTrainedModel) - added:
class DeepseekVLPreTrainedModel(PreTrainedModel) - added:
constructor DeepseekVLForConditionalGeneration(config: DeepseekVLConfig) - added:
constructor DeepseekVLModel(config) - added:
field DeepseekVLPreTrainedModel.base_model_prefix - added:
field DeepseekVLPreTrainedModel.config: DeepseekVLConfig - added:
field DeepseekVLPreTrainedModel.supports_gradient_checkpointing - added:
method DeepseekVLForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache:… - added:
method DeepseekVLForConditionalGeneration.get_input_embeddings() - added:
method DeepseekVLForConditionalGeneration.prepare_embeddings_for_image_generation() -> torch.Tensor - added:
method DeepseekVLForConditionalGeneration.prepare_inputs_for_generation(input_ids, pixel_values=…, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method DeepseekVLForConditionalGeneration.set_input_embeddings(value) - added:
method DeepseekVLModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, logits_to_keep: Union[int, torch.Tensor… - added:
method DeepseekVLModel.get_image_features(pixel_values) - added:
method DeepseekVLModel.get_input_embeddings() - added:
method DeepseekVLModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method DeepseekVLModel.set_input_embeddings(value)
- added:
class DeepseekVLConfig(PretrainedConfig) - added:
class DeepseekVLForConditionalGeneration(JanusForConditionalGeneration) - added:
class DeepseekVLImageProcessor(JanusImageProcessor) - added:
class DeepseekVLImageProcessorFast(JanusImageProcessorFast) - added:
class DeepseekVLModel(JanusModel) - added:
class DeepseekVLPreTrainedModel(JanusPreTrainedModel) - added:
class DeepseekVLProcessor(ProcessorMixin) - added:
constructor DeepseekVLConfig(text_config: Optional[AutoConfig]=…, vision_config: Optional[AutoConfig]=…, image_token_id: int=…, **kwargs) - added:
constructor DeepseekVLImageProcessor(**super_kwargs) - added:
constructor DeepseekVLImageProcessorFast(**super_kwargs) - added:
constructor DeepseekVLModel(config) - added:
constructor DeepseekVLProcessor(image_processor, tokenizer, chat_template=…, num_image_tokens=…) - added:
field DeepseekVLConfig.model_type - added:
field DeepseekVLConfig.sub_configs - added:
field DeepseekVLProcessor.attributes - added:
field DeepseekVLProcessor.image_processor_class - added:
field DeepseekVLProcessor.tokenizer_class - added:
field DeepseekVLProcessor.valid_kwargs - added:
method DeepseekVLForConditionalGeneration.decode_image_tokens() - added:
method DeepseekVLForConditionalGeneration.generate() - added:
method DeepseekVLForConditionalGeneration.prepare_embeddings_for_image_generation() - added:
method DeepseekVLImageProcessor.postprocess() - added:
method DeepseekVLImageProcessor.unnormalize() - added:
method DeepseekVLImageProcessorFast.postprocess() - added:
method DeepseekVLProcessor.batch_decode(*args, **kwargs) - added:
method DeepseekVLProcessor.decode(*args, **kwargs) - added:
property DeepseekVLProcessor.model_input_names
- added:
class DeepseekVLProcessor(ProcessorMixin) - added:
constructor DeepseekVLProcessor(image_processor, tokenizer, chat_template=…, num_image_tokens=…) - added:
field DeepseekVLProcessor.attributes - added:
field DeepseekVLProcessor.image_processor_class - added:
field DeepseekVLProcessor.tokenizer_class - added:
field DeepseekVLProcessor.valid_kwargs - added:
method DeepseekVLProcessor.batch_decode(*args, **kwargs) - added:
method DeepseekVLProcessor.decode(*args, **kwargs) - added:
property DeepseekVLProcessor.model_input_names
- added:
export define_import_structure
- added:
class DeepseekVLHybridConfig(PretrainedConfig) - added:
constructor DeepseekVLHybridConfig(text_config: Optional[AutoConfig]=…, vision_config: Optional[AutoConfig]=…, high_res_vision_config: Optional[AutoConfig]=…, image_token_id: int=…, **kwargs) - added:
field DeepseekVLHybridConfig.model_type - added:
field DeepseekVLHybridConfig.sub_configs
- added:
class DeepseekVLHybridImageProcessor(BaseImageProcessor) - added:
constructor DeepseekVLHybridImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, high_res_size: Optional[dict[str, int]]=…, min_size: int=…, resample: PILImageResampling=…, high_res_resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[fl… - added:
field DeepseekVLHybridImageProcessor.model_input_names - added:
method DeepseekVLHybridImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DeepseekVLHybridImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, high_res_size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, high_res_resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float,… - added:
method DeepseekVLHybridImageProcessor.resize(image: np.ndarray, size: Union[dict[str, int], int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DeepseekVLHybridImageProcessorFast(BaseImageProcessorFast) - added:
constructor DeepseekVLHybridImageProcessorFast(**kwargs: Unpack[DeepseekVLHybridFastImageProcessorKwargs]) - added:
field DeepseekVLHybridImageProcessorFast.do_normalize - added:
field DeepseekVLHybridImageProcessorFast.do_pad - added:
field DeepseekVLHybridImageProcessorFast.do_rescale - added:
field DeepseekVLHybridImageProcessorFast.do_resize - added:
field DeepseekVLHybridImageProcessorFast.high_res_image_mean - added:
field DeepseekVLHybridImageProcessorFast.high_res_image_std - added:
field DeepseekVLHybridImageProcessorFast.high_res_resample - added:
field DeepseekVLHybridImageProcessorFast.high_res_size - added:
field DeepseekVLHybridImageProcessorFast.image_mean - added:
field DeepseekVLHybridImageProcessorFast.image_std - added:
field DeepseekVLHybridImageProcessorFast.min_size - added:
field DeepseekVLHybridImageProcessorFast.model_input_names - added:
field DeepseekVLHybridImageProcessorFast.resample - added:
field DeepseekVLHybridImageProcessorFast.size - added:
field DeepseekVLHybridImageProcessorFast.valid_kwargs - added:
method DeepseekVLHybridImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method DeepseekVLHybridImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, min_size: int, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor
- added:
class DeepseekVLHybridForConditionalGeneration(DeepseekVLHybridPreTrainedModel, GenerationMixin) - added:
class DeepseekVLHybridModel(DeepseekVLHybridPreTrainedModel) - added:
class DeepseekVLHybridPreTrainedModel(PreTrainedModel) - added:
constructor DeepseekVLHybridForConditionalGeneration(config: DeepseekVLHybridConfig) - added:
constructor DeepseekVLHybridModel(config) - added:
field DeepseekVLHybridPreTrainedModel.base_model_prefix - added:
field DeepseekVLHybridPreTrainedModel.config: DeepseekVLHybridConfig - added:
field DeepseekVLHybridPreTrainedModel.supports_gradient_checkpointing - added:
method DeepseekVLHybridForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, high_res_pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Float… - added:
method DeepseekVLHybridForConditionalGeneration.get_input_embeddings() - added:
method DeepseekVLHybridForConditionalGeneration.prepare_embeddings_for_image_generation() -> torch.Tensor - added:
method DeepseekVLHybridForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, high_res_pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method DeepseekVLHybridForConditionalGeneration.set_input_embeddings(value) - added:
method DeepseekVLHybridModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, high_res_pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cach… - added:
method DeepseekVLHybridModel.get_high_res_image_features(pixel_values) - added:
method DeepseekVLHybridModel.get_image_features(pixel_values, high_res_pixel_values) - added:
method DeepseekVLHybridModel.get_input_embeddings() - added:
method DeepseekVLHybridModel.get_low_res_image_features(pixel_values) - added:
method DeepseekVLHybridModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method DeepseekVLHybridModel.set_input_embeddings(value)
- added:
class DeepseekVLHybridConfig(DeepseekVLConfig) - added:
class DeepseekVLHybridForConditionalGeneration(DeepseekVLForConditionalGeneration) - added:
class DeepseekVLHybridImageProcessor(DeepseekVLImageProcessor) - added:
class DeepseekVLHybridImageProcessorFast(DeepseekVLImageProcessorFast) - added:
class DeepseekVLHybridModel(DeepseekVLModel) - added:
class DeepseekVLHybridPreTrainedModel(DeepseekVLPreTrainedModel) - added:
class DeepseekVLHybridProcessor(DeepseekVLProcessor) - added:
constructor DeepseekVLHybridConfig(text_config: Optional[AutoConfig]=…, vision_config: Optional[AutoConfig]=…, high_res_vision_config: Optional[AutoConfig]=…, image_token_id: int=…, **kwargs) - added:
constructor DeepseekVLHybridImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, high_res_size: Optional[dict[str, int]]=…, min_size: int=…, resample: PILImageResampling=…, high_res_resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[fl… - added:
constructor DeepseekVLHybridImageProcessorFast(**kwargs: Unpack[DeepseekVLHybridFastImageProcessorKwargs]) - added:
constructor DeepseekVLHybridModel(config) - added:
field DeepseekVLHybridConfig.model_type - added:
field DeepseekVLHybridConfig.sub_configs - added:
field DeepseekVLHybridImageProcessor.model_input_names - added:
field DeepseekVLHybridImageProcessorFast.high_res_image_mean - added:
field DeepseekVLHybridImageProcessorFast.high_res_image_std - added:
field DeepseekVLHybridImageProcessorFast.high_res_resample - added:
field DeepseekVLHybridImageProcessorFast.high_res_size - added:
field DeepseekVLHybridImageProcessorFast.model_input_names - added:
method DeepseekVLHybridForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, high_res_pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Float… - added:
method DeepseekVLHybridForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, high_res_pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method DeepseekVLHybridImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, high_res_size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, high_res_resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float,… - added:
method DeepseekVLHybridModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, high_res_pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cach… - added:
method DeepseekVLHybridModel.get_high_res_image_features(pixel_values) - added:
method DeepseekVLHybridModel.get_image_features(pixel_values, high_res_pixel_values) - added:
method DeepseekVLHybridModel.get_low_res_image_features(pixel_values)
- added:
class DeepseekVLHybridProcessor(ProcessorMixin) - added:
constructor DeepseekVLHybridProcessor(image_processor, tokenizer, chat_template=…, num_image_tokens=…) - added:
field DeepseekVLHybridProcessor.attributes - added:
field DeepseekVLHybridProcessor.image_processor_class - added:
field DeepseekVLHybridProcessor.tokenizer_class - added:
field DeepseekVLHybridProcessor.valid_kwargs - added:
method DeepseekVLHybridProcessor.batch_decode(*args, **kwargs) - added:
method DeepseekVLHybridProcessor.decode(*args, **kwargs) - added:
property DeepseekVLHybridProcessor.model_input_names
- added:
export define_import_structure
- added:
class DeformableDetrConfig(PretrainedConfig) - added:
constructor DeformableDetrConfig(use_timm_backbone=…, backbone_config=…, num_channels=…, num_queries=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=… - added:
field DeformableDetrConfig.attribute_map - added:
field DeformableDetrConfig.model_type - added:
property DeformableDetrConfig.hidden_size -> int - added:
property DeformableDetrConfig.num_attention_heads -> int - added:
property DeformableDetrConfig.sub_configs
- added:
class DeformableDetrFeatureExtractor(DeformableDetrImageProcessor) - added:
constructor DeformableDetrFeatureExtractor(*args, **kwargs) -> None
- added:
class DeformableDetrImageProcessor(BaseImageProcessor) - added:
constructor DeformableDetrImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: Optional[bool]=…, do_pad: boo… - added:
field DeformableDetrImageProcessor.model_input_names - added:
method DeformableDetrImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method DeformableDetrImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_siz… - added:
method DeformableDetrImageProcessor.post_process(outputs, target_sizes) - added:
method DeformableDetrImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method DeformableDetrImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method DeformableDetrImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: O… - added:
method DeformableDetrImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DeformableDetrImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method DeformableDetrImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict - added:
static method DeformableDetrImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class DeformableDetrImageProcessorFast(BaseImageProcessorFast) - added:
constructor DeformableDetrImageProcessorFast(**kwargs: Unpack[DeformableDetrFastImageProcessorKwargs]) -> None - added:
field DeformableDetrImageProcessorFast.default_to_square - added:
field DeformableDetrImageProcessorFast.do_normalize - added:
field DeformableDetrImageProcessorFast.do_pad - added:
field DeformableDetrImageProcessorFast.do_rescale - added:
field DeformableDetrImageProcessorFast.do_resize - added:
field DeformableDetrImageProcessorFast.format - added:
field DeformableDetrImageProcessorFast.image_mean - added:
field DeformableDetrImageProcessorFast.image_std - added:
field DeformableDetrImageProcessorFast.model_input_names - added:
field DeformableDetrImageProcessorFast.resample - added:
field DeformableDetrImageProcessorFast.size - added:
field DeformableDetrImageProcessorFast.valid_kwargs - added:
method DeformableDetrImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method DeformableDetrImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method DeformableDetrImageProcessorFast.post_process(outputs, target_sizes) - added:
method DeformableDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method DeformableDetrImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method DeformableDetrImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[DeformableDetrFastImageProcessorKwargs]) -> BatchFeature - added:
method DeformableDetrImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method DeformableDetrImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…) - added:
static method DeformableDetrImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class DeformableDetrForObjectDetection(DeformableDetrPreTrainedModel) - added:
class DeformableDetrModel(DeformableDetrPreTrainedModel) - added:
class DeformableDetrPreTrainedModel(PreTrainedModel) - added:
constructor DeformableDetrForObjectDetection(config: DeformableDetrConfig) - added:
constructor DeformableDetrModel(config: DeformableDetrConfig) - added:
field DeformableDetrPreTrainedModel.base_model_prefix - added:
field DeformableDetrPreTrainedModel.config: DeformableDetrConfig - added:
field DeformableDetrPreTrainedModel.main_input_name - added:
field DeformableDetrPreTrainedModel.supports_gradient_checkpointing - added:
method DeformableDetrForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, outp… - added:
method DeformableDetrModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_di… - added:
method DeformableDetrModel.freeze_backbone() - added:
method DeformableDetrModel.gen_encoder_output_proposals(enc_output, padding_mask, spatial_shapes) - added:
method DeformableDetrModel.get_encoder() - added:
method DeformableDetrModel.get_proposal_pos_embed(proposals) - added:
method DeformableDetrModel.get_valid_ratio(mask, dtype=…) - added:
method DeformableDetrModel.unfreeze_backbone()
- added:
class DeformableDetrImageProcessorFast(DetrImageProcessorFast) - added:
method DeformableDetrImageProcessorFast.post_process(outputs, target_sizes) - added:
method DeformableDetrImageProcessorFast.post_process_instance() - added:
method DeformableDetrImageProcessorFast.post_process_instance_segmentation() - added:
method DeformableDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method DeformableDetrImageProcessorFast.post_process_panoptic() - added:
method DeformableDetrImageProcessorFast.post_process_panoptic_segmentation() - added:
method DeformableDetrImageProcessorFast.post_process_segmentation() - added:
method DeformableDetrImageProcessorFast.post_process_semantic_segmentation()
- added:
export define_import_structure
- added:
class DeiTConfig(PretrainedConfig) - added:
class DeiTOnnxConfig(OnnxConfig) - added:
constructor DeiTConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, encoder_stride=…, pooler_output_size=…, pooler_act=…, **kwargs) - added:
field DeiTConfig.model_type - added:
field DeiTOnnxConfig.torch_onnx_minimum_version - added:
property DeiTOnnxConfig.atol_for_validation -> float - added:
property DeiTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class DeiTFeatureExtractor(DeiTImageProcessor) - added:
constructor DeiTFeatureExtractor(*args, **kwargs) -> None
- added:
class DeiTImageProcessor(BaseImageProcessor) - added:
constructor DeiTImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, rescale_factor: Union[int, float]=…, do_rescale: bool=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field DeiTImageProcessor.model_input_names - added:
method DeiTImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]… - added:
method DeiTImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DeiTImageProcessorFast(BaseImageProcessorFast) - added:
field DeiTImageProcessorFast.crop_size - added:
field DeiTImageProcessorFast.do_center_crop - added:
field DeiTImageProcessorFast.do_normalize - added:
field DeiTImageProcessorFast.do_rescale - added:
field DeiTImageProcessorFast.do_resize - added:
field DeiTImageProcessorFast.image_mean - added:
field DeiTImageProcessorFast.image_std - added:
field DeiTImageProcessorFast.resample - added:
field DeiTImageProcessorFast.size
- added:
class DeiTForImageClassification(DeiTPreTrainedModel) - added:
class DeiTForImageClassificationWithTeacher(DeiTPreTrainedModel) - added:
class DeiTForMaskedImageModeling(DeiTPreTrainedModel) - added:
class DeiTModel(DeiTPreTrainedModel) - added:
class DeiTPreTrainedModel(PreTrainedModel) - added:
constructor DeiTForImageClassification(config: DeiTConfig) -> None - added:
constructor DeiTForImageClassificationWithTeacher(config: DeiTConfig) -> None - added:
constructor DeiTForMaskedImageModeling(config: DeiTConfig) -> None - added:
constructor DeiTModel(config: DeiTConfig, add_pooling_layer: bool=…, use_mask_token: bool=…) -> None - added:
field DeiTPreTrainedModel.base_model_prefix - added:
field DeiTPreTrainedModel.config: DeiTConfig - added:
field DeiTPreTrainedModel.main_input_name - added:
field DeiTPreTrainedModel.supports_gradient_checkpointing - added:
method DeiTForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method DeiTForImageClassificationWithTeacher.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> DeiTForImageClassificationWithTeacherOutput - added:
method DeiTForMaskedImageModeling.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> MaskedImageModelingOutput - added:
method DeiTModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method DeiTModel.get_input_embeddings() -> DeiTPatchEmbeddings
- added:
class TFDeiTForImageClassification(TFDeiTPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFDeiTForImageClassificationWithTeacher(TFDeiTPreTrainedModel) - added:
class TFDeiTForMaskedImageModeling(TFDeiTPreTrainedModel) - added:
class TFDeiTModel(TFDeiTPreTrainedModel) - added:
class TFDeiTPreTrainedModel(TFPreTrainedModel) - added:
constructor TFDeiTForImageClassification(config: DeiTConfig) - added:
constructor TFDeiTForImageClassificationWithTeacher(config: DeiTConfig) -> None - added:
constructor TFDeiTForMaskedImageModeling(config: DeiTConfig) -> None - added:
constructor TFDeiTModel(config: DeiTConfig, add_pooling_layer: bool=…, use_mask_token: bool=…, **kwargs) -> None - added:
field TFDeiTPreTrainedModel.base_model_prefix - added:
field TFDeiTPreTrainedModel.config_class - added:
field TFDeiTPreTrainedModel.main_input_name - added:
method TFDeiTForImageClassification.build(input_shape=…) - added:
method TFDeiTForImageClassification.call(pixel_values: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, interpolate_pos_encoding: bool=…, training: bool=…) -> tf.Tensor | TFImageClassifierOutput - added:
method TFDeiTForImageClassificationWithTeacher.build(input_shape=…) - added:
method TFDeiTForImageClassificationWithTeacher.call(pixel_values: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, interpolate_pos_encoding: bool=…, training: bool=…) -> tuple | TFDeiTForImageClassificationWithTeacherOutput - added:
method TFDeiTForMaskedImageModeling.build(input_shape=…) - added:
method TFDeiTForMaskedImageModeling.call(pixel_values: tf.Tensor | None=…, bool_masked_pos: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, interpolate_pos_encoding: bool=…, training: bool=…) -> tuple | TFMaskedImageModelingOutput - added:
method TFDeiTModel.build(input_shape=…) - added:
method TFDeiTModel.call(pixel_values: tf.Tensor | None=…, bool_masked_pos: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, interpolate_pos_encoding: bool=…, training: bool=…) -> tuple | TFBaseModelOutputWithPooling
- added:
export define_import_structure
- added:
export define_import_structure
- added:
class DetaConfig(PretrainedConfig) - added:
constructor DetaConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, num_queries=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…,… - added:
field DetaConfig.attribute_map - added:
field DetaConfig.model_type - added:
property DetaConfig.hidden_size -> int - added:
property DetaConfig.num_attention_heads -> int - added:
property DetaConfig.sub_configs
- added:
class DetaImageProcessor(BaseImageProcessor) - added:
constructor DetaImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: bool=…, do_pad: bool=…, pad_size: Optio… - added:
field DetaImageProcessor.model_input_names - added:
method DetaImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method DetaImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_size: Optiona… - added:
method DetaImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, nms_threshold: float=…) - added:
method DetaImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method DetaImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[list[dict], list[list[dict]]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[bool]=…, i… - added:
method DetaImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DetaImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method DetaImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict
- added:
class DetaForObjectDetection(DetaPreTrainedModel) - added:
class DetaModel(DetaPreTrainedModel) - added:
class DetaPreTrainedModel(PreTrainedModel) - added:
constructor DetaForObjectDetection(config: DetaConfig) - added:
constructor DetaModel(config: DetaConfig) - added:
field DetaPreTrainedModel.base_model_prefix - added:
field DetaPreTrainedModel.config: DetaConfig - added:
field DetaPreTrainedModel.main_input_name - added:
field DetaPreTrainedModel.supports_gradient_checkpointing - added:
method DetaForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_… - added:
method DetaModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Option… - added:
method DetaModel.freeze_backbone() - added:
method DetaModel.gen_encoder_output_proposals(enc_output, padding_mask, spatial_shapes) - added:
method DetaModel.get_encoder() - added:
method DetaModel.get_proposal_pos_embed(proposals) - added:
method DetaModel.get_valid_ratio(mask, dtype=…) - added:
method DetaModel.unfreeze_backbone()
- added:
export define_import_structure
- added:
class EfficientFormerConfig(PretrainedConfig) - added:
constructor EfficientFormerConfig(depths: list[int]=…, hidden_sizes: list[int]=…, downsamples: list[bool]=…, dim: int=…, key_dim: int=…, attention_ratio: int=…, resolution: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, mlp_expansion_ratio: int=…, hidden_dropout_prob: float=…, patch_size: int=…, num_channels: int=…, pool_size: int=…, downsample_patch_size: int=…, downsample_stride:… - added:
field EfficientFormerConfig.model_type
- added:
class EfficientFormerImageProcessor(BaseImageProcessor) - added:
constructor EfficientFormerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, crop_size: Optional[dict[str, int]]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field EfficientFormerImageProcessor.model_input_names - added:
method EfficientFormerImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optio… - added:
method EfficientFormerImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class EfficientFormerForImageClassification(EfficientFormerPreTrainedModel) - added:
class EfficientFormerForImageClassificationWithTeacher(EfficientFormerPreTrainedModel) - added:
class EfficientFormerModel(EfficientFormerPreTrainedModel) - added:
class EfficientFormerPreTrainedModel(PreTrainedModel) - added:
constructor EfficientFormerForImageClassification(config: EfficientFormerConfig) - added:
constructor EfficientFormerForImageClassificationWithTeacher(config: EfficientFormerConfig) - added:
constructor EfficientFormerModel(config: EfficientFormerConfig) - added:
field EfficientFormerPreTrainedModel.base_model_prefix - added:
field EfficientFormerPreTrainedModel.config: EfficientFormerConfig - added:
field EfficientFormerPreTrainedModel.main_input_name - added:
field EfficientFormerPreTrainedModel.supports_gradient_checkpointing - added:
method EfficientFormerForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method EfficientFormerForImageClassificationWithTeacher.forward(pixel_values: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, EfficientFormerForImageClassificationWithTeacherOutput] - added:
method EfficientFormerModel.forward(pixel_values: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
class TFEfficientFormerForImageClassification(TFEfficientFormerPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFEfficientFormerForImageClassificationWithTeacher(TFEfficientFormerPreTrainedModel) - added:
class TFEfficientFormerModel(TFEfficientFormerPreTrainedModel) - added:
class TFEfficientFormerPreTrainedModel(TFPreTrainedModel) - added:
constructor TFEfficientFormerForImageClassification(config: EfficientFormerConfig) - added:
constructor TFEfficientFormerForImageClassificationWithTeacher(config: EfficientFormerConfig) -> None - added:
constructor TFEfficientFormerModel(config: EfficientFormerConfig, **kwargs) -> None - added:
field TFEfficientFormerPreTrainedModel.base_model_prefix - added:
field TFEfficientFormerPreTrainedModel.config_class - added:
field TFEfficientFormerPreTrainedModel.main_input_name - added:
method TFEfficientFormerForImageClassification.build(input_shape=…) - added:
method TFEfficientFormerForImageClassification.call(pixel_values: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tf.Tensor, TFImageClassifierOutput] - added:
method TFEfficientFormerForImageClassificationWithTeacher.build(input_shape=…) - added:
method TFEfficientFormerForImageClassificationWithTeacher.call(pixel_values: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tuple, TFEfficientFormerForImageClassificationWithTeacherOutput] - added:
method TFEfficientFormerModel.build(input_shape=…) - added:
method TFEfficientFormerModel.call(pixel_values: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tuple, TFBaseModelOutput]
- added:
export define_import_structure
- added:
class ErnieMConfig(PretrainedConfig) - added:
constructor ErnieMConfig(vocab_size: int=…, hidden_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, intermediate_size: int=…, hidden_act: str=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…, max_position_embeddings: int=…, initializer_range: float=…, pad_token_id: int=…, layer_norm_eps: float=…, classifier_dropout=…, act_dropout=…, **kwargs) - added:
field ErnieMConfig.attribute_map: dict[str, str] - added:
field ErnieMConfig.model_type
- added:
class ErnieMForInformationExtraction(ErnieMPreTrainedModel) - added:
class ErnieMForMultipleChoice(ErnieMPreTrainedModel) - added:
class ErnieMForQuestionAnswering(ErnieMPreTrainedModel) - added:
class ErnieMForSequenceClassification(ErnieMPreTrainedModel) - added:
class ErnieMForTokenClassification(ErnieMPreTrainedModel) - added:
class ErnieMModel(ErnieMPreTrainedModel) - added:
class ErnieMPreTrainedModel(PreTrainedModel) - added:
constructor ErnieMForInformationExtraction(config) - added:
constructor ErnieMForMultipleChoice(config) - added:
constructor ErnieMForQuestionAnswering(config) - added:
constructor ErnieMForSequenceClassification(config) - added:
constructor ErnieMForTokenClassification(config) - added:
constructor ErnieMModel(config, add_pooling_layer=…) - added:
field ErnieMPreTrainedModel.base_model_prefix - added:
field ErnieMPreTrainedModel.config: ErnieMConfig - added:
method ErnieMForInformationExtraction.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool… - added:
method ErnieMForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.… - added:
method ErnieMForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…,… - added:
method ErnieMForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict:… - added:
method ErnieMForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Option… - added:
method ErnieMModel.forward(input_ids: Optional[tensor]=…, position_ids: Optional[tensor]=…, attention_mask: Optional[tensor]=…, head_mask: Optional[tensor]=…, inputs_embeds: Optional[tensor]=…, past_key_values: Optional[tuple[tuple[tensor]]]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[to… - added:
method ErnieMModel.get_input_embeddings() - added:
method ErnieMModel.set_input_embeddings(value)
- added:
class ErnieMTokenizer(PreTrainedTokenizer) - added:
constructor ErnieMTokenizer(sentencepiece_model_ckpt, vocab_file=…, do_lower_case=…, encoding=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field ErnieMTokenizer.model_input_names: list[str] - added:
field ErnieMTokenizer.resource_files_names - added:
field ErnieMTokenizer.vocab_files_names - added:
method ErnieMTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method ErnieMTokenizer.build_offset_mapping_with_special_tokens(offset_mapping_0, offset_mapping_1=…) - added:
method ErnieMTokenizer.clean_text(text) - added:
method ErnieMTokenizer.convert_ids_to_string(ids) - added:
method ErnieMTokenizer.convert_tokens_to_string(tokens) - added:
method ErnieMTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ErnieMTokenizer.get_offset_mapping(text) - added:
method ErnieMTokenizer.get_special_tokens_mask(token_ids_0, token_ids_1=…, already_has_special_tokens=…) - added:
method ErnieMTokenizer.get_vocab() - added:
method ErnieMTokenizer.is_alpha(char) - added:
method ErnieMTokenizer.is_ch_char(char) - added:
method ErnieMTokenizer.is_punct(char) - added:
method ErnieMTokenizer.is_whitespace(char) - added:
method ErnieMTokenizer.load_vocab(filepath) - added:
method ErnieMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ErnieMTokenizer.vocab_size
- added:
export define_import_structure
- added:
class GPTSanJapaneseConfig(PretrainedConfig) - added:
constructor GPTSanJapaneseConfig(vocab_size=…, max_position_embeddings=…, d_model=…, d_ff=…, d_ext=…, d_spout=…, num_switch_layers=…, num_ext_layers=…, num_heads=…, num_experts=…, expert_capacity=…, dropout_rate=…, layer_norm_epsilon=…, router_bias=…, router_jitter_noise=…, router_dtype=…, router_ignore_padding_tokens=…, output_hidden_states=…, output_attentions=…, initializer_factor=…, output_ro… - added:
field GPTSanJapaneseConfig.attribute_map - added:
field GPTSanJapaneseConfig.keys_to_ignore_at_inference - added:
field GPTSanJapaneseConfig.model_type
- added:
class GPTSanJapaneseForConditionalGeneration(GPTSanJapanesePreTrainedModel) - added:
class GPTSanJapaneseModel(GPTSanJapanesePreTrainedModel) - added:
class GPTSanJapanesePreTrainedModel(PreTrainedModel) - added:
constructor GPTSanJapaneseForConditionalGeneration(config: GPTSanJapaneseConfig) - added:
constructor GPTSanJapaneseModel(config: GPTSanJapaneseConfig) - added:
field GPTSanJapanesePreTrainedModel.base_model_prefix - added:
field GPTSanJapanesePreTrainedModel.config: GPTSanJapaneseConfig - added:
field GPTSanJapanesePreTrainedModel.supports_gradient_checkpointing - added:
method GPTSanJapaneseForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.FloatTensor]=…, spout: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, head_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embe… - added:
method GPTSanJapaneseForConditionalGeneration.get_input_embeddings() - added:
method GPTSanJapaneseForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method GPTSanJapaneseForConditionalGeneration.prepare_inputs_for_generation(input_ids: torch.LongTensor, attention_mask: torch.FloatTensor, token_type_ids: Optional[torch.FloatTensor]=…, spout: Optional[Union[list, torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, **kwargs) - added:
method GPTSanJapaneseForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method GPTSanJapaneseForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method GPTSanJapaneseModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.FloatTensor]=…, spout: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, head_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.… - added:
method GPTSanJapaneseModel.set_input_embeddings(new_embeddings) - added:
property GPTSanJapanesePreTrainedModel.dummy_inputs
- added:
class GPTSanJapaneseTokenizer(PreTrainedTokenizer) - added:
constructor GPTSanJapaneseTokenizer(vocab_file, emoji_file, unk_token=…, pad_token=…, bos_token=…, eos_token=…, sep_token=…, do_clean_text=…, **kwargs) - added:
field GPTSanJapaneseTokenizer.model_input_names - added:
field GPTSanJapaneseTokenizer.vocab_files_names - added:
method GPTSanJapaneseTokenizer.convert_tokens_to_string(tokens) - added:
method GPTSanJapaneseTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method GPTSanJapaneseTokenizer.get_vocab() - added:
method GPTSanJapaneseTokenizer.prepare_for_tokenization(text, prefix_text=…, add_sep_token=…, **kwargs) - added:
method GPTSanJapaneseTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property GPTSanJapaneseTokenizer.vocab_size
- added:
export define_import_structure
- added:
class GraphormerDataCollator - added:
constructor GraphormerDataCollator(spatial_pos_max=…, on_the_fly_processing=…) - added:
function convert_to_single_emb(x, offset: int=…) - added:
function preprocess_item(item, keep_features=…)
- added:
class GraphormerConfig(PretrainedConfig) - added:
constructor GraphormerConfig(num_classes: int=…, num_atoms: int=…, num_edges: int=…, num_in_degree: int=…, num_out_degree: int=…, num_spatial: int=…, num_edge_dis: int=…, multi_hop_max_dist: int=…, spatial_pos_max: int=…, edge_type: str=…, max_nodes: int=…, share_input_output_embed: bool=…, num_hidden_layers: int=…, embedding_dim: int=…, ffn_embedding_dim: int=…, num_attention_heads: int=…, dropo… - added:
field GraphormerConfig.keys_to_ignore_at_inference - added:
field GraphormerConfig.model_type
- added:
class GraphormerForGraphClassification(GraphormerPreTrainedModel) - added:
class GraphormerModel(GraphormerPreTrainedModel) - added:
class GraphormerPreTrainedModel(PreTrainedModel) - added:
constructor GraphormerForGraphClassification(config: GraphormerConfig) - added:
constructor GraphormerModel(config: GraphormerConfig) - added:
field GraphormerPreTrainedModel.base_model_prefix - added:
field GraphormerPreTrainedModel.config: GraphormerConfig - added:
field GraphormerPreTrainedModel.main_input_name_edges - added:
field GraphormerPreTrainedModel.main_input_name_nodes - added:
method GraphormerForGraphClassification.forward(input_nodes: torch.LongTensor, input_edges: torch.LongTensor, attn_bias: torch.Tensor, in_degree: torch.LongTensor, out_degree: torch.LongTensor, spatial_pos: torch.LongTensor, attn_edge_type: torch.LongTensor, labels: Optional[torch.LongTensor]=…, return_dict: Optional[bool]=…, **unused) -> Union[tuple[torch.Tensor], SequenceClassifierOutput] - added:
method GraphormerModel.forward(input_nodes: torch.LongTensor, input_edges: torch.LongTensor, attn_bias: torch.Tensor, in_degree: torch.LongTensor, out_degree: torch.LongTensor, spatial_pos: torch.LongTensor, attn_edge_type: torch.LongTensor, perturb: Optional[torch.FloatTensor]=…, masked_tokens: None=…, return_dict: Optional[bool]=…, **unused) -> Union[tuple[torch.LongTensor], BaseModelOutputWith… - added:
method GraphormerModel.max_nodes() - added:
method GraphormerModel.reset_output_layer_parameters() - added:
method GraphormerPreTrainedModel.init_graphormer_params(module: Union[nn.Linear, nn.Embedding, GraphormerMultiheadAttention]) - added:
method GraphormerPreTrainedModel.normal_(data: torch.Tensor)
- added:
export define_import_structure
- added:
class JukeboxConfig(PretrainedConfig) - added:
class JukeboxPriorConfig(PretrainedConfig) - added:
class JukeboxVQVAEConfig(PretrainedConfig) - added:
constructor JukeboxConfig(vqvae_config=…, prior_config_list=…, nb_priors=…, sampling_rate=…, timing_dims=…, min_duration=…, max_duration=…, max_nb_genres=…, metadata_conditioning=…, **kwargs) - added:
constructor JukeboxPriorConfig(act_fn=…, level=…, alignment_head=…, alignment_layer=…, attention_multiplier=…, attention_pattern=…, attn_dropout=…, attn_res_scale=…, blocks=…, conv_res_scale=…, num_layers=…, emb_dropout=…, encoder_config=…, encoder_loss_fraction=…, hidden_size=…, init_scale=…, is_encoder_decoder=…, lyric_vocab_size=…, mask=…, max_duration=…, max_nb_genres=…, merged_decoder=…, met… - added:
constructor JukeboxVQVAEConfig(act_fn=…, nb_discrete_codes=…, commit=…, conv_input_shape=…, conv_res_scale=…, embed_dim=…, hop_fraction=…, levels=…, lmu=…, multipliers=…, res_conv_depth=…, res_conv_width=…, res_convolution_multiplier=…, res_dilation_cycle=…, res_dilation_growth_rate=…, res_downs_t=…, res_strides_t=…, sample_length=…, init_scale=…, zero_out=…, **kwargs) - added:
field JukeboxConfig.model_type - added:
field JukeboxPriorConfig.attribute_map - added:
field JukeboxPriorConfig.model_type - added:
field JukeboxVQVAEConfig.model_type - added:
method JukeboxConfig.to_dict() - added:
static method JukeboxConfig.from_configs(prior_configs: list[JukeboxPriorConfig], vqvae_config: JukeboxVQVAEConfig, **kwargs) - added:
static method JukeboxPriorConfig.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], level=…, **kwargs) - added:
static method JukeboxVQVAEConfig.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs)
- added:
class JukeboxModel(JukeboxPreTrainedModel) - added:
class JukeboxPreTrainedModel(PreTrainedModel) - added:
class JukeboxPrior(PreTrainedModel) - added:
class JukeboxVQVAE(PreTrainedModel) - added:
constructor JukeboxModel(config) - added:
constructor JukeboxPreTrainedModel(*inputs, **kwargs) - added:
constructor JukeboxPrior(config: JukeboxPriorConfig, level=…, nb_priors=…, vqvae_encoder=…, vqvae_decoder=…) - added:
constructor JukeboxVQVAE(config: JukeboxVQVAEConfig) - added:
field JukeboxPreTrainedModel.base_model_prefix - added:
field JukeboxPreTrainedModel.config: JukeboxConfig - added:
field JukeboxPreTrainedModel.supports_gradient_checkpointing - added:
field JukeboxPrior.config: JukeboxPriorConfig - added:
field JukeboxVQVAE.base_model_prefix - added:
field JukeboxVQVAE.config: JukeboxVQVAEConfig - added:
method JukeboxModel.ancestral_sample(labels, n_samples=…, **sampling_kwargs) -> list[torch.LongTensor] - added:
method JukeboxModel.continue_sample(music_tokens, labels, **sampling_kwargs) -> list[torch.LongTensor] - added:
method JukeboxModel.decode(music_tokens, start_level=…, end_level=…, bs_chunks=…) - added:
method JukeboxModel.encode(input_audio, start_level=…, end_level=…, bs_chunks=…) - added:
method JukeboxModel.primed_sample(raw_audio, labels, **sampling_kwargs) -> list[torch.LongTensor] - added:
method JukeboxModel.sample_level(music_tokens, labels, offset, sampling_kwargs, level, total_length, hop_length, max_batch_size) - added:
method JukeboxModel.sample_partial_window(music_tokens, labels, offset, sampling_kwargs, level, tokens_to_sample, max_batch_size) - added:
method JukeboxModel.sample_single_window(music_tokens, labels, offset, sampling_kwargs, level, start, max_batch_size) - added:
method JukeboxModel.set_shared_params(model_config) - added:
method JukeboxModel.split_batch(obj, n_samples, split_size) - added:
method JukeboxModel.upsample(music_tokens, labels, **sampling_kwargs) -> list[torch.LongTensor] - added:
method JukeboxPrior.decode(music_tokens, start_level=…, end_level=…, bs_chunks=…) - added:
method JukeboxPrior.embed_tokens(music_tokens_conds) - added:
method JukeboxPrior.encode(hidden_states, start_level=…, end_level=…, bs_chunks=…) - added:
method JukeboxPrior.forward(hidden_states: torch.Tensor, metadata: Optional[list[torch.LongTensor]], decode: Optional[bool]=…, get_preds: Optional[bool]=…) -> list[torch.Tensor] - added:
method JukeboxPrior.forward_tokens(music_tokens, music_tokens_conds=…, metadata=…, get_preds=…, get_attn_weights=…) - added:
method JukeboxPrior.get_cond(music_tokens_conds, metadata) - added:
method JukeboxPrior.get_encoder_loss(last_encoder_hidden_states, target_lyrics) - added:
method JukeboxPrior.get_encoder_states(lyric_tokens, sample=…) - added:
method JukeboxPrior.get_metadata(labels, start, total_length, offset, get_indices=…) - added:
method JukeboxPrior.get_music_tokens_conds(music_tokens, start, end) - added:
method JukeboxPrior.prior_postprocess(tokens) - added:
method JukeboxPrior.prior_preprocess(tokens, conds) - added:
method JukeboxPrior.sample(n_samples, music_tokens=…, music_tokens_conds=…, metadata=…, temp=…, top_k=…, top_p=…, chunk_size=…, sample_tokens=…) - added:
method JukeboxPrior.set_metadata_lyric_tokens(labels) - added:
method JukeboxVQVAE.decode(music_tokens, start_level=…, end_level=…, bs_chunks=…) -> torch.Tensor - added:
method JukeboxVQVAE.encode(input_audio, start_level=…, end_level=…, bs_chunks=…) - added:
method JukeboxVQVAE.forward(raw_audio: torch.FloatTensor) -> tuple[torch.Tensor, torch.Tensor] - added:
method JukeboxVQVAE.sample(n_samples)
- added:
class JukeboxTokenizer(PreTrainedTokenizer) - added:
constructor JukeboxTokenizer(artists_file, genres_file, lyrics_file, version=…, max_n_lyric_tokens=…, n_genres=…, unk_token=…, **kwargs) - added:
field JukeboxTokenizer.model_input_names - added:
field JukeboxTokenizer.vocab_files_names - added:
method JukeboxTokenizer.convert_lyric_tokens_to_string(lyrics: list[str]) -> str - added:
method JukeboxTokenizer.convert_to_tensors(inputs, tensor_type: Optional[Union[str, TensorType]]=…, prepend_batch_axis: bool=…) - added:
method JukeboxTokenizer.get_vocab() - added:
method JukeboxTokenizer.prepare_for_tokenization(artists: str, genres: str, lyrics: str, is_split_into_words: bool=…) -> tuple[str, str, str, dict[str, Any]] - added:
method JukeboxTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method JukeboxTokenizer.tokenize(artist, genre, lyrics, **kwargs) - added:
property JukeboxTokenizer.vocab_size
- added:
export define_import_structure
- added:
class MCTCTConfig(PretrainedConfig) - added:
constructor MCTCTConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, intermediate_size=…, num_attention_heads=…, attention_head_dim=…, max_position_embeddings=…, layer_norm_eps=…, layerdrop=…, hidden_act=…, initializer_range=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, conv_glu_dim=…, conv_dropout=…, num_conv_layers=…, conv_kernel=… - added:
field MCTCTConfig.model_type
- added:
class MCTCTFeatureExtractor(SequenceFeatureExtractor) - added:
constructor MCTCTFeatureExtractor(feature_size=…, sampling_rate=…, padding_value=…, hop_length=…, win_length=…, win_function=…, frame_signal_scale=…, preemphasis_coeff=…, mel_floor=…, normalize_means=…, normalize_vars=…, return_attention_mask=…, **kwargs) - added:
field MCTCTFeatureExtractor.model_input_names - added:
method MCTCTFeatureExtractor.normalize(input_features: list[np.ndarray], attention_mask: Optional[np.ndarray]=…) -> list[np.ndarray]
- added:
class MCTCTForCTC(MCTCTPreTrainedModel) - added:
class MCTCTModel(MCTCTPreTrainedModel) - added:
class MCTCTPreTrainedModel(PreTrainedModel) - added:
constructor MCTCTForCTC(config) - added:
constructor MCTCTModel(config) - added:
field MCTCTPreTrainedModel.base_model_prefix - added:
field MCTCTPreTrainedModel.config: MCTCTConfig - added:
field MCTCTPreTrainedModel.main_input_name - added:
field MCTCTPreTrainedModel.supports_gradient_checkpointing - added:
method MCTCTForCTC.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.LongTensor]=…) -> Union[tuple, CausalLMOutput] - added:
method MCTCTModel.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
class MCTCTProcessor(ProcessorMixin) - added:
constructor MCTCTProcessor(feature_extractor, tokenizer) - added:
field MCTCTProcessor.feature_extractor_class - added:
field MCTCTProcessor.tokenizer_class - added:
method MCTCTProcessor.as_target_processor() - added:
method MCTCTProcessor.decode(*args, **kwargs) - added:
method MCTCTProcessor.pad(*args, **kwargs)
- added:
export define_import_structure
- added:
class MegaConfig(PretrainedConfig) - added:
class MegaOnnxConfig(OnnxConfig) - added:
constructor MegaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, intermediate_size=…, ema_projection_size=…, bidirectional=…, shared_representation_size=…, use_chunking=…, chunk_size=…, truncation=…, normalize_before_mega=…, normalization_type=…, norm_affine=…, activation=…, attention_activation=…, dropout_prob=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, use_feature_dropout=… - added:
field MegaConfig.model_type - added:
property MegaOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class MegaForCausalLM(MegaPreTrainedModel) - added:
class MegaForMaskedLM(MegaPreTrainedModel) - added:
class MegaForMultipleChoice(MegaPreTrainedModel) - added:
class MegaForQuestionAnswering(MegaPreTrainedModel) - added:
class MegaForSequenceClassification(MegaPreTrainedModel) - added:
class MegaForTokenClassification(MegaPreTrainedModel) - added:
class MegaModel(MegaPreTrainedModel) - added:
class MegaPreTrainedModel(PreTrainedModel) - added:
constructor MegaForCausalLM(config: MegaConfig) - added:
constructor MegaForMaskedLM(config: MegaConfig) - added:
constructor MegaForMultipleChoice(config) - added:
constructor MegaForQuestionAnswering(config) - added:
constructor MegaForSequenceClassification(config) - added:
constructor MegaForTokenClassification(config) - added:
constructor MegaModel(config: MegaConfig, add_pooling_layer=…) - added:
field MegaPreTrainedModel.base_model_prefix - added:
field MegaPreTrainedModel.config: MegaConfig - added:
field MegaPreTrainedModel.supports_gradient_checkpointing - added:
method MegaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, use_c… - added:
method MegaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method MegaForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outp… - added:
method MegaForMaskedLM.get_output_embeddings() - added:
method MegaForMaskedLM.set_output_embeddings(new_embeddings) - added:
method MegaForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], Multip… - added:
method MegaForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic… - added:
method MegaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor]… - added:
method MegaForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], T… - added:
method MegaModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method MegaModel.get_input_embeddings() - added:
method MegaModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class MMBTConfig - added:
constructor MMBTConfig(config, num_labels=…, modal_hidden_size=…)
- added:
class MMBTForClassification(nn.Module) - added:
class MMBTModel(nn.Module, ModuleUtilsMixin) - added:
class ModalEmbeddings(nn.Module) - added:
constructor MMBTForClassification(config, transformer, encoder) - added:
constructor MMBTModel(config, transformer, encoder) - added:
constructor ModalEmbeddings(config, encoder, embeddings) - added:
method MMBTForClassification.forward(input_modal, input_ids=…, modal_start_tokens=…, modal_end_tokens=…, attention_mask=…, token_type_ids=…, modal_token_type_ids=…, position_ids=…, modal_position_ids=…, head_mask=…, inputs_embeds=…, labels=…, return_dict=…) - added:
method MMBTModel.forward(input_modal, input_ids=…, modal_start_tokens=…, modal_end_tokens=…, attention_mask=…, token_type_ids=…, modal_token_type_ids=…, position_ids=…, modal_position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method MMBTModel.get_input_embeddings() - added:
method MMBTModel.set_input_embeddings(value) - added:
method ModalEmbeddings.forward(input_modal, start_token=…, end_token=…, position_ids=…, token_type_ids=…)
- added:
export define_import_structure
- added:
class NatConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor NatConfig(patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, kernel_size=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, initializer_range=…, layer_norm_eps=…, layer_scale_init_value=…, out_features=…, out_indices=…, **kwargs) - added:
field NatConfig.attribute_map - added:
field NatConfig.model_type
- added:
class NatBackbone(NatPreTrainedModel, BackboneMixin) - added:
class NatForImageClassification(NatPreTrainedModel) - added:
class NatModel(NatPreTrainedModel) - added:
class NatPreTrainedModel(PreTrainedModel) - added:
constructor NatBackbone(config) - added:
constructor NatForImageClassification(config) - added:
constructor NatModel(config, add_pooling_layer=…) - added:
field NatPreTrainedModel.base_model_prefix - added:
field NatPreTrainedModel.config: NatConfig - added:
field NatPreTrainedModel.main_input_name - added:
method NatBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method NatBackbone.get_input_embeddings() - added:
method NatForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, NatImageClassifierOutput] - added:
method NatModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, NatModelOutput] - added:
method NatModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class NezhaConfig(PretrainedConfig) - added:
constructor NezhaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, max_relative_position=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, classifier_dropout=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, use_cache=…, **kwargs) - added:
field NezhaConfig.model_type
- added:
class NezhaForMaskedLM(NezhaPreTrainedModel) - added:
class NezhaForMultipleChoice(NezhaPreTrainedModel) - added:
class NezhaForNextSentencePrediction(NezhaPreTrainedModel) - added:
class NezhaForPreTraining(NezhaPreTrainedModel) - added:
class NezhaForQuestionAnswering(NezhaPreTrainedModel) - added:
class NezhaForSequenceClassification(NezhaPreTrainedModel) - added:
class NezhaForTokenClassification(NezhaPreTrainedModel) - added:
class NezhaModel(NezhaPreTrainedModel) - added:
class NezhaPreTrainedModel(PreTrainedModel) - added:
constructor NezhaForMaskedLM(config) - added:
constructor NezhaForMultipleChoice(config) - added:
constructor NezhaForNextSentencePrediction(config) - added:
constructor NezhaForPreTraining(config) - added:
constructor NezhaForQuestionAnswering(config) - added:
constructor NezhaForSequenceClassification(config) - added:
constructor NezhaForTokenClassification(config) - added:
constructor NezhaModel(config, add_pooling_layer=…) - added:
field NezhaPreTrainedModel.base_model_prefix - added:
field NezhaPreTrainedModel.config: NezhaConfig - added:
field NezhaPreTrainedModel.load_tf_weights - added:
field NezhaPreTrainedModel.supports_gradient_checkpointing - added:
method NezhaForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…… - added:
method NezhaForMaskedLM.get_output_embeddings() - added:
method NezhaForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method NezhaForMaskedLM.set_output_embeddings(new_embeddings) - added:
method NezhaForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch… - added:
method NezhaForNextSentencePrediction.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) ->… - added:
method NezhaForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, next_sentence_label: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_… - added:
method NezhaForPreTraining.get_output_embeddings() - added:
method NezhaForPreTraining.set_output_embeddings(new_embeddings) - added:
method NezhaForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…… - added:
method NezhaForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tup… - added:
method NezhaForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[… - added:
method NezhaModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_att… - added:
method NezhaModel.get_input_embeddings() - added:
method NezhaModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class OpenLlamaConfig(PretrainedConfig) - added:
constructor OpenLlamaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, use_memory_efficient_attention=…, hidden_dropout_prob=…, attention_dropout_prob=…, use_stable_embedding=…, shared_i… - added:
field OpenLlamaConfig.model_type
- added:
class OpenLlamaForCausalLM(OpenLlamaPreTrainedModel) - added:
class OpenLlamaForSequenceClassification(OpenLlamaPreTrainedModel) - added:
class OpenLlamaModel(OpenLlamaPreTrainedModel) - added:
class OpenLlamaPreTrainedModel(PreTrainedModel) - added:
constructor OpenLlamaForCausalLM(config) - added:
constructor OpenLlamaForSequenceClassification(config) - added:
constructor OpenLlamaModel(config: OpenLlamaConfig) - added:
field OpenLlamaPreTrainedModel.base_model_prefix - added:
field OpenLlamaPreTrainedModel.config: OpenLlamaConfig - added:
field OpenLlamaPreTrainedModel.supports_gradient_checkpointing - added:
method OpenLlamaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dic… - added:
method OpenLlamaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, **kwargs) - added:
method OpenLlamaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]… - added:
method OpenLlamaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseMod…
- added:
export define_import_structure
- added:
class QDQBertConfig(PretrainedConfig) - added:
constructor QDQBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field QDQBertConfig.model_type
- added:
class QDQBertForMaskedLM(QDQBertPreTrainedModel) - added:
class QDQBertForMultipleChoice(QDQBertPreTrainedModel) - added:
class QDQBertForNextSentencePrediction(QDQBertPreTrainedModel) - added:
class QDQBertForQuestionAnswering(QDQBertPreTrainedModel) - added:
class QDQBertForSequenceClassification(QDQBertPreTrainedModel) - added:
class QDQBertForTokenClassification(QDQBertPreTrainedModel) - added:
class QDQBertLMHeadModel(QDQBertPreTrainedModel) - added:
class QDQBertLayer(GradientCheckpointingLayer) - added:
class QDQBertModel(QDQBertPreTrainedModel) - added:
class QDQBertPreTrainedModel(PreTrainedModel) - added:
constructor QDQBertForMaskedLM(config) - added:
constructor QDQBertForMultipleChoice(config) - added:
constructor QDQBertForNextSentencePrediction(config) - added:
constructor QDQBertForQuestionAnswering(config) - added:
constructor QDQBertForSequenceClassification(config) - added:
constructor QDQBertForTokenClassification(config) - added:
constructor QDQBertLMHeadModel(config) - added:
constructor QDQBertLayer(config) - added:
constructor QDQBertModel(config, add_pooling_layer: bool=…) - added:
field QDQBertPreTrainedModel.base_model_prefix - added:
field QDQBertPreTrainedModel.config: QDQBertConfig - added:
field QDQBertPreTrainedModel.load_tf_weights - added:
field QDQBertPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_qdqbert(model, tf_checkpoint_path) - added:
method QDQBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method QDQBertForMaskedLM.get_output_embeddings() - added:
method QDQBertForMaskedLM.prepare_inputs_for_generation(input_ids: torch.LongTensor, attention_mask: Optional[torch.FloatTensor]=…, **model_kwargs) - added:
method QDQBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method QDQBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method QDQBertForNextSentencePrediction.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method QDQBertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…… - added:
method QDQBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method QDQBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method QDQBertLMHeadModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, label… - added:
method QDQBertLMHeadModel.get_output_embeddings() - added:
method QDQBertLMHeadModel.prepare_inputs_for_generation(input_ids: Optional[torch.LongTensor], past_key_values=…, attention_mask: Optional[torch.Tensor]=…, **model_kwargs) - added:
method QDQBertLMHeadModel.set_output_embeddings(new_embeddings) - added:
method QDQBertLayer.feed_forward_chunk(attention_output) - added:
method QDQBertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, past_key_values=…, output_attentions=…) - added:
method QDQBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]… - added:
method QDQBertModel.get_input_embeddings() - added:
method QDQBertModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class RealmConfig(PretrainedConfig) - added:
constructor RealmConfig(vocab_size=…, hidden_size=…, retriever_proj_size=…, num_hidden_layers=…, num_attention_heads=…, num_candidates=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, span_hidden_size=…, max_span_width=…, reader_layer_norm_eps=…, reader_beam_size=…, rea… - added:
field RealmConfig.model_type
- added:
class RealmEmbedder(RealmPreTrainedModel) - added:
class RealmForOpenQA(RealmPreTrainedModel) - added:
class RealmKnowledgeAugEncoder(RealmPreTrainedModel) - added:
class RealmPreTrainedModel(PreTrainedModel) - added:
class RealmReader(RealmPreTrainedModel) - added:
class RealmScorer(RealmPreTrainedModel) - added:
constructor RealmEmbedder(config) - added:
constructor RealmForOpenQA(config, retriever=…) - added:
constructor RealmKnowledgeAugEncoder(config) - added:
constructor RealmReader(config) - added:
constructor RealmScorer(config, query_embedder=…) - added:
field RealmPreTrainedModel.base_model_prefix - added:
field RealmPreTrainedModel.config: RealmConfig - added:
field RealmPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_realm(model, config, tf_checkpoint_path) - added:
method RealmEmbedder.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]… - added:
method RealmEmbedder.get_input_embeddings() - added:
method RealmEmbedder.set_input_embeddings(value) - added:
method RealmForOpenQA.block_embedding_to(device) - added:
method RealmForOpenQA.forward(input_ids: Optional[torch.LongTensor], attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, answer_ids: Optional[torch.LongTensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, RealmForOpenQAOutput] - added:
method RealmKnowledgeAugEncoder.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, relevance_score: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, mlm_mas… - added:
method RealmKnowledgeAugEncoder.get_input_embeddings() - added:
method RealmKnowledgeAugEncoder.get_output_embeddings() - added:
method RealmKnowledgeAugEncoder.set_input_embeddings(value) - added:
method RealmKnowledgeAugEncoder.set_output_embeddings(new_embeddings) - added:
method RealmReader.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, relevance_score: Optional[torch.FloatTensor]=…, block_mask: Optional[torch.BoolTensor]=…, start_positions:… - added:
method RealmScorer.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, candidate_input_ids: Optional[torch.LongTensor]=…, candidate_attention_mask: Optional[torch.FloatTensor]=…, candidate_token_type_ids: Optional[torch.LongTensor]=…, candidate_inputs_embeds: Optio… - added:
property RealmForOpenQA.searcher_beam_size
- added:
class RealmRetriever - added:
constructor RealmRetriever(block_records, tokenizer) - added:
method RealmRetriever.block_has_answer(concat_inputs, answer_ids) - added:
method RealmRetriever.save_pretrained(save_directory) - added:
static method RealmRetriever.from_pretrained(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *init_inputs, **kwargs)
- added:
class RealmTokenizer(PreTrainedTokenizer) - added:
constructor RealmTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RealmTokenizer.vocab_files_names - added:
method RealmTokenizer.batch_encode_candidates(text, **kwargs) - added:
method RealmTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RealmTokenizer.convert_tokens_to_string(tokens) - added:
method RealmTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RealmTokenizer.get_vocab() - added:
method RealmTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RealmTokenizer.do_lower_case - added:
property RealmTokenizer.vocab_size
- added:
class RealmTokenizerFast(PreTrainedTokenizerFast) - added:
constructor RealmTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RealmTokenizerFast.slow_tokenizer_class - added:
field RealmTokenizerFast.vocab_files_names - added:
method RealmTokenizerFast.batch_encode_candidates(text, **kwargs) - added:
method RealmTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method RealmTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class RetriBertConfig(PretrainedConfig) - added:
constructor RetriBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, share_encoders=…, projection_dim=…, pad_token_id=…, **kwargs) - added:
field RetriBertConfig.model_type
- added:
class RetriBertModel(RetriBertPreTrainedModel) - added:
class RetriBertPreTrainedModel(PreTrainedModel) - added:
constructor RetriBertModel(config: RetriBertConfig) -> None - added:
field RetriBertPreTrainedModel.base_model_prefix - added:
field RetriBertPreTrainedModel.config: RetriBertConfig - added:
field RetriBertPreTrainedModel.load_tf_weights - added:
method RetriBertModel.embed_answers(input_ids, attention_mask=…, checkpoint_batch_size=…) - added:
method RetriBertModel.embed_questions(input_ids, attention_mask=…, checkpoint_batch_size=…) - added:
method RetriBertModel.embed_sentences_checkpointed(input_ids, attention_mask, sent_encoder, checkpoint_batch_size=…) - added:
method RetriBertModel.forward(input_ids_query: torch.LongTensor, attention_mask_query: Optional[torch.FloatTensor], input_ids_doc: torch.LongTensor, attention_mask_doc: Optional[torch.FloatTensor], checkpoint_batch_size: int=…) -> torch.FloatTensor
- added:
class RetriBertTokenizer(PreTrainedTokenizer) - added:
constructor RetriBertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RetriBertTokenizer.model_input_names - added:
field RetriBertTokenizer.vocab_files_names - added:
method RetriBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RetriBertTokenizer.convert_tokens_to_string(tokens) - added:
method RetriBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RetriBertTokenizer.get_vocab() - added:
method RetriBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RetriBertTokenizer.do_lower_case - added:
property RetriBertTokenizer.vocab_size
- added:
class RetriBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor RetriBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RetriBertTokenizerFast.model_input_names - added:
field RetriBertTokenizerFast.slow_tokenizer_class - added:
field RetriBertTokenizerFast.vocab_files_names - added:
method RetriBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method RetriBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class Speech2Text2Config(PretrainedConfig) - added:
constructor Speech2Text2Config(vocab_size=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, decoder_layerdrop=…, use_cache=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…, scale_embedding=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_target_positions=…, **kwargs) - added:
field Speech2Text2Config.attribute_map - added:
field Speech2Text2Config.keys_to_ignore_at_inference - added:
field Speech2Text2Config.model_type
- added:
class Speech2Text2ForCausalLM(Speech2Text2PreTrainedModel) - added:
class Speech2Text2PreTrainedModel(PreTrainedModel) - added:
constructor Speech2Text2ForCausalLM(config) - added:
field Speech2Text2PreTrainedModel.base_model_prefix - added:
field Speech2Text2PreTrainedModel.config: Speech2Text2Config - added:
field Speech2Text2PreTrainedModel.supports_gradient_checkpointing - added:
method Speech2Text2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, l… - added:
method Speech2Text2ForCausalLM.get_decoder() - added:
method Speech2Text2ForCausalLM.get_input_embeddings() - added:
method Speech2Text2ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, use_cache=…, **kwargs) - added:
method Speech2Text2ForCausalLM.set_decoder(decoder) - added:
method Speech2Text2ForCausalLM.set_input_embeddings(value)
- added:
class Speech2Text2Processor(ProcessorMixin) - added:
constructor Speech2Text2Processor(feature_extractor, tokenizer) - added:
field Speech2Text2Processor.feature_extractor_class - added:
field Speech2Text2Processor.tokenizer_class - added:
method Speech2Text2Processor.as_target_processor()
- added:
class Speech2Text2Tokenizer(PreTrainedTokenizer) - added:
constructor Speech2Text2Tokenizer(vocab_file, bos_token=…, pad_token=…, eos_token=…, unk_token=…, do_lower_case=…, merges_file=…, **kwargs) - added:
field Speech2Text2Tokenizer.model_input_names - added:
field Speech2Text2Tokenizer.vocab_files_names - added:
method Speech2Text2Tokenizer.bpe(token) - added:
method Speech2Text2Tokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method Speech2Text2Tokenizer.get_vocab() -> dict - added:
method Speech2Text2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property Speech2Text2Tokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class TapexTokenizer(PreTrainedTokenizer) - added:
constructor TapexTokenizer(vocab_file, merges_file, do_lower_case=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, max_cell_length=…, **kwargs) - added:
field TapexTokenizer.model_input_names - added:
field TapexTokenizer.vocab_files_names - added:
method TapexTokenizer.batch_encode_plus(table: Union[pd.DataFrame, list[pd.DataFrame]], query: Optional[list[TextInput]]=…, answer: Optional[list[str]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Optional[Union[bool, str]]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_tok… - added:
method TapexTokenizer.bpe(token) - added:
method TapexTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method TapexTokenizer.convert_tokens_to_string(tokens) - added:
method TapexTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method TapexTokenizer.delete_unrelated_rows(table_content: dict, question: str, answer: list, delete_ratio: float) - added:
method TapexTokenizer.encode(table: pd.DataFrame, query: Optional[TextInput]=…, answer: Optional[str]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, TapexTruncationStrategy]=…, max_length: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, **kwargs) -> list[int] - added:
method TapexTokenizer.encode_plus(table: pd.DataFrame, query: Optional[TextInput]=…, answer: Optional[str]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Optional[Union[bool, str]]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_token_type_ids: Optional[bool]=…, return_attenti… - added:
method TapexTokenizer.estimate_delete_ratio(table_content: dict, question: str, max_length=…) - added:
method TapexTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method TapexTokenizer.get_vocab() - added:
method TapexTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method TapexTokenizer.prepare_table_query(table, query, answer=…, truncation_strategy=…, max_length=…) - added:
method TapexTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method TapexTokenizer.source_call_func(table: Union[pd.DataFrame, list[pd.DataFrame]], query: Optional[Union[TextInput, list[TextInput]]]=…, answer: Optional[Union[str, list[str]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, return_t… - added:
method TapexTokenizer.target_batch_encode_plus(answer: list[str], add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Optional[Union[bool, str]]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_token_type_ids: Optional[bool]=…, return_attention_mask: Optional[bool]=…, return_overflowin… - added:
method TapexTokenizer.target_call_func(answer: Union[str, list[str]], add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_token_type_ids: Optional[bool]=…, return_attention_mask: Option… - added:
method TapexTokenizer.target_encode(answer: str, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, TapexTruncationStrategy]=…, max_length: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, **kwargs) -> list[int] - added:
method TapexTokenizer.target_encode_plus(answer: str, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Optional[Union[bool, str]]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_token_type_ids: Optional[bool]=…, return_attention_mask: Optional[bool]=…, return_special_tokens_mask: b… - added:
method TapexTokenizer.truncate_cell(cell_value) - added:
method TapexTokenizer.truncate_table_cells(table_content: dict, question: str, answer: list) - added:
method TapexTokenizer.truncate_table_rows(table_content: dict, question: str, answer: Optional[Union[str, list[str]]]=…, max_length=…) - added:
property TapexTokenizer.vocab_size
- added:
export define_import_structure
- added:
class TrajectoryTransformerConfig(PretrainedConfig) - added:
constructor TrajectoryTransformerConfig(vocab_size=…, action_weight=…, reward_weight=…, value_weight=…, block_size=…, action_dim=…, observation_dim=…, transition_dim=…, n_layer=…, n_head=…, n_embd=…, embd_pdrop=…, attn_pdrop=…, resid_pdrop=…, learning_rate=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, kaiming_initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…,… - added:
field TrajectoryTransformerConfig.attribute_map - added:
field TrajectoryTransformerConfig.keys_to_ignore_at_inference - added:
field TrajectoryTransformerConfig.model_type
- added:
class TrajectoryTransformerModel(TrajectoryTransformerPreTrainedModel) - added:
class TrajectoryTransformerPreTrainedModel(PreTrainedModel) - added:
constructor TrajectoryTransformerModel(config) - added:
field TrajectoryTransformerPreTrainedModel.base_model_prefix - added:
field TrajectoryTransformerPreTrainedModel.config: TrajectoryTransformerConfig - added:
field TrajectoryTransformerPreTrainedModel.load_tf_weights - added:
field TrajectoryTransformerPreTrainedModel.main_input_name - added:
field TrajectoryTransformerPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_trajectory_transformer(model, config, tf_checkpoint_path) - added:
method TrajectoryTransformerModel.forward(trajectories: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, targets: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], TrajectoryTransformerOu… - added:
method TrajectoryTransformerModel.get_block_size() - added:
method TrajectoryTransformerModel.offset_tokens(trajectories) - added:
method TrajectoryTransformerModel.pad_to_full_observation(hidden_states)
- added:
export define_import_structure
- added:
class TransfoXLConfig(PretrainedConfig) - added:
constructor TransfoXLConfig(vocab_size=…, cutoffs=…, d_model=…, d_embed=…, n_head=…, d_head=…, d_inner=…, div_val=…, pre_lnorm=…, n_layer=…, mem_len=…, clamp_len=…, same_length=…, proj_share_all_but_first=…, attn_type=…, sample_softmax=…, adaptive=…, dropout=…, dropatt=…, untie_r=…, init=…, init_range=…, proj_init_std=…, init_std=…, layer_norm_epsilon=…, eos_token_id=…, **kwargs) - added:
field TransfoXLConfig.attribute_map - added:
field TransfoXLConfig.keys_to_ignore_at_inference - added:
field TransfoXLConfig.model_type - added:
property TransfoXLConfig.max_position_embeddings
- added:
class TFAdaptiveEmbedding(keras.layers.Layer) - added:
class TFTransfoXLForSequenceClassification(TFTransfoXLPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFTransfoXLLMHeadModel(TFTransfoXLPreTrainedModel) - added:
class TFTransfoXLMainLayer(keras.layers.Layer) - added:
class TFTransfoXLModel(TFTransfoXLPreTrainedModel) - added:
class TFTransfoXLPreTrainedModel(TFPreTrainedModel) - added:
constructor TFAdaptiveEmbedding(n_token, d_embed, d_proj, cutoffs, div_val=…, init_std=…, sample_softmax=…, **kwargs) - added:
constructor TFTransfoXLForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFTransfoXLLMHeadModel(config) - added:
constructor TFTransfoXLMainLayer(config, **kwargs) - added:
constructor TFTransfoXLModel(config, *inputs, **kwargs) - added:
field TFTransfoXLMainLayer.config_class - added:
field TFTransfoXLPreTrainedModel.base_model_prefix - added:
field TFTransfoXLPreTrainedModel.config_class - added:
method TFAdaptiveEmbedding.build(input_shape) - added:
method TFAdaptiveEmbedding.call(inp) - added:
method TFTransfoXLForSequenceClassification.call(input_ids: TFModelInputType | None=…, mems: list[tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> tuple | TFTransfoXLS… - added:
method TFTransfoXLForSequenceClassification.get_output_embeddings() - added:
method TFTransfoXLLMHeadModel.call(input_ids: TFModelInputType | None=…, mems: list[tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) -> TFTransfoXLLMHeadModelOutput | tuple[tf.T… - added:
method TFTransfoXLLMHeadModel.get_output_embeddings() - added:
method TFTransfoXLLMHeadModel.init_mems(bsz) - added:
method TFTransfoXLLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, **model_kwargs) - added:
method TFTransfoXLLMHeadModel.reset_memory_length(mem_len) - added:
method TFTransfoXLLMHeadModel.tf_to_pt_weight_rename(tf_weight) - added:
method TFTransfoXLMainLayer.backward_compatible() - added:
method TFTransfoXLMainLayer.build(input_shape) - added:
method TFTransfoXLMainLayer.call(input_ids: TFModelInputType | None=…, mems: list[tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) - added:
method TFTransfoXLMainLayer.get_input_embeddings() - added:
method TFTransfoXLMainLayer.init_mems(bsz) - added:
method TFTransfoXLMainLayer.reset_memory_length(mem_len) - added:
method TFTransfoXLMainLayer.set_input_embeddings(value) - added:
method TFTransfoXLModel.call(input_ids: TFModelInputType | None=…, mems: list[tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFTransfoXLModelOutput | tuple[tf.Tensor]
- added:
class TFAdaptiveSoftmaxMask(keras.layers.Layer) - added:
constructor TFAdaptiveSoftmaxMask(vocab_size, d_embed, d_proj, cutoffs, div_val=…, keep_order=…, **kwargs) - added:
method TFAdaptiveSoftmaxMask.build(input_shape) - added:
method TFAdaptiveSoftmaxMask.call(hidden, target, return_mean=…, training=…)
- added:
class AdaptiveEmbedding(nn.Module) - added:
class TransfoXLForSequenceClassification(TransfoXLPreTrainedModel) - added:
class TransfoXLLMHeadModel(TransfoXLPreTrainedModel) - added:
class TransfoXLModel(TransfoXLPreTrainedModel) - added:
class TransfoXLPreTrainedModel(PreTrainedModel) - added:
constructor AdaptiveEmbedding(n_token, d_embed, d_proj, cutoffs, div_val=…, sample_softmax=…) - added:
constructor TransfoXLForSequenceClassification(config) - added:
constructor TransfoXLLMHeadModel(config) - added:
constructor TransfoXLModel(config) - added:
field TransfoXLPreTrainedModel.base_model_prefix - added:
field TransfoXLPreTrainedModel.config: TransfoXLConfig - added:
field TransfoXLPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_transfo_xl(model, config, tf_path) - added:
method AdaptiveEmbedding.forward(inp) - added:
method TransfoXLForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, mems: Optional[list[torch.FloatTensor]]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TransfoXLSequen… - added:
method TransfoXLLMHeadModel.forward(input_ids: Optional[torch.LongTensor]=…, mems: Optional[list[torch.FloatTensor]]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TransfoXLLMHeadModelOutput] - added:
method TransfoXLLMHeadModel.get_output_embeddings() - added:
method TransfoXLLMHeadModel.init_mems(bsz) - added:
method TransfoXLLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, **model_kwargs) - added:
method TransfoXLLMHeadModel.reset_memory_length(mem_len) - added:
method TransfoXLLMHeadModel.tie_weights() - added:
method TransfoXLModel.backward_compatible() - added:
method TransfoXLModel.forward(input_ids: Optional[torch.LongTensor]=…, mems: Optional[list[torch.FloatTensor]]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TransfoXLModelOutput] - added:
method TransfoXLModel.get_input_embeddings() - added:
method TransfoXLModel.init_mems(bsz) - added:
method TransfoXLModel.reset_memory_length(mem_len) - added:
method TransfoXLModel.set_input_embeddings(new_embeddings) - added:
method TransfoXLPreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…, layer: Optional[int]=…)
- added:
class ProjectedAdaptiveLogSoftmax(nn.Module) - added:
constructor ProjectedAdaptiveLogSoftmax(n_token, d_embed, d_proj, cutoffs, div_val=…, keep_order=…) - added:
method ProjectedAdaptiveLogSoftmax.forward(hidden, labels=…, keep_order=…) - added:
method ProjectedAdaptiveLogSoftmax.log_prob(hidden)
- added:
class TransfoXLCorpus - added:
class TransfoXLTokenizer(PreTrainedTokenizer) - added:
constructor TransfoXLCorpus(*args, **kwargs) - added:
constructor TransfoXLTokenizer(special=…, min_freq=…, max_size=…, lower_case=…, delimiter=…, vocab_file=…, pretrained_vocab_file: Optional[str]=…, never_split=…, unk_token=…, eos_token=…, additional_special_tokens=…, language=…, **kwargs) - added:
field TransfoXLTokenizer.model_input_names - added:
field TransfoXLTokenizer.vocab_files_names - added:
method TransfoXLCorpus.build_corpus(path, dataset) - added:
method TransfoXLCorpus.get_iterator(split, *args, **kwargs) - added:
method TransfoXLTokenizer.add_special(sym) - added:
method TransfoXLTokenizer.add_symbol(sym) - added:
method TransfoXLTokenizer.build_vocab() - added:
method TransfoXLTokenizer.convert_to_tensor(symbols) - added:
method TransfoXLTokenizer.convert_tokens_to_string(tokens) - added:
method TransfoXLTokenizer.count_file(path, verbose=…, add_eos=…) - added:
method TransfoXLTokenizer.count_sents(sents, verbose=…) - added:
method TransfoXLTokenizer.encode_file(path, ordered=…, verbose=…, add_eos=…, add_double_eos=…) - added:
method TransfoXLTokenizer.encode_sents(sents, ordered=…, verbose=…) - added:
method TransfoXLTokenizer.get_vocab() - added:
method TransfoXLTokenizer.moses_pipeline(text: str) -> list[str] - added:
method TransfoXLTokenizer.moses_punct_norm(text) - added:
method TransfoXLTokenizer.moses_tokenize(text) - added:
method TransfoXLTokenizer.move_added_token(token: str, target_idx: int) - added:
method TransfoXLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property TransfoXLTokenizer.do_lower_case - added:
property TransfoXLTokenizer.vocab_size - added:
static method TransfoXLCorpus.from_pretrained(pretrained_model_name_or_path, cache_dir=…, *inputs, **kwargs)
- added:
export define_import_structure
- added:
class TvltConfig(PretrainedConfig) - added:
constructor TvltConfig(image_size=…, spectrogram_length=…, frequency_length=…, image_patch_size=…, audio_patch_size=…, num_image_channels=…, num_audio_channels=…, num_frames=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, use_mean_pooling=…, de… - added:
field TvltConfig.model_type
- added:
class TvltFeatureExtractor(SequenceFeatureExtractor) - added:
constructor TvltFeatureExtractor(spectrogram_length=…, num_channels=…, patch_size=…, feature_size=…, sampling_rate=…, hop_length_to_sampling_rate=…, n_fft=…, padding_value=…, **kwargs) - added:
field TvltFeatureExtractor.model_input_names
- added:
class TvltImageProcessor(BaseImageProcessor) - added:
constructor TvltImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, patch_size: list[int]=…, num_frames: int=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]… - added:
field TvltImageProcessor.model_input_names - added:
method TvltImageProcessor.preprocess(videos: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, patch_size: Optional[list[int]]=…, num_frames: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, ima… - added:
method TvltImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class TvltForAudioVisualClassification(TvltPreTrainedModel) - added:
class TvltForPreTraining(TvltPreTrainedModel) - added:
class TvltModel(TvltPreTrainedModel) - added:
class TvltPreTrainedModel(PreTrainedModel) - added:
constructor TvltForAudioVisualClassification(config) - added:
constructor TvltForPreTraining(config) - added:
constructor TvltModel(config) - added:
field TvltPreTrainedModel.base_model_prefix - added:
field TvltPreTrainedModel.config: TvltConfig - added:
field TvltPreTrainedModel.main_input_name - added:
field TvltPreTrainedModel.supports_gradient_checkpointing - added:
method TvltForAudioVisualClassification.forward(pixel_values: torch.FloatTensor, audio_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, audio_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.LongTensor]=…) -> Union[tuple[torch.FloatTensor], SequenceClassifier… - added:
method TvltForPreTraining.audio_mae_loss(audio_values, audio_predictions, mask) - added:
method TvltForPreTraining.concatenate_mask(mask_token, sequence, ids_restore) - added:
method TvltForPreTraining.forward(pixel_values: torch.FloatTensor, audio_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, audio_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values_mixed: Optional[torch.FloatTensor]=…, pixel_mask_mixed: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…… - added:
method TvltForPreTraining.patchify_audio(audio_values) - added:
method TvltForPreTraining.patchify_pixel(pixel_values) - added:
method TvltForPreTraining.pixel_mae_loss(pixel_values, pixel_predictions, mask) - added:
method TvltModel.forward(pixel_values: torch.FloatTensor, audio_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, audio_mask: Optional[torch.FloatTensor]=…, mask_pixel: bool=…, mask_audio: bool=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], TvltModelOutput] - added:
method TvltModel.get_input_embeddings()
- added:
class TvltProcessor(ProcessorMixin) - added:
constructor TvltProcessor(image_processor, feature_extractor) - added:
field TvltProcessor.attributes - added:
field TvltProcessor.feature_extractor_class - added:
field TvltProcessor.image_processor_class
- added:
export define_import_structure
- added:
class VanConfig(PretrainedConfig) - added:
constructor VanConfig(image_size=…, num_channels=…, patch_sizes=…, strides=…, hidden_sizes=…, depths=…, mlp_ratios=…, hidden_act=…, initializer_range=…, layer_norm_eps=…, layer_scale_init_value=…, drop_path_rate=…, dropout_rate=…, **kwargs) - added:
field VanConfig.model_type
- added:
class VanForImageClassification(VanPreTrainedModel) - added:
class VanModel(VanPreTrainedModel) - added:
class VanPreTrainedModel(PreTrainedModel) - added:
constructor VanForImageClassification(config) - added:
constructor VanModel(config) - added:
field VanPreTrainedModel.base_model_prefix - added:
field VanPreTrainedModel.config: VanConfig - added:
field VanPreTrainedModel.main_input_name - added:
field VanPreTrainedModel.supports_gradient_checkpointing - added:
method VanForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method VanModel.forward(pixel_values: Optional[torch.FloatTensor], output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class ViTHybridConfig(PretrainedConfig) - added:
constructor ViTHybridConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, backbone_featmap_shape=…, qkv_bias=…, **k… - added:
field ViTHybridConfig.model_type - added:
property ViTHybridConfig.sub_configs
- added:
class ViTHybridImageProcessor(BaseImageProcessor) - added:
constructor ViTHybridImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -… - added:
field ViTHybridImageProcessor.model_input_names - added:
method ViTHybridImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Un… - added:
method ViTHybridImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class ViTHybridForImageClassification(ViTHybridPreTrainedModel) - added:
class ViTHybridModel(ViTHybridPreTrainedModel) - added:
class ViTHybridPreTrainedModel(PreTrainedModel) - added:
constructor ViTHybridForImageClassification(config: ViTHybridConfig) -> None - added:
constructor ViTHybridModel(config: ViTHybridConfig, add_pooling_layer: bool=…, use_mask_token: bool=…) - added:
field ViTHybridPreTrainedModel.base_model_prefix - added:
field ViTHybridPreTrainedModel.config: ViTHybridConfig - added:
field ViTHybridPreTrainedModel.main_input_name - added:
field ViTHybridPreTrainedModel.supports_gradient_checkpointing - added:
method ViTHybridForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method ViTHybridModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method ViTHybridModel.get_input_embeddings() -> ViTHybridPatchEmbeddings
- added:
export define_import_structure
- added:
class XLMProphetNetConfig(PretrainedConfig) - added:
constructor XLMProphetNetConfig(activation_dropout: Optional[float]=…, activation_function: Optional[Union[str, Callable]]=…, vocab_size: Optional[int]=…, hidden_size: Optional[int]=…, encoder_ffn_dim: Optional[int]=…, num_encoder_layers: Optional[int]=…, num_encoder_attention_heads: Optional[int]=…, decoder_ffn_dim: Optional[int]=…, num_decoder_layers: Optional[int]=…, num_decoder_attention_head… - added:
field XLMProphetNetConfig.attribute_map - added:
field XLMProphetNetConfig.keys_to_ignore_at_inference - added:
field XLMProphetNetConfig.model_type - added:
property XLMProphetNetConfig.num_hidden_layers -> int
- added:
class XLMProphetNetDecoder(XLMProphetNetPreTrainedModel) - added:
class XLMProphetNetEncoder(XLMProphetNetPreTrainedModel) - added:
class XLMProphetNetForCausalLM(XLMProphetNetPreTrainedModel) - added:
class XLMProphetNetForConditionalGeneration(XLMProphetNetPreTrainedModel) - added:
class XLMProphetNetModel(XLMProphetNetPreTrainedModel) - added:
class XLMProphetNetPreTrainedModel(PreTrainedModel) - added:
constructor XLMProphetNetDecoder(config: XLMProphetNetConfig, word_embeddings: Optional[nn.Embedding]=…) - added:
constructor XLMProphetNetEncoder(config: XLMProphetNetConfig, word_embeddings: Optional[nn.Embedding]=…) - added:
constructor XLMProphetNetForCausalLM(config: XLMProphetNetConfig) - added:
constructor XLMProphetNetForConditionalGeneration(config: XLMProphetNetConfig) - added:
constructor XLMProphetNetModel(config: XLMProphetNetConfig) - added:
field XLMProphetNetPreTrainedModel.base_model_prefix - added:
field XLMProphetNetPreTrainedModel.config: XLMProphetNetConfig - added:
field XLMProphetNetPreTrainedModel.supports_gradient_checkpointing - added:
method XLMProphetNetDecoder.compute_buffered_relative_buckets(position_ids) - added:
method XLMProphetNetDecoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[boo… - added:
method XLMProphetNetDecoder.get_input_embeddings() - added:
method XLMProphetNetDecoder.prepare_attention_mask(hidden_states, attention_mask) - added:
method XLMProphetNetDecoder.prepare_predict_attention_mask(hidden_states, attention_mask) - added:
method XLMProphetNetDecoder.set_input_embeddings(value) - added:
method XLMProphetNetEncoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method XLMProphetNetEncoder.get_input_embeddings() - added:
method XLMProphetNetEncoder.set_input_embeddings(value) - added:
method XLMProphetNetForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[to… - added:
method XLMProphetNetForCausalLM.get_decoder() - added:
method XLMProphetNetForCausalLM.get_input_embeddings() - added:
method XLMProphetNetForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, head_mask=…, use_cache=…, **kwargs) - added:
method XLMProphetNetForCausalLM.set_decoder(decoder) - added:
method XLMProphetNetForCausalLM.set_input_embeddings(value) - added:
method XLMProphetNetForConditionalGeneration.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[torch.Tensor… - added:
method XLMProphetNetForConditionalGeneration.get_decoder() - added:
method XLMProphetNetForConditionalGeneration.get_encoder() - added:
method XLMProphetNetForConditionalGeneration.get_input_embeddings() - added:
method XLMProphetNetForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method XLMProphetNetForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method XLMProphetNetModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple]=…, past_key_values: Opti… - added:
method XLMProphetNetModel.get_encoder() - added:
method XLMProphetNetModel.get_input_embeddings() - added:
method XLMProphetNetModel.set_input_embeddings(value)
- added:
class XLMProphetNetTokenizer(PreTrainedTokenizer) - added:
constructor XLMProphetNetTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, unk_token=…, pad_token=…, cls_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field XLMProphetNetTokenizer.model_input_names - added:
field XLMProphetNetTokenizer.vocab_files_names - added:
method XLMProphetNetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMProphetNetTokenizer.convert_tokens_to_string(tokens) - added:
method XLMProphetNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMProphetNetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method XLMProphetNetTokenizer.get_vocab() - added:
method XLMProphetNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property XLMProphetNetTokenizer.vocab_size
- added:
export define_import_structure
- added:
class DepthAnythingConfig(PretrainedConfig) - added:
constructor DepthAnythingConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, patch_size=…, initializer_range=…, reassemble_hidden_size=…, reassemble_factors=…, neck_hidden_sizes=…, fusion_hidden_size=…, head_in_index=…, head_hidden_size=…, depth_estimation_type=…, max_depth=…, **kwargs) - added:
field DepthAnythingConfig.model_type - added:
method DepthAnythingConfig.to_dict() - added:
property DepthAnythingConfig.sub_configs
- added:
class DepthAnythingForDepthEstimation(DepthAnythingPreTrainedModel) - added:
class DepthAnythingPreTrainedModel(PreTrainedModel) - added:
constructor DepthAnythingForDepthEstimation(config) - added:
field DepthAnythingPreTrainedModel.base_model_prefix - added:
field DepthAnythingPreTrainedModel.config: DepthAnythingConfig - added:
field DepthAnythingPreTrainedModel.main_input_name - added:
field DepthAnythingPreTrainedModel.supports_gradient_checkpointing - added:
method DepthAnythingForDepthEstimation.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthEstimatorOutput]
- added:
export define_import_structure
- added:
class DepthProConfig(PretrainedConfig) - added:
constructor DepthProConfig(fusion_hidden_size=…, patch_size=…, initializer_range=…, intermediate_hook_ids=…, intermediate_feature_dims=…, scaled_images_ratios=…, scaled_images_overlap_ratios=…, scaled_images_feature_dims=…, merge_padding_value=…, use_batch_norm_in_fusion_residual=…, use_bias_in_fusion_residual=…, use_fov_model=…, num_fov_head_layers=…, image_model_config=…, patch_model_config=…,… - added:
field DepthProConfig.model_type - added:
field DepthProConfig.sub_configs
- added:
class DepthProImageProcessor(BaseImageProcessor) - added:
constructor DepthProImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) - added:
field DepthProImageProcessor.model_input_names - added:
method DepthProImageProcessor.post_process_depth_estimation(outputs: DepthProDepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method DepthProImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str,… - added:
method DepthProImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DepthProImageProcessorFast(BaseImageProcessorFast) - added:
field DepthProImageProcessorFast.do_normalize - added:
field DepthProImageProcessorFast.do_rescale - added:
field DepthProImageProcessorFast.do_resize - added:
field DepthProImageProcessorFast.image_mean - added:
field DepthProImageProcessorFast.image_std - added:
field DepthProImageProcessorFast.resample - added:
field DepthProImageProcessorFast.size - added:
method DepthProImageProcessorFast.post_process_depth_estimation(outputs: DepthProDepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]]
- added:
class DepthProForDepthEstimation(DepthProPreTrainedModel) - added:
class DepthProModel(DepthProPreTrainedModel) - added:
class DepthProPreTrainedModel(PreTrainedModel) - added:
constructor DepthProForDepthEstimation(config, use_fov_model=…) - added:
constructor DepthProModel(config) - added:
field DepthProPreTrainedModel.base_model_prefix - added:
field DepthProPreTrainedModel.config: DepthProConfig - added:
field DepthProPreTrainedModel.main_input_name - added:
field DepthProPreTrainedModel.supports_gradient_checkpointing - added:
method DepthProForDepthEstimation.forward(pixel_values: torch.FloatTensor, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthProDepthEstimatorOutput] - added:
method DepthProModel.forward(pixel_values: torch.FloatTensor, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, DepthProOutput] - added:
method DepthProModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class DetrConfig(PretrainedConfig) - added:
class DetrOnnxConfig(OnnxConfig) - added:
constructor DetrConfig(use_timm_backbone=…, backbone_config=…, num_channels=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, init_xavier_s… - added:
field DetrConfig.attribute_map - added:
field DetrConfig.keys_to_ignore_at_inference - added:
field DetrConfig.model_type - added:
field DetrOnnxConfig.torch_onnx_minimum_version - added:
property DetrConfig.hidden_size -> int - added:
property DetrConfig.num_attention_heads -> int - added:
property DetrConfig.sub_configs - added:
property DetrOnnxConfig.atol_for_validation -> float - added:
property DetrOnnxConfig.default_onnx_opset -> int - added:
property DetrOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
static method DetrConfig.from_backbone_config(backbone_config: PretrainedConfig, **kwargs)
- added:
class DetrFeatureExtractor(DetrImageProcessor) - added:
constructor DetrFeatureExtractor(*args, **kwargs) -> None
- added:
class DetrImageProcessor(BaseImageProcessor) - added:
constructor DetrImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: Optional[bool]=…, do_pad: bool=…, pad_s… - added:
field DetrImageProcessor.model_input_names - added:
method DetrImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method DetrImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_size: Optiona… - added:
method DetrImageProcessor.post_process(outputs, target_sizes) - added:
method DetrImageProcessor.post_process_instance(results, outputs, orig_target_sizes, max_target_sizes, threshold=…) - added:
method DetrImageProcessor.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) -> list[dict] - added:
method DetrImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…) - added:
method DetrImageProcessor.post_process_panoptic(outputs, processed_sizes, target_sizes=…, is_thing_map=…, threshold=…) - added:
method DetrImageProcessor.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method DetrImageProcessor.post_process_segmentation(outputs, target_sizes, threshold=…, mask_threshold=…) - added:
method DetrImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) - added:
method DetrImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method DetrImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[bo… - added:
method DetrImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DetrImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method DetrImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict - added:
static method DetrImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class DetrImageProcessorFast(BaseImageProcessorFast) - added:
constructor DetrImageProcessorFast(**kwargs: Unpack[DetrFastImageProcessorKwargs]) -> None - added:
field DetrImageProcessorFast.default_to_square - added:
field DetrImageProcessorFast.do_normalize - added:
field DetrImageProcessorFast.do_pad - added:
field DetrImageProcessorFast.do_rescale - added:
field DetrImageProcessorFast.do_resize - added:
field DetrImageProcessorFast.format - added:
field DetrImageProcessorFast.image_mean - added:
field DetrImageProcessorFast.image_std - added:
field DetrImageProcessorFast.model_input_names - added:
field DetrImageProcessorFast.resample - added:
field DetrImageProcessorFast.size - added:
field DetrImageProcessorFast.valid_kwargs - added:
method DetrImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method DetrImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method DetrImageProcessorFast.post_process(outputs, target_sizes) - added:
method DetrImageProcessorFast.post_process_instance(results, outputs, orig_target_sizes, max_target_sizes, threshold=…) - added:
method DetrImageProcessorFast.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) -> list[dict] - added:
method DetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…) - added:
method DetrImageProcessorFast.post_process_panoptic(outputs, processed_sizes, target_sizes=…, is_thing_map=…, threshold=…) - added:
method DetrImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method DetrImageProcessorFast.post_process_segmentation(outputs, target_sizes, threshold=…, mask_threshold=…) - added:
method DetrImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) - added:
method DetrImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method DetrImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[DetrFastImageProcessorKwargs]) -> BatchFeature - added:
method DetrImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method DetrImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…) - added:
static method DetrImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class DetrForObjectDetection(DetrPreTrainedModel) - added:
class DetrForSegmentation(DetrPreTrainedModel) - added:
class DetrModel(DetrPreTrainedModel) - added:
class DetrPreTrainedModel(PreTrainedModel) - added:
constructor DetrForObjectDetection(config: DetrConfig) - added:
constructor DetrForSegmentation(config: DetrConfig) - added:
constructor DetrModel(config: DetrConfig) - added:
field DetrPreTrainedModel.base_model_prefix - added:
field DetrPreTrainedModel.config: DetrConfig - added:
field DetrPreTrainedModel.main_input_name - added:
method DetrForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_… - added:
method DetrForSegmentation.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_sta… - added:
method DetrModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Option… - added:
method DetrModel.freeze_backbone() - added:
method DetrModel.get_encoder() - added:
method DetrModel.unfreeze_backbone()
- added:
export define_import_structure
- added:
class DiaConfig(PretrainedConfig) - added:
class DiaDecoderConfig(PretrainedConfig) - added:
class DiaEncoderConfig(PretrainedConfig) - added:
constructor DiaConfig(encoder_config: Optional[DiaEncoderConfig]=…, decoder_config: Optional[DiaDecoderConfig]=…, norm_eps: float=…, is_encoder_decoder: bool=…, pad_token_id: int=…, eos_token_id: int=…, bos_token_id: int=…, delay_pattern: Optional[list[int]]=…, initializer_range: float=…, use_cache: bool=…, **kwargs) - added:
constructor DiaDecoderConfig(max_position_embeddings: int=…, num_hidden_layers: int=…, hidden_size: int=…, intermediate_size: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, head_dim: int=…, cross_num_attention_heads: int=…, cross_head_dim: int=…, cross_num_key_value_heads: int=…, cross_hidden_size: int=…, norm_eps: float=…, vocab_size: int=…, hidden_act: str=…, num_channels: int=…… - added:
constructor DiaEncoderConfig(max_position_embeddings: int=…, num_hidden_layers: int=…, hidden_size: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, head_dim: int=…, intermediate_size: int=…, norm_eps: float=…, vocab_size: int=…, hidden_act: str=…, rope_theta: float=…, rope_scaling: Optional[dict]=…, initializer_range: float=…, **kwargs) - added:
field DiaConfig.keys_to_ignore_at_inference - added:
field DiaConfig.model_type - added:
field DiaConfig.sub_configs - added:
field DiaDecoderConfig.model_type - added:
field DiaEncoderConfig.model_type - added:
method DiaConfig.get_text_config(*args, **kwargs)
- added:
class DiaFeatureExtractor(SequenceFeatureExtractor) - added:
constructor DiaFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, hop_length: int=…, **kwargs) - added:
field DiaFeatureExtractor.model_input_names
- added:
class DiaGenerationMixin(GenerationMixin) - added:
method DiaGenerationMixin.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], list[int]]]=…, synced_gpus: Optional[bool]=…, assistant_model: Optional[PreTrainedModel]=…, streamer: Optional[B… - added:
method DiaGenerationMixin.prepare_inputs_for_generation(input_ids, encoder_outputs=…, decoder_delay_mask=…, **kwargs) - added:
static method DiaGenerationMixin.apply_delay_mask(input_ids: torch.Tensor, pad_id: int, delay_mask: Optional[torch.Tensor]) -> torch.Tensor - added:
variable logger
- added:
class DiaForConditionalGeneration(DiaPreTrainedModel, DiaGenerationMixin) - added:
class DiaModel(DiaPreTrainedModel) - added:
class DiaPreTrainedModel(PreTrainedModel) - added:
constructor DiaForConditionalGeneration(config: DiaConfig) - added:
constructor DiaModel(config: DiaConfig) - added:
field DiaForConditionalGeneration.base_model_prefix - added:
field DiaPreTrainedModel.base_model_prefix - added:
field DiaPreTrainedModel.config: DiaConfig - added:
field DiaPreTrainedModel.main_input_name - added:
field DiaPreTrainedModel.supports_gradient_checkpointing - added:
method DiaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[Union[BaseModelOutput, tuple]]=…, past_key_values: Optional[EncoderDecoderCache]=…, use_… - added:
method DiaForConditionalGeneration.get_decoder() - added:
method DiaForConditionalGeneration.get_encoder() - added:
method DiaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[Union[BaseModelOutput, tuple]]=…, past_key_values: Optional[EncoderDecoderCache]=…, use_cache: Optional[boo… - added:
method DiaModel.get_encoder()
- added:
class DiaForConditionalGeneration(DiaPreTrainedModel, DiaGenerationMixin) - added:
class DiaModel(DiaPreTrainedModel) - added:
class DiaPreTrainedModel(PreTrainedModel) - added:
constructor DiaForConditionalGeneration(config: DiaConfig) - added:
constructor DiaModel(config: DiaConfig) - added:
field DiaForConditionalGeneration.base_model_prefix - added:
field DiaPreTrainedModel.base_model_prefix - added:
field DiaPreTrainedModel.config: DiaConfig - added:
field DiaPreTrainedModel.main_input_name - added:
field DiaPreTrainedModel.supports_gradient_checkpointing - added:
method DiaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[Union[BaseModelOutput, tuple]]=…, past_key_values: Optional[EncoderDecoderCache]=…, use_… - added:
method DiaForConditionalGeneration.get_decoder() - added:
method DiaForConditionalGeneration.get_encoder() - added:
method DiaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[Union[BaseModelOutput, tuple]]=…, past_key_values: Optional[EncoderDecoderCache]=…, use_cache: Optional[boo… - added:
method DiaModel.get_encoder()
- added:
class DiaProcessor(ProcessorMixin) - added:
constructor DiaProcessor(feature_extractor, tokenizer, audio_tokenizer) - added:
field DiaProcessor.audio_tokenizer_class - added:
field DiaProcessor.feature_extractor_class - added:
field DiaProcessor.tokenizer_class - added:
method DiaProcessor.batch_decode(decoder_input_ids: torch.Tensor, audio_prompt_len: Optional[int]=…, **kwargs: Unpack[DiaProcessorKwargs]) -> list[torch.Tensor] - added:
method DiaProcessor.decode(decoder_input_ids: torch.Tensor, audio_prompt_len: Optional[int]=…, **kwargs: Unpack[DiaProcessorKwargs]) -> torch.Tensor - added:
method DiaProcessor.get_audio_prompt_len(decoder_attention_mask: torch.Tensor, **kwargs: Unpack[DiaProcessorKwargs]) -> int - added:
method DiaProcessor.save_audio(audio: AudioInput, saving_path: Union[str, Path, list[Union[str, Path]]], **kwargs: Unpack[DiaProcessorKwargs]) - added:
static method DiaProcessor.apply_audio_delay(audio: torch.Tensor, pad_token_id: int, bos_token_id: int, precomputed_idx: tuple[torch.Tensor, torch.Tensor]) -> torch.Tensor - added:
static method DiaProcessor.build_indices(bsz: int, seq_len: int, num_channels: int, delay_pattern: list[int], revert: bool=…) -> tuple[torch.Tensor, torch.Tensor]
- added:
class DiaTokenizer(PreTrainedTokenizer) - added:
constructor DiaTokenizer(pad_token: Optional[str]=…, unk_token: Optional[str]=…, max_length: Optional[int]=…, offset: int=…, **kwargs) - added:
field DiaTokenizer.model_input_names - added:
method DiaTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method DiaTokenizer.get_vocab() - added:
method DiaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property DiaTokenizer.vocab_size
- added:
export define_import_structure
- added:
class DiffLlamaConfig(PretrainedConfig) - added:
constructor DiffLlamaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, lambda_std_dev=…, head_… - added:
field DiffLlamaConfig.keys_to_ignore_at_inference - added:
field DiffLlamaConfig.model_type
- added:
class DiffLlamaForCausalLM(DiffLlamaPreTrainedModel, GenerationMixin) - added:
class DiffLlamaForQuestionAnswering(GenericForQuestionAnswering, DiffLlamaPreTrainedModel) - added:
class DiffLlamaForSequenceClassification(GenericForSequenceClassification, DiffLlamaPreTrainedModel) - added:
class DiffLlamaForTokenClassification(GenericForTokenClassification, DiffLlamaPreTrainedModel) - added:
class DiffLlamaModel(DiffLlamaPreTrainedModel) - added:
class DiffLlamaPreTrainedModel(PreTrainedModel) - added:
constructor DiffLlamaForCausalLM(config) - added:
constructor DiffLlamaModel(config: DiffLlamaConfig) - added:
field DiffLlamaForQuestionAnswering.base_model_prefix - added:
field DiffLlamaPreTrainedModel.base_model_prefix - added:
field DiffLlamaPreTrainedModel.config: DiffLlamaConfig - added:
field DiffLlamaPreTrainedModel.supports_gradient_checkpointing - added:
method DiffLlamaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=… - added:
method DiffLlamaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class DiffLlamaForCausalLM(GemmaForCausalLM) - added:
class DiffLlamaForQuestionAnswering(LlamaForQuestionAnswering) - added:
class DiffLlamaForSequenceClassification(LlamaForSequenceClassification) - added:
class DiffLlamaForTokenClassification(LlamaForTokenClassification) - added:
class DiffLlamaModel(LlamaModel) - added:
class DiffLlamaPreTrainedModel(LlamaPreTrainedModel)
- added:
export define_import_structure
- added:
class DinatConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor DinatConfig(patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, kernel_size=…, dilations=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, initializer_range=…, layer_norm_eps=…, layer_scale_init_value=…, out_features=…, out_indices=…, **kwargs) - added:
field DinatConfig.attribute_map - added:
field DinatConfig.model_type
- added:
class DinatBackbone(DinatPreTrainedModel, BackboneMixin) - added:
class DinatForImageClassification(DinatPreTrainedModel) - added:
class DinatModel(DinatPreTrainedModel) - added:
class DinatPreTrainedModel(PreTrainedModel) - added:
constructor DinatBackbone(config) - added:
constructor DinatForImageClassification(config) - added:
constructor DinatModel(config, add_pooling_layer=…) - added:
field DinatPreTrainedModel.base_model_prefix - added:
field DinatPreTrainedModel.config: DinatConfig - added:
field DinatPreTrainedModel.main_input_name - added:
method DinatBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method DinatBackbone.get_input_embeddings() - added:
method DinatForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, DinatImageClassifierOutput] - added:
method DinatModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, DinatModelOutput] - added:
method DinatModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class Dinov2Config(BackboneConfigMixin, PretrainedConfig) - added:
class Dinov2OnnxConfig(OnnxConfig) - added:
constructor Dinov2Config(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, layerscale_value=…, drop_path_rate=…, use_swiglu_ffn=…, out_features=…, out_indices=…, apply_layernorm=…, reshape_hidden_states=…, use_mas… - added:
field Dinov2Config.model_type - added:
field Dinov2OnnxConfig.torch_onnx_minimum_version - added:
property Dinov2OnnxConfig.atol_for_validation -> float - added:
property Dinov2OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class Dinov2Backbone(Dinov2PreTrainedModel, BackboneMixin) - added:
class Dinov2ForImageClassification(Dinov2PreTrainedModel) - added:
class Dinov2Model(Dinov2PreTrainedModel) - added:
class Dinov2PreTrainedModel(PreTrainedModel) - added:
constructor Dinov2Backbone(config) - added:
constructor Dinov2ForImageClassification(config: Dinov2Config) -> None - added:
constructor Dinov2Model(config: Dinov2Config) - added:
field Dinov2PreTrainedModel.base_model_prefix - added:
field Dinov2PreTrainedModel.config: Dinov2Config - added:
field Dinov2PreTrainedModel.main_input_name - added:
field Dinov2PreTrainedModel.supports_gradient_checkpointing - added:
method Dinov2Backbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, **kwargs) -> BackboneOutput - added:
method Dinov2Backbone.get_input_embeddings() -> Dinov2PatchEmbeddings - added:
method Dinov2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method Dinov2Model.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> BaseModelOutputWithPooling - added:
method Dinov2Model.get_input_embeddings() -> Dinov2PatchEmbeddings
- added:
class FlaxDinov2ForImageClassification(FlaxDinov2PreTrainedModel) - added:
class FlaxDinov2Model(FlaxDinov2PreTrainedModel) - added:
class FlaxDinov2PreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxDinov2PreTrainedModel(config: Dinov2Config, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxDinov2ForImageClassification.module_class - added:
field FlaxDinov2Model.module_class - added:
field FlaxDinov2PreTrainedModel.base_model_prefix - added:
field FlaxDinov2PreTrainedModel.config_class - added:
field FlaxDinov2PreTrainedModel.main_input_name - added:
field FlaxDinov2PreTrainedModel.module_class: nn.Module - added:
method FlaxDinov2PreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
export define_import_structure
- added:
class Dinov2WithRegistersConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor Dinov2WithRegistersConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, layerscale_value=…, drop_path_rate=…, use_swiglu_ffn=…, num_register_tokens=…, out_features=…, out_indices=…, apply_layernorm… - added:
field Dinov2WithRegistersConfig.model_type
- added:
class Dinov2WithRegistersBackbone(Dinov2WithRegistersPreTrainedModel, BackboneMixin) - added:
class Dinov2WithRegistersForImageClassification(Dinov2WithRegistersPreTrainedModel) - added:
class Dinov2WithRegistersModel(Dinov2WithRegistersPreTrainedModel) - added:
class Dinov2WithRegistersPreTrainedModel(PreTrainedModel) - added:
constructor Dinov2WithRegistersBackbone(config) - added:
constructor Dinov2WithRegistersForImageClassification(config: Dinov2WithRegistersConfig) -> None - added:
constructor Dinov2WithRegistersModel(config: Dinov2WithRegistersConfig) - added:
field Dinov2WithRegistersPreTrainedModel.base_model_prefix - added:
field Dinov2WithRegistersPreTrainedModel.config: Dinov2WithRegistersConfig - added:
field Dinov2WithRegistersPreTrainedModel.main_input_name - added:
field Dinov2WithRegistersPreTrainedModel.supports_gradient_checkpointing - added:
method Dinov2WithRegistersBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, **kwargs) -> BackboneOutput - added:
method Dinov2WithRegistersBackbone.get_input_embeddings() -> Dinov2WithRegistersPatchEmbeddings - added:
method Dinov2WithRegistersForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method Dinov2WithRegistersModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> BaseModelOutputWithPooling - added:
method Dinov2WithRegistersModel.get_input_embeddings() -> Dinov2WithRegistersPatchEmbeddings
- added:
class Dinov2WithRegistersBackbone(Dinov2Backbone) - added:
class Dinov2WithRegistersConfig(BackboneConfigMixin, PretrainedConfig) - added:
class Dinov2WithRegistersForImageClassification(Dinov2ForImageClassification) - added:
class Dinov2WithRegistersModel(Dinov2Model) - added:
class Dinov2WithRegistersPreTrainedModel(Dinov2PreTrainedModel) - added:
constructor Dinov2WithRegistersBackbone(config) - added:
constructor Dinov2WithRegistersConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, layerscale_value=…, drop_path_rate=…, use_swiglu_ffn=…, num_register_tokens=…, out_features=…, out_indices=…, apply_layernorm… - added:
field Dinov2WithRegistersConfig.model_type - added:
method Dinov2WithRegistersBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, **kwargs) -> BackboneOutput - added:
method Dinov2WithRegistersBackbone.get_input_embeddings() -> Dinov2WithRegistersPatchEmbeddings - added:
method Dinov2WithRegistersForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput
- added:
export define_import_structure
- added:
class DINOv3ConvNextConfig(PretrainedConfig) - added:
constructor DINOv3ConvNextConfig(num_channels: int=…, hidden_sizes: Optional[list[int]]=…, depths: Optional[list[int]]=…, hidden_act: str=…, initializer_range: float=…, layer_norm_eps: float=…, layer_scale_init_value: float=…, drop_path_rate: float=…, image_size: int=…, **kwargs) - added:
field DINOv3ConvNextConfig.model_type - added:
property DINOv3ConvNextConfig.num_stages -> int
- added:
class DINOv3ConvNextModel(DINOv3ConvNextPreTrainedModel) - added:
class DINOv3ConvNextPreTrainedModel(PreTrainedModel) - added:
constructor DINOv3ConvNextModel(config: DINOv3ConvNextConfig) - added:
field DINOv3ConvNextPreTrainedModel.base_model_prefix - added:
field DINOv3ConvNextPreTrainedModel.config: DINOv3ConvNextConfig - added:
field DINOv3ConvNextPreTrainedModel.main_input_name - added:
method DINOv3ConvNextModel.forward(pixel_values: torch.FloatTensor, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
export define_import_structure
- added:
class DINOv3ViTConfig(PretrainedConfig) - added:
constructor DINOv3ViTConfig(patch_size: int=…, hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, hidden_act: str=…, attention_dropout: float=…, initializer_range: float=…, layer_norm_eps: float=…, rope_theta: float=…, image_size: int=…, num_channels: int=…, query_bias: bool=…, key_bias: bool=…, value_bias: bool=…, proj_bias: bool=…, mlp_bias: bool… - added:
field DINOv3ViTConfig.model_type
- added:
class DINOv3ViTImageProcessorFast(BaseImageProcessorFast) - added:
field DINOv3ViTImageProcessorFast.do_normalize - added:
field DINOv3ViTImageProcessorFast.do_rescale - added:
field DINOv3ViTImageProcessorFast.do_resize - added:
field DINOv3ViTImageProcessorFast.image_mean - added:
field DINOv3ViTImageProcessorFast.image_std - added:
field DINOv3ViTImageProcessorFast.resample - added:
field DINOv3ViTImageProcessorFast.size
- added:
class DINOv3ViTModel(DINOv3ViTPreTrainedModel) - added:
class DINOv3ViTPreTrainedModel(PreTrainedModel) - added:
constructor DINOv3ViTModel(config: DINOv3ViTConfig) - added:
field DINOv3ViTPreTrainedModel.base_model_prefix - added:
field DINOv3ViTPreTrainedModel.config: DINOv3ViTConfig - added:
field DINOv3ViTPreTrainedModel.main_input_name - added:
field DINOv3ViTPreTrainedModel.supports_gradient_checkpointing - added:
method DINOv3ViTModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method DINOv3ViTModel.get_input_embeddings()
- added:
class DINOv3ViTModel(DINOv3ViTPreTrainedModel) - added:
class DINOv3ViTPreTrainedModel(Dinov2PreTrainedModel) - added:
constructor DINOv3ViTModel(config: DINOv3ViTConfig) - added:
method DINOv3ViTModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method DINOv3ViTModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class DistilBertConfig(PretrainedConfig) - added:
class DistilBertOnnxConfig(OnnxConfig) - added:
constructor DistilBertConfig(vocab_size=…, max_position_embeddings=…, sinusoidal_pos_embds=…, n_layers=…, n_heads=…, dim=…, hidden_dim=…, dropout=…, attention_dropout=…, activation=…, initializer_range=…, qa_dropout=…, seq_classif_dropout=…, pad_token_id=…, **kwargs) - added:
field DistilBertConfig.attribute_map - added:
field DistilBertConfig.model_type - added:
property DistilBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class DistilBertForMaskedLM(DistilBertPreTrainedModel) - added:
class DistilBertForMultipleChoice(DistilBertPreTrainedModel) - added:
class DistilBertForQuestionAnswering(DistilBertPreTrainedModel) - added:
class DistilBertForSequenceClassification(DistilBertPreTrainedModel) - added:
class DistilBertForTokenClassification(DistilBertPreTrainedModel) - added:
class DistilBertModel(DistilBertPreTrainedModel) - added:
class DistilBertPreTrainedModel(PreTrainedModel) - added:
constructor DistilBertForMaskedLM(config: PretrainedConfig) - added:
constructor DistilBertForMultipleChoice(config: PretrainedConfig) - added:
constructor DistilBertForQuestionAnswering(config: PretrainedConfig) - added:
constructor DistilBertForSequenceClassification(config: PretrainedConfig) - added:
constructor DistilBertForTokenClassification(config: PretrainedConfig) - added:
constructor DistilBertModel(config: PretrainedConfig) - added:
field DistilBertPreTrainedModel.base_model_prefix - added:
field DistilBertPreTrainedModel.config: DistilBertConfig - added:
field DistilBertPreTrainedModel.load_tf_weights - added:
field DistilBertPreTrainedModel.supports_gradient_checkpointing - added:
method DistilBertForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[MaskedLMOutput, tuple[torch.Tensor, ...]] - added:
method DistilBertForMaskedLM.get_output_embeddings() -> nn.Module - added:
method DistilBertForMaskedLM.get_position_embeddings() -> nn.Embedding - added:
method DistilBertForMaskedLM.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertForMaskedLM.set_output_embeddings(new_embeddings: nn.Module) - added:
method DistilBertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[MultipleChoiceModelOutput, tuple[torch.Tenso… - added:
method DistilBertForMultipleChoice.get_position_embeddings() -> nn.Embedding - added:
method DistilBertForMultipleChoice.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> U… - added:
method DistilBertForQuestionAnswering.get_position_embeddings() -> nn.Embedding - added:
method DistilBertForQuestionAnswering.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[SequenceClassifierOutput, tuple[torc… - added:
method DistilBertForSequenceClassification.get_position_embeddings() -> nn.Embedding - added:
method DistilBertForSequenceClassification.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[TokenClassifierOutput, tuple[torch.Tens… - added:
method DistilBertForTokenClassification.get_position_embeddings() -> nn.Embedding - added:
method DistilBertForTokenClassification.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[BaseModelOutput, tuple[torch.Tensor, ...]] - added:
method DistilBertModel.get_input_embeddings() -> nn.Embedding - added:
method DistilBertModel.get_position_embeddings() -> nn.Embedding - added:
method DistilBertModel.resize_position_embeddings(new_num_position_embeddings: int) - added:
method DistilBertModel.set_input_embeddings(new_embeddings: nn.Embedding)
- added:
class FlaxDistilBertForMaskedLM(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertForMultipleChoice(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertForQuestionAnswering(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertForSequenceClassification(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertForTokenClassification(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertModel(FlaxDistilBertPreTrainedModel) - added:
class FlaxDistilBertPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxDistilBertPreTrainedModel(config: DistilBertConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxDistilBertForMaskedLM.module_class - added:
field FlaxDistilBertForMultipleChoice.module_class - added:
field FlaxDistilBertForQuestionAnswering.module_class - added:
field FlaxDistilBertForSequenceClassification.module_class - added:
field FlaxDistilBertForTokenClassification.module_class - added:
field FlaxDistilBertModel.module_class - added:
field FlaxDistilBertPreTrainedModel.base_model_prefix - added:
field FlaxDistilBertPreTrainedModel.config_class - added:
field FlaxDistilBertPreTrainedModel.module_class: nn.Module - added:
method FlaxDistilBertPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFDistilBertMainLayer(keras.layers.Layer) - added:
class TFDistilBertModel(TFDistilBertPreTrainedModel) - added:
class TFDistilBertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFDistilBertForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFDistilBertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFDistilBertForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFDistilBertForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFDistilBertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFDistilBertMainLayer(config, **kwargs) - added:
constructor TFDistilBertModel(config, *inputs, **kwargs) - added:
field TFDistilBertMainLayer.config_class - added:
field TFDistilBertPreTrainedModel.base_model_prefix - added:
field TFDistilBertPreTrainedModel.config_class - added:
method TFDistilBertForMaskedLM.build(input_shape=…) - added:
method TFDistilBertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFMaskedLMOutput… - added:
method TFDistilBertForMaskedLM.get_lm_head() - added:
method TFDistilBertForMaskedLM.get_prefix_bias_name() - added:
method TFDistilBertForMultipleChoice.build(input_shape=…) - added:
method TFDistilBertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFMultiple… - added:
method TFDistilBertForQuestionAnswering.build(input_shape=…) - added:
method TFDistilBertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: np.ndarray | tf.Tensor | None=…, end_positions: np.ndarray… - added:
method TFDistilBertForSequenceClassification.build(input_shape=…) - added:
method TFDistilBertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TF… - added:
method TFDistilBertForTokenClassification.build(input_shape=…) - added:
method TFDistilBertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFTok… - added:
method TFDistilBertMainLayer.build(input_shape=…) - added:
method TFDistilBertMainLayer.call(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFDistilBertMainLayer.get_input_embeddings() - added:
method TFDistilBertMainLayer.set_input_embeddings(value) - added:
method TFDistilBertModel.build(input_shape=…) - added:
method TFDistilBertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOutput | tuple[tf.Tensor]
- added:
class DistilBertTokenizer(PreTrainedTokenizer) - added:
constructor DistilBertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field DistilBertTokenizer.model_input_names - added:
field DistilBertTokenizer.vocab_files_names - added:
method DistilBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method DistilBertTokenizer.convert_tokens_to_string(tokens) - added:
method DistilBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method DistilBertTokenizer.get_vocab() - added:
method DistilBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property DistilBertTokenizer.do_lower_case - added:
property DistilBertTokenizer.vocab_size
- added:
class DistilBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor DistilBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field DistilBertTokenizerFast.model_input_names - added:
field DistilBertTokenizerFast.slow_tokenizer_class - added:
field DistilBertTokenizerFast.vocab_files_names - added:
method DistilBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method DistilBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class DogeConfig(PretrainedConfig) - added:
constructor DogeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, hidden_dropout=…, hidden_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, num_attention_heads=…, num_key_value_heads=…, attention_bias=…, attention_dropout=…, mlp_bias=…, sliding_window=…, keep_window_size=…, is_moe=…… - added:
field DogeConfig.base_model_pp_plan - added:
field DogeConfig.base_model_tp_plan - added:
field DogeConfig.keys_to_ignore_at_inference - added:
field DogeConfig.model_type
- added:
class DogeForCausalLM(DogePreTrainedModel, GenerationMixin) - added:
class DogeForSequenceClassification(GenericForSequenceClassification, DogePreTrainedModel) - added:
class DogeModel(DogePreTrainedModel) - added:
class DogePreTrainedModel(PreTrainedModel) - added:
constructor DogeForCausalLM(config) - added:
constructor DogeModel(config: DogeConfig) - added:
field DogePreTrainedModel.base_model_prefix - added:
field DogePreTrainedModel.config: DogeConfig - added:
field DogePreTrainedModel.supports_gradient_checkpointing - added:
method DogeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, ou… - added:
method DogeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class DogeConfig(PretrainedConfig) - added:
class DogeForCausalLM(MixtralForCausalLM) - added:
class DogeForSequenceClassification(LlamaForSequenceClassification) - added:
class DogeModel(MixtralModel) - added:
class DogePreTrainedModel(LlamaPreTrainedModel) - added:
constructor DogeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, hidden_dropout=…, hidden_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, num_attention_heads=…, num_key_value_heads=…, attention_bias=…, attention_dropout=…, mlp_bias=…, sliding_window=…, keep_window_size=…, is_moe=…… - added:
constructor DogeForCausalLM(config) - added:
field DogeConfig.base_model_pp_plan - added:
field DogeConfig.base_model_tp_plan - added:
field DogeConfig.keys_to_ignore_at_inference - added:
field DogeConfig.model_type - added:
method DogeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, ou…
- added:
export define_import_structure
- added:
class DonutSwinConfig(PretrainedConfig) - added:
constructor DonutSwinConfig(image_size=…, patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, window_size=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, use_absolute_embeddings=…, initializer_range=…, layer_norm_eps=…, **kwargs) - added:
field DonutSwinConfig.attribute_map - added:
field DonutSwinConfig.model_type
- added:
class DonutFeatureExtractor(DonutImageProcessor) - added:
constructor DonutFeatureExtractor(*args, **kwargs) -> None
- added:
class DonutImageProcessor(BaseImageProcessor) - added:
constructor DonutImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_thumbnail: bool=…, do_align_long_axis: bool=…, do_pad: bool=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field DonutImageProcessor.model_input_names - added:
method DonutImageProcessor.align_long_axis(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DonutImageProcessor.pad_image(image: np.ndarray, size: dict[str, int], random_padding: bool=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method DonutImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_thumbnail: Optional[bool]=…, do_align_long_axis: Optional[bool]=…, do_pad: Optional[bool]=…, random_padding: bool=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optiona… - added:
method DonutImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method DonutImageProcessor.thumbnail(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DonutImageProcessorFast(BaseImageProcessorFast) - added:
constructor DonutImageProcessorFast(**kwargs: Unpack[DonutFastImageProcessorKwargs]) - added:
field DonutImageProcessorFast.do_align_long_axis - added:
field DonutImageProcessorFast.do_normalize - added:
field DonutImageProcessorFast.do_pad - added:
field DonutImageProcessorFast.do_rescale - added:
field DonutImageProcessorFast.do_resize - added:
field DonutImageProcessorFast.do_thumbnail - added:
field DonutImageProcessorFast.image_mean - added:
field DonutImageProcessorFast.image_std - added:
field DonutImageProcessorFast.resample - added:
field DonutImageProcessorFast.size - added:
field DonutImageProcessorFast.valid_kwargs - added:
method DonutImageProcessorFast.align_long_axis(image: torch.Tensor, size: SizeDict) -> torch.Tensor - added:
method DonutImageProcessorFast.pad(*args, **kwargs) - added:
method DonutImageProcessorFast.pad_image(image: torch.Tensor, size: SizeDict, random_padding: bool=…) -> torch.Tensor - added:
method DonutImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[DonutFastImageProcessorKwargs]) -> BatchFeature - added:
method DonutImageProcessorFast.thumbnail(image: torch.Tensor, size: SizeDict) -> torch.Tensor
- added:
class DonutSwinForImageClassification(DonutSwinPreTrainedModel) - added:
class DonutSwinModel(DonutSwinPreTrainedModel) - added:
class DonutSwinPreTrainedModel(PreTrainedModel) - added:
constructor DonutSwinForImageClassification(config) - added:
constructor DonutSwinModel(config, add_pooling_layer=…, use_mask_token=…) - added:
field DonutSwinPreTrainedModel.base_model_prefix - added:
field DonutSwinPreTrainedModel.config: DonutSwinConfig - added:
field DonutSwinPreTrainedModel.main_input_name - added:
field DonutSwinPreTrainedModel.supports_gradient_checkpointing - added:
method DonutSwinForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, DonutSwinImageClassifierOutput] - added:
method DonutSwinModel.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, DonutSwinModelOutput] - added:
method DonutSwinModel.get_input_embeddings()
- added:
class DonutProcessor(ProcessorMixin) - added:
constructor DonutProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field DonutProcessor.attributes - added:
field DonutProcessor.image_processor_class - added:
field DonutProcessor.tokenizer_class - added:
method DonutProcessor.as_target_processor() - added:
method DonutProcessor.token2json(tokens, is_inner_value=…, added_vocab=…) - added:
property DonutProcessor.feature_extractor - added:
property DonutProcessor.feature_extractor_class - added:
property DonutProcessor.model_input_names
- added:
export define_import_structure
- added:
class Dots1Config(PretrainedConfig) - added:
constructor Dots1Config(vocab_size=…, hidden_size=…, intermediate_size=…, moe_intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, n_shared_experts=…, n_routed_experts=…, n_group=…, topk_group=…, num_experts_per_tok=…, first_k_dense_replace=…, norm_topk_prob=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_em… - added:
field Dots1Config.base_model_pp_plan - added:
field Dots1Config.base_model_tp_plan - added:
field Dots1Config.keys_to_ignore_at_inference - added:
field Dots1Config.model_type
- added:
class Dots1ForCausalLM(Dots1PreTrainedModel, GenerationMixin) - added:
class Dots1Model(Dots1PreTrainedModel) - added:
class Dots1PreTrainedModel(PreTrainedModel) - added:
constructor Dots1ForCausalLM(config) - added:
constructor Dots1Model(config: Dots1Config) - added:
field Dots1PreTrainedModel.base_model_prefix - added:
field Dots1PreTrainedModel.config: Dots1Config - added:
field Dots1PreTrainedModel.supports_gradient_checkpointing - added:
method Dots1ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method Dots1Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Dots1ForCausalLM(Qwen3ForCausalLM) - added:
class Dots1Model(Qwen3Model) - added:
class Dots1PreTrainedModel(DeepseekV3PreTrainedModel) - added:
method Dots1ForCausalLM.forward(**super_kwargs: Unpack[TransformersKwargs]) -> CausalLMOutputWithPast
- added:
export define_import_structure
- added:
class DPRConfig(PretrainedConfig) - added:
constructor DPRConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, projection_dim: int=…, **kwargs) - added:
field DPRConfig.model_type
- added:
class DPRContextEncoder(DPRPretrainedContextEncoder) - added:
class DPRPreTrainedModel(PreTrainedModel) - added:
class DPRPretrainedContextEncoder(DPRPreTrainedModel) - added:
class DPRPretrainedQuestionEncoder(DPRPreTrainedModel) - added:
class DPRPretrainedReader(DPRPreTrainedModel) - added:
class DPRQuestionEncoder(DPRPretrainedQuestionEncoder) - added:
class DPRReader(DPRPretrainedReader) - added:
constructor DPRContextEncoder(config: DPRConfig) - added:
constructor DPRQuestionEncoder(config: DPRConfig) - added:
constructor DPRReader(config: DPRConfig) - added:
field DPRPretrainedContextEncoder.base_model_prefix - added:
field DPRPretrainedContextEncoder.config: DPRConfig - added:
field DPRPretrainedContextEncoder.load_tf_weights - added:
field DPRPretrainedQuestionEncoder.base_model_prefix - added:
field DPRPretrainedQuestionEncoder.config: DPRConfig - added:
field DPRPretrainedQuestionEncoder.load_tf_weights - added:
field DPRPretrainedReader.base_model_prefix - added:
field DPRPretrainedReader.config: DPRConfig - added:
field DPRPretrainedReader.load_tf_weights - added:
method DPRContextEncoder.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, token_type_ids: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[DPRContextEncoderOutput, tuple[Tensor, ...]] - added:
method DPRQuestionEncoder.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, token_type_ids: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[DPRQuestionEncoderOutput, tuple[Tensor, ...]] - added:
method DPRReader.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[DPRReaderOutput, tuple[Tensor, ...]]
- added:
class TFDPRContextEncoder(TFDPRPretrainedContextEncoder) - added:
class TFDPRPretrainedContextEncoder(TFPreTrainedModel) - added:
class TFDPRPretrainedQuestionEncoder(TFPreTrainedModel) - added:
class TFDPRPretrainedReader(TFPreTrainedModel) - added:
class TFDPRQuestionEncoder(TFDPRPretrainedQuestionEncoder) - added:
class TFDPRReader(TFDPRPretrainedReader) - added:
constructor TFDPRContextEncoder(config: DPRConfig, *args, **kwargs) - added:
constructor TFDPRQuestionEncoder(config: DPRConfig, *args, **kwargs) - added:
constructor TFDPRReader(config: DPRConfig, *args, **kwargs) - added:
field TFDPRPretrainedContextEncoder.base_model_prefix - added:
field TFDPRPretrainedContextEncoder.config_class - added:
field TFDPRPretrainedQuestionEncoder.base_model_prefix - added:
field TFDPRPretrainedQuestionEncoder.config_class - added:
field TFDPRPretrainedReader.base_model_prefix - added:
field TFDPRPretrainedReader.config_class - added:
method TFDPRContextEncoder.build(input_shape=…) - added:
method TFDPRContextEncoder.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFDPRContextEncoderOutput | tuple[tf.Tensor, ...] - added:
method TFDPRContextEncoder.get_input_embeddings() - added:
method TFDPRQuestionEncoder.build(input_shape=…) - added:
method TFDPRQuestionEncoder.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFDPRQuestionEncoderOutput | tuple[tf.Tensor, ...] - added:
method TFDPRQuestionEncoder.get_input_embeddings() - added:
method TFDPRReader.build(input_shape=…) - added:
method TFDPRReader.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFDPRReaderOutput | tuple[tf.Tensor, ...] - added:
method TFDPRReader.get_input_embeddings()
- added:
class DPRContextEncoderTokenizer(BertTokenizer) - added:
class DPRQuestionEncoderTokenizer(BertTokenizer) - added:
class DPRReaderTokenizer(CustomDPRReaderTokenizerMixin, BertTokenizer) - added:
field DPRContextEncoderTokenizer.vocab_files_names - added:
field DPRQuestionEncoderTokenizer.vocab_files_names - added:
field DPRReaderTokenizer.model_input_names - added:
field DPRReaderTokenizer.vocab_files_names - added:
variable DPRReaderOutput
- added:
class DPRContextEncoderTokenizerFast(BertTokenizerFast) - added:
class DPRQuestionEncoderTokenizerFast(BertTokenizerFast) - added:
class DPRReaderTokenizerFast(CustomDPRReaderTokenizerMixin, BertTokenizerFast) - added:
field DPRContextEncoderTokenizerFast.slow_tokenizer_class - added:
field DPRContextEncoderTokenizerFast.vocab_files_names - added:
field DPRQuestionEncoderTokenizerFast.slow_tokenizer_class - added:
field DPRQuestionEncoderTokenizerFast.vocab_files_names - added:
field DPRReaderTokenizerFast.model_input_names - added:
field DPRReaderTokenizerFast.slow_tokenizer_class - added:
field DPRReaderTokenizerFast.vocab_files_names
- added:
export define_import_structure
- added:
class DPTConfig(PretrainedConfig) - added:
constructor DPTConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, is_hybrid=…, qkv_bias=…, backbone_out_indices=…, readout_type=…, reassemble_factors=…, neck_hidden_sizes=…, fusion_hidden_size=…, head_in_index=…… - added:
field DPTConfig.model_type - added:
method DPTConfig.to_dict() - added:
property DPTConfig.sub_configs
- added:
class DPTFeatureExtractor(DPTImageProcessor) - added:
constructor DPTFeatureExtractor(*args, **kwargs) -> None
- added:
class DPTImageProcessor(BaseImageProcessor) - added:
constructor DPTImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, size_divisor: Optional[int]=…, do_… - added:
field DPTImageProcessor.model_input_names - added:
method DPTImageProcessor.pad_image(image: np.ndarray, size_divisor: int, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method DPTImageProcessor.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method DPTImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method DPTImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[int]=…, keep_aspect_ratio: Optional[bool]=…, ensure_multiple_of: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, l… - added:
method DPTImageProcessor.reduce_label(label: ImageInput) -> np.ndarray - added:
method DPTImageProcessor.resize(image: np.ndarray, size: dict[str, int], keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class DPTImageProcessorFast(BaseImageProcessorFast) - added:
constructor DPTImageProcessorFast(**kwargs: Unpack[DPTFastImageProcessorKwargs]) - added:
field DPTImageProcessorFast.crop_size - added:
field DPTImageProcessorFast.default_to_square - added:
field DPTImageProcessorFast.do_center_crop - added:
field DPTImageProcessorFast.do_normalize - added:
field DPTImageProcessorFast.do_pad - added:
field DPTImageProcessorFast.do_reduce_labels - added:
field DPTImageProcessorFast.do_rescale - added:
field DPTImageProcessorFast.do_resize - added:
field DPTImageProcessorFast.ensure_multiple_of - added:
field DPTImageProcessorFast.image_mean - added:
field DPTImageProcessorFast.image_std - added:
field DPTImageProcessorFast.keep_aspect_ratio - added:
field DPTImageProcessorFast.resample - added:
field DPTImageProcessorFast.rescale_factor - added:
field DPTImageProcessorFast.size - added:
field DPTImageProcessorFast.valid_kwargs - added:
method DPTImageProcessorFast.pad_image(image: torch.Tensor, size_divisor: int=…) -> torch.Tensor - added:
method DPTImageProcessorFast.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method DPTImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method DPTImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[DPTFastImageProcessorKwargs]) -> BatchFeature - added:
method DPTImageProcessorFast.reduce_label(labels: list[torch.Tensor]) - added:
method DPTImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, ensure_multiple_of: Optional[int]=…, keep_aspect_ratio: bool=…) -> torch.Tensor
- added:
class DPTForDepthEstimation(DPTPreTrainedModel) - added:
class DPTForSemanticSegmentation(DPTPreTrainedModel) - added:
class DPTModel(DPTPreTrainedModel) - added:
class DPTPreTrainedModel(PreTrainedModel) - added:
constructor DPTForDepthEstimation(config) - added:
constructor DPTForSemanticSegmentation(config: DPTConfig) - added:
constructor DPTModel(config: DPTConfig, add_pooling_layer: bool=…) - added:
field DPTPreTrainedModel.base_model_prefix - added:
field DPTPreTrainedModel.config: DPTConfig - added:
field DPTPreTrainedModel.main_input_name - added:
field DPTPreTrainedModel.supports_gradient_checkpointing - added:
method DPTForDepthEstimation.forward(pixel_values: torch.FloatTensor, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> DepthEstimatorOutput - added:
method DPTForSemanticSegmentation.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> SemanticSegmenterOutput - added:
method DPTModel.forward(pixel_values: torch.FloatTensor, head_mask: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> BaseModelOutputWithPoolingAndIntermediateActivations - added:
method DPTModel.get_input_embeddings()
- added:
class DPTImageProcessorFast(BeitImageProcessorFast) - added:
field DPTImageProcessorFast.crop_size - added:
field DPTImageProcessorFast.do_center_crop - added:
field DPTImageProcessorFast.do_normalize - added:
field DPTImageProcessorFast.do_pad - added:
field DPTImageProcessorFast.do_reduce_labels - added:
field DPTImageProcessorFast.do_rescale - added:
field DPTImageProcessorFast.do_resize - added:
field DPTImageProcessorFast.ensure_multiple_of - added:
field DPTImageProcessorFast.image_mean - added:
field DPTImageProcessorFast.image_std - added:
field DPTImageProcessorFast.keep_aspect_ratio - added:
field DPTImageProcessorFast.resample - added:
field DPTImageProcessorFast.rescale_factor - added:
field DPTImageProcessorFast.size - added:
field DPTImageProcessorFast.valid_kwargs - added:
method DPTImageProcessorFast.pad_image(image: torch.Tensor, size_divisor: int=…) -> torch.Tensor - added:
method DPTImageProcessorFast.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method DPTImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, ensure_multiple_of: Optional[int]=…, keep_aspect_ratio: bool=…) -> torch.Tensor
- added:
export define_import_structure
- added:
class EdgeTamConfig(PretrainedConfig) - added:
class EdgeTamMaskDecoderConfig(PretrainedConfig) - added:
class EdgeTamPromptEncoderConfig(PretrainedConfig) - added:
class EdgeTamVisionConfig(PretrainedConfig) - added:
constructor EdgeTamConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, **kwargs) - added:
constructor EdgeTamMaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, dynamic_multimask_via_stability=…, dynamic_multimask_stability_delta=…, dynamic_multimask_stability_thresh=…, **kwargs) - added:
constructor EdgeTamPromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, scale=…, **kwargs) - added:
constructor EdgeTamVisionConfig(backbone_config=…, backbone_channel_list=…, backbone_feature_sizes=…, fpn_hidden_size=…, fpn_kernel_size=…, fpn_stride=…, fpn_padding=…, fpn_top_down_levels=…, num_feature_levels=…, hidden_act=…, layer_norm_eps=…, initializer_range=…, **kwargs) - added:
field EdgeTamConfig.model_type - added:
field EdgeTamConfig.sub_configs - added:
field EdgeTamMaskDecoderConfig.base_config_key - added:
field EdgeTamPromptEncoderConfig.base_config_key - added:
field EdgeTamVisionConfig.base_config_key - added:
field EdgeTamVisionConfig.model_type - added:
field EdgeTamVisionConfig.sub_configs
- added:
class EdgeTamModel(EdgeTamPreTrainedModel) - added:
class EdgeTamPreTrainedModel(PreTrainedModel) - added:
class EdgeTamVisionModel(EdgeTamPreTrainedModel) - added:
constructor EdgeTamModel(config: EdgeTamConfig) - added:
constructor EdgeTamVisionModel(config: EdgeTamVisionConfig) - added:
field EdgeTamPreTrainedModel.base_model_prefix - added:
field EdgeTamPreTrainedModel.config_class - added:
field EdgeTamPreTrainedModel.main_input_name - added:
field EdgeTamVisionModel.config_class - added:
field EdgeTamVisionModel.main_input_name - added:
method EdgeTamModel.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, target_embedding: Opti… - added:
method EdgeTamModel.get_image_embeddings(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> list[torch.Tensor] - added:
method EdgeTamModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> tuple[list[torch.Tensor], list[torch.Tensor], Optional[tuple[torch.FloatTensor, ...]], Optional[tuple[torch.FloatTensor, ...]]] - added:
method EdgeTamModel.get_image_wide_positional_embeddings() -> torch.Tensor - added:
method EdgeTamModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) - added:
method EdgeTamVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, EdgeTamVisionEncoderOutput]
- added:
class EdgeTamConfig(Sam2Config) - added:
class EdgeTamMaskDecoderConfig(Sam2MaskDecoderConfig) - added:
class EdgeTamModel(Sam2Model) - added:
class EdgeTamPreTrainedModel(Sam2PreTrainedModel) - added:
class EdgeTamPromptEncoderConfig(Sam2PromptEncoderConfig) - added:
class EdgeTamVisionConfig(PretrainedConfig) - added:
class EdgeTamVisionModel(Sam2VisionModel) - added:
constructor EdgeTamVisionConfig(backbone_config=…, backbone_channel_list=…, backbone_feature_sizes=…, fpn_hidden_size=…, fpn_kernel_size=…, fpn_stride=…, fpn_padding=…, fpn_top_down_levels=…, num_feature_levels=…, hidden_act=…, layer_norm_eps=…, initializer_range=…, **kwargs) - added:
field EdgeTamVisionConfig.base_config_key - added:
field EdgeTamVisionConfig.model_type - added:
field EdgeTamVisionConfig.sub_configs - added:
field EdgeTamVisionModel.config_class - added:
field EdgeTamVisionModel.main_input_name - added:
method EdgeTamModel.get_input_embeddings() - added:
method EdgeTamVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, EdgeTamVisionEncoderOutput] - added:
method EdgeTamVisionModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class EdgeTamVideoConfig(PretrainedConfig) - added:
class EdgeTamVideoMaskDecoderConfig(PretrainedConfig) - added:
class EdgeTamVideoPromptEncoderConfig(PretrainedConfig) - added:
constructor EdgeTamVideoConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, num_maskmem=…, image_size=…, sigmoid_scale_for_mem_enc=…, sigmoid_bias_for_mem_enc=…, enable_occlusion_spatial_embedding=…, multimask_output_in_sam=…, multimask_min_pt_num=…, multimask_max_pt_num=…, multimask_output_for_tracking=…, max_object_pointers_in_encoder=…, enable_temporal_… - added:
constructor EdgeTamVideoMaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, dynamic_multimask_via_stability=…, dynamic_multimask_stability_delta=…, dynamic_multimask_stability_thresh=…, **kwargs) - added:
constructor EdgeTamVideoPromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, scale=…, **kwargs) - added:
field EdgeTamVideoConfig.model_type - added:
field EdgeTamVideoConfig.sub_configs - added:
field EdgeTamVideoMaskDecoderConfig.base_config_key - added:
field EdgeTamVideoPromptEncoderConfig.base_config_key
- added:
class EdgeTamVideoInferenceSession - added:
class EdgeTamVideoModel(EdgeTamVideoPreTrainedModel) - added:
class EdgeTamVideoPreTrainedModel(PreTrainedModel) - added:
constructor EdgeTamVideoInferenceSession(video: Optional[torch.FloatTensor]=…, video_height: Optional[int]=…, video_width: Optional[int]=…, inference_device: Union[torch.device, str]=…, inference_state_device: Union[torch.device, str]=…, video_storage_device: Union[torch.device, str]=…, dtype: Union[torch.dtype, str]=…, max_vision_features_cache_size: int=…) - added:
constructor EdgeTamVideoModel(config: EdgeTamVideoConfig) - added:
field EdgeTamVideoPreTrainedModel.base_model_prefix - added:
field EdgeTamVideoPreTrainedModel.config_class - added:
field EdgeTamVideoPreTrainedModel.main_input_name - added:
method EdgeTamVideoInferenceSession.add_mask_inputs(obj_idx: int, frame_idx: int, inputs: torch.Tensor) - added:
method EdgeTamVideoInferenceSession.add_new_frame(pixel_values: torch.Tensor, frame_idx: Optional[int]=…) -> int - added:
method EdgeTamVideoInferenceSession.add_point_inputs(obj_idx: int, frame_idx: int, inputs: dict) - added:
method EdgeTamVideoInferenceSession.get_frame(frame_idx: int) -> torch.Tensor - added:
method EdgeTamVideoInferenceSession.get_obj_num() -> int - added:
method EdgeTamVideoInferenceSession.get_output(obj_idx: int, frame_idx: int, output_key: str, is_conditioning_frame: bool=…) - added:
method EdgeTamVideoInferenceSession.obj_id_to_idx(obj_id: int) -> int - added:
method EdgeTamVideoInferenceSession.obj_idx_to_id(obj_idx: int) -> int - added:
method EdgeTamVideoInferenceSession.remove_mask_inputs(obj_idx: int, frame_idx: int) - added:
method EdgeTamVideoInferenceSession.remove_point_inputs(obj_idx: int, frame_idx: int) - added:
method EdgeTamVideoInferenceSession.reset_inference_session() - added:
method EdgeTamVideoInferenceSession.reset_tracking_data() - added:
method EdgeTamVideoInferenceSession.store_output(obj_idx: int, frame_idx: int, output_key: Optional[str]=…, output_value: Optional[Union[torch.Tensor, dict]]=…, is_conditioning_frame: bool=…) - added:
method EdgeTamVideoModel.forward(inference_session: EdgeTamVideoInferenceSession, frame_idx: Optional[int]=…, frame: Optional[torch.Tensor]=…, reverse: bool=…) -> EdgeTamVideoSegmentationOutput - added:
method EdgeTamVideoModel.get_image_embeddings(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> list[torch.Tensor] - added:
method EdgeTamVideoModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> tuple[list[torch.Tensor], list[torch.Tensor], Optional[tuple[torch.FloatTensor, ...]], Optional[tuple[torch.FloatTensor, ...]]] - added:
method EdgeTamVideoModel.get_image_wide_positional_embeddings() -> torch.Tensor - added:
method EdgeTamVideoModel.get_input_embeddings() - added:
method EdgeTamVideoModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method EdgeTamVideoModel.propagate_in_video_iterator(inference_session: EdgeTamVideoInferenceSession, start_frame_idx: Optional[int]=…, max_frame_num_to_track: Optional[int]=…, reverse: bool=…) -> Iterator[EdgeTamVideoSegmentationOutput] - added:
property EdgeTamVideoInferenceSession.num_frames -> Optional[int]
- added:
class EdgeTamVideoConfig(Sam2VideoConfig) - added:
class EdgeTamVideoInferenceSession(Sam2VideoInferenceSession) - added:
class EdgeTamVideoMaskDecoderConfig(Sam2VideoMaskDecoderConfig) - added:
class EdgeTamVideoModel(Sam2VideoModel) - added:
class EdgeTamVideoPreTrainedModel(Sam2VideoPreTrainedModel) - added:
class EdgeTamVideoPromptEncoderConfig(Sam2VideoPromptEncoderConfig) - added:
constructor EdgeTamVideoConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, num_maskmem=…, image_size=…, sigmoid_scale_for_mem_enc=…, sigmoid_bias_for_mem_enc=…, enable_occlusion_spatial_embedding=…, multimask_output_in_sam=…, multimask_min_pt_num=…, multimask_max_pt_num=…, multimask_output_for_tracking=…, max_object_pointers_in_encoder=…, enable_temporal_… - added:
constructor EdgeTamVideoModel(config: EdgeTamVideoConfig) - added:
field EdgeTamVideoConfig.model_type - added:
field EdgeTamVideoConfig.sub_configs
- added:
export define_import_structure
- added:
class EfficientLoFTRConfig(PretrainedConfig) - added:
constructor EfficientLoFTRConfig(stage_num_blocks: Optional[list[int]]=…, out_features: Optional[list[int]]=…, stage_stride: Optional[list[int]]=…, hidden_size: int=…, activation_function: str=…, q_aggregation_kernel_size: int=…, kv_aggregation_kernel_size: int=…, q_aggregation_stride: int=…, kv_aggregation_stride: int=…, num_attention_layers: int=…, num_attention_heads: int=…, attention_dropout:… - added:
field EfficientLoFTRConfig.model_type
- added:
class EfficientLoFTRImageProcessor(BaseImageProcessor) - added:
constructor EfficientLoFTRImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_grayscale: bool=…, **kwargs) -> None - added:
field EfficientLoFTRImageProcessor.model_input_names - added:
method EfficientLoFTRImageProcessor.post_process_keypoint_matching(outputs: KeypointMatchingOutput, target_sizes: Union[TensorType, list[tuple]], threshold: float=…) -> list[dict[str, torch.Tensor]] - added:
method EfficientLoFTRImageProcessor.preprocess(images, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_grayscale: Optional[bool]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]… - added:
method EfficientLoFTRImageProcessor.resize(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method EfficientLoFTRImageProcessor.visualize_keypoint_matching(images: ImageInput, keypoint_matching_output: list[dict[str, torch.Tensor]]) -> list[Image.Image]
- added:
class EfficientLoFTRImageProcessorFast(BaseImageProcessorFast) - added:
constructor EfficientLoFTRImageProcessorFast(**kwargs: Unpack[EfficientLoFTRFastImageProcessorKwargs]) - added:
field EfficientLoFTRImageProcessorFast.default_to_square - added:
field EfficientLoFTRImageProcessorFast.do_normalize - added:
field EfficientLoFTRImageProcessorFast.do_rescale - added:
field EfficientLoFTRImageProcessorFast.do_resize - added:
field EfficientLoFTRImageProcessorFast.resample - added:
field EfficientLoFTRImageProcessorFast.rescale_factor - added:
field EfficientLoFTRImageProcessorFast.size - added:
field EfficientLoFTRImageProcessorFast.valid_kwargs - added:
method EfficientLoFTRImageProcessorFast.post_process_keypoint_matching(outputs: KeypointMatchingOutput, target_sizes: Union[TensorType, list[tuple]], threshold: float=…) -> list[dict[str, torch.Tensor]] - added:
method EfficientLoFTRImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[EfficientLoFTRFastImageProcessorKwargs]) -> BatchFeature - added:
method EfficientLoFTRImageProcessorFast.visualize_keypoint_matching(images, keypoint_matching_output: list[dict[str, torch.Tensor]]) -> list[Image.Image]
- added:
class EfficientLoFTRForKeypointMatching(EfficientLoFTRPreTrainedModel) - added:
class EfficientLoFTRModel(EfficientLoFTRPreTrainedModel) - added:
class EfficientLoFTRPreTrainedModel(PreTrainedModel) - added:
constructor EfficientLoFTRForKeypointMatching(config: EfficientLoFTRConfig) - added:
constructor EfficientLoFTRModel(config: EfficientLoFTRConfig) - added:
field EfficientLoFTRPreTrainedModel.base_model_prefix - added:
field EfficientLoFTRPreTrainedModel.config_class - added:
field EfficientLoFTRPreTrainedModel.main_input_name - added:
field EfficientLoFTRPreTrainedModel.supports_gradient_checkpointing - added:
method EfficientLoFTRForKeypointMatching.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> KeypointMatchingOutput - added:
method EfficientLoFTRModel.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BackboneOutput - added:
method EfficientLoFTRPreTrainedModel.extract_one_channel_pixel_values(pixel_values: torch.FloatTensor) -> torch.FloatTensor
- added:
export define_import_structure
- added:
class EfficientNetConfig(PretrainedConfig) - added:
class EfficientNetOnnxConfig(OnnxConfig) - added:
constructor EfficientNetConfig(num_channels: int=…, image_size: int=…, width_coefficient: float=…, depth_coefficient: float=…, depth_divisor: int=…, kernel_sizes: list[int]=…, in_channels: list[int]=…, out_channels: list[int]=…, depthwise_padding: list[int]=…, strides: list[int]=…, num_block_repeats: list[int]=…, expand_ratios: list[int]=…, squeeze_expansion_ratio: float=…, hidden_act: str=…, hid… - added:
field EfficientNetConfig.model_type - added:
field EfficientNetOnnxConfig.torch_onnx_minimum_version - added:
property EfficientNetOnnxConfig.atol_for_validation -> float - added:
property EfficientNetOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class EfficientNetImageProcessor(BaseImageProcessor) - added:
constructor EfficientNetImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, rescale_factor: Union[int, float]=…, rescale_offset: bool=…, do_rescale: bool=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, include_… - added:
field EfficientNetImageProcessor.model_input_names - added:
method EfficientNetImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, rescale_offset: Optional[bool]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, imag… - added:
method EfficientNetImageProcessor.rescale(image: np.ndarray, scale: Union[int, float], offset: bool=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method EfficientNetImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class EfficientNetImageProcessorFast(BaseImageProcessorFast) - added:
constructor EfficientNetImageProcessorFast(**kwargs: Unpack[EfficientNetFastImageProcessorKwargs]) - added:
field EfficientNetImageProcessorFast.crop_size - added:
field EfficientNetImageProcessorFast.do_center_crop - added:
field EfficientNetImageProcessorFast.do_normalize - added:
field EfficientNetImageProcessorFast.do_rescale - added:
field EfficientNetImageProcessorFast.do_resize - added:
field EfficientNetImageProcessorFast.image_mean - added:
field EfficientNetImageProcessorFast.image_std - added:
field EfficientNetImageProcessorFast.include_top - added:
field EfficientNetImageProcessorFast.resample - added:
field EfficientNetImageProcessorFast.rescale_factor - added:
field EfficientNetImageProcessorFast.rescale_offset - added:
field EfficientNetImageProcessorFast.size - added:
field EfficientNetImageProcessorFast.valid_kwargs - added:
method EfficientNetImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[EfficientNetFastImageProcessorKwargs]) -> BatchFeature - added:
method EfficientNetImageProcessorFast.rescale(image: torch.Tensor, scale: float, offset: Optional[bool]=…, **kwargs) -> torch.Tensor - added:
method EfficientNetImageProcessorFast.rescale_and_normalize(images: torch.Tensor, do_rescale: bool, rescale_factor: float, do_normalize: bool, image_mean: Union[float, list[float]], image_std: Union[float, list[float]], rescale_offset: bool=…) -> torch.Tensor
- added:
class EfficientNetForImageClassification(EfficientNetPreTrainedModel) - added:
class EfficientNetModel(EfficientNetPreTrainedModel) - added:
class EfficientNetPreTrainedModel(PreTrainedModel) - added:
constructor EfficientNetForImageClassification(config) - added:
constructor EfficientNetModel(config: EfficientNetConfig) - added:
field EfficientNetPreTrainedModel.base_model_prefix - added:
field EfficientNetPreTrainedModel.config: EfficientNetConfig - added:
field EfficientNetPreTrainedModel.main_input_name - added:
method EfficientNetForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method EfficientNetModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class ElectraConfig(PretrainedConfig) - added:
class ElectraOnnxConfig(OnnxConfig) - added:
constructor ElectraConfig(vocab_size=…, embedding_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_last_dropout=…, pad_token_id=…, position_… - added:
field ElectraConfig.model_type - added:
property ElectraOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ElectraForCausalLM(ElectraPreTrainedModel, GenerationMixin) - added:
class ElectraForMaskedLM(ElectraPreTrainedModel) - added:
class ElectraForMultipleChoice(ElectraPreTrainedModel) - added:
class ElectraForPreTraining(ElectraPreTrainedModel) - added:
class ElectraForQuestionAnswering(ElectraPreTrainedModel) - added:
class ElectraForSequenceClassification(ElectraPreTrainedModel) - added:
class ElectraForTokenClassification(ElectraPreTrainedModel) - added:
class ElectraModel(ElectraPreTrainedModel) - added:
class ElectraPreTrainedModel(PreTrainedModel) - added:
constructor ElectraForCausalLM(config) - added:
constructor ElectraForMaskedLM(config) - added:
constructor ElectraForMultipleChoice(config) - added:
constructor ElectraForPreTraining(config) - added:
constructor ElectraForQuestionAnswering(config) - added:
constructor ElectraForSequenceClassification(config) - added:
constructor ElectraForTokenClassification(config) - added:
constructor ElectraModel(config) - added:
field ElectraForQuestionAnswering.base_model_prefix - added:
field ElectraForQuestionAnswering.config: ElectraConfig - added:
field ElectraPreTrainedModel.base_model_prefix - added:
field ElectraPreTrainedModel.config: ElectraConfig - added:
field ElectraPreTrainedModel.load_tf_weights - added:
field ElectraPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_electra(model, config, tf_checkpoint_path, discriminator_or_generator=…) - added:
method ElectraForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tens… - added:
method ElectraForCausalLM.get_output_embeddings() - added:
method ElectraForCausalLM.set_output_embeddings(new_embeddings) - added:
method ElectraForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Op… - added:
method ElectraForMaskedLM.get_output_embeddings() - added:
method ElectraForMaskedLM.set_output_embeddings(word_embeddings) - added:
method ElectraForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_di… - added:
method ElectraForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
method ElectraForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]… - added:
method ElectraForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, r… - added:
method ElectraForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retu… - added:
method ElectraModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=… - added:
method ElectraModel.get_input_embeddings() - added:
method ElectraModel.set_input_embeddings(value)
- added:
class FlaxElectraForCausalLM(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForMaskedLM(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForMultipleChoice(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForPreTraining(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForQuestionAnswering(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForSequenceClassification(FlaxElectraPreTrainedModel) - added:
class FlaxElectraForTokenClassification(FlaxElectraPreTrainedModel) - added:
class FlaxElectraModel(FlaxElectraPreTrainedModel) - added:
class FlaxElectraPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxElectraPreTrainedModel(config: ElectraConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxElectraForCausalLM.module_class - added:
field FlaxElectraForMaskedLM.module_class - added:
field FlaxElectraForMultipleChoice.module_class - added:
field FlaxElectraForPreTraining.module_class - added:
field FlaxElectraForQuestionAnswering.module_class - added:
field FlaxElectraForSequenceClassification.module_class - added:
field FlaxElectraForTokenClassification.module_class - added:
field FlaxElectraModel.module_class - added:
field FlaxElectraPreTrainedModel.base_model_prefix - added:
field FlaxElectraPreTrainedModel.config_class - added:
field FlaxElectraPreTrainedModel.module_class: nn.Module - added:
method FlaxElectraForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxElectraForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxElectraPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxElectraPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxElectraPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFElectraForMaskedLM(TFElectraPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFElectraForMultipleChoice(TFElectraPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFElectraForPreTraining(TFElectraPreTrainedModel) - added:
class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFElectraForSequenceClassification(TFElectraPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassificationLoss) - added:
class TFElectraModel(TFElectraPreTrainedModel) - added:
class TFElectraPreTrainedModel(TFPreTrainedModel) - added:
constructor TFElectraForMaskedLM(config, **kwargs) - added:
constructor TFElectraForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFElectraForPreTraining(config, **kwargs) - added:
constructor TFElectraForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFElectraForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFElectraForTokenClassification(config, **kwargs) - added:
constructor TFElectraModel(config, *inputs, **kwargs) - added:
field TFElectraPreTrainedModel.base_model_prefix - added:
field TFElectraPreTrainedModel.config_class - added:
method TFElectraForMaskedLM.build(input_shape=…) - added:
method TFElectraForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool |… - added:
method TFElectraForMaskedLM.get_lm_head() - added:
method TFElectraForMaskedLM.get_prefix_bias_name() - added:
method TFElectraForMultipleChoice.build(input_shape=…) - added:
method TFElectraForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: b… - added:
method TFElectraForPreTraining.build(input_shape=…) - added:
method TFElectraForPreTraining.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFElectraForQuestionAnswering.build(input_shape=…) - added:
method TFElectraForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict… - added:
method TFElectraForSequenceClassification.build(input_shape=…) - added:
method TFElectraForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return… - added:
method TFElectraForTokenClassification.build(input_shape=…) - added:
method TFElectraForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFElectraModel.build(input_shape=…) - added:
method TFElectraModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.T…
- added:
class ElectraTokenizer(PreTrainedTokenizer) - added:
constructor ElectraTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field ElectraTokenizer.vocab_files_names - added:
method ElectraTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ElectraTokenizer.convert_tokens_to_string(tokens) - added:
method ElectraTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method ElectraTokenizer.get_vocab() - added:
method ElectraTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ElectraTokenizer.do_lower_case - added:
property ElectraTokenizer.vocab_size
- added:
class ElectraTokenizerFast(PreTrainedTokenizerFast) - added:
constructor ElectraTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field ElectraTokenizerFast.slow_tokenizer_class - added:
field ElectraTokenizerFast.vocab_files_names - added:
method ElectraTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method ElectraTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class Emu3Config(PretrainedConfig) - added:
class Emu3TextConfig(PretrainedConfig) - added:
class Emu3VQVAEConfig(PretrainedConfig) - added:
constructor Emu3Config(vq_config: Union[dict, Emu3VQVAEConfig]=…, text_config: Union[dict, Emu3TextConfig]=…, vocabulary_map: Optional[dict[int, int]]=…, **kwargs) - added:
constructor Emu3TextConfig(vocab_size: int=…, hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads: Optional[int]=…, hidden_act: str=…, max_position_embeddings: int=…, rms_norm_eps: float=…, use_cache: bool=…, pad_token_id: int=…, bos_token_id: int=…, eos_token_id: int=…, tie_word_embeddings: bool=…, rope_theta: float=…, rope_scal… - added:
constructor Emu3VQVAEConfig(codebook_size: int=…, embed_dim: int=…, latent_channels: int=…, double_latent: bool=…, in_channels: int=…, out_channels: int=…, temporal_downsample_factor: int=…, base_channels: int=…, channel_multiplier: list[int]=…, num_res_blocks: int=…, attn_resolutions: list[int]=…, hidden_size: int=…, num_attention_heads: int=…, attention_dropout: float=…, **kwargs) - added:
field Emu3Config.keys_to_ignore_at_inference - added:
field Emu3Config.model_type - added:
field Emu3Config.sub_configs - added:
field Emu3TextConfig.base_config_key - added:
field Emu3TextConfig.keys_to_ignore_at_inference - added:
field Emu3TextConfig.model_type - added:
field Emu3VQVAEConfig.base_config_key - added:
field Emu3VQVAEConfig.model_type
- added:
class Emu3ImageProcessor(BaseImageProcessor) - added:
constructor Emu3ImageProcessor(do_resize: bool=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, do_pad: bool=…, min_pixels: int=…, max_pixels: int=…, spatial_factor: int=…, **kwargs) -> None - added:
field Emu3ImageProcessor.model_input_names - added:
method Emu3ImageProcessor.postprocess(images: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Union[str, TensorType]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method Emu3ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, do_pa… - added:
method Emu3ImageProcessor.unnormalize(image: np.ndarray, image_mean: Union[float, Iterable[float]], image_std: Union[float, Iterable[float]], input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray
- added:
class Emu3ForCausalLM(Emu3PreTrainedModel, GenerationMixin) - added:
class Emu3ForConditionalGeneration(Emu3PreTrainedModel, GenerationMixin) - added:
class Emu3Model(Emu3PreTrainedModel) - added:
class Emu3PreTrainedModel(PreTrainedModel) - added:
class Emu3TextModel(Emu3PreTrainedModel) - added:
class Emu3VQVAE(PreTrainedModel) - added:
constructor Emu3ForCausalLM(config) - added:
constructor Emu3ForConditionalGeneration(config) - added:
constructor Emu3Model(config) - added:
constructor Emu3TextModel(config: Emu3Config) - added:
constructor Emu3VQVAE(config: Emu3VQVAEConfig) - added:
field Emu3ForCausalLM.config: Emu3TextConfig - added:
field Emu3ForConditionalGeneration.base_model_prefix - added:
field Emu3PreTrainedModel.base_model_prefix - added:
field Emu3PreTrainedModel.config: Emu3Config - added:
field Emu3PreTrainedModel.supports_gradient_checkpointing - added:
field Emu3VQVAE.base_model_prefix - added:
field Emu3VQVAE.config: Emu3VQVAEConfig - added:
field Emu3VQVAE.main_input_name - added:
method Emu3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **… - added:
method Emu3ForConditionalGeneration.decode_image_tokens(**kwargs) - added:
method Emu3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.Lo… - added:
method Emu3ForConditionalGeneration.get_decoder() - added:
method Emu3ForConditionalGeneration.get_input_embeddings() - added:
method Emu3ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Emu3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, **kwargs) - added:
method Emu3ForConditionalGeneration.set_decoder(decoder) - added:
method Emu3ForConditionalGeneration.set_input_embeddings(value) - added:
method Emu3Model.decode_image_tokens(image_tokens: torch.LongTensor, height: int, width: int) - added:
method Emu3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwar… - added:
method Emu3Model.get_decoder() - added:
method Emu3Model.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.LongTensor) - added:
method Emu3Model.get_image_tokens(pixel_values: torch.FloatTensor, image_sizes: torch.LongTensor) - added:
method Emu3Model.get_input_embeddings() - added:
method Emu3Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Emu3Model.set_decoder(decoder) - added:
method Emu3Model.set_input_embeddings(value) - added:
method Emu3TextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method Emu3VQVAE.decode(hidden_states: torch.Tensor) - added:
method Emu3VQVAE.encode(pixel_values: torch.Tensor, image_sizes: torch.Tensor) - added:
property Emu3ForConditionalGeneration.text_model - added:
property Emu3ForConditionalGeneration.vocabulary_mapping - added:
property Emu3ForConditionalGeneration.vqmodel
- added:
class Emu3ForCausalLM(LlamaForCausalLM, Emu3PreTrainedModel, GenerationMixin) - added:
class Emu3ForConditionalGeneration(Emu3PreTrainedModel, GenerationMixin) - added:
class Emu3Model(Emu3PreTrainedModel) - added:
class Emu3PreTrainedModel(ChameleonPreTrainedModel, Emu3VQVAE) - added:
class Emu3TextModel(LlamaModel, Emu3PreTrainedModel) - added:
class Emu3VQVAE(PreTrainedModel) - added:
constructor Emu3ForCausalLM(config) - added:
constructor Emu3ForConditionalGeneration(config) - added:
constructor Emu3Model(config) - added:
constructor Emu3TextModel(config: Emu3Config) - added:
constructor Emu3VQVAE(config: Emu3VQVAEConfig) - added:
field Emu3ForCausalLM.config: Emu3TextConfig - added:
field Emu3ForConditionalGeneration.base_model_prefix - added:
field Emu3VQVAE.base_model_prefix - added:
field Emu3VQVAE.config: Emu3VQVAEConfig - added:
field Emu3VQVAE.main_input_name - added:
method Emu3ForCausalLM.forward(**super_kwargs) - added:
method Emu3ForConditionalGeneration.decode_image_tokens(**kwargs) - added:
method Emu3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.Lo… - added:
method Emu3ForConditionalGeneration.get_decoder() - added:
method Emu3ForConditionalGeneration.get_input_embeddings() - added:
method Emu3ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Emu3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, **kwargs) - added:
method Emu3ForConditionalGeneration.set_decoder(decoder) - added:
method Emu3ForConditionalGeneration.set_input_embeddings(value) - added:
method Emu3Model.decode_image_tokens(image_tokens: torch.LongTensor, height: int, width: int) - added:
method Emu3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwar… - added:
method Emu3Model.get_decoder() - added:
method Emu3Model.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.LongTensor) - added:
method Emu3Model.get_image_tokens(pixel_values: torch.FloatTensor, image_sizes: torch.LongTensor) - added:
method Emu3Model.get_input_embeddings() - added:
method Emu3Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Emu3Model.set_decoder(decoder) - added:
method Emu3Model.set_input_embeddings(value) - added:
method Emu3VQVAE.decode(hidden_states: torch.Tensor) - added:
method Emu3VQVAE.encode(pixel_values: torch.Tensor, image_sizes: torch.Tensor) - added:
property Emu3ForConditionalGeneration.text_model - added:
property Emu3ForConditionalGeneration.vocabulary_mapping - added:
property Emu3ForConditionalGeneration.vqmodel
- added:
class Emu3Processor(ProcessorMixin) - added:
constructor Emu3Processor(image_processor, tokenizer, chat_template=…, **kwargs) - added:
field Emu3Processor.attributes - added:
field Emu3Processor.image_processor_class - added:
field Emu3Processor.tokenizer_class - added:
method Emu3Processor.calculate_generate_size(ratio, image_area, spatial_factor) - added:
method Emu3Processor.postprocess(images: ImageInput, **kwargs)
- added:
export define_import_structure
- added:
class EncodecConfig(PretrainedConfig) - added:
constructor EncodecConfig(target_bandwidths=…, sampling_rate=…, audio_channels=…, normalize=…, chunk_length_s=…, overlap=…, hidden_size=…, num_filters=…, num_residual_layers=…, upsampling_ratios=…, norm_type=…, kernel_size=…, last_kernel_size=…, residual_kernel_size=…, dilation_growth_rate=…, use_causal_conv=…, pad_mode=…, compress=…, num_lstm_layers=…, trim_right_ratio=…, codebook_size=…, codebo… - added:
field EncodecConfig.model_type - added:
property EncodecConfig.chunk_length -> Optional[int] - added:
property EncodecConfig.chunk_stride -> Optional[int] - added:
property EncodecConfig.codebook_nbits -> int - added:
property EncodecConfig.frame_rate -> int - added:
property EncodecConfig.hop_length -> int - added:
property EncodecConfig.num_quantizers -> int
- added:
class EncodecFeatureExtractor(SequenceFeatureExtractor) - added:
constructor EncodecFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, chunk_length_s: Optional[float]=…, overlap: Optional[float]=…, **kwargs) - added:
field EncodecFeatureExtractor.model_input_names - added:
property EncodecFeatureExtractor.chunk_length -> Optional[int] - added:
property EncodecFeatureExtractor.chunk_stride -> Optional[int]
- added:
class EncodecModel(EncodecPreTrainedModel) - added:
class EncodecPreTrainedModel(PreTrainedAudioTokenizerBase) - added:
constructor EncodecModel(config: EncodecConfig) - added:
field EncodecPreTrainedModel.base_model_prefix - added:
field EncodecPreTrainedModel.config: EncodecConfig - added:
field EncodecPreTrainedModel.main_input_name - added:
method EncodecModel.decode(audio_codes: torch.LongTensor, audio_scales: torch.Tensor, padding_mask: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, last_frame_pad_length: Optional[int]=…) -> Union[tuple[torch.Tensor, torch.Tensor], EncodecDecoderOutput] - added:
method EncodecModel.encode(input_values: torch.Tensor, padding_mask: Optional[torch.Tensor]=…, bandwidth: Optional[float]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor, Optional[torch.Tensor], int], EncodecEncoderOutput] - added:
method EncodecModel.forward(input_values: torch.FloatTensor, padding_mask: Optional[torch.BoolTensor]=…, bandwidth: Optional[float]=…, audio_codes: Optional[torch.LongTensor]=…, audio_scales: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, last_frame_pad_length: Optional[int]=…) -> Union[tuple[torch.Tensor, torch.Tensor], EncodecOutput] - added:
method EncodecModel.get_encoder()
- added:
export define_import_structure
- added:
class EncoderDecoderConfig(PretrainedConfig) - added:
constructor EncoderDecoderConfig(**kwargs) - added:
field EncoderDecoderConfig.has_no_defaults_at_init - added:
field EncoderDecoderConfig.model_type - added:
field EncoderDecoderConfig.sub_configs - added:
static method EncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs) -> PretrainedConfig
- added:
class EncoderDecoderModel(PreTrainedModel, GenerationMixin) - added:
constructor EncoderDecoderModel(config: Optional[PretrainedConfig]=…, encoder: Optional[PreTrainedModel]=…, decoder: Optional[PreTrainedModel]=…) - added:
field EncoderDecoderModel.base_model_prefix - added:
field EncoderDecoderModel.config: EncoderDecoderConfig - added:
field EncoderDecoderModel.main_input_name - added:
field EncoderDecoderModel.supports_gradient_checkpointing - added:
method EncoderDecoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torc… - added:
method EncoderDecoderModel.get_encoder() - added:
method EncoderDecoderModel.get_input_embeddings() - added:
method EncoderDecoderModel.get_output_embeddings() - added:
method EncoderDecoderModel.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method EncoderDecoderModel.resize_token_embeddings(*args, **kwargs) - added:
method EncoderDecoderModel.set_output_embeddings(new_embeddings) - added:
method EncoderDecoderModel.tie_weights() - added:
static method EncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[str]=…, decoder_pretrained_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel - added:
static method EncoderDecoderModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
- added:
class FlaxEncoderDecoderModel(FlaxPreTrainedModel) - added:
constructor FlaxEncoderDecoderModel(config: EncoderDecoderConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxEncoderDecoderModel.base_model_prefix - added:
field FlaxEncoderDecoderModel.config_class - added:
field FlaxEncoderDecoderModel.module_class - added:
method FlaxEncoderDecoderModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]… - added:
method FlaxEncoderDecoderModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxEncoderDecoderModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxEncoderDecoderModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxEncoderDecoderModel.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxEncoderDecoderModel.update_inputs_for_generation(model_outputs, model_kwargs) - added:
static method FlaxEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, decoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, *model_args, **kwargs) -> FlaxPreTrainedModel
- added:
class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss) - added:
constructor TFEncoderDecoderModel(config: PretrainedConfig | None=…, encoder: TFPreTrainedModel | None=…, decoder: TFPreTrainedModel | None=…) - added:
field TFEncoderDecoderModel.base_model_prefix - added:
field TFEncoderDecoderModel.config_class - added:
field TFEncoderDecoderModel.load_weight_prefix - added:
method TFEncoderDecoderModel.build(input_shape=…) - added:
method TFEncoderDecoderModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[tf.Tensor]] | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, decoder_inputs_embed… - added:
method TFEncoderDecoderModel.get_encoder() - added:
method TFEncoderDecoderModel.get_input_embeddings() - added:
method TFEncoderDecoderModel.get_output_embeddings() - added:
method TFEncoderDecoderModel.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFEncoderDecoderModel.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFEncoderDecoderModel.resize_token_embeddings(*args, **kwargs) - added:
method TFEncoderDecoderModel.set_output_embeddings(new_embeddings) - added:
method TFEncoderDecoderModel.tf_to_pt_weight_rename(tf_weight) - added:
static method TFEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: str | None=…, decoder_pretrained_model_name_or_path: str | None=…, *model_args, **kwargs) -> TFPreTrainedModel
- added:
export define_import_structure
- added:
class EomtConfig(PretrainedConfig) - added:
constructor EomtConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, hidden_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, layerscale_value=…, drop_path_rate=…, num_upscale_blocks=…, attention_dropout=…, use_swiglu_ffn=…, num_blocks=…, no_object_weight: float=…, class_weight: float=…, mask_weight: float=…,… - added:
field EomtConfig.model_type
- added:
class EomtImageProcessor(BaseImageProcessor) - added:
constructor EomtImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, do_split_image: bool=…, do_pad: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, ignore_index: Optional[int]=…, num_labels: Optional[int]=…, **kwargs) - added:
field EomtImageProcessor.model_input_names - added:
method EomtImageProcessor.encode_inputs(pixel_values_list: list[ImageInput], segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, ignore_index: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method EomtImageProcessor.merge_image_patches(segmentation_logits: torch.Tensor, patch_offsets: list[tuple[int, int, int]], target_sizes: list[tuple[int, int]], size: dict[str, int]) -> list[torch.Tensor] - added:
method EomtImageProcessor.post_process_instance_segmentation(outputs, target_sizes: list[tuple[int, int]], threshold: float=…, size: Optional[dict[str, int]]=…) - added:
method EomtImageProcessor.post_process_panoptic_segmentation(outputs, target_sizes: list[tuple[int, int]], threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, stuff_classes: Optional[list[int]]=…, size: Optional[dict[str, int]]=…) - added:
method EomtImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: list[tuple[int, int]], size: Optional[dict[str, int]]=…) -> np.ndarray - added:
method EomtImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, instance_id_to_semantic_id: Optional[dict[int, int]]=…, do_split_image: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_nor… - added:
method EomtImageProcessor.resize(image: np.ndarray, size: dict, resample: PILImageResampling=…, data_format=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method EomtImageProcessor.unpad_image(segmentation_logits: torch.Tensor, target_sizes: list[tuple[int, int]], size: dict[str, int]) -> list[torch.Tensor]
- added:
class EomtImageProcessorFast(BaseImageProcessorFast) - added:
constructor EomtImageProcessorFast(**kwargs: Unpack[EomtImageProcessorFastKwargs]) - added:
field EomtImageProcessorFast.default_to_square - added:
field EomtImageProcessorFast.do_normalize - added:
field EomtImageProcessorFast.do_pad - added:
field EomtImageProcessorFast.do_rescale - added:
field EomtImageProcessorFast.do_resize - added:
field EomtImageProcessorFast.do_split_image - added:
field EomtImageProcessorFast.ignore_index - added:
field EomtImageProcessorFast.image_mean - added:
field EomtImageProcessorFast.image_std - added:
field EomtImageProcessorFast.resample - added:
field EomtImageProcessorFast.size - added:
field EomtImageProcessorFast.valid_kwargs - added:
method EomtImageProcessorFast.merge_image_patches(segmentation_logits: torch.Tensor, patch_offsets: list[tuple[int, int, int]], target_sizes: list[tuple[int, int]], size: dict[str, int]) -> list[torch.Tensor] - added:
method EomtImageProcessorFast.post_process_instance_segmentation(outputs, target_sizes: list[tuple[int, int]], threshold: float=…, size: Optional[dict[str, int]]=…) - added:
method EomtImageProcessorFast.post_process_panoptic_segmentation(outputs, target_sizes: list[tuple[int, int]], threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, stuff_classes: Optional[list[int]]=…, size: Optional[dict[str, int]]=…) - added:
method EomtImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: list[tuple[int, int]], size: Optional[dict[str, int]]=…) -> np.ndarray - added:
method EomtImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[list[torch.Tensor]]=…, instance_id_to_semantic_id: Optional[dict[int, int]]=…, **kwargs: Unpack[EomtImageProcessorFastKwargs]) -> BatchFeature - added:
method EomtImageProcessorFast.unpad_image(segmentation_logits: torch.Tensor, target_sizes: list[tuple[int, int]], size: dict[str, int]) -> list[torch.Tensor]
- added:
class EomtForUniversalSegmentation(EomtPreTrainedModel) - added:
class EomtPreTrainedModel(PreTrainedModel) - added:
constructor EomtForUniversalSegmentation(config: EomtConfig) - added:
field EomtForUniversalSegmentation.main_input_name - added:
field EomtPreTrainedModel.base_model_prefix - added:
field EomtPreTrainedModel.config: EomtConfig - added:
field EomtPreTrainedModel.main_input_name - added:
field EomtPreTrainedModel.supports_gradient_checkpointing - added:
method EomtForUniversalSegmentation.forward(pixel_values: Tensor, mask_labels: Optional[list[Tensor]]=…, class_labels: Optional[list[Tensor]]=…, patch_offsets: Optional[list[Tensor]]=…, **kwargs: Unpack[TransformersKwargs]) -> EomtForUniversalSegmentationOutput - added:
method EomtForUniversalSegmentation.get_input_embeddings() - added:
method EomtForUniversalSegmentation.get_loss(loss_dict: dict[str, Tensor]) -> Tensor - added:
method EomtForUniversalSegmentation.get_loss_dict(masks_queries_logits: Tensor, class_queries_logits: Tensor, mask_labels: Tensor, class_labels: Tensor, auxiliary_predictions: dict[str, Tensor]) -> dict[str, Tensor] - added:
method EomtForUniversalSegmentation.predict(logits: torch.Tensor)
- added:
class EomtConfig(ViTConfig) - added:
class EomtForUniversalSegmentation(Mask2FormerForUniversalSegmentation) - added:
class EomtPreTrainedModel(PreTrainedModel) - added:
constructor EomtConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, hidden_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, layerscale_value=…, drop_path_rate=…, num_upscale_blocks=…, attention_dropout=…, use_swiglu_ffn=…, num_blocks=…, no_object_weight: float=…, class_weight: float=…, mask_weight: float=…,… - added:
constructor EomtForUniversalSegmentation(config: EomtConfig) - added:
field EomtConfig.model_type - added:
field EomtPreTrainedModel.base_model_prefix - added:
field EomtPreTrainedModel.config: EomtConfig - added:
field EomtPreTrainedModel.main_input_name - added:
field EomtPreTrainedModel.supports_gradient_checkpointing - added:
method EomtForUniversalSegmentation.forward(pixel_values: Tensor, mask_labels: Optional[list[Tensor]]=…, class_labels: Optional[list[Tensor]]=…, patch_offsets: Optional[list[Tensor]]=…, **kwargs: Unpack[TransformersKwargs]) -> EomtForUniversalSegmentationOutput - added:
method EomtForUniversalSegmentation.get_auxiliary_logits() - added:
method EomtForUniversalSegmentation.get_input_embeddings() - added:
method EomtForUniversalSegmentation.predict(logits: torch.Tensor)
- added:
export define_import_structure
- added:
class ErnieConfig(PretrainedConfig) - added:
class ErnieOnnxConfig(OnnxConfig) - added:
constructor ErnieConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, task_type_vocab_size=…, use_task_id=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field ErnieConfig.model_type - added:
property ErnieOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ErnieForCausalLM(ErniePreTrainedModel, GenerationMixin) - added:
class ErnieForMaskedLM(ErniePreTrainedModel) - added:
class ErnieForMultipleChoice(ErniePreTrainedModel) - added:
class ErnieForNextSentencePrediction(ErniePreTrainedModel) - added:
class ErnieForPreTraining(ErniePreTrainedModel) - added:
class ErnieForQuestionAnswering(ErniePreTrainedModel) - added:
class ErnieForSequenceClassification(ErniePreTrainedModel) - added:
class ErnieForTokenClassification(ErniePreTrainedModel) - added:
class ErnieModel(ErniePreTrainedModel) - added:
class ErniePreTrainedModel(PreTrainedModel) - added:
constructor ErnieForCausalLM(config) - added:
constructor ErnieForMaskedLM(config) - added:
constructor ErnieForMultipleChoice(config) - added:
constructor ErnieForNextSentencePrediction(config) - added:
constructor ErnieForPreTraining(config) - added:
constructor ErnieForQuestionAnswering(config) - added:
constructor ErnieForSequenceClassification(config) - added:
constructor ErnieForTokenClassification(config) - added:
constructor ErnieModel(config, add_pooling_layer=…) - added:
field ErniePreTrainedModel.base_model_prefix - added:
field ErniePreTrainedModel.config: ErnieConfig - added:
field ErniePreTrainedModel.supports_gradient_checkpointing - added:
method ErnieForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch… - added:
method ErnieForCausalLM.get_output_embeddings() - added:
method ErnieForCausalLM.set_output_embeddings(new_embeddings) - added:
method ErnieForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch… - added:
method ErnieForMaskedLM.get_output_embeddings() - added:
method ErnieForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method ErnieForMaskedLM.set_output_embeddings(new_embeddings) - added:
method ErnieForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method ErnieForNextSentencePrediction.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, ou… - added:
method ErnieForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, next_sentence_label: Optional[torch.Tensor]=…, out… - added:
method ErnieForPreTraining.get_output_embeddings() - added:
method ErnieForPreTraining.set_output_embeddings(new_embeddings) - added:
method ErnieForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tenso… - added:
method ErnieForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, ou… - added:
method ErnieForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, outpu… - added:
method ErnieModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, task_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tenso… - added:
method ErnieModel.get_input_embeddings() - added:
method ErnieModel.set_input_embeddings(value) - added:
static method ErnieForMaskedLM.can_generate() -> bool
- added:
export define_import_structure
- added:
class Ernie4_5Config(PretrainedConfig) - added:
constructor Ernie4_5Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, use_bias=…, head_dim=…, **kwargs) - added:
field Ernie4_5Config.base_model_pp_plan - added:
field Ernie4_5Config.base_model_tp_plan - added:
field Ernie4_5Config.keys_to_ignore_at_inference - added:
field Ernie4_5Config.model_type
- added:
class Ernie4_5ForCausalLM(Ernie4_5PreTrainedModel, GenerationMixin) - added:
class Ernie4_5Model(Ernie4_5PreTrainedModel) - added:
class Ernie4_5PreTrainedModel(PreTrainedModel) - added:
constructor Ernie4_5ForCausalLM(config) - added:
constructor Ernie4_5Model(config: Ernie4_5Config) - added:
field Ernie4_5PreTrainedModel.base_model_prefix - added:
field Ernie4_5PreTrainedModel.config: Ernie4_5Config - added:
field Ernie4_5PreTrainedModel.supports_gradient_checkpointing - added:
method Ernie4_5ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…… - added:
method Ernie4_5Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Ernie4_5ForCausalLM(LlamaForCausalLM) - added:
method Ernie4_5ForCausalLM.forward(**super_kwargs)
- added:
export define_import_structure
- added:
class Ernie4_5_MoeConfig(PretrainedConfig) - added:
constructor Ernie4_5_MoeConfig(vocab_size=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, use_bias=…, moe_intermediate_size=…, moe_k=…, moe_num_exper… - added:
field Ernie4_5_MoeConfig.attribute_map - added:
field Ernie4_5_MoeConfig.base_model_pp_plan - added:
field Ernie4_5_MoeConfig.base_model_tp_plan - added:
field Ernie4_5_MoeConfig.keys_to_ignore_at_inference - added:
field Ernie4_5_MoeConfig.model_type
- added:
class Ernie4_5_MoeForCausalLM(Ernie4_5_MoePreTrainedModel, GenerationMixin) - added:
class Ernie4_5_MoeModel(Ernie4_5_MoePreTrainedModel) - added:
class Ernie4_5_MoePreTrainedModel(PreTrainedModel) - added:
constructor Ernie4_5_MoeForCausalLM(config) - added:
constructor Ernie4_5_MoeModel(config: Ernie4_5_MoeConfig) - added:
field Ernie4_5_MoePreTrainedModel.base_model_prefix - added:
field Ernie4_5_MoePreTrainedModel.config: Ernie4_5_MoeConfig - added:
field Ernie4_5_MoePreTrainedModel.supports_gradient_checkpointing - added:
method Ernie4_5_MoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…… - added:
method Ernie4_5_MoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class Ernie4_5_MoeForCausalLM(MixtralForCausalLM) - added:
class Ernie4_5_MoeModel(Ernie4_5_MoePreTrainedModel) - added:
class Ernie4_5_MoePreTrainedModel(MixtralPreTrainedModel) - added:
constructor Ernie4_5_MoeForCausalLM(config) - added:
constructor Ernie4_5_MoeModel(config: Ernie4_5_MoeConfig) - added:
field Ernie4_5_MoePreTrainedModel.config: Ernie4_5_MoeConfig - added:
method Ernie4_5_MoeForCausalLM.forward(**super_kwargs) - added:
method Ernie4_5_MoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
export define_import_structure
- added:
class EsmConfig(PretrainedConfig) - added:
constructor EsmConfig(vocab_size=…, mask_token_id=…, pad_token_id=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, position_embedding_type=…, use_cache=…, emb_layer_norm_before=…, token_dropout=…, is_folding_model=…, esmfold_config=…, vocab_lis… - added:
field EsmConfig.model_type - added:
method EsmConfig.to_dict()
- added:
class EsmForMaskedLM(EsmPreTrainedModel) - added:
class EsmForSequenceClassification(EsmPreTrainedModel) - added:
class EsmForTokenClassification(EsmPreTrainedModel) - added:
class EsmModel(EsmPreTrainedModel) - added:
class EsmPreTrainedModel(PreTrainedModel) - added:
constructor EsmForMaskedLM(config) - added:
constructor EsmForSequenceClassification(config) - added:
constructor EsmForTokenClassification(config) - added:
constructor EsmModel(config, add_pooling_layer=…) - added:
field EsmPreTrainedModel.accepts_loss_kwargs - added:
field EsmPreTrainedModel.base_model_prefix - added:
field EsmPreTrainedModel.config: EsmConfig - added:
field EsmPreTrainedModel.supports_gradient_checkpointing - added:
method EsmForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, **kwargs: Unpack[… - added:
method EsmForMaskedLM.get_output_embeddings() - added:
method EsmForMaskedLM.predict_contacts(tokens, attention_mask) - added:
method EsmForMaskedLM.set_output_embeddings(new_embeddings) - added:
method EsmForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, SequenceClassifierOutput] - added:
method EsmForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, TokenClassifierOutput] - added:
method EsmModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple[torch.Tensor], BaseModelOu… - added:
method EsmModel.get_input_embeddings() - added:
method EsmModel.predict_contacts(tokens, attention_mask) - added:
method EsmModel.set_input_embeddings(value) - added:
method EsmPreTrainedModel.get_output_embeddings()
- added:
class EsmFoldPreTrainedModel(EsmPreTrainedModel) - added:
class EsmForProteinFolding(EsmPreTrainedModel) - added:
constructor EsmForProteinFolding(config) - added:
method EsmForProteinFolding.af2_idx_to_esm_idx(aa, mask) - added:
method EsmForProteinFolding.bert_mask(aa, esmaa, mask, pattern) - added:
method EsmForProteinFolding.compute_language_model_representations(esmaa: torch.Tensor) -> torch.Tensor - added:
method EsmForProteinFolding.forward(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, masking_pattern: Optional[torch.Tensor]=…, num_recycles: Optional[int]=…, output_hidden_states: Optional[bool]=…) -> EsmForProteinFoldingOutput - added:
method EsmForProteinFolding.infer(seqs: Union[str, list[str]], position_ids=…) - added:
method EsmForProteinFolding.infer_pdb(seqs, *args, **kwargs) -> str - added:
method EsmForProteinFolding.infer_pdbs(seqs: list[str], *args, **kwargs) -> list[str] - added:
static method EsmForProteinFolding.output_to_pdb(output: dict) -> list[str]
- added:
class TFEsmForMaskedLM(TFEsmPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFEsmForSequenceClassification(TFEsmPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFEsmForTokenClassification(TFEsmPreTrainedModel, TFTokenClassificationLoss) - added:
class TFEsmModel(TFEsmPreTrainedModel) - added:
class TFEsmPreTrainedModel(TFPreTrainedModel) - added:
constructor TFEsmForMaskedLM(config) - added:
constructor TFEsmForSequenceClassification(config) - added:
constructor TFEsmForTokenClassification(config) - added:
constructor TFEsmModel(config: EsmConfig, add_pooling_layer=…, *inputs, **kwargs) - added:
field TFEsmPreTrainedModel.base_model_prefix - added:
field TFEsmPreTrainedModel.config_class - added:
method TFEsmForMaskedLM.build(input_shape=…) - added:
method TFEsmForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor… - added:
method TFEsmForMaskedLM.get_lm_head() - added:
method TFEsmForMaskedLM.get_output_embeddings() - added:
method TFEsmForMaskedLM.predict_contacts(tokens, attention_mask) - added:
method TFEsmForMaskedLM.set_output_embeddings(new_embeddings) - added:
method TFEsmForSequenceClassification.build(input_shape=…) - added:
method TFEsmForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFEsmForTokenClassification.build(input_shape=…) - added:
method TFEsmForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | N… - added:
method TFEsmModel.build(input_shape=…) - added:
method TFEsmModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.ndarr… - added:
method TFEsmModel.predict_contacts(tokens, attention_mask)
- added:
export OFProtein - added:
export Rigid - added:
export Rotation - added:
export atom14_to_atom37 - added:
export chunk_layer - added:
export compute_predicted_aligned_error - added:
export compute_tm - added:
export dict_multimap - added:
export flatten_final_dims - added:
export frames_and_literature_positions_to_atom14_pos - added:
export make_atom14_masks - added:
export permute_final_dims - added:
export to_pdb - added:
export torsion_angles_to_frames
- added:
class ChunkSizeTuner - added:
constructor ChunkSizeTuner(max_chunk_size: int=…) - added:
function chunk_layer(layer: Callable, inputs: dict[str, Any], chunk_size: int, no_batch_dims: int, low_mem: bool=…, _out: Any=…, _add_into_out: bool=…) -> Any - added:
method ChunkSizeTuner.tune_chunk_size(representative_fn: Callable, args: tuple, min_chunk_size: int) -> int
- added:
function make_atom14_masks(protein: dict[str, torch.Tensor]) -> dict[str, torch.Tensor] - added:
function make_atom14_masks_np(batch: dict[str, torch.Tensor]) -> dict[str, np.ndarray]
- added:
function atom14_to_atom37(atom14: torch.Tensor, batch: dict[str, torch.Tensor]) -> torch.Tensor - added:
function build_extra_msa_feat(batch: dict[str, torch.Tensor]) -> torch.Tensor - added:
function build_template_angle_feat(template_feats: dict[str, torch.Tensor]) -> torch.Tensor - added:
function build_template_pair_feat(batch: dict[str, torch.Tensor], min_bin: torch.types.Number, max_bin: torch.types.Number, no_bins: int, use_unit_vector: bool=…, eps: float=…, inf: float=…) -> torch.Tensor - added:
function frames_and_literature_positions_to_atom14_pos(r: Rigid, aatype: torch.Tensor, default_frames: torch.Tensor, group_idx: torch.Tensor, atom_mask: torch.Tensor, lit_positions: torch.Tensor) -> torch.Tensor - added:
function pseudo_beta_fn(aatype: torch.Tensor, all_atom_positions: torch.Tensor, all_atom_masks: None) -> torch.Tensor | (aatype: torch.Tensor, all_atom_positions: torch.Tensor, all_atom_masks: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor] - added:
function torsion_angles_to_frames(r: Rigid, alpha: torch.Tensor, aatype: torch.Tensor, rrgdf: torch.Tensor) -> Rigid
- added:
function compute_predicted_aligned_error(logits: torch.Tensor, max_bin: int=…, no_bins: int=…, **kwargs) -> dict[str, torch.Tensor] - added:
function compute_tm(logits: torch.Tensor, residue_weights: Optional[torch.Tensor]=…, max_bin: int=…, no_bins: int=…, eps: float=…, **kwargs) -> torch.Tensor
- added:
class Protein - added:
field Protein.aatype: np.ndarray - added:
field Protein.atom_mask: np.ndarray - added:
field Protein.atom_positions: np.ndarray - added:
field Protein.b_factors: np.ndarray - added:
field Protein.chain_index: Optional[np.ndarray] - added:
field Protein.parents: Optional[Sequence[str]] - added:
field Protein.parents_chain_index: Optional[Sequence[int]] - added:
field Protein.remark: Optional[str] - added:
field Protein.residue_index: np.ndarray - added:
function add_pdb_headers(prot: Protein, pdb_str: str) -> str - added:
function from_prediction(features: FeatureDict, result: ModelOutput, b_factors: Optional[np.ndarray]=…, chain_index: Optional[np.ndarray]=…, remark: Optional[str]=…, parents: Optional[Sequence[str]]=…, parents_chain_index: Optional[Sequence[int]]=…) -> Protein - added:
function from_proteinnet_string(proteinnet_str: str) -> Protein - added:
function get_pdb_headers(prot: Protein, chain_id: int=…) -> list[str] - added:
function ideal_atom_mask(prot: Protein) -> np.ndarray - added:
function to_pdb(prot: Protein) -> str - added:
variable FeatureDict - added:
variable ModelOutput - added:
variable PICO_TO_ANGSTROM
- added:
function aatype_to_str_sequence(aatype: Sequence[int]) -> str - added:
function chi_angle_atom(atom_index: int) -> np.ndarray - added:
function load_stereo_chemical_props() -> tuple[Mapping[str, list[Bond]], Mapping[str, list[Bond]], Mapping[str, list[BondAngle]]] - added:
function make_atom14_dists_bounds(overlap_tolerance: float=…, bond_length_tolerance_factor: int=…) -> dict[str, np.ndarray] - added:
function map_structure_with_atom_order(in_list: list, first_call: bool=…) -> list - added:
function sequence_to_onehot(sequence: str, mapping: Mapping[str, int], map_unknown_to_x: bool=…) -> np.ndarray - added:
variable Bond - added:
variable BondAngle - added:
variable HHBLITS_AA_TO_ID: dict[str, int] - added:
variable ID_TO_HHBLITS_AA: dict[int, str] - added:
variable MAP_HHBLITS_AATYPE_TO_OUR_AATYPE: tuple[int, ...] - added:
variable STANDARD_ATOM_MASK - added:
variable atom_order: dict[str, int] - added:
variable atom_type_num - added:
variable atom_types: list[str] - added:
variable between_res_bond_length_c_n: tuple[float, float] - added:
variable between_res_bond_length_stddev_c_n: tuple[float, float] - added:
variable between_res_cos_angles_c_n_ca: tuple[float, float] - added:
variable between_res_cos_angles_ca_c_n: tuple[float, float] - added:
variable ca_ca - added:
variable chi_angles_atom_indices - added:
variable chi_angles_atom_indices_list: list[list[list[str]]] - added:
variable chi_angles_atom_indices_ours: list - added:
variable chi_angles_atoms: dict[str, list[list[str]]] - added:
variable chi_angles_mask: list[list[float]] - added:
variable chi_atom_1_one_hot - added:
variable chi_atom_2_one_hot - added:
variable chi_groups_for_atom - added:
variable chi_groups_for_atom: dict[tuple[str, str], list[tuple[int, int]]] - added:
variable chi_pi_periodic: list[list[float]] - added:
variable residue_atom_renaming_swaps: dict[str, dict[str, str]] - added:
variable residue_atoms: dict[str, list[str]] - added:
variable resname_to_idx: dict[str, int] - added:
variable resnames: list[str] - added:
variable restype_1to3: dict[str, str] - added:
variable restype_3to1: dict[str, str] - added:
variable restype_atom14_ambiguous_atoms - added:
variable restype_atom14_ambiguous_atoms_swap_idx: np.ndarray - added:
variable restype_atom14_mask - added:
variable restype_atom14_rigid_group_positions - added:
variable restype_atom14_to_rigid_group - added:
variable restype_atom37_mask - added:
variable restype_atom37_rigid_group_positions - added:
variable restype_atom37_to_rigid_group - added:
variable restype_name_to_atom14_names: dict[str, list[str]] - added:
variable restype_num - added:
variable restype_order: dict[str, int] - added:
variable restype_order_with_x: dict[str, int] - added:
variable restype_rigid_group_default_frame - added:
variable restypes: list[str] - added:
variable restypes_with_x: list[str] - added:
variable restypes_with_x_and_gap: list[str] - added:
variable rigid_group_atom_positions: dict[str, list[tuple[str, int, tuple[float, float, float]]]] - added:
variable unk_restype - added:
variable unk_restype_index - added:
variable van_der_waals_radius: dict[str, float]
- added:
class Rigid - added:
class Rotation - added:
constructor Rigid(rots: Rotation | None, trans: torch.Tensor | None) - added:
constructor Rotation(rot_mats: torch.Tensor | None=…, quats: torch.Tensor | None=…, normalize_quats: bool=…) - added:
function identity_quats(batch_dims: tuple[int, ...], dtype: torch.dtype | None=…, device: torch.device | None=…, requires_grad: bool=…) -> torch.Tensor - added:
function identity_rot_mats(batch_dims: tuple[int, ...], dtype: torch.dtype | None=…, device: torch.device | None=…, requires_grad: bool=…) -> torch.Tensor - added:
function identity_trans(batch_dims: tuple[int, ...], dtype: torch.dtype | None=…, device: torch.device | None=…, requires_grad: bool=…) -> torch.Tensor - added:
function invert_quat(quat: torch.Tensor) -> torch.Tensor - added:
function invert_rot_mat(rot_mat: torch.Tensor) -> torch.Tensor - added:
function quat_multiply(quat1: torch.Tensor, quat2: torch.Tensor) -> torch.Tensor - added:
function quat_multiply_by_vec(quat: torch.Tensor, vec: torch.Tensor) -> torch.Tensor - added:
function quat_to_rot(quat: torch.Tensor) -> torch.Tensor - added:
function rot_matmul(a: torch.Tensor, b: torch.Tensor) -> torch.Tensor - added:
function rot_to_quat(rot: torch.Tensor) -> torch.Tensor - added:
function rot_vec_mul(r: torch.Tensor, t: torch.Tensor) -> torch.Tensor - added:
method Rigid.apply(pts: torch.Tensor) -> torch.Tensor - added:
method Rigid.apply_rot_fn(fn: Callable[[Rotation], Rotation]) -> Rigid - added:
method Rigid.apply_trans_fn(fn: Callable[[torch.Tensor], torch.Tensor]) -> Rigid - added:
method Rigid.compose(r: Rigid) -> Rigid - added:
method Rigid.compose_q_update_vec(q_update_vec: torch.Tensor) -> Rigid - added:
method Rigid.cuda() -> Rigid - added:
method Rigid.get_rots() -> Rotation - added:
method Rigid.get_trans() -> torch.Tensor - added:
method Rigid.invert() -> Rigid - added:
method Rigid.invert_apply(pts: torch.Tensor) -> torch.Tensor - added:
method Rigid.map_tensor_fn(fn: Callable[[torch.Tensor], torch.Tensor]) -> Rigid - added:
method Rigid.scale_translation(trans_scale_factor: float) -> Rigid - added:
method Rigid.stop_rot_gradient() -> Rigid - added:
method Rigid.to_tensor_4x4() -> torch.Tensor - added:
method Rigid.to_tensor_7() -> torch.Tensor - added:
method Rigid.unsqueeze(dim: int) -> Rigid - added:
method Rotation.apply(pts: torch.Tensor) -> torch.Tensor - added:
method Rotation.compose_q(r: Rotation, normalize_quats: bool=…) -> Rotation - added:
method Rotation.compose_q_update_vec(q_update_vec: torch.Tensor, normalize_quats: bool=…) -> Rotation - added:
method Rotation.compose_r(r: Rotation) -> Rotation - added:
method Rotation.cuda() -> Rotation - added:
method Rotation.detach() -> Rotation - added:
method Rotation.get_cur_rot() -> torch.Tensor - added:
method Rotation.get_quats() -> torch.Tensor - added:
method Rotation.get_rot_mats() -> torch.Tensor - added:
method Rotation.invert() -> Rotation - added:
method Rotation.invert_apply(pts: torch.Tensor) -> torch.Tensor - added:
method Rotation.map_tensor_fn(fn: Callable[[torch.Tensor], torch.Tensor]) -> Rotation - added:
method Rotation.to(device: torch.device | None, dtype: torch.dtype | None) -> Rotation - added:
method Rotation.unsqueeze(dim: int) -> Rotation - added:
property Rigid.device -> torch.device - added:
property Rigid.shape -> torch.Size - added:
property Rotation.device -> torch.device - added:
property Rotation.dtype -> torch.dtype - added:
property Rotation.requires_grad -> bool - added:
property Rotation.shape -> torch.Size - added:
static method Rigid.cat(ts: Sequence[Rigid], dim: int) -> Rigid - added:
static method Rigid.from_3_points(p_neg_x_axis: torch.Tensor, origin: torch.Tensor, p_xy_plane: torch.Tensor, eps: float=…) -> Rigid - added:
static method Rigid.from_tensor_4x4(t: torch.Tensor) -> Rigid - added:
static method Rigid.from_tensor_7(t: torch.Tensor, normalize_quats: bool=…) -> Rigid - added:
static method Rigid.identity(shape: tuple[int, ...], dtype: torch.dtype | None=…, device: torch.device | None=…, requires_grad: bool=…, fmt: str=…) -> Rigid - added:
static method Rigid.make_transform_from_reference(n_xyz: torch.Tensor, ca_xyz: torch.Tensor, c_xyz: torch.Tensor, eps: float=…) -> Rigid - added:
static method Rotation.cat(rs: Sequence[Rotation], dim: int) -> Rotation - added:
static method Rotation.identity(shape, dtype: torch.dtype | None=…, device: torch.device | None=…, requires_grad: bool=…, fmt: str=…) -> Rotation
- added:
function add(m1: torch.Tensor, m2: torch.Tensor, inplace: bool) -> torch.Tensor - added:
function batched_gather(data: torch.Tensor, inds: torch.Tensor, dim: int=…, no_batch_dims: int=…) -> torch.Tensor - added:
function dict_map(fn: Callable[[T], Any], dic: dict[Any, Union[dict, list, tuple, T]], leaf_type: type[T]) -> dict[Any, Union[dict, list, tuple, Any]] - added:
function dict_multimap(fn: Callable[[list], Any], dicts: list[dict]) -> dict - added:
function flatten_final_dims(t: torch.Tensor, no_dims: int) -> torch.Tensor - added:
function masked_mean(mask: torch.Tensor, value: torch.Tensor, dim: int, eps: float=…) -> torch.Tensor - added:
function one_hot(x: torch.Tensor, v_bins: torch.Tensor) -> torch.Tensor - added:
function permute_final_dims(tensor: torch.Tensor, inds: list[int]) -> torch.Tensor - added:
function pts_to_distogram(pts: torch.Tensor, min_bin: torch.types.Number=…, max_bin: torch.types.Number=…, no_bins: int=…) -> torch.Tensor - added:
function tree_map(fn: Callable[[T], Any], tree: T, leaf_type: type[T]) -> Any | (fn: Callable[[T], Any], tree: dict, leaf_type: type[T]) -> dict | (fn: Callable[[T], Any], tree: list, leaf_type: type[T]) -> list | (fn: Callable[[T], Any], tree: tuple, leaf_type: type[T]) -> tuple - added:
variable T - added:
variable tensor_tree_map
- added:
class EsmTokenizer(PreTrainedTokenizer) - added:
constructor EsmTokenizer(vocab_file, unk_token=…, cls_token=…, pad_token=…, mask_token=…, eos_token=…, **kwargs) - added:
field EsmTokenizer.model_input_names - added:
field EsmTokenizer.vocab_files_names - added:
method EsmTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method EsmTokenizer.get_special_tokens_mask(token_ids_0: list, token_ids_1: Optional[list]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method EsmTokenizer.get_vocab() - added:
method EsmTokenizer.id_to_token(index: int) -> str - added:
method EsmTokenizer.save_vocabulary(save_directory, filename_prefix) - added:
method EsmTokenizer.token_to_id(token: str) -> int - added:
property EsmTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class EvollaConfig(PretrainedConfig) - added:
constructor EvollaConfig(protein_encoder_config=…, vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, rms_norm_eps=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, aligner_ffn_mult=…, aligner_enable_bias=…, aligner_attention_probs_dropout_prob=…, aligner_nu… - added:
field EvollaConfig.model_type - added:
field EvollaConfig.sub_configs
- added:
class EvollaForProteinText2Text(EvollaPreTrainedModel, GenerationMixin) - added:
class EvollaModel(EvollaPreTrainedModel) - added:
class EvollaPreTrainedModel(PreTrainedModel) - added:
constructor EvollaForProteinText2Text(config) - added:
constructor EvollaModel(config: EvollaConfig) - added:
field EvollaPreTrainedModel.base_model_prefix - added:
field EvollaPreTrainedModel.config: EvollaConfig - added:
field EvollaPreTrainedModel.supports_gradient_checkpointing - added:
method EvollaForProteinText2Text.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, protein_input_ids: Optional[torch.LongTensor]=…, protein_attention_mask: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, **kwargs) - added:
method EvollaForProteinText2Text.get_input_embeddings() - added:
method EvollaForProteinText2Text.set_input_embeddings(value) - added:
method EvollaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, protein_input_ids: Optional[torch.LongTensor]=…, protein_attention_mask: Optional[torch.… - added:
method EvollaModel.get_input_embeddings() - added:
method EvollaModel.set_input_embeddings(value)
- added:
class EvollaForProteinText2Text(EvollaPreTrainedModel, GenerationMixin) - added:
class EvollaModel(EvollaPreTrainedModel) - added:
class EvollaPreTrainedModel(LlamaPreTrainedModel) - added:
constructor EvollaForProteinText2Text(config) - added:
constructor EvollaModel(config: EvollaConfig) - added:
method EvollaForProteinText2Text.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, protein_input_ids: Optional[torch.LongTensor]=…, protein_attention_mask: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, **kwargs) - added:
method EvollaForProteinText2Text.get_input_embeddings() - added:
method EvollaForProteinText2Text.set_input_embeddings(value) - added:
method EvollaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, protein_input_ids: Optional[torch.LongTensor]=…, protein_attention_mask: Optional[torch.… - added:
method EvollaModel.get_input_embeddings() - added:
method EvollaModel.set_input_embeddings(value)
- added:
class EvollaProcessor(ProcessorMixin) - added:
constructor EvollaProcessor(protein_tokenizer, tokenizer=…, protein_max_length=…, text_max_length=…, **kwargs) - added:
field EvollaProcessor.attributes - added:
field EvollaProcessor.protein_tokenizer_class - added:
field EvollaProcessor.protein_tokenizer_dir_name - added:
field EvollaProcessor.tokenizer_class - added:
field EvollaProcessor.valid_kwargs - added:
method EvollaProcessor.batch_decode(*args, **kwargs) - added:
method EvollaProcessor.decode(*args, **kwargs) - added:
method EvollaProcessor.process_proteins(proteins, protein_max_length=…) - added:
method EvollaProcessor.process_text(texts, text_max_length: int=…) - added:
method EvollaProcessor.protein_batch_decode(*args, **kwargs) - added:
method EvollaProcessor.protein_decode(*args, **kwargs) - added:
method EvollaProcessor.save_pretrained(save_directory, **kwargs) - added:
static method EvollaProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
export define_import_structure
- added:
class Exaone4Config(PretrainedConfig) - added:
constructor Exaone4Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_dropout=…, sliding_window=…, sliding_window_pattern=…, layer_types=…,… - added:
field Exaone4Config.base_model_pp_plan - added:
field Exaone4Config.base_model_tp_plan - added:
field Exaone4Config.keys_to_ignore_at_inference - added:
field Exaone4Config.model_type
- added:
class Exaone4ForCausalLM(Exaone4PreTrainedModel, GenerationMixin) - added:
class Exaone4ForQuestionAnswering(GenericForQuestionAnswering, Exaone4PreTrainedModel) - added:
class Exaone4ForSequenceClassification(GenericForSequenceClassification, Exaone4PreTrainedModel) - added:
class Exaone4ForTokenClassification(GenericForTokenClassification, Exaone4PreTrainedModel) - added:
class Exaone4Model(Exaone4PreTrainedModel) - added:
class Exaone4PreTrainedModel(PreTrainedModel) - added:
constructor Exaone4ForCausalLM(config) - added:
constructor Exaone4Model(config: Exaone4Config) - added:
field Exaone4ForQuestionAnswering.base_model_prefix - added:
field Exaone4PreTrainedModel.base_model_prefix - added:
field Exaone4PreTrainedModel.config: Exaone4Config - added:
field Exaone4PreTrainedModel.config_class - added:
field Exaone4PreTrainedModel.supports_gradient_checkpointing - added:
method Exaone4ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method Exaone4Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPast]
- added:
class Exaone4Config(PretrainedConfig) - added:
class Exaone4ForCausalLM(LlamaForCausalLM) - added:
class Exaone4ForQuestionAnswering(LlamaForQuestionAnswering) - added:
class Exaone4ForSequenceClassification(LlamaForSequenceClassification) - added:
class Exaone4ForTokenClassification(LlamaForTokenClassification) - added:
class Exaone4Model(Exaone4PreTrainedModel, LlamaModel) - added:
class Exaone4PreTrainedModel(LlamaPreTrainedModel) - added:
constructor Exaone4Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_dropout=…, sliding_window=…, sliding_window_pattern=…, layer_types=…,… - added:
constructor Exaone4Model(config: Exaone4Config) - added:
field Exaone4Config.base_model_pp_plan - added:
field Exaone4Config.base_model_tp_plan - added:
field Exaone4Config.keys_to_ignore_at_inference - added:
field Exaone4Config.model_type - added:
field Exaone4PreTrainedModel.config_class - added:
method Exaone4ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method Exaone4Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPast]
- added:
export define_import_structure
- added:
class FalconConfig(PretrainedConfig) - added:
constructor FalconConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_ln_in_parallel_attn=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, hidden_dropout=…, attention_dropout=…, num_kv_heads=…, alibi=…, new_decoder_architecture=…, multi_query=…, parallel_attn=…, bias=…, max_position_embeddings=…, rope_theta=…, rope_scaling=…, bos_token_id=…, eos_token_id=… - added:
field FalconConfig.keys_to_ignore_at_inference - added:
field FalconConfig.model_type - added:
property FalconConfig.head_dim - added:
property FalconConfig.rotary
- added:
class FalconForCausalLM(FalconPreTrainedModel, GenerationMixin) - added:
class FalconForQuestionAnswering(FalconPreTrainedModel) - added:
class FalconForSequenceClassification(FalconPreTrainedModel) - added:
class FalconForTokenClassification(FalconPreTrainedModel) - added:
class FalconModel(FalconPreTrainedModel) - added:
class FalconPreTrainedModel(PreTrainedModel) - added:
constructor FalconForCausalLM(config: FalconConfig) - added:
constructor FalconForQuestionAnswering(config) - added:
constructor FalconForSequenceClassification(config: FalconConfig) - added:
constructor FalconForTokenClassification(config: FalconConfig) - added:
constructor FalconModel(config: FalconConfig) - added:
constructor FalconPreTrainedModel(*inputs, **kwargs) - added:
field FalconPreTrainedModel.base_model_prefix - added:
field FalconPreTrainedModel.config: FalconConfig - added:
field FalconPreTrainedModel.supports_gradient_checkpointing - added:
method FalconForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_a… - added:
method FalconForCausalLM.set_output_embeddings(new_embeddings: torch.Tensor) - added:
method FalconForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
method FalconForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Op… - added:
method FalconForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optio… - added:
method FalconModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor, torch.Tensor], ...]]]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, out… - added:
method FalconModel.get_input_embeddings() - added:
method FalconModel.set_input_embeddings(new_embeddings: torch.Tensor)
- added:
export define_import_structure
- added:
class FalconH1Config(PretrainedConfig) - added:
constructor FalconH1Config(vocab_size=…, tie_word_embeddings=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, num_logits_to_keep=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_position_embeddings=…, attention_dropout=…, mamba_d_ssm=…, mamba_n_heads=…, mamba_d_head=…, ma… - added:
field FalconH1Config.keys_to_ignore_at_inference - added:
field FalconH1Config.model_type - added:
property FalconH1Config.layers_block_type
- added:
class FalconH1ForCausalLM(FalconH1PreTrainedModel, GenerationMixin) - added:
class FalconH1Model(FalconH1PreTrainedModel) - added:
class FalconH1PreTrainedModel(PreTrainedModel) - added:
constructor FalconH1ForCausalLM(config) - added:
constructor FalconH1Model(config: FalconH1Config) - added:
field FalconH1PreTrainedModel.base_model_prefix - added:
field FalconH1PreTrainedModel.config: FalconH1Config - added:
field FalconH1PreTrainedModel.supports_gradient_checkpointing - added:
method FalconH1ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[FalconHybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method FalconH1ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method FalconH1Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[FalconHybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Option…
- added:
class FalconH1ForCausalLM(LlamaForCausalLM) - added:
class FalconH1Model(FalconH1PreTrainedModel) - added:
class FalconH1PreTrainedModel(PreTrainedModel) - added:
constructor FalconH1Model(config: FalconH1Config) - added:
field FalconH1PreTrainedModel.base_model_prefix - added:
field FalconH1PreTrainedModel.config: FalconH1Config - added:
field FalconH1PreTrainedModel.supports_gradient_checkpointing - added:
method FalconH1ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[FalconHybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method FalconH1ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method FalconH1Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[FalconHybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Option…
- added:
export define_import_structure
- added:
class FalconMambaConfig(PretrainedConfig) - added:
constructor FalconMambaConfig(vocab_size=…, hidden_size=…, state_size=…, num_hidden_layers=…, layer_norm_epsilon=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, expand=…, conv_kernel=…, use_bias=…, use_conv_bias=…, hidden_act=…, initializer_range=…, residual_in_fp32=…, time_step_rank=…, time_step_scale=…, time_step_min=…, time_step_max=…, time_step_init_scheme=…, time_step_floor=…, rescale_pre… - added:
field FalconMambaConfig.model_type
- added:
class FalconMambaCache - added:
class FalconMambaForCausalLM(FalconMambaPreTrainedModel, GenerationMixin) - added:
class FalconMambaModel(FalconMambaPreTrainedModel) - added:
class FalconMambaPreTrainedModel(PreTrainedModel) - added:
constructor FalconMambaCache(config: PretrainedConfig, max_batch_size: int, dtype: torch.dtype=…, device: Union[torch.device, str, None]=…) - added:
constructor FalconMambaForCausalLM(config) - added:
constructor FalconMambaModel(config) - added:
field FalconMambaCache.is_compileable - added:
field FalconMambaPreTrainedModel.base_model_prefix - added:
field FalconMambaPreTrainedModel.config: FalconMambaConfig - added:
field FalconMambaPreTrainedModel.supports_gradient_checkpointing - added:
method FalconMambaCache.reset() - added:
method FalconMambaCache.update_conv_state(layer_idx: int, new_conv_state: torch.Tensor, cache_position: torch.LongTensor) -> torch.Tensor - added:
method FalconMambaCache.update_ssm_state(layer_idx: int, new_ssm_state: torch.Tensor) - added:
method FalconMambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_params: Optional[FalconMambaCache]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…, **kw… - added:
method FalconMambaForCausalLM.get_input_embeddings() - added:
method FalconMambaForCausalLM.prepare_inputs_for_generation(input_ids, inputs_embeds=…, use_cache=…, cache_params: Optional[FalconMambaCache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, **kwargs) - added:
method FalconMambaForCausalLM.set_input_embeddings(new_embeddings) - added:
method FalconMambaModel.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, cache_params: Optional[FalconMambaCache]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…) -> Union[tuple, FalconMambaOutput] - added:
method FalconMambaModel.get_input_embeddings() - added:
method FalconMambaModel.set_input_embeddings(new_embeddings)
- added:
class FalconMambaCache(MambaCache) - added:
class FalconMambaConfig(MambaConfig) - added:
class FalconMambaForCausalLM(MambaForCausalLM) - added:
class FalconMambaModel(MambaModel, FalconMambaPreTrainedModel) - added:
class FalconMambaPreTrainedModel(MambaPreTrainedModel) - added:
constructor FalconMambaConfig(vocab_size=…, hidden_size=…, state_size=…, num_hidden_layers=…, layer_norm_epsilon=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, expand=…, conv_kernel=…, use_bias=…, use_conv_bias=…, hidden_act=…, initializer_range=…, residual_in_fp32=…, time_step_rank=…, time_step_scale=…, time_step_min=…, time_step_max=…, time_step_init_scheme=…, time_step_floor=…, rescale_pre… - added:
constructor FalconMambaModel(config) - added:
method FalconMambaModel.load_hook(state_dict, prefix, *args)
- added:
export define_import_structure
- added:
class FastSpeech2ConformerConfig(PretrainedConfig) - added:
class FastSpeech2ConformerHifiGanConfig(PretrainedConfig) - added:
class FastSpeech2ConformerWithHifiGanConfig(PretrainedConfig) - added:
constructor FastSpeech2ConformerConfig(hidden_size=…, vocab_size=…, num_mel_bins=…, encoder_num_attention_heads=…, encoder_layers=…, encoder_linear_units=…, decoder_layers=…, decoder_num_attention_heads=…, decoder_linear_units=…, speech_decoder_postnet_layers=…, speech_decoder_postnet_units=…, speech_decoder_postnet_kernel=…, positionwise_conv_kernel_size=…, encoder_normalize_before=…, decoder_no… - added:
constructor FastSpeech2ConformerHifiGanConfig(model_in_dim=…, upsample_initial_channel=…, upsample_rates=…, upsample_kernel_sizes=…, resblock_kernel_sizes=…, resblock_dilation_sizes=…, initializer_range=…, leaky_relu_slope=…, normalize_before=…, **kwargs) - added:
constructor FastSpeech2ConformerWithHifiGanConfig(model_config: Optional[dict]=…, vocoder_config: Optional[dict]=…, **kwargs) - added:
field FastSpeech2ConformerConfig.attribute_map - added:
field FastSpeech2ConformerConfig.base_config_key - added:
field FastSpeech2ConformerConfig.model_type - added:
field FastSpeech2ConformerHifiGanConfig.base_config_key - added:
field FastSpeech2ConformerHifiGanConfig.model_type - added:
field FastSpeech2ConformerWithHifiGanConfig.model_type - added:
field FastSpeech2ConformerWithHifiGanConfig.sub_configs
- added:
class FastSpeech2ConformerHifiGan(PreTrainedModel) - added:
class FastSpeech2ConformerModel(FastSpeech2ConformerPreTrainedModel) - added:
class FastSpeech2ConformerPreTrainedModel(PreTrainedModel) - added:
class FastSpeech2ConformerWithHifiGan(PreTrainedModel) - added:
constructor FastSpeech2ConformerHifiGan(config: FastSpeech2ConformerHifiGanConfig) - added:
constructor FastSpeech2ConformerModel(config: FastSpeech2ConformerConfig) - added:
constructor FastSpeech2ConformerWithHifiGan(config: FastSpeech2ConformerWithHifiGanConfig) - added:
field FastSpeech2ConformerHifiGan.config: FastSpeech2ConformerHifiGanConfig - added:
field FastSpeech2ConformerHifiGan.main_input_name - added:
field FastSpeech2ConformerPreTrainedModel.base_model_prefix - added:
field FastSpeech2ConformerPreTrainedModel.config: FastSpeech2ConformerConfig - added:
field FastSpeech2ConformerPreTrainedModel.main_input_name - added:
field FastSpeech2ConformerWithHifiGan.config: FastSpeech2ConformerWithHifiGanConfig - added:
method FastSpeech2ConformerHifiGan.apply_weight_norm() - added:
method FastSpeech2ConformerHifiGan.forward(spectrogram: torch.FloatTensor) -> torch.FloatTensor - added:
method FastSpeech2ConformerHifiGan.remove_weight_norm() - added:
method FastSpeech2ConformerModel.forward(input_ids: torch.LongTensor, attention_mask: Optional[torch.LongTensor]=…, spectrogram_labels: Optional[torch.FloatTensor]=…, duration_labels: Optional[torch.LongTensor]=…, pitch_labels: Optional[torch.FloatTensor]=…, energy_labels: Optional[torch.FloatTensor]=…, speaker_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, speaker_emb… - added:
method FastSpeech2ConformerWithHifiGan.forward(input_ids: torch.LongTensor, attention_mask: Optional[torch.LongTensor]=…, spectrogram_labels: Optional[torch.FloatTensor]=…, duration_labels: Optional[torch.LongTensor]=…, pitch_labels: Optional[torch.FloatTensor]=…, energy_labels: Optional[torch.FloatTensor]=…, speaker_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, speak…
- added:
class FastSpeech2ConformerTokenizer(PreTrainedTokenizer) - added:
constructor FastSpeech2ConformerTokenizer(vocab_file, bos_token=…, eos_token=…, pad_token=…, unk_token=…, should_strip_spaces=…, **kwargs) - added:
field FastSpeech2ConformerTokenizer.model_input_names - added:
field FastSpeech2ConformerTokenizer.vocab_files_names - added:
method FastSpeech2ConformerTokenizer.convert_tokens_to_string(tokens, **kwargs) - added:
method FastSpeech2ConformerTokenizer.decode(token_ids, **kwargs) - added:
method FastSpeech2ConformerTokenizer.get_vocab() - added:
method FastSpeech2ConformerTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method FastSpeech2ConformerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property FastSpeech2ConformerTokenizer.vocab_size
- added:
export define_import_structure
- added:
class FlaubertConfig(PretrainedConfig) - added:
class FlaubertOnnxConfig(OnnxConfig) - added:
constructor FlaubertConfig(pre_norm=…, layerdrop=…, vocab_size=…, emb_dim=…, n_layers=…, n_heads=…, dropout=…, attention_dropout=…, gelu_activation=…, sinusoidal_embeddings=…, causal=…, asm=…, n_langs=…, use_lang_emb=…, max_position_embeddings=…, embed_init_std=…, layer_norm_eps=…, init_std=…, bos_index=…, eos_index=…, pad_index=…, unk_index=…, mask_index=…, is_encoder=…, summary_type=…, summary_… - added:
field FlaubertConfig.attribute_map - added:
field FlaubertConfig.model_type - added:
property FlaubertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaubertForMultipleChoice(FlaubertPreTrainedModel) - added:
class FlaubertForQuestionAnswering(FlaubertPreTrainedModel) - added:
class FlaubertForQuestionAnsweringSimple(FlaubertPreTrainedModel) - added:
class FlaubertForSequenceClassification(FlaubertPreTrainedModel) - added:
class FlaubertForTokenClassification(FlaubertPreTrainedModel) - added:
class FlaubertModel(FlaubertPreTrainedModel) - added:
class FlaubertPreTrainedModel(PreTrainedModel) - added:
class FlaubertWithLMHeadModel(FlaubertPreTrainedModel, GenerationMixin) - added:
constructor FlaubertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor FlaubertForQuestionAnswering(config) - added:
constructor FlaubertForQuestionAnsweringSimple(config) - added:
constructor FlaubertForSequenceClassification(config) - added:
constructor FlaubertForTokenClassification(config) - added:
constructor FlaubertModel(config) - added:
constructor FlaubertPreTrainedModel(*inputs, **kwargs) - added:
constructor FlaubertWithLMHeadModel(config) - added:
field FlaubertPreTrainedModel.base_model_prefix - added:
field FlaubertPreTrainedModel.config: FlaubertConfig - added:
field FlaubertPreTrainedModel.load_tf_weights - added:
method FlaubertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels:… - added:
method FlaubertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, star… - added:
method FlaubertForQuestionAnsweringSimple.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…… - added:
method FlaubertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…,… - added:
method FlaubertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, la… - added:
method FlaubertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, lengths: Optional[torch.LongTensor]=…, cache: Optional[dict[str, torch.FloatTensor]]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch… - added:
method FlaubertModel.get_input_embeddings() - added:
method FlaubertModel.set_input_embeddings(new_embeddings) - added:
method FlaubertWithLMHeadModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: O… - added:
method FlaubertWithLMHeadModel.get_output_embeddings() - added:
method FlaubertWithLMHeadModel.prepare_inputs_for_generation(input_ids, **kwargs) - added:
method FlaubertWithLMHeadModel.set_output_embeddings(new_embeddings) - added:
property FlaubertPreTrainedModel.dummy_inputs
- added:
class TFFlaubertForMultipleChoice(TFFlaubertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFFlaubertForQuestionAnsweringSimple(TFFlaubertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFFlaubertForSequenceClassification(TFFlaubertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFFlaubertForTokenClassification(TFFlaubertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFFlaubertModel(TFFlaubertPreTrainedModel) - added:
class TFFlaubertPreTrainedModel(TFPreTrainedModel) - added:
class TFFlaubertWithLMHeadModel(TFFlaubertPreTrainedModel) - added:
constructor TFFlaubertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFFlaubertForQuestionAnsweringSimple(config, *inputs, **kwargs) - added:
constructor TFFlaubertForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFFlaubertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFFlaubertModel(config, *inputs, **kwargs) - added:
constructor TFFlaubertWithLMHeadModel(config, *inputs, **kwargs) - added:
field TFFlaubertPreTrainedModel.base_model_prefix - added:
field TFFlaubertPreTrainedModel.config_class - added:
method TFFlaubertForMultipleChoice.build(input_shape=…) - added:
method TFFlaubertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embed… - added:
method TFFlaubertForQuestionAnsweringSimple.build(input_shape=…) - added:
method TFFlaubertForQuestionAnsweringSimple.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inp… - added:
method TFFlaubertForSequenceClassification.build(input_shape=…) - added:
method TFFlaubertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inpu… - added:
method TFFlaubertForTokenClassification.build(input_shape=…) - added:
method TFFlaubertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_… - added:
method TFFlaubertModel.build(input_shape=…) - added:
method TFFlaubertModel.call(input_ids: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.… - added:
method TFFlaubertWithLMHeadModel.build(input_shape=…) - added:
method TFFlaubertWithLMHeadModel.call(input_ids: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_e… - added:
method TFFlaubertWithLMHeadModel.get_lm_head() - added:
method TFFlaubertWithLMHeadModel.get_prefix_bias_name() - added:
method TFFlaubertWithLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - added:
property TFFlaubertForMultipleChoice.dummy_inputs - added:
property TFFlaubertPreTrainedModel.dummy_inputs
- added:
class FlaubertTokenizer(PreTrainedTokenizer) - added:
constructor FlaubertTokenizer(vocab_file, merges_file, do_lowercase=…, unk_token=…, bos_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, lang2id=…, id2lang=…, **kwargs) - added:
field FlaubertTokenizer.vocab_files_names - added:
method FlaubertTokenizer.bpe(token) - added:
method FlaubertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FlaubertTokenizer.convert_tokens_to_string(tokens) - added:
method FlaubertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method FlaubertTokenizer.get_vocab() - added:
method FlaubertTokenizer.ja_tokenize(text) - added:
method FlaubertTokenizer.moses_pipeline(text, lang) - added:
method FlaubertTokenizer.moses_punct_norm(text, lang) - added:
method FlaubertTokenizer.moses_tokenize(text, lang) - added:
method FlaubertTokenizer.preprocess_text(text) - added:
method FlaubertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property FlaubertTokenizer.do_lower_case - added:
property FlaubertTokenizer.vocab_size
- added:
export define_import_structure
- added:
class FlavaConfig(PretrainedConfig) - added:
class FlavaImageCodebookConfig(PretrainedConfig) - added:
class FlavaImageConfig(PretrainedConfig) - added:
class FlavaMultimodalConfig(PretrainedConfig) - added:
class FlavaTextConfig(PretrainedConfig) - added:
constructor FlavaConfig(image_config: Optional[dict[str, Any]]=…, text_config: Optional[dict[str, Any]]=…, multimodal_config: Optional[dict[str, Any]]=…, image_codebook_config: Optional[dict[str, Any]]=…, hidden_size: int=…, layer_norm_eps: float=…, projection_dim: int=…, init_codebook: bool=…, logit_scale_init_value: float=…, initializer_range: float=…, ce_ignore_index: int=…, mim_weight: float=… - added:
constructor FlavaImageCodebookConfig(num_groups: int=…, input_channels: int=…, num_blocks_per_group: int=…, hidden_size: int=…, vocab_size: int=…, freeze: int=…, initializer_range: float=…, **kwargs) - added:
constructor FlavaImageConfig(hidden_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, intermediate_size: int=…, hidden_act: int=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…, initializer_range: float=…, layer_norm_eps: float=…, image_size: int=…, patch_size: int=…, num_channels: int=…, qkv_bias: bool=…, mask_token: bool=…, vocab_size: int=…, **kwargs) - added:
constructor FlavaMultimodalConfig(hidden_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, intermediate_size: int=…, hidden_act: int=…, hidden_dropout_prob: int=…, attention_probs_dropout_prob: int=…, initializer_range: float=…, layer_norm_eps: float=…, qkv_bias: bool=…, use_cls_token: bool=…, **kwargs) - added:
constructor FlavaTextConfig(vocab_size: int=…, type_vocab_size: int=…, max_position_embeddings: int=…, position_embedding_type: str=…, hidden_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, intermediate_size: int=…, hidden_act: str=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…, initializer_range: float=…, layer_norm_eps: float=…, pad_token_id: int=…, qk… - added:
field FlavaConfig.model_type - added:
field FlavaConfig.sub_configs - added:
field FlavaImageCodebookConfig.base_config_key - added:
field FlavaImageCodebookConfig.model_type - added:
field FlavaImageConfig.base_config_key - added:
field FlavaImageConfig.model_type - added:
field FlavaMultimodalConfig.base_config_key - added:
field FlavaMultimodalConfig.model_type - added:
field FlavaTextConfig.base_config_key - added:
field FlavaTextConfig.model_type - added:
static method FlavaConfig.from_configs(image_config: FlavaImageConfig, text_config: FlavaTextConfig, multimodal_config: FlavaMultimodalConfig, image_codebook_config: FlavaImageCodebookConfig, **kwargs)
- added:
class FlavaFeatureExtractor(FlavaImageProcessor) - added:
constructor FlavaFeatureExtractor(*args, **kwargs) -> None
- added:
class FlavaImageProcessor(BaseImageProcessor) - added:
constructor FlavaImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, Iterable[float]]]=…, image_std: Optional[Union[float, Iterable[float]]]=…, return_image_mask: bool=…, inpu… - added:
field FlavaImageProcessor.model_input_names - added:
method FlavaImageProcessor.map_pixels(image: np.ndarray) -> np.ndarray - added:
method FlavaImageProcessor.masking_generator(input_size_patches, total_mask_patches, mask_group_min_patches, mask_group_max_patches, mask_group_min_aspect_ratio, mask_group_max_aspect_ratio) -> FlavaMaskingGenerator - added:
method FlavaImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Opti… - added:
method FlavaImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
static method FlavaImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class FlavaImageProcessorFast(BaseImageProcessorFast) - added:
constructor FlavaImageProcessorFast(**kwargs: Unpack[FlavaFastImageProcessorKwargs]) - added:
field FlavaImageProcessorFast.codebook_crop_size - added:
field FlavaImageProcessorFast.codebook_do_center_crop - added:
field FlavaImageProcessorFast.codebook_do_map_pixels - added:
field FlavaImageProcessorFast.codebook_do_normalize - added:
field FlavaImageProcessorFast.codebook_do_rescale - added:
field FlavaImageProcessorFast.codebook_do_resize - added:
field FlavaImageProcessorFast.codebook_image_mean - added:
field FlavaImageProcessorFast.codebook_image_std - added:
field FlavaImageProcessorFast.codebook_resample - added:
field FlavaImageProcessorFast.codebook_rescale_factor - added:
field FlavaImageProcessorFast.codebook_size - added:
field FlavaImageProcessorFast.crop_size - added:
field FlavaImageProcessorFast.do_center_crop - added:
field FlavaImageProcessorFast.do_normalize - added:
field FlavaImageProcessorFast.do_rescale - added:
field FlavaImageProcessorFast.do_resize - added:
field FlavaImageProcessorFast.image_mean - added:
field FlavaImageProcessorFast.image_std - added:
field FlavaImageProcessorFast.input_size_patches - added:
field FlavaImageProcessorFast.mask_group_max_aspect_ratio - added:
field FlavaImageProcessorFast.mask_group_max_patches - added:
field FlavaImageProcessorFast.mask_group_min_aspect_ratio - added:
field FlavaImageProcessorFast.mask_group_min_patches - added:
field FlavaImageProcessorFast.resample - added:
field FlavaImageProcessorFast.return_codebook_pixels - added:
field FlavaImageProcessorFast.return_image_mask - added:
field FlavaImageProcessorFast.size - added:
field FlavaImageProcessorFast.total_mask_patches - added:
field FlavaImageProcessorFast.valid_kwargs - added:
method FlavaImageProcessorFast.map_pixels(image: torch.Tensor) -> torch.Tensor - added:
method FlavaImageProcessorFast.masking_generator(input_size_patches, total_mask_patches, mask_group_min_patches, mask_group_max_patches, mask_group_min_aspect_ratio, mask_group_max_aspect_ratio) -> FlavaMaskingGenerator - added:
method FlavaImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[DefaultFastImageProcessorKwargs]) -> BatchFeature - added:
static method FlavaImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class FlavaForPreTraining(FlavaPreTrainedModel) - added:
class FlavaImageCodebook(FlavaPreTrainedModel) - added:
class FlavaImageModel(FlavaPreTrainedModel) - added:
class FlavaModel(FlavaPreTrainedModel) - added:
class FlavaMultimodalModel(FlavaPreTrainedModel) - added:
class FlavaPreTrainedModel(PreTrainedModel) - added:
class FlavaTextModel(FlavaPreTrainedModel) - added:
constructor FlavaForPreTraining(config: FlavaConfig, image_codebook: Optional[nn.Module]=…) - added:
constructor FlavaImageCodebook(config: FlavaImageCodebookConfig, **kwargs: Any) - added:
constructor FlavaImageModel(config: FlavaImageConfig, add_pooling_layer: bool=…) - added:
constructor FlavaModel(config: FlavaConfig) - added:
constructor FlavaMultimodalModel(config: FlavaMultimodalConfig, add_pooling_layer=…) - added:
constructor FlavaTextModel(config: FlavaTextConfig, add_pooling_layer: bool=…) - added:
field FlavaImageCodebook.base_model_prefix - added:
field FlavaImageCodebook.config: FlavaImageCodebookConfig - added:
field FlavaImageCodebook.main_input_name - added:
field FlavaImageCodebook.supports_gradient_checkpointing - added:
field FlavaImageModel.base_model_prefix - added:
field FlavaImageModel.config: FlavaImageConfig - added:
field FlavaImageModel.main_input_name - added:
field FlavaModel.config: FlavaConfig - added:
field FlavaMultimodalModel.base_model_prefix - added:
field FlavaMultimodalModel.config: FlavaMultimodalConfig - added:
field FlavaMultimodalModel.main_input_name - added:
field FlavaPreTrainedModel.base_model_prefix - added:
field FlavaPreTrainedModel.config: FlavaConfig - added:
field FlavaPreTrainedModel.supports_gradient_checkpointing - added:
field FlavaTextModel.base_model_prefix - added:
field FlavaTextModel.config: FlavaTextConfig - added:
method FlavaForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, input_ids_masked: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, codebook_pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, image… - added:
method FlavaImageCodebook.forward(pixel_values: torch.FloatTensor) -> torch.Tensor - added:
method FlavaImageCodebook.get_codebook_indices(pixel_values: torch.Tensor) -> torch.Tensor - added:
method FlavaImageCodebook.get_codebook_probs(pixel_values: torch.Tensor) -> torch.Tensor - added:
method FlavaImageModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, interpolate_pos_encoding: Optional[bool]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method FlavaImageModel.get_input_embeddings() -> nn.Module - added:
method FlavaImageModel.set_input_embeddings(value: nn.Module) - added:
method FlavaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, image_attention_mask: Optional[torch.Tensor]=…, skip_multimodal_encoder: Optional[bool]=…, output_attentions: Option… - added:
method FlavaModel.get_image_features(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…, interpolate_pos_encoding: Optional[bool]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method FlavaModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method FlavaMultimodalModel.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method FlavaTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method FlavaTextModel.get_input_embeddings() -> PatchEmbeddings - added:
method FlavaTextModel.set_input_embeddings(value: nn.Module)
- added:
class FlavaProcessor(ProcessorMixin) - added:
constructor FlavaProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field FlavaProcessor.attributes - added:
field FlavaProcessor.image_processor_class - added:
field FlavaProcessor.tokenizer_class - added:
field FlavaProcessor.valid_processor_kwargs - added:
property FlavaProcessor.feature_extractor - added:
property FlavaProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class FlexOlmoConfig(PretrainedConfig) - added:
constructor FlexOlmoConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, num_experts_per_tok=…, n… - added:
field FlexOlmoConfig.base_model_pp_plan - added:
field FlexOlmoConfig.base_model_tp_plan - added:
field FlexOlmoConfig.keys_to_ignore_at_inference - added:
field FlexOlmoConfig.model_type
- added:
class FlexOlmoForCausalLM(FlexOlmoPreTrainedModel, GenerationMixin) - added:
class FlexOlmoModel(FlexOlmoPreTrainedModel) - added:
class FlexOlmoPreTrainedModel(PreTrainedModel) - added:
constructor FlexOlmoForCausalLM(config) - added:
constructor FlexOlmoModel(config: FlexOlmoConfig) - added:
field FlexOlmoPreTrainedModel.base_model_prefix - added:
field FlexOlmoPreTrainedModel.config: FlexOlmoConfig - added:
field FlexOlmoPreTrainedModel.supports_gradient_checkpointing - added:
method FlexOlmoForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_rout… - added:
method FlexOlmoModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class FlexOlmoConfig(OlmoeConfig) - added:
class FlexOlmoForCausalLM(OlmoeForCausalLM) - added:
class FlexOlmoModel(MixtralModel) - added:
class FlexOlmoPreTrainedModel(MixtralPreTrainedModel) - added:
constructor FlexOlmoConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, num_experts_per_tok=…, n… - added:
field FlexOlmoConfig.base_model_pp_plan - added:
field FlexOlmoConfig.base_model_tp_plan - added:
field FlexOlmoConfig.keys_to_ignore_at_inference - added:
field FlexOlmoConfig.model_type - added:
method FlexOlmoModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
export define_import_structure
- added:
class Florence2Config(PretrainedConfig) - added:
class Florence2VisionConfig(PretrainedConfig) - added:
constructor Florence2Config(text_config=…, vision_config=…, image_token_id=…, is_encoder_decoder=…, **kwargs) - added:
constructor Florence2VisionConfig(in_channels=…, depths=…, patch_size=…, patch_stride=…, patch_padding=…, patch_prenorm=…, embed_dim=…, num_heads=…, num_groups=…, window_size=…, drop_path_rate=…, mlp_ratio=…, qkv_bias=…, activation_function=…, projection_dim=…, max_temporal_embeddings=…, max_position_embeddings=…, initializer_range=…, **kwargs) - added:
field Florence2Config.model_type - added:
field Florence2Config.sub_configs - added:
field Florence2VisionConfig.model_type
- added:
class Florence2ForConditionalGeneration(Florence2PreTrainedModel, GenerationMixin) - added:
class Florence2Model(Florence2PreTrainedModel) - added:
class Florence2PreTrainedModel(PreTrainedModel) - added:
class Florence2VisionBackbone(Florence2VisionPreTrainedModel) - added:
class Florence2VisionPreTrainedModel(PreTrainedModel) - added:
constructor Florence2ForConditionalGeneration(config: Florence2Config) - added:
constructor Florence2Model(config: Florence2Config) - added:
constructor Florence2VisionBackbone(config: Florence2VisionConfig) - added:
field Florence2PreTrainedModel.base_model_prefix - added:
field Florence2PreTrainedModel.config: Florence2Config - added:
field Florence2PreTrainedModel.config_class - added:
field Florence2PreTrainedModel.supports_gradient_checkpointing - added:
field Florence2VisionPreTrainedModel.config_class - added:
field Florence2VisionPreTrainedModel.main_input_name - added:
method Florence2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.… - added:
method Florence2ForConditionalGeneration.get_decoder() - added:
method Florence2ForConditionalGeneration.get_encoder() - added:
method Florence2ForConditionalGeneration.get_image_features(pixel_values: torch.Tensor, **kwargs) - added:
method Florence2ForConditionalGeneration.get_input_embeddings() - added:
method Florence2ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Florence2ForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Florence2ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Florence2ForConditionalGeneration.set_decoder(decoder) - added:
method Florence2ForConditionalGeneration.set_input_embeddings(value) - added:
method Florence2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, decoder_… - added:
method Florence2Model.get_decoder() - added:
method Florence2Model.get_encoder() - added:
method Florence2Model.get_image_features(pixel_values: torch.Tensor, **kwargs) - added:
method Florence2Model.get_input_embeddings() - added:
method Florence2Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Florence2Model.set_decoder(decoder) - added:
method Florence2Model.set_input_embeddings(value) - added:
method Florence2VisionBackbone.forward(hidden_states: torch.Tensor) - added:
property Florence2ForConditionalGeneration.language_model - added:
property Florence2ForConditionalGeneration.multi_modal_projector - added:
property Florence2ForConditionalGeneration.vision_tower
- added:
class Florence2Config(PretrainedConfig) - added:
class Florence2ForConditionalGeneration(LlavaForConditionalGeneration) - added:
class Florence2Model(LlavaModel) - added:
class Florence2PreTrainedModel(LlavaPreTrainedModel) - added:
class Florence2Processor(ProcessorMixin) - added:
class Florence2VisionBackbone(Florence2VisionPreTrainedModel) - added:
class Florence2VisionConfig(PretrainedConfig) - added:
class Florence2VisionPreTrainedModel(PreTrainedModel) - added:
constructor Florence2Config(text_config=…, vision_config=…, image_token_id=…, is_encoder_decoder=…, **kwargs) - added:
constructor Florence2Model(config: Florence2Config) - added:
constructor Florence2Processor(image_processor=…, tokenizer=…, num_additional_image_tokens: int=…, post_processor_config: Optional[dict]=…, **kwargs) - added:
constructor Florence2VisionBackbone(config: Florence2VisionConfig) - added:
constructor Florence2VisionConfig(in_channels=…, depths=…, patch_size=…, patch_stride=…, patch_padding=…, patch_prenorm=…, embed_dim=…, num_heads=…, num_groups=…, window_size=…, drop_path_rate=…, mlp_ratio=…, qkv_bias=…, activation_function=…, projection_dim=…, max_temporal_embeddings=…, max_position_embeddings=…, initializer_range=…, **kwargs) - added:
field Florence2Config.model_type - added:
field Florence2Config.sub_configs - added:
field Florence2PreTrainedModel.config_class - added:
field Florence2Processor.attributes - added:
field Florence2Processor.image_processor_class - added:
field Florence2Processor.tokenizer_class - added:
field Florence2VisionConfig.model_type - added:
field Florence2VisionPreTrainedModel.config_class - added:
field Florence2VisionPreTrainedModel.main_input_name - added:
method Florence2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.… - added:
method Florence2ForConditionalGeneration.get_encoder() - added:
method Florence2ForConditionalGeneration.get_image_features(pixel_values: torch.Tensor, **kwargs) - added:
method Florence2ForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Florence2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, decoder_… - added:
method Florence2Model.get_decoder() - added:
method Florence2Model.get_encoder() - added:
method Florence2Model.get_image_features(pixel_values: torch.Tensor, **kwargs) - added:
method Florence2Processor.batch_decode(*args, **kwargs) - added:
method Florence2Processor.decode(*args, **kwargs) - added:
method Florence2Processor.post_process_generation(text=…, sequence=…, task=…, image_size=…) -> dict[str, Any] - added:
method Florence2Processor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, **kwargs) - added:
method Florence2VisionBackbone.forward(hidden_states: torch.Tensor) - added:
property Florence2Processor.model_input_names
- added:
class Florence2Processor(ProcessorMixin) - added:
constructor Florence2Processor(image_processor=…, tokenizer=…, num_additional_image_tokens: int=…, post_processor_config: Optional[dict]=…, **kwargs) - added:
field Florence2Processor.attributes - added:
field Florence2Processor.image_processor_class - added:
field Florence2Processor.tokenizer_class - added:
method Florence2Processor.batch_decode(*args, **kwargs) - added:
method Florence2Processor.decode(*args, **kwargs) - added:
method Florence2Processor.post_process_generation(text=…, sequence=…, task=…, image_size=…) -> dict[str, Any] - added:
method Florence2Processor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, **kwargs) - added:
property Florence2Processor.model_input_names
- added:
export define_import_structure
- added:
class FNetConfig(PretrainedConfig) - added:
constructor FNetConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_tpu_fourier_optimizations=…, tpu_short_seq_length=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field FNetConfig.model_type
- added:
class FNetForMaskedLM(FNetPreTrainedModel) - added:
class FNetForMultipleChoice(FNetPreTrainedModel) - added:
class FNetForNextSentencePrediction(FNetPreTrainedModel) - added:
class FNetForPreTraining(FNetPreTrainedModel) - added:
class FNetForQuestionAnswering(FNetPreTrainedModel) - added:
class FNetForSequenceClassification(FNetPreTrainedModel) - added:
class FNetForTokenClassification(FNetPreTrainedModel) - added:
class FNetLayer(GradientCheckpointingLayer) - added:
class FNetModel(FNetPreTrainedModel) - added:
class FNetPreTrainedModel(PreTrainedModel) - added:
constructor FNetForMaskedLM(config) - added:
constructor FNetForMultipleChoice(config) - added:
constructor FNetForNextSentencePrediction(config) - added:
constructor FNetForPreTraining(config) - added:
constructor FNetForQuestionAnswering(config) - added:
constructor FNetForSequenceClassification(config) - added:
constructor FNetForTokenClassification(config) - added:
constructor FNetLayer(config) - added:
constructor FNetModel(config, add_pooling_layer=…) - added:
field FNetPreTrainedModel.base_model_prefix - added:
field FNetPreTrainedModel.config: FNetConfig - added:
field FNetPreTrainedModel.supports_gradient_checkpointing - added:
method FNetForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, MaskedLMOutput] - added:
method FNetForMaskedLM.get_output_embeddings() - added:
method FNetForMaskedLM.set_output_embeddings(new_embeddings) - added:
method FNetForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, MultipleChoiceModelOutput] - added:
method FNetForNextSentencePrediction.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> Union[tuple, NextSentencePredictorOutput] - added:
method FNetForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, next_sentence_label: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, FNetForPreTrainingOutput] - added:
method FNetForPreTraining.get_output_embeddings() - added:
method FNetForPreTraining.set_output_embeddings(new_embeddings) - added:
method FNetForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, QuestionAnsweringModelOutput] - added:
method FNetForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method FNetForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method FNetLayer.feed_forward_chunk(fourier_output) - added:
method FNetLayer.forward(hidden_states) - added:
method FNetModel.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method FNetModel.get_input_embeddings() - added:
method FNetModel.set_input_embeddings(value)
- added:
class FNetTokenizer(PreTrainedTokenizer) - added:
constructor FNetTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field FNetTokenizer.model_input_names - added:
field FNetTokenizer.vocab_files_names - added:
method FNetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FNetTokenizer.convert_tokens_to_string(tokens) - added:
method FNetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method FNetTokenizer.get_vocab() - added:
method FNetTokenizer.preprocess_text(inputs) - added:
method FNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property FNetTokenizer.vocab_size
- added:
class FNetTokenizerFast(PreTrainedTokenizerFast) - added:
constructor FNetTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field FNetTokenizerFast.model_input_names - added:
field FNetTokenizerFast.slow_tokenizer_class - added:
field FNetTokenizerFast.vocab_files_names - added:
method FNetTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FNetTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class FocalNetConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor FocalNetConfig(image_size=…, patch_size=…, num_channels=…, embed_dim=…, use_conv_embed=…, hidden_sizes=…, depths=…, focal_levels=…, focal_windows=…, hidden_act=…, mlp_ratio=…, hidden_dropout_prob=…, drop_path_rate=…, use_layerscale=…, layerscale_value=…, use_post_layernorm=…, use_post_layernorm_in_modulation=…, normalize_modulator=…, initializer_range=…, layer_norm_eps=…, encoder_stri… - added:
field FocalNetConfig.model_type
- added:
class FocalNetBackbone(FocalNetPreTrainedModel, BackboneMixin) - added:
class FocalNetForImageClassification(FocalNetPreTrainedModel) - added:
class FocalNetForMaskedImageModeling(FocalNetPreTrainedModel) - added:
class FocalNetModel(FocalNetPreTrainedModel) - added:
class FocalNetPreTrainedModel(PreTrainedModel) - added:
constructor FocalNetBackbone(config: FocalNetConfig) - added:
constructor FocalNetForImageClassification(config) - added:
constructor FocalNetForMaskedImageModeling(config) - added:
constructor FocalNetModel(config, add_pooling_layer=…, use_mask_token=…) - added:
field FocalNetBackbone.has_attentions - added:
field FocalNetPreTrainedModel.base_model_prefix - added:
field FocalNetPreTrainedModel.config: FocalNetConfig - added:
field FocalNetPreTrainedModel.main_input_name - added:
field FocalNetPreTrainedModel.supports_gradient_checkpointing - added:
method FocalNetBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method FocalNetForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, FocalNetImageClassifierOutput] - added:
method FocalNetForMaskedImageModeling.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, FocalNetMaskedImageModelingOutput] - added:
method FocalNetModel.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, FocalNetModelOutput] - added:
method FocalNetModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class FSMTConfig(PretrainedConfig) - added:
constructor FSMTConfig(langs=…, src_vocab_size=…, tgt_vocab_size=…, activation_function=…, d_model=…, max_length=…, max_position_embeddings=…, encoder_ffn_dim=…, encoder_layers=…, encoder_attention_heads=…, encoder_layerdrop=…, decoder_ffn_dim=…, decoder_layers=…, decoder_attention_heads=…, decoder_layerdrop=…, attention_dropout=…, dropout=…, activation_dropout=…, init_std=…, decoder_start_token_… - added:
field FSMTConfig.attribute_map - added:
field FSMTConfig.model_type - added:
field FSMTConfig.sub_configs
- added:
class FSMTForConditionalGeneration(PretrainedFSMTModel, GenerationMixin) - added:
class FSMTModel(PretrainedFSMTModel) - added:
class PretrainedFSMTModel(PreTrainedModel) - added:
constructor FSMTForConditionalGeneration(config: FSMTConfig) - added:
constructor FSMTModel(config: FSMTConfig) - added:
field FSMTForConditionalGeneration.base_model_prefix - added:
field PretrainedFSMTModel.base_model_prefix - added:
field PretrainedFSMTModel.config: FSMTConfig - added:
method FSMTForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.F… - added:
method FSMTForConditionalGeneration.get_decoder() - added:
method FSMTForConditionalGeneration.get_encoder() - added:
method FSMTForConditionalGeneration.get_output_embeddings() - added:
method FSMTForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method FSMTForConditionalGeneration.set_output_embeddings(value) - added:
method FSMTModel.forward(input_ids: torch.LongTensor, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values… - added:
method FSMTModel.get_encoder() - added:
method FSMTModel.get_input_embeddings() - added:
method FSMTModel.get_output_embeddings() - added:
method FSMTModel.set_input_embeddings(value) - added:
method FSMTModel.set_output_embeddings(value) - added:
property PretrainedFSMTModel.dummy_inputs
- added:
class FSMTTokenizer(PreTrainedTokenizer) - added:
constructor FSMTTokenizer(langs=…, src_vocab_file=…, tgt_vocab_file=…, merges_file=…, do_lower_case=…, unk_token=…, bos_token=…, sep_token=…, pad_token=…, **kwargs) - added:
field FSMTTokenizer.model_input_names - added:
field FSMTTokenizer.vocab_files_names - added:
method FSMTTokenizer.bpe(token) - added:
method FSMTTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FSMTTokenizer.convert_tokens_to_string(tokens) - added:
method FSMTTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method FSMTTokenizer.get_src_vocab() - added:
method FSMTTokenizer.get_tgt_vocab() - added:
method FSMTTokenizer.get_vocab() -> dict[str, int] - added:
method FSMTTokenizer.moses_detokenize(tokens, lang) - added:
method FSMTTokenizer.moses_pipeline(text, lang) - added:
method FSMTTokenizer.moses_punct_norm(text, lang) - added:
method FSMTTokenizer.moses_tokenize(text, lang) - added:
method FSMTTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property FSMTTokenizer.src_vocab_size - added:
property FSMTTokenizer.tgt_vocab_size - added:
property FSMTTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class FunnelConfig(PretrainedConfig) - added:
constructor FunnelConfig(vocab_size=…, block_sizes=…, block_repeats=…, num_decoder_layers=…, d_model=…, n_head=…, d_head=…, d_inner=…, hidden_act=…, hidden_dropout=…, attention_dropout=…, activation_dropout=…, initializer_range=…, initializer_std=…, layer_norm_eps=…, pooling_type=…, attention_type=…, separate_cls=…, truncate_seq=…, pool_q_only=…, **kwargs) - added:
field FunnelConfig.attribute_map - added:
field FunnelConfig.model_type - added:
property FunnelConfig.num_blocks - added:
property FunnelConfig.num_hidden_layers
- added:
class FunnelBaseModel(FunnelPreTrainedModel) - added:
class FunnelForMaskedLM(FunnelPreTrainedModel) - added:
class FunnelForMultipleChoice(FunnelPreTrainedModel) - added:
class FunnelForPreTraining(FunnelPreTrainedModel) - added:
class FunnelForQuestionAnswering(FunnelPreTrainedModel) - added:
class FunnelForSequenceClassification(FunnelPreTrainedModel) - added:
class FunnelForTokenClassification(FunnelPreTrainedModel) - added:
class FunnelModel(FunnelPreTrainedModel) - added:
class FunnelPreTrainedModel(PreTrainedModel) - added:
constructor FunnelBaseModel(config: FunnelConfig) -> None - added:
constructor FunnelForMaskedLM(config: FunnelConfig) -> None - added:
constructor FunnelForMultipleChoice(config: FunnelConfig) -> None - added:
constructor FunnelForPreTraining(config: FunnelConfig) -> None - added:
constructor FunnelForQuestionAnswering(config: FunnelConfig) -> None - added:
constructor FunnelForSequenceClassification(config: FunnelConfig) -> None - added:
constructor FunnelForTokenClassification(config: FunnelConfig) -> None - added:
constructor FunnelModel(config: FunnelConfig) -> None - added:
field FunnelPreTrainedModel.base_model_prefix - added:
field FunnelPreTrainedModel.config: FunnelConfig - added:
field FunnelPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_funnel(model, config, tf_checkpoint_path) - added:
method FunnelBaseModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseM… - added:
method FunnelBaseModel.get_input_embeddings() -> nn.Embedding - added:
method FunnelBaseModel.set_input_embeddings(new_embeddings: nn.Embedding) -> None - added:
method FunnelForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, MaskedLMOutput] - added:
method FunnelForMaskedLM.get_output_embeddings() -> nn.Linear - added:
method FunnelForMaskedLM.set_output_embeddings(new_embeddings: nn.Embedding) -> None - added:
method FunnelForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, MultipleChoiceModelOutput] - added:
method FunnelForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, FunnelForPreTrainingOutput] - added:
method FunnelForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) ->… - added:
method FunnelForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method FunnelForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method FunnelModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method FunnelModel.get_input_embeddings() -> nn.Embedding - added:
method FunnelModel.set_input_embeddings(new_embeddings: nn.Embedding) -> None
- added:
class TFFunnelBaseModel(TFFunnelPreTrainedModel) - added:
class TFFunnelForMaskedLM(TFFunnelPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFFunnelForMultipleChoice(TFFunnelPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFFunnelForPreTraining(TFFunnelPreTrainedModel) - added:
class TFFunnelForQuestionAnswering(TFFunnelPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFFunnelForSequenceClassification(TFFunnelPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFFunnelForTokenClassification(TFFunnelPreTrainedModel, TFTokenClassificationLoss) - added:
class TFFunnelModel(TFFunnelPreTrainedModel) - added:
class TFFunnelPreTrainedModel(TFPreTrainedModel) - added:
constructor TFFunnelBaseModel(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelForMaskedLM(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelForMultipleChoice(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelForPreTraining(config: FunnelConfig, **kwargs) -> None - added:
constructor TFFunnelForQuestionAnswering(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelForSequenceClassification(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelForTokenClassification(config: FunnelConfig, *inputs, **kwargs) -> None - added:
constructor TFFunnelModel(config: FunnelConfig, *inputs, **kwargs) -> None - added:
field TFFunnelPreTrainedModel.base_model_prefix - added:
field TFFunnelPreTrainedModel.config_class - added:
method TFFunnelBaseModel.build(input_shape=…) - added:
method TFFunnelBaseModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple[tf.Tensor] | TFBaseModelOutput - added:
method TFFunnelBaseModel.serving_output(output) - added:
method TFFunnelForMaskedLM.build(input_shape=…) - added:
method TFFunnelForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) -> tuple[tf.Tensor] | TFM… - added:
method TFFunnelForMaskedLM.get_lm_head() -> TFFunnelMaskedLMHead - added:
method TFFunnelForMaskedLM.get_prefix_bias_name() -> str - added:
method TFFunnelForMaskedLM.serving_output(output: TFMaskedLMOutput) -> TFMaskedLMOutput - added:
method TFFunnelForMultipleChoice.build(input_shape=…) - added:
method TFFunnelForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) -> tuple[tf.Tensor]… - added:
method TFFunnelForMultipleChoice.serving_output(output: TFMultipleChoiceModelOutput) -> TFMultipleChoiceModelOutput - added:
method TFFunnelForPreTraining.build(input_shape=…) - added:
method TFFunnelForPreTraining.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…, **kwargs) -> tuple[tf.Tensor] | TFFunnelForPreTrainingOutput - added:
method TFFunnelForPreTraining.serving_output(output) - added:
method TFFunnelForQuestionAnswering.build(input_shape=…) - added:
method TFFunnelForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: np.ndarray | tf.Tensor | None=…, end_positions: np.ndarray… - added:
method TFFunnelForQuestionAnswering.serving_output(output: TFQuestionAnsweringModelOutput) -> TFQuestionAnsweringModelOutput - added:
method TFFunnelForSequenceClassification.build(input_shape=…) - added:
method TFFunnelForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) -> tuple[tf… - added:
method TFFunnelForSequenceClassification.serving_output(output: TFSequenceClassifierOutput) -> TFSequenceClassifierOutput - added:
method TFFunnelForTokenClassification.build(input_shape=…) - added:
method TFFunnelForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool=…) -> tuple[tf.Te… - added:
method TFFunnelForTokenClassification.serving_output(output: TFTokenClassifierOutput) -> TFTokenClassifierOutput - added:
method TFFunnelModel.build(input_shape=…) - added:
method TFFunnelModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple[tf.Tensor] | TFBaseModelOutput - added:
method TFFunnelModel.serving_output(output) - added:
property TFFunnelForMultipleChoice.dummy_inputs - added:
property TFFunnelPreTrainedModel.dummy_inputs
- added:
class FunnelTokenizer(PreTrainedTokenizer) - added:
constructor FunnelTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, bos_token=…, eos_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field FunnelTokenizer.cls_token_type_id: int - added:
field FunnelTokenizer.vocab_files_names - added:
method FunnelTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FunnelTokenizer.convert_tokens_to_string(tokens) - added:
method FunnelTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FunnelTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method FunnelTokenizer.get_vocab() - added:
method FunnelTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property FunnelTokenizer.do_lower_case - added:
property FunnelTokenizer.vocab_size
- added:
class FunnelTokenizerFast(PreTrainedTokenizerFast) - added:
constructor FunnelTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, bos_token=…, eos_token=…, clean_text=…, tokenize_chinese_chars=…, strip_accents=…, wordpieces_prefix=…, **kwargs) - added:
field FunnelTokenizerFast.cls_token_type_id: int - added:
field FunnelTokenizerFast.slow_tokenizer_class - added:
field FunnelTokenizerFast.vocab_files_names - added:
method FunnelTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method FunnelTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method FunnelTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class FuyuConfig(PretrainedConfig) - added:
constructor FuyuConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_act=…, max_position_embeddings=…, image_size=…, patch_size=…, num_channels=…, initializer_range=…, layer_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, qk_layernorm=…, hidden_dropout=…, attention_dropout=…, partial_rotary_factor=…, pad_token_id… - added:
field FuyuConfig.keys_to_ignore_at_inference - added:
field FuyuConfig.model_type - added:
field FuyuConfig.sub_configs
- added:
class FuyuImageProcessor(BaseImageProcessor) - added:
constructor FuyuImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_pad: bool=…, padding_value: float=…, padding_mode: str=…, do_normalize: bool=…, image_mean: Union[float, list[float]]=…, image_std: Union[float, list[float]]=…, do_rescale: bool=…, rescale_factor: float=…, patch_size: Optional[dict[str, int]]=…, **kwargs) - added:
field FuyuImageProcessor.model_input_names - added:
method FuyuImageProcessor.get_num_patches(image_height: int, image_width: int, patch_size: Optional[dict[str, int]]=…) -> int - added:
method FuyuImageProcessor.pad_image(image: np.ndarray, size: dict[str, int], mode: str=…, constant_values: float=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method FuyuImageProcessor.patchify_image(image: torch.Tensor, patch_size: Optional[dict[str, int]]=…) -> torch.Tensor - added:
method FuyuImageProcessor.preprocess(images, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_pad: Optional[bool]=…, padding_value: Optional[float]=…, padding_mode: Optional[str]=…, do_normalize: Optional[bool]=…, image_mean: Optional[float]=…, image_std: Optional[float]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…,… - added:
method FuyuImageProcessor.preprocess_with_tokenizer_info(image_input: torch.Tensor, image_present: torch.Tensor, image_unpadded_h: torch.Tensor, image_unpadded_w: torch.Tensor, image_placeholder_id: int, image_newline_id: int, variable_sized: bool, patch_size: Optional[dict[str, int]]=…) -> FuyuBatchFeature - added:
method FuyuImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class FuyuForCausalLM(FuyuPreTrainedModel, GenerationMixin) - added:
class FuyuModel(FuyuPreTrainedModel) - added:
class FuyuPreTrainedModel(PreTrainedModel) - added:
constructor FuyuForCausalLM(config: FuyuConfig) - added:
constructor FuyuModel(config: FuyuConfig) - added:
field FuyuPreTrainedModel.base_model_prefix - added:
field FuyuPreTrainedModel.config: FuyuConfig - added:
field FuyuPreTrainedModel.supports_gradient_checkpointing - added:
method FuyuForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, image_patches: Optional[torch.Tensor]=…, image_patches_indices: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, labels: Optional[torch.Tensor]=…, output… - added:
method FuyuForCausalLM.get_decoder() - added:
method FuyuForCausalLM.get_input_embeddings() - added:
method FuyuForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, image_patches=…, image_patches_indices=…, cache_position=…, **kwargs) - added:
method FuyuForCausalLM.set_decoder(decoder) - added:
method FuyuForCausalLM.set_input_embeddings(value) - added:
method FuyuModel.forward(input_ids: Optional[torch.LongTensor]=…, image_patches: Optional[torch.Tensor]=…, image_patches_indices: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hi… - added:
method FuyuModel.gather_continuous_embeddings(word_embeddings: torch.Tensor, continuous_embeddings: list[torch.Tensor], image_patch_input_indices: torch.Tensor) -> torch.Tensor - added:
method FuyuModel.get_decoder() - added:
method FuyuModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs) - added:
method FuyuModel.get_input_embeddings() - added:
method FuyuModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method FuyuModel.set_decoder(decoder) - added:
method FuyuModel.set_input_embeddings(value)
- added:
class FuyuProcessor(ProcessorMixin) - added:
constructor FuyuProcessor(image_processor, tokenizer, **kwargs) - added:
field FuyuProcessor.attributes - added:
field FuyuProcessor.image_processor_class - added:
field FuyuProcessor.tokenizer_class - added:
method FuyuProcessor.get_sample_encoding(prompts, scale_factors, image_unpadded_heights, image_unpadded_widths, image_placeholder_id, image_newline_id, tensor_batch_images) - added:
method FuyuProcessor.post_process_box_coordinates(outputs, target_sizes=…) - added:
method FuyuProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, **kwargs) - added:
property FuyuProcessor.model_input_names
- added:
export define_import_structure
- added:
class GemmaConfig(PretrainedConfig) - added:
constructor GemmaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
field GemmaConfig.base_model_pp_plan - added:
field GemmaConfig.base_model_tp_plan - added:
field GemmaConfig.keys_to_ignore_at_inference - added:
field GemmaConfig.model_type
- added:
class FlaxGemmaForCausalLM(FlaxGemmaPreTrainedModel) - added:
class FlaxGemmaModel(FlaxGemmaPreTrainedModel) - added:
class FlaxGemmaPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxGemmaPreTrainedModel(config: GemmaConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxGemmaForCausalLM.module_class - added:
field FlaxGemmaModel.module_class - added:
field FlaxGemmaPreTrainedModel.base_model_prefix - added:
field FlaxGemmaPreTrainedModel.config_class - added:
field FlaxGemmaPreTrainedModel.module_class: nn.Module - added:
method FlaxGemmaForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxGemmaForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxGemmaPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxGemmaPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class GemmaForCausalLM(GemmaPreTrainedModel, GenerationMixin) - added:
class GemmaForSequenceClassification(GenericForSequenceClassification, GemmaPreTrainedModel) - added:
class GemmaForTokenClassification(GenericForTokenClassification, GemmaPreTrainedModel) - added:
class GemmaModel(GemmaPreTrainedModel) - added:
class GemmaPreTrainedModel(PreTrainedModel) - added:
constructor GemmaForCausalLM(config) - added:
constructor GemmaModel(config: GemmaConfig) - added:
field GemmaPreTrainedModel.base_model_prefix - added:
field GemmaPreTrainedModel.config: GemmaConfig - added:
field GemmaPreTrainedModel.supports_gradient_checkpointing - added:
method GemmaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method GemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class GemmaConfig(PretrainedConfig) - added:
class GemmaForCausalLM(LlamaForCausalLM) - added:
class GemmaForSequenceClassification(LlamaForSequenceClassification) - added:
class GemmaForTokenClassification(LlamaForTokenClassification) - added:
class GemmaModel(LlamaModel) - added:
class GemmaPreTrainedModel(LlamaPreTrainedModel) - added:
class GemmaTokenizer(LlamaTokenizer, PreTrainedTokenizer) - added:
constructor GemmaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
constructor GemmaTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, pad_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, add_bos_token=…, add_eos_token=…, clean_up_tokenization_spaces=…, use_default_system_prompt=…, spaces_between_special_tokens=…, **kwargs) - added:
field GemmaConfig.base_model_pp_plan - added:
field GemmaConfig.base_model_tp_plan - added:
field GemmaConfig.keys_to_ignore_at_inference - added:
field GemmaConfig.model_type - added:
method GemmaForCausalLM.forward(**super_kwargs) - added:
method GemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method GemmaTokenizer.convert_tokens_to_string(tokens) - added:
method GemmaTokenizer.get_spm_processor() - added:
method GemmaTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
method GemmaTokenizer.unk_token_length()
- added:
class GemmaTokenizer(PreTrainedTokenizer) - added:
constructor GemmaTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, pad_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, add_bos_token=…, add_eos_token=…, clean_up_tokenization_spaces=…, use_default_system_prompt=…, spaces_between_special_tokens=…, **kwargs) - added:
field GemmaTokenizer.model_input_names - added:
field GemmaTokenizer.vocab_files_names - added:
method GemmaTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method GemmaTokenizer.convert_tokens_to_string(tokens) - added:
method GemmaTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method GemmaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method GemmaTokenizer.get_vocab() - added:
method GemmaTokenizer.save_vocabulary(save_directory, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method GemmaTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
property GemmaTokenizer.vocab_size
- added:
class GemmaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor GemmaTokenizerFast(vocab_file=…, tokenizer_file=…, clean_up_tokenization_spaces=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_bos_token=…, add_eos_token=…, **kwargs) - added:
field GemmaTokenizerFast.model_input_names - added:
field GemmaTokenizerFast.padding_side - added:
field GemmaTokenizerFast.slow_tokenizer_class - added:
field GemmaTokenizerFast.vocab_files_names - added:
method GemmaTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method GemmaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method GemmaTokenizerFast.update_post_processor() - added:
property GemmaTokenizerFast.add_bos_token - added:
property GemmaTokenizerFast.add_eos_token
- added:
export define_import_structure
- added:
class Gemma2Config(PretrainedConfig) - added:
constructor Gemma2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_scalar=…… - added:
field Gemma2Config.base_model_pp_plan - added:
field Gemma2Config.base_model_tp_plan - added:
field Gemma2Config.keys_to_ignore_at_inference - added:
field Gemma2Config.model_type
- added:
class Gemma2ForCausalLM(Gemma2PreTrainedModel, GenerationMixin) - added:
class Gemma2ForSequenceClassification(GenericForSequenceClassification, Gemma2PreTrainedModel) - added:
class Gemma2ForTokenClassification(GenericForTokenClassification, Gemma2PreTrainedModel) - added:
class Gemma2Model(Gemma2PreTrainedModel) - added:
class Gemma2PreTrainedModel(PreTrainedModel) - added:
constructor Gemma2ForCausalLM(config) - added:
constructor Gemma2Model(config: Gemma2Config) - added:
field Gemma2PreTrainedModel.base_model_prefix - added:
field Gemma2PreTrainedModel.config: Gemma2Config - added:
field Gemma2PreTrainedModel.supports_gradient_checkpointing - added:
method Gemma2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_positio… - added:
method Gemma2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: U…
- added:
class Gemma2Config(PretrainedConfig) - added:
class Gemma2ForCausalLM(GemmaForCausalLM) - added:
class Gemma2ForSequenceClassification(GemmaForSequenceClassification) - added:
class Gemma2ForTokenClassification(GemmaForTokenClassification) - added:
class Gemma2Model(GemmaModel) - added:
class Gemma2PreTrainedModel(GemmaPreTrainedModel) - added:
constructor Gemma2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_scalar=…… - added:
constructor Gemma2ForCausalLM(config) - added:
constructor Gemma2Model(config: Gemma2Config) - added:
field Gemma2Config.base_model_pp_plan - added:
field Gemma2Config.base_model_tp_plan - added:
field Gemma2Config.keys_to_ignore_at_inference - added:
field Gemma2Config.model_type - added:
method Gemma2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_positio… - added:
method Gemma2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: U…
- added:
export define_import_structure
- added:
class Gemma3Config(PretrainedConfig) - added:
class Gemma3TextConfig(PretrainedConfig) - added:
constructor Gemma3Config(text_config: Optional[Union[Gemma3TextConfig, dict[str, Any]]]=…, vision_config: Optional[Union[SiglipVisionConfig, dict[str, Any]]]=…, mm_tokens_per_image: int=…, boi_token_index: int=…, eoi_token_index: int=…, image_token_index: int=…, initializer_range: float=…, **kwargs) - added:
constructor Gemma3TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_scal… - added:
field Gemma3Config.attribute_map - added:
field Gemma3Config.model_type - added:
field Gemma3Config.sub_configs - added:
field Gemma3TextConfig.base_model_pp_plan - added:
field Gemma3TextConfig.base_model_tp_plan - added:
field Gemma3TextConfig.keys_to_ignore_at_inference - added:
field Gemma3TextConfig.model_type
- added:
class Gemma3ImageProcessor(BaseImageProcessor) - added:
constructor Gemma3ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, do_pan_and_scan: Optional[bool]=…, pan_and_scan_min_crop_size: Opti… - added:
field Gemma3ImageProcessor.model_input_names - added:
method Gemma3ImageProcessor.pan_and_scan(image: np.ndarray, pan_and_scan_min_crop_size: int, pan_and_scan_max_num_crops: int, pan_and_scan_min_ratio_to_activate: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method Gemma3ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, T…
- added:
class Gemma3ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Gemma3ImageProcessorFast(**kwargs: Unpack[Gemma3FastImageProcessorKwargs]) - added:
field Gemma3ImageProcessorFast.default_to_square - added:
field Gemma3ImageProcessorFast.do_convert_rgb - added:
field Gemma3ImageProcessorFast.do_normalize - added:
field Gemma3ImageProcessorFast.do_pan_and_scan - added:
field Gemma3ImageProcessorFast.do_rescale - added:
field Gemma3ImageProcessorFast.do_resize - added:
field Gemma3ImageProcessorFast.image_mean - added:
field Gemma3ImageProcessorFast.image_std - added:
field Gemma3ImageProcessorFast.pan_and_scan_max_num_crops - added:
field Gemma3ImageProcessorFast.pan_and_scan_min_crop_size - added:
field Gemma3ImageProcessorFast.pan_and_scan_min_ratio_to_activate - added:
field Gemma3ImageProcessorFast.resample - added:
field Gemma3ImageProcessorFast.size - added:
field Gemma3ImageProcessorFast.valid_kwargs - added:
method Gemma3ImageProcessorFast.pan_and_scan_batched(images: torch.Tensor, pan_and_scan_min_crop_size: int, pan_and_scan_max_num_crops: int, pan_and_scan_min_ratio_to_activate: float) - added:
method Gemma3ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Gemma3FastImageProcessorKwargs]) -> BatchFeature
- added:
class Gemma3ForCausalLM(Gemma3PreTrainedModel, GenerationMixin) - added:
class Gemma3ForConditionalGeneration(Gemma3PreTrainedModel, GenerationMixin) - added:
class Gemma3ForSequenceClassification(Gemma3PreTrainedModel) - added:
class Gemma3Model(Gemma3PreTrainedModel) - added:
class Gemma3PreTrainedModel(PreTrainedModel) - added:
class Gemma3TextForSequenceClassification(GenericForSequenceClassification, Gemma3PreTrainedModel) - added:
class Gemma3TextModel(Gemma3PreTrainedModel) - added:
constructor Gemma3ForCausalLM(config: Gemma3TextConfig) - added:
constructor Gemma3ForConditionalGeneration(config: Gemma3Config) - added:
constructor Gemma3ForSequenceClassification(config) - added:
constructor Gemma3Model(config: Gemma3Config) - added:
constructor Gemma3TextModel(config: Gemma3TextConfig) - added:
field Gemma3ForCausalLM.base_model_prefix - added:
field Gemma3ForCausalLM.config: Gemma3TextConfig - added:
field Gemma3ForConditionalGeneration.accepts_loss_kwargs - added:
field Gemma3Model.accepts_loss_kwargs - added:
field Gemma3PreTrainedModel.base_model_prefix - added:
field Gemma3PreTrainedModel.config: Gemma3Config - added:
field Gemma3PreTrainedModel.supports_gradient_checkpointing - added:
field Gemma3TextForSequenceClassification.config: Gemma3TextConfig - added:
field Gemma3TextModel.config: Gemma3TextConfig - added:
method Gemma3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_positio… - added:
method Gemma3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method Gemma3ForConditionalGeneration.get_decoder() - added:
method Gemma3ForConditionalGeneration.get_image_features(pixel_values) - added:
method Gemma3ForConditionalGeneration.get_input_embeddings() - added:
method Gemma3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, cache_position=…, position_ids=…, pixel_values=…, attention_mask=…, token_type_ids=…, use_cache=…, logits_to_keep=…, labels=…, **kwargs) - added:
method Gemma3ForConditionalGeneration.set_decoder(decoder) - added:
method Gemma3ForConditionalGeneration.set_input_embeddings(value) - added:
method Gemma3ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Opt… - added:
method Gemma3ForSequenceClassification.get_input_embeddings() - added:
method Gemma3ForSequenceClassification.set_input_embeddings(value) - added:
method Gemma3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.Lon… - added:
method Gemma3Model.get_decoder() - added:
method Gemma3Model.get_image_features(pixel_values: torch.Tensor) -> torch.Tensor - added:
method Gemma3Model.get_input_embeddings() - added:
method Gemma3Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Gemma3Model.set_decoder(decoder) - added:
method Gemma3Model.set_input_embeddings(value) - added:
method Gemma3TextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwarg… - added:
property Gemma3ForConditionalGeneration.language_model - added:
property Gemma3ForConditionalGeneration.multi_modal_projector - added:
property Gemma3ForConditionalGeneration.vision_tower - added:
static method Gemma3ForConditionalGeneration.create_masks_for_generate(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor], token_type_ids: Optional[torch.Tensor]=…, **kwargs) -> dict
- added:
class Gemma3Config(PretrainedConfig) - added:
class Gemma3ForCausalLM(Gemma2ForCausalLM) - added:
class Gemma3ForConditionalGeneration(PaliGemmaForConditionalGeneration) - added:
class Gemma3ForSequenceClassification(Gemma3PreTrainedModel) - added:
class Gemma3Model(PaliGemmaModel) - added:
class Gemma3PreTrainedModel(Gemma2PreTrainedModel) - added:
class Gemma3TextConfig(Gemma2Config, PretrainedConfig) - added:
class Gemma3TextForSequenceClassification(GenericForSequenceClassification, Gemma3PreTrainedModel) - added:
class Gemma3TextModel(Gemma2Model) - added:
constructor Gemma3Config(text_config: Optional[Union[Gemma3TextConfig, dict[str, Any]]]=…, vision_config: Optional[Union[SiglipVisionConfig, dict[str, Any]]]=…, mm_tokens_per_image: int=…, boi_token_index: int=…, eoi_token_index: int=…, image_token_index: int=…, initializer_range: float=…, **kwargs) - added:
constructor Gemma3ForCausalLM(config: Gemma3TextConfig) - added:
constructor Gemma3ForSequenceClassification(config) - added:
constructor Gemma3Model(config: Gemma3Config) - added:
constructor Gemma3TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_scal… - added:
constructor Gemma3TextModel(config: Gemma3TextConfig) - added:
field Gemma3Config.attribute_map - added:
field Gemma3Config.model_type - added:
field Gemma3Config.sub_configs - added:
field Gemma3ForCausalLM.base_model_prefix - added:
field Gemma3ForCausalLM.config: Gemma3TextConfig - added:
field Gemma3ForConditionalGeneration.accepts_loss_kwargs - added:
field Gemma3Model.accepts_loss_kwargs - added:
field Gemma3PreTrainedModel.base_model_prefix - added:
field Gemma3TextConfig.model_type - added:
field Gemma3TextForSequenceClassification.config: Gemma3TextConfig - added:
field Gemma3TextModel.config: Gemma3TextConfig - added:
method Gemma3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method Gemma3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, cache_position=…, position_ids=…, pixel_values=…, attention_mask=…, token_type_ids=…, use_cache=…, logits_to_keep=…, labels=…, **kwargs) - added:
method Gemma3ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Opt… - added:
method Gemma3ForSequenceClassification.get_input_embeddings() - added:
method Gemma3ForSequenceClassification.set_input_embeddings(value) - added:
method Gemma3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.Lon… - added:
method Gemma3Model.get_image_features(pixel_values: torch.Tensor) -> torch.Tensor - added:
method Gemma3TextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwarg… - added:
static method Gemma3ForConditionalGeneration.create_masks_for_generate(config: PretrainedConfig, input_embeds: torch.Tensor, attention_mask: Optional[torch.Tensor], cache_position: torch.Tensor, past_key_values: Optional[Cache], position_ids: Optional[torch.Tensor], token_type_ids: Optional[torch.Tensor]=…, **kwargs) -> dict
- added:
class Gemma3Processor(ProcessorMixin) - added:
constructor Gemma3Processor(image_processor, tokenizer, chat_template=…, image_seq_length: int=…, **kwargs) - added:
field Gemma3Processor.attributes - added:
field Gemma3Processor.image_processor_class - added:
field Gemma3Processor.tokenizer_class - added:
property Gemma3Processor.model_input_names
- added:
export define_import_structure
- added:
class Gemma3nAudioConfig(PretrainedConfig) - added:
class Gemma3nConfig(PretrainedConfig) - added:
class Gemma3nTextConfig(PretrainedConfig) - added:
class Gemma3nVisionConfig(PretrainedConfig) - added:
constructor Gemma3nAudioConfig(vocab_size: int=…, vocab_offset: int=…, input_feat_size: int=…, hidden_size: int=…, rms_norm_eps: float=…, gradient_clipping: float=…, conf_attention_chunk_size: int=…, conf_attention_context_left: int=…, conf_attention_context_right: int=…, conf_attention_logit_cap: float=…, conf_num_attention_heads: int=…, conf_num_hidden_layers: int=…, conf_conv_kernel_size: int=… - added:
constructor Gemma3nConfig(text_config: Optional[Union[Gemma3nTextConfig, dict[str, Any]]]=…, vision_config: Optional[Union[Gemma3nVisionConfig, dict[str, Any]]]=…, audio_config: Optional[Union[Gemma3nAudioConfig, dict[str, Any]]]=…, audio_soft_tokens_per_image: int=…, vision_soft_tokens_per_image: int=…, boi_token_id: int=…, eoi_token_id: int=…, image_token_id: int=…, boa_token_id: int=…, eoa_tok… - added:
constructor Gemma3nTextConfig(vocab_size: int=…, vocab_size_per_layer_input: int=…, hidden_size: int=…, hidden_size_per_layer_input: int=…, intermediate_size: Union[int, Sequence[int]]=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, head_dim: int=…, hidden_activation: str=…, max_position_embeddings: int=…, initializer_range: float=…, rms_norm_eps: float=…, use… - added:
constructor Gemma3nVisionConfig(initializer_range: float=…, do_pooling: bool=…, architecture: str=…, hidden_size: int=…, vocab_size: int=…, vocab_offset: int=…, rms_norm_eps: float=…, model_args: Optional[dict]=…, **kwargs) - added:
field Gemma3nAudioConfig.model_type - added:
field Gemma3nConfig.model_type - added:
field Gemma3nConfig.sub_configs - added:
field Gemma3nTextConfig.base_model_pp_plan - added:
field Gemma3nTextConfig.base_model_tp_plan - added:
field Gemma3nTextConfig.keys_to_ignore_at_inference - added:
field Gemma3nTextConfig.model_type - added:
field Gemma3nVisionConfig.model_type - added:
method Gemma3nVisionConfig.to_dict() -> dict[str, Any] - added:
static method Gemma3nVisionConfig.from_dict(config_dict: dict[str, Any], **kwargs)
- added:
class Gemma3nAudioFeatureExtractor(SequenceFeatureExtractor) - added:
constructor Gemma3nAudioFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, return_attention_mask: bool=…, frame_length_ms: float=…, hop_length_ms: float=…, min_frequency: float=…, max_frequency: float=…, preemphasis: float=…, preemphasis_htk_flavor: bool=…, fft_overdrive: bool=…, dither: float=…, input_scale_factor: float=…, mel_floor: float=…, per_bin_mean: Optio… - added:
field Gemma3nAudioFeatureExtractor.model_input_names
- added:
class Gemma3nAudioEncoder(PreTrainedModel) - added:
class Gemma3nForCausalLM(Gemma3nPreTrainedModel, GenerationMixin) - added:
class Gemma3nForConditionalGeneration(Gemma3nPreTrainedModel, GenerationMixin) - added:
class Gemma3nModel(Gemma3nPreTrainedModel) - added:
class Gemma3nPreTrainedModel(PreTrainedModel) - added:
class Gemma3nTextModel(Gemma3nPreTrainedModel) - added:
constructor Gemma3nAudioEncoder(config: Gemma3nAudioConfig) - added:
constructor Gemma3nForCausalLM(config: Gemma3nTextConfig) - added:
constructor Gemma3nForConditionalGeneration(config: Gemma3nConfig) - added:
constructor Gemma3nModel(config: Gemma3nConfig) - added:
constructor Gemma3nTextModel(config: Gemma3nTextConfig) - added:
field Gemma3nAudioEncoder.config: Gemma3nAudioConfig - added:
field Gemma3nAudioEncoder.main_input_name - added:
field Gemma3nForCausalLM.base_model_prefix - added:
field Gemma3nForCausalLM.config: Gemma3nTextConfig - added:
field Gemma3nForConditionalGeneration.base_model_prefix - added:
field Gemma3nModel.accepts_loss_kwargs - added:
field Gemma3nPreTrainedModel.base_model_prefix - added:
field Gemma3nPreTrainedModel.config: Gemma3nConfig - added:
field Gemma3nPreTrainedModel.supports_gradient_checkpointing - added:
field Gemma3nTextModel.config: Gemma3nTextConfig - added:
method Gemma3nAudioEncoder.forward(audio_mel: torch.Tensor, audio_mel_mask: torch.BoolTensor) -> tuple[torch.Tensor, torch.BoolTensor] - added:
method Gemma3nForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_positi… - added:
method Gemma3nForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, input_features_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cach… - added:
method Gemma3nForConditionalGeneration.get_decoder() - added:
method Gemma3nForConditionalGeneration.get_image_features(pixel_values) - added:
method Gemma3nForConditionalGeneration.get_input_embeddings() - added:
method Gemma3nForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, cache_position=…, position_ids=…, pixel_values=…, input_features=…, attention_mask=…, input_features_mask=…, token_type_ids=…, use_cache=…, logits_to_keep=…, labels=…, **kwargs) - added:
method Gemma3nForConditionalGeneration.set_decoder(decoder) - added:
method Gemma3nForConditionalGeneration.set_input_embeddings(value) - added:
method Gemma3nModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, input_features_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optiona… - added:
method Gemma3nModel.get_audio_features(input_features: torch.Tensor, input_features_mask: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor] - added:
method Gemma3nModel.get_decoder() - added:
method Gemma3nModel.get_image_features(pixel_values: torch.Tensor) -> torch.Tensor - added:
method Gemma3nModel.get_input_embeddings() - added:
method Gemma3nModel.get_placeholder_mask(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_features: Optional[torch.FloatTensor]=…, audio_features: Optional[torch.FloatTensor]=…) - added:
method Gemma3nModel.set_decoder(decoder) - added:
method Gemma3nModel.set_input_embeddings(value) - added:
method Gemma3nTextModel.forward(input_ids: Optional[torch.LongTensor]=…, per_layer_inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_po… - added:
method Gemma3nTextModel.get_per_layer_inputs(input_ids: torch.LongTensor) -> torch.Tensor - added:
method Gemma3nTextModel.project_per_layer_inputs(inputs_embeds: torch.Tensor, per_layer_inputs: Optional[torch.Tensor]=…) -> torch.Tensor - added:
property Gemma3nForConditionalGeneration.audio_tower - added:
property Gemma3nForConditionalGeneration.language_model - added:
property Gemma3nForConditionalGeneration.multi_modal_projector - added:
property Gemma3nForConditionalGeneration.vision_tower
- added:
class Gemma3nAudioConfig(PretrainedConfig) - added:
class Gemma3nAudioEncoder(PreTrainedModel) - added:
class Gemma3nConfig(PretrainedConfig) - added:
class Gemma3nForCausalLM(Gemma3ForCausalLM) - added:
class Gemma3nForConditionalGeneration(PaliGemmaForConditionalGeneration) - added:
class Gemma3nModel(PaliGemmaModel) - added:
class Gemma3nPreTrainedModel(Gemma2PreTrainedModel) - added:
class Gemma3nTextConfig(Gemma2Config, PretrainedConfig) - added:
class Gemma3nTextModel(Gemma3TextModel) - added:
class Gemma3nVisionConfig(TimmWrapperConfig) - added:
constructor Gemma3nAudioConfig(vocab_size: int=…, vocab_offset: int=…, input_feat_size: int=…, hidden_size: int=…, rms_norm_eps: float=…, gradient_clipping: float=…, conf_attention_chunk_size: int=…, conf_attention_context_left: int=…, conf_attention_context_right: int=…, conf_attention_logit_cap: float=…, conf_num_attention_heads: int=…, conf_num_hidden_layers: int=…, conf_conv_kernel_size: int=… - added:
constructor Gemma3nAudioEncoder(config: Gemma3nAudioConfig) - added:
constructor Gemma3nConfig(text_config: Optional[Union[Gemma3nTextConfig, dict[str, Any]]]=…, vision_config: Optional[Union[Gemma3nVisionConfig, dict[str, Any]]]=…, audio_config: Optional[Union[Gemma3nAudioConfig, dict[str, Any]]]=…, audio_soft_tokens_per_image: int=…, vision_soft_tokens_per_image: int=…, boi_token_id: int=…, eoi_token_id: int=…, image_token_id: int=…, boa_token_id: int=…, eoa_tok… - added:
constructor Gemma3nModel(config: Gemma3nConfig) - added:
constructor Gemma3nTextConfig(vocab_size: int=…, vocab_size_per_layer_input: int=…, hidden_size: int=…, hidden_size_per_layer_input: int=…, intermediate_size: Union[int, Sequence[int]]=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, head_dim: int=…, hidden_activation: str=…, max_position_embeddings: int=…, initializer_range: float=…, rms_norm_eps: float=…, use… - added:
constructor Gemma3nTextModel(config: Gemma3nTextConfig) - added:
constructor Gemma3nVisionConfig(initializer_range: float=…, do_pooling: bool=…, architecture: str=…, hidden_size: int=…, vocab_size: int=…, vocab_offset: int=…, rms_norm_eps: float=…, model_args: Optional[dict]=…, **kwargs) - added:
field Gemma3nAudioConfig.model_type - added:
field Gemma3nAudioEncoder.config: Gemma3nAudioConfig - added:
field Gemma3nAudioEncoder.main_input_name - added:
field Gemma3nConfig.model_type - added:
field Gemma3nConfig.sub_configs - added:
field Gemma3nForCausalLM.base_model_prefix - added:
field Gemma3nForConditionalGeneration.base_model_prefix - added:
field Gemma3nPreTrainedModel.base_model_prefix - added:
field Gemma3nPreTrainedModel.config: Gemma3nConfig - added:
field Gemma3nTextConfig.model_type - added:
field Gemma3nTextModel.config: Gemma3nTextConfig - added:
field Gemma3nVisionConfig.model_type - added:
method Gemma3nAudioEncoder.forward(audio_mel: torch.Tensor, audio_mel_mask: torch.BoolTensor) -> tuple[torch.Tensor, torch.BoolTensor] - added:
method Gemma3nForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, input_features_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cach… - added:
method Gemma3nForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, cache_position=…, position_ids=…, pixel_values=…, input_features=…, attention_mask=…, input_features_mask=…, token_type_ids=…, use_cache=…, logits_to_keep=…, labels=…, **kwargs) - added:
method Gemma3nModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, input_features_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optiona… - added:
method Gemma3nModel.get_audio_features(input_features: torch.Tensor, input_features_mask: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor] - added:
method Gemma3nModel.get_image_features(pixel_values: torch.Tensor) -> torch.Tensor - added:
method Gemma3nModel.get_placeholder_mask(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_features: Optional[torch.FloatTensor]=…, audio_features: Optional[torch.FloatTensor]=…) - added:
method Gemma3nTextModel.forward(input_ids: Optional[torch.LongTensor]=…, per_layer_inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_po… - added:
method Gemma3nTextModel.get_per_layer_inputs(input_ids: torch.LongTensor) -> torch.Tensor - added:
method Gemma3nTextModel.project_per_layer_inputs(inputs_embeds: torch.Tensor, per_layer_inputs: Optional[torch.Tensor]=…) -> torch.Tensor - added:
property Gemma3nForConditionalGeneration.audio_tower - added:
property Gemma3nForConditionalGeneration.multi_modal_projector
- added:
class Gemma3nProcessor(ProcessorMixin) - added:
constructor Gemma3nProcessor(feature_extractor, image_processor, tokenizer, chat_template=…, audio_seq_length: int=…, image_seq_length: int=…, **kwargs) - added:
field Gemma3nProcessor.attributes - added:
field Gemma3nProcessor.feature_extractor_class - added:
field Gemma3nProcessor.image_processor_class - added:
field Gemma3nProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class GitConfig(PretrainedConfig) - added:
class GitVisionConfig(PretrainedConfig) - added:
constructor GitConfig(vision_config=…, vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, tie_word_embeddings=…, bos_token_id=…, eos_token_id=…, num_image_with_embed… - added:
constructor GitVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, **kwargs) - added:
field GitConfig.model_type - added:
field GitConfig.sub_configs - added:
field GitVisionConfig.base_config_key - added:
field GitVisionConfig.model_type
- added:
class GitForCausalLM(GitPreTrainedModel, GenerationMixin) - added:
class GitModel(GitPreTrainedModel) - added:
class GitPreTrainedModel(PreTrainedModel) - added:
class GitVisionModel(GitPreTrainedModel) - added:
constructor GitForCausalLM(config) - added:
constructor GitModel(config) - added:
constructor GitVisionModel(config: GitVisionConfig) - added:
field GitPreTrainedModel.base_model_prefix - added:
field GitPreTrainedModel.config: GitConfig - added:
field GitPreTrainedModel.supports_gradient_checkpointing - added:
field GitVisionModel.config: GitVisionConfig - added:
field GitVisionModel.main_input_name - added:
method GitForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, list[torch.Tensor]]]=…, use_cache: Optional[bool]=…, output… - added:
method GitForCausalLM.get_output_embeddings() - added:
method GitForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, use_cache=…, **kwargs) - added:
method GitForCausalLM.set_output_embeddings(new_embeddings) - added:
method GitModel.create_attention_mask(tgt, memory, tgt_mask, past_key_values_length, memory_key_padding_mask=…) - added:
method GitModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, outp… - added:
method GitModel.get_input_embeddings() - added:
method GitModel.set_input_embeddings(value) - added:
method GitVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method GitVisionModel.get_input_embeddings() -> nn.Module
- added:
class GitProcessor(ProcessorMixin) - added:
constructor GitProcessor(image_processor, tokenizer) - added:
field GitProcessor.attributes - added:
field GitProcessor.image_processor_class - added:
field GitProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class GlmConfig(PretrainedConfig) - added:
constructor GlmConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, partial_rotary_factor=…, head_dim=…, hidden_act=…, attention_dropout=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, attention_bias=…, **kwargs) - added:
field GlmConfig.base_model_pp_plan - added:
field GlmConfig.base_model_tp_plan - added:
field GlmConfig.keys_to_ignore_at_inference - added:
field GlmConfig.model_type
- added:
class GlmForCausalLM(GlmPreTrainedModel, GenerationMixin) - added:
class GlmForSequenceClassification(GenericForSequenceClassification, GlmPreTrainedModel) - added:
class GlmForTokenClassification(GenericForTokenClassification, GlmPreTrainedModel) - added:
class GlmModel(GlmPreTrainedModel) - added:
class GlmPreTrainedModel(PreTrainedModel) - added:
constructor GlmForCausalLM(config) - added:
constructor GlmModel(config: GlmConfig) - added:
field GlmPreTrainedModel.base_model_prefix - added:
field GlmPreTrainedModel.config: GlmConfig - added:
field GlmPreTrainedModel.supports_gradient_checkpointing - added:
method GlmForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **k… - added:
method GlmModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class GlmForCausalLM(LlamaForCausalLM) - added:
class GlmForSequenceClassification(LlamaForSequenceClassification) - added:
class GlmForTokenClassification(LlamaForTokenClassification)
- added:
export define_import_structure
- added:
class Glm4Config(PretrainedConfig) - added:
constructor Glm4Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, partial_rotary_factor=…, head_dim=…, hidden_act=…, attention_dropout=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, attention_bias=…, **kwarg… - added:
field Glm4Config.base_model_pp_plan - added:
field Glm4Config.base_model_tp_plan - added:
field Glm4Config.keys_to_ignore_at_inference - added:
field Glm4Config.model_type
- added:
class Glm4ForCausalLM(Glm4PreTrainedModel, GenerationMixin) - added:
class Glm4ForSequenceClassification(GenericForSequenceClassification, Glm4PreTrainedModel) - added:
class Glm4ForTokenClassification(GenericForTokenClassification, Glm4PreTrainedModel) - added:
class Glm4Model(Glm4PreTrainedModel) - added:
class Glm4PreTrainedModel(PreTrainedModel) - added:
constructor Glm4ForCausalLM(config) - added:
constructor Glm4Model(config: Glm4Config) - added:
field Glm4PreTrainedModel.base_model_prefix - added:
field Glm4PreTrainedModel.config: Glm4Config - added:
field Glm4PreTrainedModel.supports_gradient_checkpointing - added:
method Glm4ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **… - added:
method Glm4Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Glm4ForCausalLM(GlmForCausalLM) - added:
class Glm4ForSequenceClassification(GlmForSequenceClassification) - added:
class Glm4ForTokenClassification(GlmForTokenClassification) - added:
method Glm4ForCausalLM.forward(**super_kwargs: Unpack[TransformersKwargs]) -> Union[tuple, CausalLMOutputWithPast]
- added:
export define_import_structure
- added:
class Glm4MoeConfig(PretrainedConfig) - added:
constructor Glm4MoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, partial_rotary_factor=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, moe_intermediate_size=…, num_experts_per_tok=…,… - added:
field Glm4MoeConfig.base_model_pp_plan - added:
field Glm4MoeConfig.base_model_tp_plan - added:
field Glm4MoeConfig.keys_to_ignore_at_inference - added:
field Glm4MoeConfig.model_type
- added:
class Glm4MoeForCausalLM(Glm4MoePreTrainedModel, GenerationMixin) - added:
class Glm4MoeModel(Glm4MoePreTrainedModel) - added:
class Glm4MoePreTrainedModel(PreTrainedModel) - added:
constructor Glm4MoeForCausalLM(config) - added:
constructor Glm4MoeModel(config: Glm4MoeConfig) - added:
field Glm4MoePreTrainedModel.base_model_prefix - added:
field Glm4MoePreTrainedModel.config: Glm4MoeConfig - added:
field Glm4MoePreTrainedModel.supports_gradient_checkpointing - added:
method Glm4MoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method Glm4MoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Glm4MoeConfig(PretrainedConfig) - added:
class Glm4MoeForCausalLM(DeepseekV3ForCausalLM) - added:
class Glm4MoeModel(DeepseekV3Model) - added:
class Glm4MoePreTrainedModel(DeepseekV3PreTrainedModel) - added:
constructor Glm4MoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, partial_rotary_factor=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, moe_intermediate_size=…, num_experts_per_tok=…,… - added:
field Glm4MoeConfig.base_model_pp_plan - added:
field Glm4MoeConfig.base_model_tp_plan - added:
field Glm4MoeConfig.keys_to_ignore_at_inference - added:
field Glm4MoeConfig.model_type
- added:
export define_import_structure
- added:
class Glm4vConfig(PretrainedConfig) - added:
class Glm4vTextConfig(PretrainedConfig) - added:
constructor Glm4vConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, image_start_token_id=…, image_end_token_id=…, video_start_token_id=…, video_end_token_id=…, **kwargs) - added:
constructor Glm4vTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, attention_dropout=…, rope_scaling=…, image_token_id=…, video_token_id=…, **kwargs) - added:
field Glm4vConfig.keys_to_ignore_at_inference - added:
field Glm4vConfig.model_type - added:
field Glm4vConfig.sub_configs - added:
field Glm4vTextConfig.base_config_key - added:
field Glm4vTextConfig.base_model_pp_plan - added:
field Glm4vTextConfig.base_model_tp_plan - added:
field Glm4vTextConfig.keys_to_ignore_at_inference - added:
field Glm4vTextConfig.model_type
- added:
class Glm4vImageProcessor(BaseImageProcessor) - added:
constructor Glm4vImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, patch_size: int=…, temporal_patch_size: int=…, merge_size: int=…, **kwargs) ->… - added:
field Glm4vImageProcessor.model_input_names - added:
method Glm4vImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Glm4vImageProcessor.preprocess(images: ImageInput, videos: Optional[VideoInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, patch_…
- added:
class Glm4vImageProcessorFast(BaseImageProcessorFast) - added:
constructor Glm4vImageProcessorFast(**kwargs: Unpack[Glm4vFastImageProcessorKwargs]) - added:
field Glm4vImageProcessorFast.do_convert_rgb - added:
field Glm4vImageProcessorFast.do_normalize - added:
field Glm4vImageProcessorFast.do_rescale - added:
field Glm4vImageProcessorFast.do_resize - added:
field Glm4vImageProcessorFast.image_mean - added:
field Glm4vImageProcessorFast.image_std - added:
field Glm4vImageProcessorFast.merge_size - added:
field Glm4vImageProcessorFast.model_input_names - added:
field Glm4vImageProcessorFast.patch_size - added:
field Glm4vImageProcessorFast.resample - added:
field Glm4vImageProcessorFast.size - added:
field Glm4vImageProcessorFast.temporal_patch_size - added:
field Glm4vImageProcessorFast.valid_kwargs - added:
method Glm4vImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Glm4vFastImageProcessorKwargs]) -> BatchFeature
- added:
class Glm4vForConditionalGeneration(Glm4vPreTrainedModel, GenerationMixin) - added:
class Glm4vModel(Glm4vPreTrainedModel) - added:
class Glm4vPreTrainedModel(PreTrainedModel) - added:
class Glm4vTextModel(Glm4vPreTrainedModel) - added:
constructor Glm4vForConditionalGeneration(config) - added:
constructor Glm4vModel(config) - added:
constructor Glm4vTextModel(config: Glm4vTextConfig) - added:
field Glm4vForConditionalGeneration.accepts_loss_kwargs - added:
field Glm4vModel.accepts_loss_kwargs - added:
field Glm4vModel.base_model_prefix - added:
field Glm4vModel.config: Glm4vConfig - added:
field Glm4vPreTrainedModel.base_model_prefix - added:
field Glm4vPreTrainedModel.config: Glm4vConfig - added:
field Glm4vPreTrainedModel.supports_gradient_checkpointing - added:
field Glm4vTextModel.config: Glm4vTextConfig - added:
method Glm4vForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw:… - added:
method Glm4vForConditionalGeneration.get_decoder() - added:
method Glm4vForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vForConditionalGeneration.get_input_embeddings() - added:
method Glm4vForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Glm4vForConditionalGeneration.set_decoder(decoder) - added:
method Glm4vForConditionalGeneration.set_input_embeddings(value) - added:
method Glm4vModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[t… - added:
method Glm4vModel.get_decoder() - added:
method Glm4vModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vModel.get_input_embeddings() - added:
method Glm4vModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Glm4vModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Glm4vModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vModel.set_decoder(decoder) - added:
method Glm4vModel.set_input_embeddings(value) - added:
method Glm4vTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[FlashAttentionKwargs]) -> Union[tuple, BaseModelOutputWithPast] - added:
property Glm4vForConditionalGeneration.language_model - added:
property Glm4vForConditionalGeneration.visual
- added:
class Glm4vConfig(PretrainedConfig) - added:
class Glm4vForConditionalGeneration(Qwen2_5_VLForConditionalGeneration) - added:
class Glm4vModel(Qwen2_5_VLModel) - added:
class Glm4vPreTrainedModel(Qwen2_5_VLPreTrainedModel) - added:
class Glm4vProcessor(Qwen2_VLProcessor) - added:
class Glm4vTextConfig(PretrainedConfig) - added:
class Glm4vTextModel(Qwen2_5_VLTextModel) - added:
constructor Glm4vConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, image_start_token_id=…, image_end_token_id=…, video_start_token_id=…, video_end_token_id=…, **kwargs) - added:
constructor Glm4vModel(config) - added:
constructor Glm4vProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
constructor Glm4vTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, attention_dropout=…, rope_scaling=…, image_token_id=…, video_token_id=…, **kwargs) - added:
constructor Glm4vTextModel(config: Glm4vTextConfig) - added:
field Glm4vConfig.keys_to_ignore_at_inference - added:
field Glm4vConfig.model_type - added:
field Glm4vConfig.sub_configs - added:
field Glm4vProcessor.tokenizer_class - added:
field Glm4vTextConfig.base_config_key - added:
field Glm4vTextConfig.base_model_pp_plan - added:
field Glm4vTextConfig.base_model_tp_plan - added:
field Glm4vTextConfig.keys_to_ignore_at_inference - added:
field Glm4vTextConfig.model_type - added:
method Glm4vForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw:… - added:
method Glm4vForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Glm4vModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[t… - added:
method Glm4vModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Glm4vModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Glm4vModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[FlashAttentionKwargs]) -> Union[tuple, BaseModelOutputWithPast]
- added:
class Glm4vProcessor(ProcessorMixin) - added:
constructor Glm4vProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
field Glm4vProcessor.attributes - added:
field Glm4vProcessor.image_processor_class - added:
field Glm4vProcessor.tokenizer_class - added:
field Glm4vProcessor.video_processor_class - added:
method Glm4vProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs)
- added:
class Glm4vVideoProcessor(BaseVideoProcessor) - added:
constructor Glm4vVideoProcessor(**kwargs: Unpack[Glm4vVideoProcessorInitKwargs]) - added:
field Glm4vVideoProcessor.do_convert_rgb - added:
field Glm4vVideoProcessor.do_normalize - added:
field Glm4vVideoProcessor.do_rescale - added:
field Glm4vVideoProcessor.do_resize - added:
field Glm4vVideoProcessor.do_sample_frames - added:
field Glm4vVideoProcessor.fps - added:
field Glm4vVideoProcessor.image_mean - added:
field Glm4vVideoProcessor.image_std - added:
field Glm4vVideoProcessor.max_duration - added:
field Glm4vVideoProcessor.max_image_size - added:
field Glm4vVideoProcessor.merge_size - added:
field Glm4vVideoProcessor.model_input_names - added:
field Glm4vVideoProcessor.num_frames - added:
field Glm4vVideoProcessor.patch_size - added:
field Glm4vVideoProcessor.resample - added:
field Glm4vVideoProcessor.size - added:
field Glm4vVideoProcessor.temporal_patch_size - added:
field Glm4vVideoProcessor.valid_kwargs - added:
method Glm4vVideoProcessor.sample_frames(metadata: VideoMetadata, fps: Optional[Union[int, float]]=…, **kwargs)
- added:
export define_import_structure
- added:
class Glm4vMoeConfig(PretrainedConfig) - added:
class Glm4vMoeTextConfig(PretrainedConfig) - added:
constructor Glm4vMoeConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, image_start_token_id=…, image_end_token_id=…, video_start_token_id=…, video_end_token_id=…, **kwargs) - added:
constructor Glm4vMoeTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, partial_rotary_factor=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, moe_intermediate_size=…, num_experts_per_to… - added:
field Glm4vMoeConfig.keys_to_ignore_at_inference - added:
field Glm4vMoeConfig.model_type - added:
field Glm4vMoeConfig.sub_configs - added:
field Glm4vMoeTextConfig.base_config_key - added:
field Glm4vMoeTextConfig.base_model_pp_plan - added:
field Glm4vMoeTextConfig.base_model_tp_plan - added:
field Glm4vMoeTextConfig.keys_to_ignore_at_inference - added:
field Glm4vMoeTextConfig.model_type
- added:
class Glm4vMoeForConditionalGeneration(Glm4vMoePreTrainedModel, GenerationMixin) - added:
class Glm4vMoeModel(Glm4vMoePreTrainedModel) - added:
class Glm4vMoePreTrainedModel(PreTrainedModel) - added:
class Glm4vMoeTextModel(Glm4vMoePreTrainedModel) - added:
constructor Glm4vMoeForConditionalGeneration(config) - added:
constructor Glm4vMoeModel(config) - added:
constructor Glm4vMoeTextModel(config: Glm4vMoeTextConfig) - added:
field Glm4vMoeForConditionalGeneration.accepts_loss_kwargs - added:
field Glm4vMoeModel.accepts_loss_kwargs - added:
field Glm4vMoeModel.base_model_prefix - added:
field Glm4vMoeModel.config: Glm4vMoeConfig - added:
field Glm4vMoePreTrainedModel.base_model_prefix - added:
field Glm4vMoePreTrainedModel.config: Glm4vMoeConfig - added:
field Glm4vMoePreTrainedModel.supports_gradient_checkpointing - added:
field Glm4vMoeTextModel.config: Glm4vMoeTextConfig - added:
method Glm4vMoeForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_t… - added:
method Glm4vMoeForConditionalGeneration.get_decoder() - added:
method Glm4vMoeForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vMoeForConditionalGeneration.get_input_embeddings() - added:
method Glm4vMoeForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vMoeForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Glm4vMoeForConditionalGeneration.set_decoder(decoder) - added:
method Glm4vMoeForConditionalGeneration.set_input_embeddings(value) - added:
method Glm4vMoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optiona… - added:
method Glm4vMoeModel.get_decoder() - added:
method Glm4vMoeModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vMoeModel.get_input_embeddings() - added:
method Glm4vMoeModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Glm4vMoeModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Glm4vMoeModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Glm4vMoeModel.set_decoder(decoder) - added:
method Glm4vMoeModel.set_input_embeddings(value) - added:
method Glm4vMoeTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[FlashAttentionKwargs]) -> Union[tuple, BaseModelOutputWithPast] - added:
property Glm4vMoeForConditionalGeneration.language_model - added:
property Glm4vMoeForConditionalGeneration.visual
- added:
class Glm4vMoeConfig(Glm4vConfig) - added:
class Glm4vMoeForConditionalGeneration(Glm4vForConditionalGeneration) - added:
class Glm4vMoePreTrainedModel(Glm4MoePreTrainedModel) - added:
class Glm4vMoeTextConfig(Glm4MoeConfig) - added:
constructor Glm4vMoeConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, image_start_token_id=…, image_end_token_id=…, video_start_token_id=…, video_end_token_id=…, **kwargs) - added:
constructor Glm4vMoeTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, partial_rotary_factor=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, moe_intermediate_size=…, num_experts_per_to… - added:
field Glm4vMoePreTrainedModel.base_model_prefix - added:
field Glm4vMoePreTrainedModel.config: Glm4vMoeConfig - added:
field Glm4vMoeTextConfig.base_config_key - added:
field Glm4vMoeTextConfig.base_model_pp_plan - added:
field Glm4vMoeTextConfig.base_model_tp_plan - added:
field Glm4vMoeTextConfig.keys_to_ignore_at_inference - added:
field Glm4vMoeTextConfig.model_type
- added:
export define_import_structure
- added:
class GLPNConfig(PretrainedConfig) - added:
constructor GLPNConfig(num_channels=…, num_encoder_blocks=…, depths=…, sr_ratios=…, hidden_sizes=…, patch_sizes=…, strides=…, num_attention_heads=…, mlp_ratios=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, drop_path_rate=…, layer_norm_eps=…, decoder_hidden_size=…, max_depth=…, head_in_index=…, **kwargs) - added:
field GLPNConfig.model_type
- added:
class GLPNFeatureExtractor(GLPNImageProcessor) - added:
constructor GLPNFeatureExtractor(*args, **kwargs) -> None
- added:
class GLPNImageProcessor(BaseImageProcessor) - added:
constructor GLPNImageProcessor(do_resize: bool=…, size_divisor: int=…, resample=…, do_rescale: bool=…, **kwargs) -> None - added:
field GLPNImageProcessor.model_input_names - added:
method GLPNImageProcessor.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method GLPNImageProcessor.preprocess(images: Union[PIL.Image.Image, TensorType, list[PIL.Image.Image], list[TensorType]], do_resize: Optional[bool]=…, size_divisor: Optional[int]=…, resample=…, do_rescale: Optional[bool]=…, return_tensors: Optional[Union[TensorType, str]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method GLPNImageProcessor.resize(image: np.ndarray, size_divisor: int, resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class GLPNForDepthEstimation(GLPNPreTrainedModel) - added:
class GLPNLayer(nn.Module) - added:
class GLPNModel(GLPNPreTrainedModel) - added:
class GLPNPreTrainedModel(PreTrainedModel) - added:
constructor GLPNForDepthEstimation(config) - added:
constructor GLPNLayer(config, hidden_size, num_attention_heads, drop_path, sequence_reduction_ratio, mlp_ratio) - added:
constructor GLPNModel(config) - added:
field GLPNPreTrainedModel.base_model_prefix - added:
field GLPNPreTrainedModel.config: GLPNConfig - added:
field GLPNPreTrainedModel.main_input_name - added:
method GLPNForDepthEstimation.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthEstimatorOutput] - added:
method GLPNLayer.forward(hidden_states, height, width, output_attentions=…) - added:
method GLPNModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
export define_import_structure
- added:
class GotOcr2Config(PretrainedConfig) - added:
class GotOcr2VisionConfig(PretrainedConfig) - added:
constructor GotOcr2Config(vision_config=…, text_config=…, image_token_index=…, image_seq_length=…, pad_token_id=…, **kwargs) - added:
constructor GotOcr2VisionConfig(hidden_size=…, output_channels=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, use_abs_pos=…, use_rel_pos=…, window_size=…, global_attn_indexes=…, mlp_dim=…, **kwargs) - added:
field GotOcr2Config.attribute_map - added:
field GotOcr2Config.model_type - added:
field GotOcr2Config.sub_configs - added:
field GotOcr2VisionConfig.base_config_key
- added:
class GotOcr2ImageProcessor(BaseImageProcessor) - added:
constructor GotOcr2ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, crop_to_patches: bool=…, min_patches: int=…, max_patches: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -… - added:
field GotOcr2ImageProcessor.model_input_names - added:
method GotOcr2ImageProcessor.crop_image_to_patches(images: np.ndarray, min_patches: int, max_patches: int, use_thumbnail: bool=…, patch_size: Optional[Union[tuple, int, dict]]=…, data_format: Optional[ChannelDimension]=…) - added:
method GotOcr2ImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method GotOcr2ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, crop_to_patches: Optional[bool]=…, min_patches: Optional[int]=…, max_patches: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[flo… - added:
method GotOcr2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class GotOcr2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor GotOcr2ImageProcessorFast(**kwargs: Unpack[GotOcr2FastImageProcessorKwargs]) - added:
field GotOcr2ImageProcessorFast.crop_to_patches - added:
field GotOcr2ImageProcessorFast.do_convert_rgb - added:
field GotOcr2ImageProcessorFast.do_normalize - added:
field GotOcr2ImageProcessorFast.do_rescale - added:
field GotOcr2ImageProcessorFast.do_resize - added:
field GotOcr2ImageProcessorFast.image_mean - added:
field GotOcr2ImageProcessorFast.image_std - added:
field GotOcr2ImageProcessorFast.max_patches - added:
field GotOcr2ImageProcessorFast.min_patches - added:
field GotOcr2ImageProcessorFast.resample - added:
field GotOcr2ImageProcessorFast.size - added:
field GotOcr2ImageProcessorFast.valid_kwargs - added:
method GotOcr2ImageProcessorFast.crop_image_to_patches(images: torch.Tensor, min_patches: int, max_patches: int, use_thumbnail: bool=…, patch_size: Optional[Union[tuple, int, dict]]=…, interpolation: Optional[F.InterpolationMode]=…) - added:
method GotOcr2ImageProcessorFast.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method GotOcr2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[GotOcr2FastImageProcessorKwargs]) -> BatchFeature
- added:
class GotOcr2ForConditionalGeneration(GotOcr2PreTrainedModel, GenerationMixin) - added:
class GotOcr2Model(GotOcr2PreTrainedModel) - added:
class GotOcr2PreTrainedModel(PreTrainedModel) - added:
constructor GotOcr2ForConditionalGeneration(config: GotOcr2Config) - added:
constructor GotOcr2Model(config: GotOcr2Config) - added:
field GotOcr2PreTrainedModel.base_model_prefix - added:
field GotOcr2PreTrainedModel.config: GotOcr2Config - added:
field GotOcr2PreTrainedModel.supports_gradient_checkpointing - added:
method GotOcr2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[boo… - added:
method GotOcr2ForConditionalGeneration.get_decoder() - added:
method GotOcr2ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method GotOcr2ForConditionalGeneration.get_input_embeddings() - added:
method GotOcr2ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method GotOcr2ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method GotOcr2ForConditionalGeneration.set_decoder(decoder) - added:
method GotOcr2ForConditionalGeneration.set_input_embeddings(value) - added:
method GotOcr2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict… - added:
method GotOcr2Model.get_decoder() - added:
method GotOcr2Model.get_image_features(pixel_values: torch.FloatTensor) - added:
method GotOcr2Model.get_input_embeddings() - added:
method GotOcr2Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method GotOcr2Model.set_decoder(decoder) - added:
method GotOcr2Model.set_input_embeddings(value) - added:
property GotOcr2ForConditionalGeneration.language_model - added:
property GotOcr2ForConditionalGeneration.multi_modal_projector - added:
property GotOcr2ForConditionalGeneration.vision_tower
- added:
class GotOcr2Config(PretrainedConfig) - added:
class GotOcr2ForConditionalGeneration(LlavaForConditionalGeneration) - added:
class GotOcr2Model(LlavaModel) - added:
class GotOcr2PreTrainedModel(LlavaPreTrainedModel) - added:
class GotOcr2PreTrainedModel(SamPreTrainedModel) - added:
class GotOcr2VisionConfig(PretrainedConfig) - added:
constructor GotOcr2Config(vision_config=…, text_config=…, image_token_index=…, image_seq_length=…, pad_token_id=…, **kwargs) - added:
constructor GotOcr2Model(config: GotOcr2Config) - added:
constructor GotOcr2VisionConfig(hidden_size=…, output_channels=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, use_abs_pos=…, use_rel_pos=…, window_size=…, global_attn_indexes=…, mlp_dim=…, **kwargs) - added:
field GotOcr2Config.attribute_map - added:
field GotOcr2Config.model_type - added:
field GotOcr2Config.sub_configs - added:
field GotOcr2VisionConfig.base_config_key - added:
method GotOcr2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[boo… - added:
method GotOcr2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict… - added:
method GotOcr2Model.get_image_features(pixel_values: torch.FloatTensor)
- added:
class GotOcr2Processor(ProcessorMixin) - added:
constructor GotOcr2Processor(image_processor=…, tokenizer=…, chat_template=…, **kwargs) - added:
field GotOcr2Processor.attributes - added:
field GotOcr2Processor.image_processor_class - added:
field GotOcr2Processor.tokenizer_class
- added:
export define_import_structure
- added:
class GPT2Config(PretrainedConfig) - added:
class GPT2OnnxConfig(OnnxConfigWithPast) - added:
constructor GPT2Config(vocab_size=…, n_positions=…, n_embd=…, n_layer=…, n_head=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dropout=…, scale_attn_weights=…, use_cache=…, bos_token_id=…, eos_token_id=…, scale_attn_by_invers… - added:
constructor GPT2OnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…, use_past: bool=…) - added:
field GPT2Config.attribute_map - added:
field GPT2Config.keys_to_ignore_at_inference - added:
field GPT2Config.model_type - added:
method GPT2OnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property GPT2OnnxConfig.default_onnx_opset -> int - added:
property GPT2OnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property GPT2OnnxConfig.num_attention_heads -> int - added:
property GPT2OnnxConfig.num_layers -> int
- added:
class FlaxGPT2LMHeadModel(FlaxGPT2PreTrainedModel) - added:
class FlaxGPT2Model(FlaxGPT2PreTrainedModel) - added:
class FlaxGPT2PreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxGPT2PreTrainedModel(config: GPT2Config, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxGPT2LMHeadModel.module_class - added:
field FlaxGPT2Model.module_class - added:
field FlaxGPT2PreTrainedModel.base_model_prefix - added:
field FlaxGPT2PreTrainedModel.config_class - added:
field FlaxGPT2PreTrainedModel.module_class: nn.Module - added:
method FlaxGPT2LMHeadModel.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxGPT2LMHeadModel.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxGPT2PreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxGPT2PreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class GPT2DoubleHeadsModel(GPT2PreTrainedModel, GenerationMixin) - added:
class GPT2ForQuestionAnswering(GPT2PreTrainedModel) - added:
class GPT2ForSequenceClassification(GPT2PreTrainedModel) - added:
class GPT2ForTokenClassification(GPT2PreTrainedModel) - added:
class GPT2LMHeadModel(GPT2PreTrainedModel, GenerationMixin) - added:
class GPT2Model(GPT2PreTrainedModel) - added:
class GPT2PreTrainedModel(PreTrainedModel) - added:
constructor GPT2DoubleHeadsModel(config) - added:
constructor GPT2ForQuestionAnswering(config) - added:
constructor GPT2ForSequenceClassification(config) - added:
constructor GPT2ForTokenClassification(config) - added:
constructor GPT2LMHeadModel(config) - added:
constructor GPT2Model(config) - added:
constructor GPT2PreTrainedModel(*inputs, **kwargs) - added:
field GPT2PreTrainedModel.base_model_prefix - added:
field GPT2PreTrainedModel.config: GPT2Config - added:
field GPT2PreTrainedModel.is_parallelizable - added:
field GPT2PreTrainedModel.load_tf_weights - added:
field GPT2PreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_gpt2(model, config, gpt2_checkpoint_path) - added:
method GPT2DoubleHeadsModel.deparallelize() - added:
method GPT2DoubleHeadsModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, mc_token_ids: O… - added:
method GPT2DoubleHeadsModel.parallelize(device_map=…) - added:
method GPT2ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, o… - added:
method GPT2ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Opt… - added:
method GPT2ForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Option… - added:
method GPT2LMHeadModel.deparallelize() - added:
method GPT2LMHeadModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_state… - added:
method GPT2LMHeadModel.parallelize(device_map=…) - added:
method GPT2Model.deparallelize() - added:
method GPT2Model.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Opt… - added:
method GPT2Model.get_input_embeddings() - added:
method GPT2Model.parallelize(device_map=…) - added:
method GPT2Model.set_input_embeddings(new_embeddings)
- added:
class TFGPT2DoubleHeadsModel(TFGPT2PreTrainedModel) - added:
class TFGPT2ForSequenceClassification(TFGPT2PreTrainedModel, TFSequenceClassificationLoss) - added:
class TFGPT2LMHeadModel(TFGPT2PreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFGPT2MainLayer(keras.layers.Layer) - added:
class TFGPT2Model(TFGPT2PreTrainedModel) - added:
class TFGPT2PreTrainedModel(TFPreTrainedModel) - added:
constructor TFGPT2DoubleHeadsModel(config, *inputs, **kwargs) - added:
constructor TFGPT2ForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFGPT2LMHeadModel(config, *inputs, **kwargs) - added:
constructor TFGPT2MainLayer(config, *inputs, **kwargs) - added:
constructor TFGPT2Model(config, *inputs, **kwargs) - added:
field TFGPT2MainLayer.config_class - added:
field TFGPT2PreTrainedModel.base_model_prefix - added:
field TFGPT2PreTrainedModel.config_class - added:
method TFGPT2DoubleHeadsModel.build(input_shape=…) - added:
method TFGPT2DoubleHeadsModel.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, mc_token_ids: np.ndarray… - added:
method TFGPT2ForSequenceClassification.build(input_shape=…) - added:
method TFGPT2ForSequenceClassification.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool… - added:
method TFGPT2LMHeadModel.build(input_shape=…) - added:
method TFGPT2LMHeadModel.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndar… - added:
method TFGPT2LMHeadModel.get_output_embeddings() - added:
method TFGPT2LMHeadModel.prepare_inputs_for_generation(inputs, past_key_values=…, use_cache=…, **kwargs) - added:
method TFGPT2LMHeadModel.set_output_embeddings(value) - added:
method TFGPT2MainLayer.build(input_shape=…) - added:
method TFGPT2MainLayer.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarra… - added:
method TFGPT2MainLayer.get_input_embeddings() - added:
method TFGPT2MainLayer.set_input_embeddings(new_embeddings) - added:
method TFGPT2Model.build(input_shape=…) - added:
method TFGPT2Model.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray |… - added:
property TFGPT2DoubleHeadsModel.input_signature - added:
property TFGPT2PreTrainedModel.input_signature
- added:
class GPT2Tokenizer(PreTrainedTokenizer) - added:
constructor GPT2Tokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_prefix_space=…, add_bos_token=…, **kwargs) - added:
field GPT2Tokenizer.model_input_names - added:
field GPT2Tokenizer.vocab_files_names - added:
method GPT2Tokenizer.bpe(token) - added:
method GPT2Tokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method GPT2Tokenizer.convert_tokens_to_string(tokens) - added:
method GPT2Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method GPT2Tokenizer.get_vocab() - added:
method GPT2Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method GPT2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property GPT2Tokenizer.vocab_size
- added:
class GPT2TokenizerFast(PreTrainedTokenizerFast) - added:
constructor GPT2TokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, **kwargs) - added:
field GPT2TokenizerFast.model_input_names - added:
field GPT2TokenizerFast.slow_tokenizer_class - added:
field GPT2TokenizerFast.vocab_files_names - added:
method GPT2TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
class TFGPT2Tokenizer(keras.layers.Layer) - added:
constructor TFGPT2Tokenizer(vocab: dict[str, int], merges: list[str], max_length: Optional[int]=…, pad_token_id: Optional[int]=…) - added:
method TFGPT2Tokenizer.call(x, max_length: Optional[int]=…) - added:
method TFGPT2Tokenizer.get_config() - added:
static method TFGPT2Tokenizer.from_config(config) - added:
static method TFGPT2Tokenizer.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], *init_inputs, **kwargs) - added:
static method TFGPT2Tokenizer.from_tokenizer(tokenizer: GPT2Tokenizer, *args, **kwargs)
- added:
export define_import_structure
- added:
class GPTBigCodeConfig(PretrainedConfig) - added:
constructor GPTBigCodeConfig(vocab_size=…, n_positions=…, n_embd=…, n_layer=…, n_head=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, scale_attn_weights=…, use_cache=…, bos_token_id=…, eos_token_id=…, attention_softmax_in_fp32=…, scale_attention_softmax_in_fp32=…, multi_query=…, **kwargs) - added:
field GPTBigCodeConfig.attribute_map - added:
field GPTBigCodeConfig.keys_to_ignore_at_inference - added:
field GPTBigCodeConfig.model_type
- added:
class GPTBigCodeForCausalLM(GPTBigCodePreTrainedModel, GenerationMixin) - added:
class GPTBigCodeForSequenceClassification(GPTBigCodePreTrainedModel) - added:
class GPTBigCodeForTokenClassification(GPTBigCodePreTrainedModel) - added:
class GPTBigCodeModel(GPTBigCodePreTrainedModel) - added:
class GPTBigCodePreTrainedModel(PreTrainedModel) - added:
constructor GPTBigCodeForCausalLM(config) - added:
constructor GPTBigCodeForSequenceClassification(config) - added:
constructor GPTBigCodeForTokenClassification(config) - added:
constructor GPTBigCodeModel(config) - added:
constructor GPTBigCodePreTrainedModel(*inputs, **kwargs) - added:
field GPTBigCodePreTrainedModel.base_model_prefix - added:
field GPTBigCodePreTrainedModel.config: GPTBigCodeConfig - added:
field GPTBigCodePreTrainedModel.supports_gradient_checkpointing - added:
method GPTBigCodeForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch… - added:
method GPTBigCodeForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_att… - added:
method GPTBigCodeForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attent… - added:
method GPTBigCodeModel.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tenso… - added:
method GPTBigCodeModel.get_input_embeddings() - added:
method GPTBigCodeModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class GPTNeoConfig(PretrainedConfig) - added:
class GPTNeoOnnxConfig(OnnxConfigWithPast) - added:
constructor GPTNeoConfig(vocab_size=…, max_position_embeddings=…, hidden_size=…, num_layers=…, attention_types=…, num_heads=…, intermediate_size=…, window_size=…, activation_function=…, resid_dropout=…, embed_dropout=…, attention_dropout=…, classifier_dropout=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field GPTNeoConfig.attribute_map - added:
field GPTNeoConfig.keys_to_ignore_at_inference - added:
field GPTNeoConfig.model_type - added:
method GPTNeoOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property GPTNeoOnnxConfig.default_onnx_opset -> int - added:
property GPTNeoOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property GPTNeoOnnxConfig.num_attention_heads -> int - added:
static method GPTNeoConfig.expand_attention_types_params(attention_types)
- added:
class FlaxGPTNeoForCausalLM(FlaxGPTNeoPreTrainedModel) - added:
class FlaxGPTNeoModel(FlaxGPTNeoPreTrainedModel) - added:
class FlaxGPTNeoPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxGPTNeoPreTrainedModel(config: GPTNeoConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxGPTNeoForCausalLM.module_class - added:
field FlaxGPTNeoModel.module_class - added:
field FlaxGPTNeoPreTrainedModel.base_model_prefix - added:
field FlaxGPTNeoPreTrainedModel.config_class - added:
field FlaxGPTNeoPreTrainedModel.module_class: nn.Module - added:
method FlaxGPTNeoForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxGPTNeoForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxGPTNeoPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxGPTNeoPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class GPTNeoForCausalLM(GPTNeoPreTrainedModel, GenerationMixin) - added:
class GPTNeoForQuestionAnswering(GPTNeoPreTrainedModel) - added:
class GPTNeoForSequenceClassification(GPTNeoPreTrainedModel) - added:
class GPTNeoForTokenClassification(GPTNeoPreTrainedModel) - added:
class GPTNeoModel(GPTNeoPreTrainedModel) - added:
class GPTNeoPreTrainedModel(PreTrainedModel) - added:
constructor GPTNeoForCausalLM(config) - added:
constructor GPTNeoForQuestionAnswering(config) - added:
constructor GPTNeoForSequenceClassification(config) - added:
constructor GPTNeoForTokenClassification(config) - added:
constructor GPTNeoModel(config) - added:
constructor GPTNeoPreTrainedModel(*inputs, **kwargs) - added:
field GPTNeoPreTrainedModel.base_model_prefix - added:
field GPTNeoPreTrainedModel.config: GPTNeoConfig - added:
field GPTNeoPreTrainedModel.load_tf_weights - added:
field GPTNeoPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_gpt_neo(model, config, gpt_neo_checkpoint_path) - added:
method GPTNeoForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, tuple[torch.FloatTensor]]]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool… - added:
method GPTNeoForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…,… - added:
method GPTNeoForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, tuple[torch.FloatTensor]]]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache:… - added:
method GPTNeoForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[t… - added:
method GPTNeoModel.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, tuple[torch.FloatTensor]]]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…… - added:
method GPTNeoModel.get_input_embeddings() - added:
method GPTNeoModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class GPTNeoXConfig(PretrainedConfig) - added:
constructor GPTNeoXConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, rotary_pct=…, rotary_emb_base=…, attention_dropout=…, hidden_dropout=…, classifier_dropout=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, use_parallel_residual=…, rope_scaling… - added:
field GPTNeoXConfig.base_model_pp_plan - added:
field GPTNeoXConfig.base_model_tp_plan - added:
field GPTNeoXConfig.keys_to_ignore_at_inference - added:
field GPTNeoXConfig.model_type
- added:
class GPTNeoXForCausalLM(GPTNeoXPreTrainedModel, GenerationMixin) - added:
class GPTNeoXForQuestionAnswering(GPTNeoXPreTrainedModel) - added:
class GPTNeoXForSequenceClassification(GPTNeoXPreTrainedModel) - added:
class GPTNeoXForTokenClassification(GPTNeoXPreTrainedModel) - added:
class GPTNeoXLayer(GradientCheckpointingLayer) - added:
class GPTNeoXModel(GPTNeoXPreTrainedModel) - added:
class GPTNeoXPreTrainedModel(PreTrainedModel) - added:
constructor GPTNeoXForCausalLM(config) - added:
constructor GPTNeoXForQuestionAnswering(config) - added:
constructor GPTNeoXForSequenceClassification(config) - added:
constructor GPTNeoXForTokenClassification(config) - added:
constructor GPTNeoXLayer(config, layer_idx) - added:
constructor GPTNeoXModel(config) - added:
field GPTNeoXPreTrainedModel.base_model_prefix - added:
field GPTNeoXPreTrainedModel.config: GPTNeoXConfig - added:
field GPTNeoXPreTrainedModel.supports_gradient_checkpointing - added:
method GPTNeoXForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, ou… - added:
method GPTNeoXForCausalLM.get_output_embeddings() - added:
method GPTNeoXForCausalLM.set_output_embeddings(new_embeddings) - added:
method GPTNeoXForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…… - added:
method GPTNeoXForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Option… - added:
method GPTNeoXForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[… - added:
method GPTNeoXLayer.forward(hidden_states: Optional[torch.FloatTensor], attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, layer_past: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, position_embeddings: Optional[tuple[torch.Tensor, t… - added:
method GPTNeoXModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_pos… - added:
method GPTNeoXModel.get_input_embeddings() - added:
method GPTNeoXModel.set_input_embeddings(value)
- added:
class GPTNeoXForCausalLM(GPTNeoXPreTrainedModel, GenerationMixin) - added:
class GPTNeoXForQuestionAnswering(GPTNeoXPreTrainedModel) - added:
class GPTNeoXForSequenceClassification(GPTNeoXPreTrainedModel) - added:
class GPTNeoXForTokenClassification(GPTNeoXPreTrainedModel) - added:
class GPTNeoXLayer(GradientCheckpointingLayer) - added:
class GPTNeoXModel(LlamaModel) - added:
class GPTNeoXPreTrainedModel(LlamaPreTrainedModel) - added:
constructor GPTNeoXForCausalLM(config) - added:
constructor GPTNeoXForQuestionAnswering(config) - added:
constructor GPTNeoXForSequenceClassification(config) - added:
constructor GPTNeoXForTokenClassification(config) - added:
constructor GPTNeoXLayer(config, layer_idx) - added:
constructor GPTNeoXModel(config) - added:
field GPTNeoXPreTrainedModel.base_model_prefix - added:
method GPTNeoXForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, ou… - added:
method GPTNeoXForCausalLM.get_output_embeddings() - added:
method GPTNeoXForCausalLM.set_output_embeddings(new_embeddings) - added:
method GPTNeoXForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…… - added:
method GPTNeoXForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Option… - added:
method GPTNeoXForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[… - added:
method GPTNeoXLayer.forward(hidden_states: Optional[torch.FloatTensor], attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, layer_past: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, position_embeddings: Optional[tuple[torch.Tensor, t… - added:
method GPTNeoXModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_pos… - added:
method GPTNeoXModel.get_input_embeddings() - added:
method GPTNeoXModel.set_input_embeddings(value)
- added:
class GPTNeoXTokenizerFast(PreTrainedTokenizerFast) - added:
constructor GPTNeoXTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_bos_token=…, add_eos_token=…, add_prefix_space=…, **kwargs) - added:
field GPTNeoXTokenizerFast.model_input_names - added:
field GPTNeoXTokenizerFast.vocab_files_names - added:
method GPTNeoXTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method GPTNeoXTokenizerFast.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method GPTNeoXTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method GPTNeoXTokenizerFast.update_post_processor() - added:
property GPTNeoXTokenizerFast.add_bos_token - added:
property GPTNeoXTokenizerFast.add_eos_token
- added:
export define_import_structure
- added:
class GPTNeoXJapaneseConfig(PretrainedConfig) - added:
constructor GPTNeoXJapaneseConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_multiple_size=…, hidden_act=…, rotary_pct=…, rotary_emb_base=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, bos_token_id=…, eos_token_id=…, rope_scaling=…, attention_dropout=…, hidden_dropout=…, **kwargs) - added:
field GPTNeoXJapaneseConfig.model_type
- added:
class GPTNeoXJapaneseForCausalLM(GPTNeoXJapanesePreTrainedModel, GenerationMixin) - added:
class GPTNeoXJapaneseLayer(nn.Module) - added:
class GPTNeoXJapaneseModel(GPTNeoXJapanesePreTrainedModel) - added:
class GPTNeoXJapanesePreTrainedModel(PreTrainedModel) - added:
constructor GPTNeoXJapaneseForCausalLM(config) - added:
constructor GPTNeoXJapaneseLayer(config, layer_number) - added:
constructor GPTNeoXJapaneseModel(config) - added:
field GPTNeoXJapanesePreTrainedModel.base_model_prefix - added:
field GPTNeoXJapanesePreTrainedModel.config: GPTNeoXJapaneseConfig - added:
method GPTNeoXJapaneseForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[boo… - added:
method GPTNeoXJapaneseForCausalLM.get_output_embeddings() - added:
method GPTNeoXJapaneseForCausalLM.set_output_embeddings(new_embeddings) - added:
method GPTNeoXJapaneseLayer.forward(hidden_states: Optional[torch.FloatTensor], attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, layer_past: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, position_embeddings: Optional[tuple[torch.T… - added:
method GPTNeoXJapaneseModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.FloatTensor]]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, o… - added:
method GPTNeoXJapaneseModel.get_input_embeddings() - added:
method GPTNeoXJapaneseModel.set_input_embeddings(value)
- added:
class GPTNeoXJapaneseTokenizer(PreTrainedTokenizer) - added:
constructor GPTNeoXJapaneseTokenizer(vocab_file, emoji_file, unk_token=…, pad_token=…, bos_token=…, eos_token=…, do_clean_text=…, **kwargs) - added:
field GPTNeoXJapaneseTokenizer.model_input_names - added:
field GPTNeoXJapaneseTokenizer.vocab_files_names - added:
method GPTNeoXJapaneseTokenizer.convert_tokens_to_string(tokens) - added:
method GPTNeoXJapaneseTokenizer.get_vocab() - added:
method GPTNeoXJapaneseTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property GPTNeoXJapaneseTokenizer.vocab_size
- added:
export define_import_structure
- added:
class GptOssConfig(PretrainedConfig) - added:
constructor GptOssConfig(num_hidden_layers: int=…, num_local_experts: int=…, vocab_size: int=…, hidden_size: int=…, intermediate_size: int=…, head_dim: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, sliding_window: int=…, rope_theta: float=…, tie_word_embeddings=…, hidden_act: str=…, initializer_range: float=…, max_position_embeddings=…, rms_norm_eps: float=…, rope_scaling=…, atte… - added:
field GptOssConfig.base_model_pp_plan - added:
field GptOssConfig.base_model_tp_plan - added:
field GptOssConfig.model_type
- added:
class GptOssForCausalLM(GptOssPreTrainedModel, GenerationMixin) - added:
class GptOssForSequenceClassification(GenericForSequenceClassification, GptOssPreTrainedModel) - added:
class GptOssForTokenClassification(GenericForTokenClassification, GptOssPreTrainedModel) - added:
class GptOssModel(GptOssPreTrainedModel) - added:
class GptOssPreTrainedModel(PreTrainedModel) - added:
constructor GptOssForCausalLM(config) - added:
constructor GptOssModel(config: GptOssConfig) - added:
field GptOssPreTrainedModel.base_model_prefix - added:
field GptOssPreTrainedModel.config: GptOssConfig - added:
field GptOssPreTrainedModel.supports_gradient_checkpointing - added:
method GptOssForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logi… - added:
method GptOssModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class GptOssForCausalLM(MixtralForCausalLM) - added:
class GptOssForSequenceClassification(MixtralForSequenceClassification) - added:
class GptOssForTokenClassification(MixtralForTokenClassification) - added:
class GptOssModel(MixtralModel) - added:
class GptOssPreTrainedModel(LlamaPreTrainedModel) - added:
method GptOssModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
export define_import_structure
- added:
class GPTSw3Tokenizer(PreTrainedTokenizer) - added:
constructor GPTSw3Tokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, pad_token=…, unk_token=…, eos_token=…, bos_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field GPTSw3Tokenizer.model_input_names - added:
field GPTSw3Tokenizer.vocab_files_names - added:
method GPTSw3Tokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method GPTSw3Tokenizer.decode_fast(token_ids: Union[int, list[int]]) -> str - added:
method GPTSw3Tokenizer.encode_fast(text: Union[str, list[str]], return_tensors: Union[str, bool]=…) -> Union[list[int], list[list[int]], torch.Tensor] - added:
method GPTSw3Tokenizer.get_vocab() -> dict[str, int] - added:
method GPTSw3Tokenizer.preprocess_text(text: str) -> str - added:
method GPTSw3Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property GPTSw3Tokenizer.vocab_size -> int - added:
static method GPTSw3Tokenizer.clean_up_tokenization(out_string: str) -> str
- added:
export define_import_structure
- added:
class GPTJConfig(PretrainedConfig) - added:
class GPTJOnnxConfig(OnnxConfigWithPast) - added:
constructor GPTJConfig(vocab_size=…, n_positions=…, n_embd=…, n_layer=…, n_head=…, rotary_dim=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, **kwargs) - added:
constructor GPTJOnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…, use_past: bool=…) - added:
field GPTJConfig.attribute_map - added:
field GPTJConfig.model_type - added:
method GPTJOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property GPTJOnnxConfig.default_onnx_opset -> int - added:
property GPTJOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property GPTJOnnxConfig.num_attention_heads -> int - added:
property GPTJOnnxConfig.num_layers -> int
- added:
class FlaxGPTJForCausalLM(FlaxGPTJPreTrainedModel) - added:
class FlaxGPTJModel(FlaxGPTJPreTrainedModel) - added:
class FlaxGPTJPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxGPTJPreTrainedModel(config: GPTJConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxGPTJForCausalLM.module_class - added:
field FlaxGPTJModel.module_class - added:
field FlaxGPTJPreTrainedModel.base_model_prefix - added:
field FlaxGPTJPreTrainedModel.config_class - added:
field FlaxGPTJPreTrainedModel.module_class: nn.Module - added:
method FlaxGPTJForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxGPTJForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxGPTJPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxGPTJPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class GPTJForCausalLM(GPTJPreTrainedModel, GenerationMixin) - added:
class GPTJForQuestionAnswering(GPTJPreTrainedModel) - added:
class GPTJForSequenceClassification(GPTJPreTrainedModel) - added:
class GPTJModel(GPTJPreTrainedModel) - added:
class GPTJPreTrainedModel(PreTrainedModel) - added:
constructor GPTJForCausalLM(config) - added:
constructor GPTJForQuestionAnswering(config) - added:
constructor GPTJForSequenceClassification(config) - added:
constructor GPTJModel(config) - added:
constructor GPTJPreTrainedModel(*inputs, **kwargs) - added:
field GPTJPreTrainedModel.base_model_prefix - added:
field GPTJPreTrainedModel.config: GPTJConfig - added:
field GPTJPreTrainedModel.is_parallelizable - added:
field GPTJPreTrainedModel.supports_gradient_checkpointing - added:
method GPTJForCausalLM.deparallelize() - added:
method GPTJForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTens… - added:
method GPTJForCausalLM.parallelize(device_map=…) - added:
method GPTJForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, o… - added:
method GPTJForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Opt… - added:
method GPTJModel.deparallelize() - added:
method GPTJModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, tuple[tuple[torch.Tensor]]]]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_a… - added:
method GPTJModel.get_input_embeddings() - added:
method GPTJModel.parallelize(device_map=…) - added:
method GPTJModel.set_input_embeddings(new_embeddings)
- added:
class TFGPTJForCausalLM(TFGPTJPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFGPTJForQuestionAnswering(TFGPTJPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFGPTJForSequenceClassification(TFGPTJPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFGPTJModel(TFGPTJPreTrainedModel) - added:
class TFGPTJPreTrainedModel(TFPreTrainedModel) - added:
constructor TFGPTJForCausalLM(config, *inputs, **kwargs) - added:
constructor TFGPTJForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFGPTJForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFGPTJModel(config, *inputs, **kwargs) - added:
field TFGPTJPreTrainedModel.base_model_prefix - added:
field TFGPTJPreTrainedModel.config_class - added:
method TFGPTJForCausalLM.build(input_shape=…) - added:
method TFGPTJForCausalLM.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor… - added:
method TFGPTJForCausalLM.prepare_inputs_for_generation(inputs, past_key_values=…, use_cache=…, **kwargs) - added:
method TFGPTJForQuestionAnswering.build(input_shape=…) - added:
method TFGPTJForQuestionAnswering.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, start_positions: np.n… - added:
method TFGPTJForSequenceClassification.build(input_shape=…) - added:
method TFGPTJForSequenceClassification.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarr… - added:
method TFGPTJModel.build(input_shape=…) - added:
method TFGPTJModel.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool | None=…, output_att…
- added:
export define_import_structure
- added:
class GraniteConfig(PretrainedConfig) - added:
constructor GraniteConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=…, embedding_mul… - added:
field GraniteConfig.base_model_pp_plan - added:
field GraniteConfig.base_model_tp_plan - added:
field GraniteConfig.keys_to_ignore_at_inference - added:
field GraniteConfig.model_type
- added:
class GraniteForCausalLM(GranitePreTrainedModel, GenerationMixin) - added:
class GraniteModel(GranitePreTrainedModel) - added:
class GranitePreTrainedModel(PreTrainedModel) - added:
constructor GraniteForCausalLM(config) - added:
constructor GraniteModel(config: GraniteConfig) - added:
field GranitePreTrainedModel.base_model_prefix - added:
field GranitePreTrainedModel.config: GraniteConfig - added:
field GranitePreTrainedModel.supports_gradient_checkpointing - added:
method GraniteForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states… - added:
method GraniteModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs:…
- added:
class GraniteForCausalLM(LlamaForCausalLM) - added:
class GraniteModel(LlamaModel) - added:
class GranitePreTrainedModel(LlamaPreTrainedModel) - added:
constructor GraniteModel(config: GraniteConfig) - added:
method GraniteForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states… - added:
method GraniteModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs:…
- added:
export define_import_structure
- added:
class GraniteSpeechConfig(PretrainedConfig) - added:
class GraniteSpeechEncoderConfig(PretrainedConfig) - added:
constructor GraniteSpeechConfig(text_config=…, encoder_config=…, projector_config=…, audio_token_index=…, initializer_range=…, has_lora_adapter=…, downsample_rate=…, window_size=…, **kwargs) - added:
constructor GraniteSpeechEncoderConfig(input_dim=…, num_layers=…, hidden_dim=…, feedforward_mult=…, num_heads=…, dim_head=…, output_dim=…, context_size=…, max_pos_emb=…, dropout=…, conv_kernel_size=…, conv_expansion_factor=…, **kwargs) - added:
field GraniteSpeechConfig.attribute_map - added:
field GraniteSpeechConfig.model_type - added:
field GraniteSpeechConfig.sub_configs - added:
field GraniteSpeechEncoderConfig.model_type
- added:
class GraniteSpeechFeatureExtractor(FeatureExtractionMixin) - added:
constructor GraniteSpeechFeatureExtractor(sampling_rate: int=…, n_fft: int=…, win_length: int=…, hop_length: int=…, n_mels: int=…, projector_window_size: int=…, projector_downsample_rate: int=…, **kwargs) - added:
field GraniteSpeechFeatureExtractor.model_input_names
- added:
class GraniteSpeechCTCEncoder(nn.Module) - added:
class GraniteSpeechForConditionalGeneration(GraniteSpeechPreTrainedModel, GenerationMixin) - added:
class GraniteSpeechPreTrainedModel(PreTrainedModel) - added:
constructor GraniteSpeechCTCEncoder(config: GraniteSpeechEncoderConfig) - added:
constructor GraniteSpeechForConditionalGeneration(config: GraniteSpeechConfig) - added:
field GraniteSpeechPreTrainedModel.config: GraniteSpeechConfig - added:
method GraniteSpeechCTCEncoder.forward(hidden_states: torch.Tensor) - added:
method GraniteSpeechForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, input_features_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_c… - added:
method GraniteSpeechForConditionalGeneration.generate(*args, **kwargs) -> torch.LongTensor - added:
method GraniteSpeechForConditionalGeneration.get_audio_features(input_features: torch.Tensor) -> torch.Tensor - added:
method GraniteSpeechForConditionalGeneration.get_input_embeddings() - added:
method GraniteSpeechForConditionalGeneration.get_merged_audio_embeddings(input_ids: torch.Tensor, audio_features: torch.Tensor, input_features_mask: Optional[torch.Tensor]=…) -> torch.Tensor - added:
method GraniteSpeechForConditionalGeneration.get_output_embeddings() - added:
method GraniteSpeechForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, input_features=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method GraniteSpeechForConditionalGeneration.save_pretrained(save_directory, *args, **kwargs) - added:
method GraniteSpeechForConditionalGeneration.set_input_embeddings(value) - added:
method GraniteSpeechForConditionalGeneration.set_output_embeddings(new_embeddings)
- added:
class GraniteSpeechProcessor(ProcessorMixin) - added:
constructor GraniteSpeechProcessor(audio_processor, tokenizer, audio_token=…, chat_template=…) - added:
field GraniteSpeechProcessor.attributes - added:
field GraniteSpeechProcessor.audio_processor_class - added:
field GraniteSpeechProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class GraniteMoeConfig(PretrainedConfig) - added:
constructor GraniteMoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, embedding_multiplier=…… - added:
field GraniteMoeConfig.keys_to_ignore_at_inference - added:
field GraniteMoeConfig.model_type
- added:
class GraniteMoeForCausalLM(GraniteMoePreTrainedModel, GenerationMixin) - added:
class GraniteMoeModel(GraniteMoePreTrainedModel) - added:
class GraniteMoePreTrainedModel(PreTrainedModel) - added:
constructor GraniteMoeForCausalLM(config: GraniteMoeConfig) - added:
constructor GraniteMoeModel(config: GraniteMoeConfig) - added:
field GraniteMoePreTrainedModel.base_model_prefix - added:
field GraniteMoePreTrainedModel.config: GraniteMoeConfig - added:
field GraniteMoePreTrainedModel.supports_gradient_checkpointing - added:
method GraniteMoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_sta… - added:
method GraniteMoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits:…
- added:
export define_import_structure
- added:
class GraniteMoeHybridConfig(PretrainedConfig) - added:
constructor GraniteMoeHybridConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, embedding_multip… - added:
field GraniteMoeHybridConfig.attribute_map - added:
field GraniteMoeHybridConfig.keys_to_ignore_at_inference - added:
field GraniteMoeHybridConfig.model_type - added:
property GraniteMoeHybridConfig.layers_block_type
- added:
class GraniteMoeHybridForCausalLM(GraniteMoeHybridPreTrainedModel, GenerationMixin) - added:
class GraniteMoeHybridModel(GraniteMoeHybridPreTrainedModel) - added:
class GraniteMoeHybridPreTrainedModel(PreTrainedModel) - added:
constructor GraniteMoeHybridForCausalLM(config: GraniteMoeHybridConfig) - added:
constructor GraniteMoeHybridModel(config: GraniteMoeHybridConfig) - added:
field GraniteMoeHybridPreTrainedModel.base_model_prefix - added:
field GraniteMoeHybridPreTrainedModel.config: GraniteMoeHybridConfig - added:
field GraniteMoeHybridPreTrainedModel.supports_gradient_checkpointing - added:
method GraniteMoeHybridForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method GraniteMoeHybridForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method GraniteMoeHybridModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_l…
- added:
class GraniteMoeHybridForCausalLM(GraniteMoeSharedForCausalLM) - added:
class GraniteMoeHybridModel(GraniteMoeSharedModel) - added:
class GraniteMoeHybridPreTrainedModel(GraniteMoeSharedPreTrainedModel) - added:
constructor GraniteMoeHybridForCausalLM(config: GraniteMoeHybridConfig) - added:
constructor GraniteMoeHybridModel(config: GraniteMoeHybridConfig) - added:
field GraniteMoeHybridPreTrainedModel.config: GraniteMoeHybridConfig - added:
method GraniteMoeHybridForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method GraniteMoeHybridModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_l…
- added:
export define_import_structure
- added:
class GraniteMoeSharedConfig(PretrainedConfig) - added:
constructor GraniteMoeSharedConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, embedding_multip… - added:
field GraniteMoeSharedConfig.keys_to_ignore_at_inference - added:
field GraniteMoeSharedConfig.model_type
- added:
class GraniteMoeSharedForCausalLM(GraniteMoeSharedPreTrainedModel, GenerationMixin) - added:
class GraniteMoeSharedModel(GraniteMoeSharedPreTrainedModel) - added:
class GraniteMoeSharedPreTrainedModel(PreTrainedModel) - added:
constructor GraniteMoeSharedForCausalLM(config: GraniteMoeSharedConfig) - added:
constructor GraniteMoeSharedModel(config: GraniteMoeSharedConfig) - added:
field GraniteMoeSharedPreTrainedModel.base_model_prefix - added:
field GraniteMoeSharedPreTrainedModel.config: GraniteMoeSharedConfig - added:
field GraniteMoeSharedPreTrainedModel.supports_gradient_checkpointing - added:
method GraniteMoeSharedForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method GraniteMoeSharedModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_l…
- added:
class GraniteMoeSharedForCausalLM(GraniteMoeForCausalLM) - added:
class GraniteMoeSharedModel(GraniteMoeModel) - added:
class GraniteMoeSharedPreTrainedModel(GraniteMoePreTrainedModel) - added:
constructor GraniteMoeSharedForCausalLM(config: GraniteMoeSharedConfig) - added:
constructor GraniteMoeSharedModel(config: GraniteMoeSharedConfig) - added:
field GraniteMoeSharedPreTrainedModel.config: GraniteMoeSharedConfig
- added:
export define_import_structure
- added:
class GroundingDinoConfig(PretrainedConfig) - added:
constructor GroundingDinoConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, text_config=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, a… - added:
field GroundingDinoConfig.attribute_map - added:
field GroundingDinoConfig.model_type - added:
property GroundingDinoConfig.hidden_size -> int - added:
property GroundingDinoConfig.num_attention_heads -> int - added:
property GroundingDinoConfig.sub_configs
- added:
class GroundingDinoImageProcessor(BaseImageProcessor) - added:
constructor GroundingDinoImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: Optional[bool]=…, do_pad: bool… - added:
field GroundingDinoImageProcessor.model_input_names - added:
method GroundingDinoImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method GroundingDinoImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_size… - added:
method GroundingDinoImageProcessor.post_process_object_detection(outputs: GroundingDinoObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method GroundingDinoImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method GroundingDinoImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Op… - added:
method GroundingDinoImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method GroundingDinoImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method GroundingDinoImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict - added:
static method GroundingDinoImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class GroundingDinoImageProcessorFast(BaseImageProcessorFast) - added:
constructor GroundingDinoImageProcessorFast(**kwargs: Unpack[GroundingDinoFastImageProcessorKwargs]) -> None - added:
field GroundingDinoImageProcessorFast.default_to_square - added:
field GroundingDinoImageProcessorFast.do_normalize - added:
field GroundingDinoImageProcessorFast.do_pad - added:
field GroundingDinoImageProcessorFast.do_rescale - added:
field GroundingDinoImageProcessorFast.do_resize - added:
field GroundingDinoImageProcessorFast.format - added:
field GroundingDinoImageProcessorFast.image_mean - added:
field GroundingDinoImageProcessorFast.image_std - added:
field GroundingDinoImageProcessorFast.model_input_names - added:
field GroundingDinoImageProcessorFast.resample - added:
field GroundingDinoImageProcessorFast.size - added:
field GroundingDinoImageProcessorFast.valid_kwargs - added:
method GroundingDinoImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method GroundingDinoImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method GroundingDinoImageProcessorFast.post_process_object_detection(outputs: GroundingDinoObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method GroundingDinoImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method GroundingDinoImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[GroundingDinoFastImageProcessorKwargs]) -> BatchFeature - added:
method GroundingDinoImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method GroundingDinoImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…) - added:
static method GroundingDinoImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class GroundingDinoForObjectDetection(GroundingDinoPreTrainedModel) - added:
class GroundingDinoModel(GroundingDinoPreTrainedModel) - added:
class GroundingDinoPreTrainedModel(PreTrainedModel) - added:
constructor GroundingDinoForObjectDetection(config: GroundingDinoConfig) - added:
constructor GroundingDinoModel(config: GroundingDinoConfig) - added:
field GroundingDinoPreTrainedModel.base_model_prefix - added:
field GroundingDinoPreTrainedModel.config: GroundingDinoConfig - added:
field GroundingDinoPreTrainedModel.main_input_name - added:
method GroundingDinoForObjectDetection.forward(pixel_values: torch.FloatTensor, input_ids: torch.LongTensor, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, pixel_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[Union[GroundingDinoEncoderOutput, tuple]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_d… - added:
method GroundingDinoModel.forward(pixel_values: Tensor, input_ids: Tensor, token_type_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, pixel_mask: Optional[Tensor]=…, encoder_outputs=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method GroundingDinoModel.freeze_backbone() - added:
method GroundingDinoModel.generate_encoder_output_proposals(enc_output, padding_mask, spatial_shapes) - added:
method GroundingDinoModel.get_valid_ratio(mask) - added:
method GroundingDinoModel.unfreeze_backbone()
- added:
class GroundingDinoImageProcessorFast(DetrImageProcessorFast) - added:
method GroundingDinoImageProcessorFast.post_process() - added:
method GroundingDinoImageProcessorFast.post_process_instance() - added:
method GroundingDinoImageProcessorFast.post_process_instance_segmentation() - added:
method GroundingDinoImageProcessorFast.post_process_object_detection(outputs: GroundingDinoObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method GroundingDinoImageProcessorFast.post_process_panoptic() - added:
method GroundingDinoImageProcessorFast.post_process_panoptic_segmentation() - added:
method GroundingDinoImageProcessorFast.post_process_segmentation() - added:
method GroundingDinoImageProcessorFast.post_process_semantic_segmentation()
- added:
class GroundingDinoProcessor(ProcessorMixin) - added:
constructor GroundingDinoProcessor(image_processor, tokenizer) - added:
field GroundingDinoProcessor.attributes - added:
field GroundingDinoProcessor.image_processor_class - added:
field GroundingDinoProcessor.tokenizer_class - added:
field GroundingDinoProcessor.valid_processor_kwargs - added:
method GroundingDinoProcessor.post_process_grounded_object_detection(outputs: GroundingDinoObjectDetectionOutput, input_ids: Optional[TensorType]=…, threshold: float=…, text_threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…, text_labels: Optional[list[list[str]]]=…)
- added:
export define_import_structure
- added:
class GroupViTConfig(PretrainedConfig) - added:
class GroupViTOnnxConfig(OnnxConfig) - added:
class GroupViTTextConfig(PretrainedConfig) - added:
class GroupViTVisionConfig(PretrainedConfig) - added:
constructor GroupViTConfig(text_config=…, vision_config=…, projection_dim=…, projection_intermediate_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor GroupViTTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, dropout=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor GroupViTVisionConfig(hidden_size=…, intermediate_size=…, depths=…, num_hidden_layers=…, num_group_tokens=…, num_output_groups=…, num_attention_heads=…, image_size=…, patch_size=…, num_channels=…, hidden_act=…, layer_norm_eps=…, dropout=…, attention_dropout=…, initializer_range=…, initializer_factor=…, assign_eps=…, assign_mlp_ratio=…, **kwargs) - added:
field GroupViTConfig.model_type - added:
field GroupViTConfig.sub_configs - added:
field GroupViTTextConfig.base_config_key - added:
field GroupViTTextConfig.model_type - added:
field GroupViTVisionConfig.base_config_key - added:
field GroupViTVisionConfig.model_type - added:
method GroupViTOnnxConfig.generate_dummy_inputs(processor: ProcessorMixin, batch_size: int=…, seq_length: int=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property GroupViTOnnxConfig.atol_for_validation -> float - added:
property GroupViTOnnxConfig.default_onnx_opset -> int - added:
property GroupViTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property GroupViTOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class GroupViTModel(GroupViTPreTrainedModel) - added:
class GroupViTPreTrainedModel(PreTrainedModel) - added:
class GroupViTTextModel(GroupViTPreTrainedModel) - added:
class GroupViTVisionModel(GroupViTPreTrainedModel) - added:
constructor GroupViTModel(config: GroupViTConfig) - added:
constructor GroupViTTextModel(config: GroupViTTextConfig) - added:
constructor GroupViTVisionModel(config: GroupViTVisionConfig) - added:
field GroupViTModel.config: GroupViTConfig - added:
field GroupViTPreTrainedModel.base_model_prefix - added:
field GroupViTPreTrainedModel.config: GroupViTConfig - added:
field GroupViTPreTrainedModel.supports_gradient_checkpointing - added:
field GroupViTTextModel.config: GroupViTTextConfig - added:
field GroupViTVisionModel.config: GroupViTVisionConfig - added:
field GroupViTVisionModel.main_input_name - added:
method GroupViTModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_segmentation: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Grou… - added:
method GroupViTModel.get_image_features(pixel_values: torch.Tensor) -> torch.FloatTensor - added:
method GroupViTModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method GroupViTTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method GroupViTTextModel.get_input_embeddings() -> nn.Module - added:
method GroupViTTextModel.set_input_embeddings(value) - added:
method GroupViTVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method GroupViTVisionModel.get_input_embeddings() -> GroupViTPatchEmbeddings
- added:
class TFGroupViTModel(TFGroupViTPreTrainedModel) - added:
class TFGroupViTPreTrainedModel(TFPreTrainedModel) - added:
class TFGroupViTTextModel(TFGroupViTPreTrainedModel) - added:
class TFGroupViTVisionModel(TFGroupViTPreTrainedModel) - added:
constructor TFGroupViTModel(config: GroupViTConfig, *inputs, **kwargs) - added:
constructor TFGroupViTTextModel(config: GroupViTTextConfig, *inputs, **kwargs) - added:
constructor TFGroupViTVisionModel(config: GroupViTVisionConfig, *inputs, **kwargs) - added:
field TFGroupViTModel.config_class - added:
field TFGroupViTPreTrainedModel.base_model_prefix - added:
field TFGroupViTPreTrainedModel.config_class - added:
field TFGroupViTTextModel.config_class - added:
field TFGroupViTTextModel.main_input_name - added:
field TFGroupViTVisionModel.config_class - added:
field TFGroupViTVisionModel.main_input_name - added:
method TFGroupViTModel.build(input_shape=…) - added:
method TFGroupViTModel.call(input_ids: TFModelInputType | None=…, pixel_values: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, return_loss: bool | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, output_segmentation: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFGroupViTMo… - added:
method TFGroupViTModel.get_image_features(pixel_values: TFModelInputType | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tf.Tensor - added:
method TFGroupViTModel.get_text_features(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tf.Tensor - added:
method TFGroupViTModel.serving_output(output: TFGroupViTModelOutput) -> TFGroupViTModelOutput - added:
method TFGroupViTTextModel.build(input_shape=…) - added:
method TFGroupViTTextModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor] - added:
method TFGroupViTVisionModel.build(input_shape=…) - added:
method TFGroupViTVisionModel.call(pixel_values: TFModelInputType | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor]
- added:
export define_import_structure
- added:
class HeliumConfig(PretrainedConfig) - added:
constructor HeliumConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, attention_dropout=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, attention_bias=…, mlp_bias=…, **kwargs) - added:
field HeliumConfig.base_model_pp_plan - added:
field HeliumConfig.base_model_tp_plan - added:
field HeliumConfig.keys_to_ignore_at_inference - added:
field HeliumConfig.model_type
- added:
class HeliumForCausalLM(HeliumPreTrainedModel, GenerationMixin) - added:
class HeliumForSequenceClassification(GenericForSequenceClassification, HeliumPreTrainedModel) - added:
class HeliumForTokenClassification(GenericForTokenClassification, HeliumPreTrainedModel) - added:
class HeliumModel(HeliumPreTrainedModel) - added:
class HeliumPreTrainedModel(PreTrainedModel) - added:
constructor HeliumForCausalLM(config) - added:
constructor HeliumModel(config: HeliumConfig) - added:
field HeliumPreTrainedModel.base_model_prefix - added:
field HeliumPreTrainedModel.config: HeliumConfig - added:
field HeliumPreTrainedModel.supports_gradient_checkpointing - added:
method HeliumForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method HeliumModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class HeliumForCausalLM(GemmaForCausalLM) - added:
class HeliumForSequenceClassification(GemmaForSequenceClassification) - added:
class HeliumForTokenClassification(GemmaForTokenClassification) - added:
class HeliumModel(HeliumPreTrainedModel, LlamaModel) - added:
class HeliumPreTrainedModel(LlamaPreTrainedModel) - added:
constructor HeliumModel(config: HeliumConfig)
- added:
export define_import_structure
- added:
class HerbertTokenizer(PreTrainedTokenizer) - added:
constructor HerbertTokenizer(vocab_file, merges_file, tokenizer_file=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sep_token=…, bos_token=…, do_lowercase_and_remove_accent=…, additional_special_tokens=…, lang2id=…, id2lang=…, **kwargs) - added:
field HerbertTokenizer.vocab_files_names - added:
method HerbertTokenizer.bpe(token) - added:
method HerbertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method HerbertTokenizer.convert_tokens_to_string(tokens) - added:
method HerbertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method HerbertTokenizer.get_vocab() - added:
method HerbertTokenizer.ja_tokenize(text) - added:
method HerbertTokenizer.moses_pipeline(text, lang) - added:
method HerbertTokenizer.moses_punct_norm(text, lang) - added:
method HerbertTokenizer.moses_tokenize(text, lang) - added:
method HerbertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property HerbertTokenizer.do_lower_case - added:
property HerbertTokenizer.vocab_size
- added:
class HerbertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor HerbertTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sep_token=…, **kwargs) - added:
field HerbertTokenizerFast.slow_tokenizer_class - added:
field HerbertTokenizerFast.vocab_files_names - added:
method HerbertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method HerbertTokenizerFast.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method HerbertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class HGNetV2Config(BackboneConfigMixin, PretrainedConfig) - added:
constructor HGNetV2Config(num_channels=…, embedding_size=…, depths=…, hidden_sizes=…, hidden_act=…, out_features=…, out_indices=…, stem_channels=…, stage_in_channels=…, stage_mid_channels=…, stage_out_channels=…, stage_num_blocks=…, stage_downsample=…, stage_light_block=…, stage_kernel_size=…, stage_numb_of_layers=…, use_learnable_affine_block=…, initializer_range=…, **kwargs) - added:
field HGNetV2Config.model_type
- added:
class HGNetV2Backbone(HGNetV2PreTrainedModel, BackboneMixin) - added:
class HGNetV2ForImageClassification(HGNetV2PreTrainedModel) - added:
class HGNetV2PreTrainedModel(PreTrainedModel) - added:
constructor HGNetV2Backbone(config: HGNetV2Config) - added:
constructor HGNetV2ForImageClassification(config: HGNetV2Config) - added:
field HGNetV2Backbone.has_attentions - added:
field HGNetV2PreTrainedModel.base_model_prefix - added:
field HGNetV2PreTrainedModel.config: HGNetV2Config - added:
field HGNetV2PreTrainedModel.main_input_name - added:
method HGNetV2Backbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method HGNetV2ForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention
- added:
class HGNetV2Backbone(HGNetV2PreTrainedModel, BackboneMixin) - added:
class HGNetV2Config(BackboneConfigMixin, PretrainedConfig) - added:
class HGNetV2ForImageClassification(HGNetV2PreTrainedModel) - added:
class HGNetV2PreTrainedModel(PreTrainedModel) - added:
constructor HGNetV2Backbone(config: HGNetV2Config) - added:
constructor HGNetV2Config(num_channels=…, embedding_size=…, depths=…, hidden_sizes=…, hidden_act=…, out_features=…, out_indices=…, stem_channels=…, stage_in_channels=…, stage_mid_channels=…, stage_out_channels=…, stage_num_blocks=…, stage_downsample=…, stage_light_block=…, stage_kernel_size=…, stage_numb_of_layers=…, use_learnable_affine_block=…, initializer_range=…, **kwargs) - added:
constructor HGNetV2ForImageClassification(config: HGNetV2Config) - added:
field HGNetV2Backbone.has_attentions - added:
field HGNetV2Config.model_type - added:
field HGNetV2PreTrainedModel.base_model_prefix - added:
field HGNetV2PreTrainedModel.config: HGNetV2Config - added:
field HGNetV2PreTrainedModel.main_input_name - added:
method HGNetV2Backbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method HGNetV2ForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention
- added:
export define_import_structure
- added:
class HieraConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor HieraConfig(embed_dim=…, image_size=…, patch_size=…, patch_stride=…, patch_padding=…, mlp_ratio=…, depths=…, num_heads=…, embed_dim_multiplier=…, num_query_pool=…, query_stride=…, masked_unit_size=…, masked_unit_attention=…, drop_path_rate=…, num_channels=…, hidden_act=…, initializer_range=…, layer_norm_init=…, layer_norm_eps=…, decoder_hidden_size=…, decoder_depth=…, decoder_num_head… - added:
field HieraConfig.attribute_map - added:
field HieraConfig.model_type
- added:
class HieraBackbone(HieraPreTrainedModel, BackboneMixin) - added:
class HieraForImageClassification(HieraPreTrainedModel) - added:
class HieraForPreTraining(HieraPreTrainedModel) - added:
class HieraModel(HieraPreTrainedModel) - added:
class HieraPreTrainedModel(PreTrainedModel) - added:
constructor HieraBackbone(config: HieraConfig) - added:
constructor HieraForImageClassification(config: HieraConfig) -> None - added:
constructor HieraForPreTraining(config: HieraConfig) -> None - added:
constructor HieraModel(config: HieraConfig, add_pooling_layer: bool=…, is_mae: bool=…) - added:
field HieraPreTrainedModel.base_model_prefix - added:
field HieraPreTrainedModel.config: HieraConfig - added:
field HieraPreTrainedModel.main_input_name - added:
field HieraPreTrainedModel.supports_gradient_checkpointing - added:
method HieraBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method HieraBackbone.get_input_embeddings() - added:
method HieraForImageClassification.forward(pixel_values, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, HieraForImageClassificationOutput] - added:
method HieraForPreTraining.forward(pixel_values: Optional[torch.Tensor]=…, noise: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, HieraForPreTrainingOutput] - added:
method HieraForPreTraining.forward_loss(pixel_values: torch.Tensor, logits: torch.Tensor, bool_masked_pos: torch.BoolTensor) - added:
method HieraForPreTraining.get_pixel_label_2d(pixel_values: torch.Tensor, bool_masked_pos: torch.BoolTensor) -> torch.Tensor - added:
method HieraModel.forward(pixel_values: Optional[torch.Tensor]=…, noise: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method HieraModel.get_input_embeddings() -> HieraPatchEmbeddings
- added:
export define_import_structure
- added:
class HubertConfig(PretrainedConfig) - added:
constructor HubertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_layer_norm=…, feat_proj_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel=…, c… - added:
field HubertConfig.model_type - added:
property HubertConfig.inputs_to_logits_ratio
- added:
class HubertForCTC(HubertPreTrainedModel) - added:
class HubertForSequenceClassification(HubertPreTrainedModel) - added:
class HubertModel(HubertPreTrainedModel) - added:
class HubertPreTrainedModel(PreTrainedModel) - added:
constructor HubertForCTC(config, target_lang: Optional[str]=…) - added:
constructor HubertForSequenceClassification(config) - added:
constructor HubertModel(config: HubertConfig) - added:
field HubertPreTrainedModel.base_model_prefix - added:
field HubertPreTrainedModel.config: HubertConfig - added:
field HubertPreTrainedModel.main_input_name - added:
field HubertPreTrainedModel.supports_gradient_checkpointing - added:
method HubertForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method HubertForCTC.freeze_base_model() - added:
method HubertForCTC.freeze_feature_encoder() - added:
method HubertForCTC.freeze_feature_extractor() - added:
method HubertForCTC.tie_weights() - added:
method HubertForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method HubertForSequenceClassification.freeze_base_model() - added:
method HubertForSequenceClassification.freeze_feature_encoder() - added:
method HubertForSequenceClassification.freeze_feature_extractor() - added:
method HubertModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
class TFHubertForCTC(TFHubertPreTrainedModel) - added:
class TFHubertModel(TFHubertPreTrainedModel) - added:
class TFHubertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFHubertForCTC(config: HubertConfig, *inputs, **kwargs) - added:
constructor TFHubertModel(config: HubertConfig, *inputs, **kwargs) - added:
constructor TFHubertPreTrainedModel(config, *inputs, **kwargs) - added:
field TFHubertPreTrainedModel.base_model_prefix - added:
field TFHubertPreTrainedModel.config_class - added:
field TFHubertPreTrainedModel.main_input_name - added:
method TFHubertForCTC.build(input_shape=…) - added:
method TFHubertForCTC.call(input_values: tf.Tensor, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, labels: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFCausalLMOutput | t… - added:
method TFHubertForCTC.freeze_feature_encoder() - added:
method TFHubertForCTC.freeze_feature_extractor() - added:
method TFHubertModel.build(input_shape=…) - added:
method TFHubertModel.call(input_values: tf.Tensor, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutput | tuple[tf.Tensor] - added:
property TFHubertPreTrainedModel.input_signature
- added:
class HubertForCTC(Wav2Vec2ForCTC) - added:
class HubertForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class HubertModel(Wav2Vec2Model, HubertPreTrainedModel) - added:
class HubertPreTrainedModel(PreTrainedModel) - added:
constructor HubertModel(config: HubertConfig) - added:
field HubertPreTrainedModel.base_model_prefix - added:
field HubertPreTrainedModel.config: HubertConfig - added:
field HubertPreTrainedModel.main_input_name - added:
field HubertPreTrainedModel.supports_gradient_checkpointing - added:
method HubertModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method HubertModel.freeze_feature_encoder() - added:
method HubertModel.freeze_feature_extractor()
- added:
export define_import_structure
- added:
class HunYuanDenseV1Config(PretrainedConfig) - added:
constructor HunYuanDenseV1Config(vocab_size=…, hidden_size=…, intermediate_size: int=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, eod_token_id=…, pretraining_tp=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…,… - added:
field HunYuanDenseV1Config.keys_to_ignore_at_inference - added:
field HunYuanDenseV1Config.model_type
- added:
class HunYuanDenseV1ForCausalLM(HunYuanDenseV1PreTrainedModel, GenerationMixin) - added:
class HunYuanDenseV1ForSequenceClassification(GenericForSequenceClassification, HunYuanDenseV1PreTrainedModel) - added:
class HunYuanDenseV1Model(HunYuanDenseV1PreTrainedModel) - added:
class HunYuanDenseV1PreTrainedModel(PreTrainedModel) - added:
constructor HunYuanDenseV1ForCausalLM(config) - added:
constructor HunYuanDenseV1Model(config: HunYuanDenseV1Config) - added:
field HunYuanDenseV1PreTrainedModel.base_model_prefix - added:
field HunYuanDenseV1PreTrainedModel.config: HunYuanDenseV1Config - added:
field HunYuanDenseV1PreTrainedModel.supports_gradient_checkpointing - added:
method HunYuanDenseV1ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Ten… - added:
method HunYuanDenseV1Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class HunYuanDenseV1ForCausalLM(LlamaForCausalLM) - added:
class HunYuanDenseV1ForSequenceClassification(LlamaForSequenceClassification) - added:
class HunYuanDenseV1Model(LlamaModel) - added:
class HunYuanDenseV1PreTrainedModel(LlamaPreTrainedModel)
- added:
export define_import_structure
- added:
class HunYuanMoEV1Config(PretrainedConfig) - added:
constructor HunYuanMoEV1Config(vocab_size=…, hidden_size=…, intermediate_size: int=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, eod_token_id=…, sep_token_id=…, pretraining_tp=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, atte… - added:
field HunYuanMoEV1Config.keys_to_ignore_at_inference - added:
field HunYuanMoEV1Config.model_type
- added:
class HunYuanMoEV1ForCausalLM(HunYuanMoEV1PreTrainedModel, GenerationMixin) - added:
class HunYuanMoEV1ForSequenceClassification(GenericForSequenceClassification, HunYuanMoEV1PreTrainedModel) - added:
class HunYuanMoEV1Model(HunYuanMoEV1PreTrainedModel) - added:
class HunYuanMoEV1PreTrainedModel(PreTrainedModel) - added:
constructor HunYuanMoEV1ForCausalLM(config) - added:
constructor HunYuanMoEV1Model(config: HunYuanMoEV1Config) - added:
field HunYuanMoEV1PreTrainedModel.base_model_prefix - added:
field HunYuanMoEV1PreTrainedModel.config: HunYuanMoEV1Config - added:
field HunYuanMoEV1PreTrainedModel.supports_gradient_checkpointing - added:
method HunYuanMoEV1ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tenso… - added:
method HunYuanMoEV1Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class HunYuanMoEV1ForCausalLM(LlamaForCausalLM) - added:
class HunYuanMoEV1ForSequenceClassification(LlamaForSequenceClassification) - added:
class HunYuanMoEV1Model(LlamaModel) - added:
class HunYuanMoEV1PreTrainedModel(LlamaPreTrainedModel)
- added:
export define_import_structure
- added:
class IBertConfig(PretrainedConfig) - added:
class IBertOnnxConfig(OnnxConfig) - added:
constructor IBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, quant_mode=…, force_dequant=…, **kwargs) - added:
field IBertConfig.model_type - added:
property IBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class IBertForMaskedLM(IBertPreTrainedModel) - added:
class IBertForMultipleChoice(IBertPreTrainedModel) - added:
class IBertForQuestionAnswering(IBertPreTrainedModel) - added:
class IBertForSequenceClassification(IBertPreTrainedModel) - added:
class IBertForTokenClassification(IBertPreTrainedModel) - added:
class IBertModel(IBertPreTrainedModel) - added:
class IBertPreTrainedModel(PreTrainedModel) - added:
constructor IBertForMaskedLM(config) - added:
constructor IBertForMultipleChoice(config) - added:
constructor IBertForQuestionAnswering(config) - added:
constructor IBertForSequenceClassification(config) - added:
constructor IBertForTokenClassification(config) - added:
constructor IBertModel(config, add_pooling_layer=…) - added:
field IBertPreTrainedModel.base_model_prefix - added:
field IBertPreTrainedModel.config: IBertConfig - added:
method IBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Opti… - added:
method IBertForMaskedLM.get_output_embeddings() - added:
method IBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method IBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states… - added:
method IBertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…,… - added:
method IBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidde… - added:
method IBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_s… - added:
method IBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…)… - added:
method IBertModel.get_input_embeddings() - added:
method IBertModel.set_input_embeddings(value) - added:
method IBertPreTrainedModel.resize_token_embeddings(new_num_tokens=…)
- added:
class FixedPointMul(Function) - added:
class IntGELU(nn.Module) - added:
class IntLayerNorm(nn.Module) - added:
class IntSoftmax(nn.Module) - added:
class QuantAct(nn.Module) - added:
class QuantEmbedding(nn.Module) - added:
class QuantLinear(nn.Module) - added:
class SymmetricQuantFunction(Function) - added:
class floor_ste(Function) - added:
class round_ste(Function) - added:
constructor IntGELU(quant_mode=…, force_dequant=…) - added:
constructor IntLayerNorm(normalized_shape, eps, output_bit=…, quant_mode=…, force_dequant=…) - added:
constructor IntSoftmax(output_bit, quant_mode=…, force_dequant=…) - added:
constructor QuantAct(activation_bit, act_range_momentum=…, per_channel=…, channel_len=…, quant_mode=…) - added:
constructor QuantEmbedding(num_embeddings, embedding_dim, padding_idx=…, max_norm=…, norm_type=…, scale_grad_by_freq=…, sparse=…, _weight=…, weight_bit=…, momentum=…, quant_mode=…) - added:
constructor QuantLinear(in_features, out_features, bias=…, weight_bit=…, bias_bit=…, per_channel=…, quant_mode=…) - added:
function batch_frexp(inputs, max_bit=…) - added:
function get_percentile_min_max(input, lower_percentile, upper_percentile, output_tensor=…) - added:
function linear_quantize(input, scale, zero_point, inplace=…) - added:
function symmetric_linear_quantization_params(num_bits, saturation_min, saturation_max, per_channel=…) - added:
method IntGELU.forward(x, scaling_factor=…) - added:
method IntGELU.int_erf(x_int, scaling_factor) - added:
method IntLayerNorm.forward(x, scaling_factor=…) - added:
method IntLayerNorm.overflow_fallback(y_int) - added:
method IntLayerNorm.set_shift(y_int) - added:
method IntSoftmax.forward(x, scaling_factor) - added:
method IntSoftmax.int_exp(x_int, scaling_factor) - added:
method IntSoftmax.int_polynomial(x_int, scaling_factor) - added:
method QuantAct.forward(x, pre_act_scaling_factor=…, identity=…, identity_scaling_factor=…, specified_min=…, specified_max=…) - added:
method QuantEmbedding.forward(x, positions=…, incremental_state=…) - added:
method QuantLinear.forward(x, prev_act_scaling_factor=…) - added:
static method FixedPointMul.backward(ctx, grad_output) - added:
static method FixedPointMul.forward(ctx, pre_act, pre_act_scaling_factor, bit_num, z_scaling_factor, identity=…, identity_scaling_factor=…) - added:
static method SymmetricQuantFunction.backward(ctx, grad_output) - added:
static method SymmetricQuantFunction.forward(ctx, x, k, percentile_mode, scale) - added:
static method floor_ste.backward(ctx, grad_output) - added:
static method floor_ste.forward(ctx, x) - added:
static method round_ste.backward(ctx, grad_output) - added:
static method round_ste.forward(ctx, x) - added:
variable logger
- added:
export define_import_structure
- added:
class IdeficsConfig(PretrainedConfig) - added:
constructor IdeficsConfig(vocab_size=…, additional_vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, dropout=…, hidden_act=…, initializer_range=…, alpha_initializer=…, alphas_initializer_range=…, alpha_type=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, cross_layer_interval=…, qk_layer_norms=…, fre… - added:
field IdeficsConfig.model_type - added:
field IdeficsConfig.sub_configs
- added:
class IdeficsImageProcessor(BaseImageProcessor) - added:
constructor IdeficsImageProcessor(image_size: int=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, image_num_channels: Optional[int]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, **kwargs) -> None - added:
field IdeficsImageProcessor.model_input_names - added:
method IdeficsImageProcessor.preprocess(images: ImageInput, image_num_channels: Optional[int]=…, image_size: Optional[dict[str, int]]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, transform: Optional[Callable]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, return_tensors: Optional[Union[str, TensorType]]=…, **kwargs) -…
- added:
class IdeficsForVisionText2Text(IdeficsPreTrainedModel, GenerationMixin) - added:
class IdeficsModel(IdeficsPreTrainedModel) - added:
class IdeficsPreTrainedModel(PreTrainedModel) - added:
constructor IdeficsForVisionText2Text(config, vision_model=…) - added:
constructor IdeficsModel(config: IdeficsConfig) - added:
field IdeficsPreTrainedModel.base_model_prefix - added:
field IdeficsPreTrainedModel.config: IdeficsConfig - added:
field IdeficsPreTrainedModel.supports_gradient_checkpointing - added:
method IdeficsForVisionText2Text.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_encoder_embeddings: Optional[torch.FloatTensor]=…, perceiver_embeddings: Optional[torch.FloatTenso… - added:
method IdeficsForVisionText2Text.prepare_inputs_for_generation(input_ids, attention_mask=…, position_ids=…, inputs_embeds=…, past_key_values=…, cache_position=…, pixel_values=…, image_hidden_states=…, image_attention_mask=…, use_cache=…, **kwargs) - added:
method IdeficsForVisionText2Text.tie_weights() - added:
method IdeficsModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_encoder_embeddings: Optional[torch.FloatTensor]=…, perceiver_embeddings: Optional[torch.FloatTensor]=…, image_a… - added:
method IdeficsModel.freeze_relevant_params(config=…) - added:
method IdeficsModel.freeze_text_layers(module_exceptions=…) - added:
method IdeficsModel.freeze_vision_layers(module_exceptions=…)
- added:
class TFIdeficsForVisionText2Text(TFPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFIdeficsModel(TFIdeficsPreTrainedModel) - added:
class TFIdeficsPreTrainedModel(TFPreTrainedModel) - added:
constructor TFIdeficsForVisionText2Text(config, vision_model=…, **kwargs) - added:
constructor TFIdeficsModel(config: IdeficsConfig, *inputs, **kwargs) - added:
field TFIdeficsForVisionText2Text.config_class - added:
field TFIdeficsPreTrainedModel.base_model_prefix - added:
field TFIdeficsPreTrainedModel.config_class - added:
field TFIdeficsPreTrainedModel.supports_gradient_checkpointing - added:
method TFIdeficsForVisionText2Text.build(input_shape=…) - added:
method TFIdeficsForVisionText2Text.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, past_key_values: list[tf.Tensor] | None=…, inputs_embeds: tf.Tensor | None=…, pixel_values: tf.Tensor | None=…, image_encoder_embeddings: tf.Tensor | None=…, perceiver_embeddings: tf.Tensor | None=…, image_attention_mask: tf.Tensor | None=…, labels: t… - added:
method TFIdeficsForVisionText2Text.prepare_inputs_for_generation(input_ids, past=…, **kwargs) - added:
method TFIdeficsForVisionText2Text.tie_weights() - added:
method TFIdeficsModel.build(input_shape=…) - added:
method TFIdeficsModel.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, past_key_values: list[tf.Tensor] | None=…, inputs_embeds: tf.Tensor | None=…, pixel_values: tf.Tensor | None=…, image_encoder_embeddings: tf.Tensor | None=…, perceiver_embeddings: tf.Tensor | None=…, image_attention_mask: tf.Tensor | None=…, use_cache: bool | None…
- added:
class IdeficsMLP(nn.Module) - added:
class IdeficsPerceiverAttention(nn.Module) - added:
class IdeficsPerceiverResampler(nn.Module) - added:
constructor IdeficsMLP(intermediate_size, config: IdeficsConfig) - added:
constructor IdeficsPerceiverAttention(embed_dim: int, n_heads: int, head_dim: int, qk_layer_norms: bool) -> None - added:
constructor IdeficsPerceiverResampler(config: IdeficsConfig, embed_dim: int, depth: int, n_heads: int, head_dim: int, n_latents: int) -> None - added:
method IdeficsMLP.forward(hidden_states: Optional[tuple[torch.FloatTensor]]) -> torch.FloatTensor - added:
method IdeficsPerceiverAttention.forward(context: torch.Tensor, latents: torch.Tensor) -> torch.Tensor - added:
method IdeficsPerceiverResampler.forward(context: torch.Tensor) -> torch.Tensor
- added:
class TFIdeficsMLP(tf.keras.layers.Layer) - added:
class TFIdeficsPerceiverAttention(tf.keras.layers.Layer) - added:
class TFIdeficsPerceiverResampler(tf.keras.layers.Layer) - added:
constructor TFIdeficsMLP(intermediate_size, config: IdeficsConfig, **kwargs) - added:
constructor TFIdeficsPerceiverAttention(embed_dim: int, n_heads: int, head_dim: int, qk_layer_norms: bool, **kwargs) -> None - added:
constructor TFIdeficsPerceiverResampler(config: IdeficsConfig, embed_dim: int, depth: int, n_heads: int, head_dim: int, n_latents: int, **kwargs) -> None - added:
method TFIdeficsMLP.call(hidden_states: Optional[tuple[tf.Tensor]]) -> tf.Tensor - added:
method TFIdeficsPerceiverAttention.call(context: tf.Tensor, latents: tf.Tensor) -> tf.Tensor - added:
method TFIdeficsPerceiverResampler.build(input_shape) - added:
method TFIdeficsPerceiverResampler.call(context: tf.Tensor) -> tf.Tensor
- added:
class IdeficsProcessor(ProcessorMixin) - added:
constructor IdeficsProcessor(image_processor, tokenizer=…, image_size=…, add_end_of_utterance_token=…, **kwargs) - added:
field IdeficsProcessor.attributes - added:
field IdeficsProcessor.image_processor_class - added:
field IdeficsProcessor.tokenizer_class - added:
property IdeficsProcessor.model_input_names
- added:
class IdeficsVisionAttention(nn.Module) - added:
class IdeficsVisionEmbeddings(nn.Module) - added:
class IdeficsVisionEncoder(nn.Module) - added:
class IdeficsVisionEncoderLayer(GradientCheckpointingLayer) - added:
class IdeficsVisionMLP(nn.Module) - added:
class IdeficsVisionModelOutput(ModelOutput) - added:
class IdeficsVisionTransformer(nn.Module) - added:
constructor IdeficsVisionAttention(config: IdeficsVisionConfig) - added:
constructor IdeficsVisionEmbeddings(config: IdeficsVisionConfig) - added:
constructor IdeficsVisionEncoder(config: IdeficsVisionConfig) - added:
constructor IdeficsVisionEncoderLayer(config: IdeficsVisionConfig) - added:
constructor IdeficsVisionMLP(config) - added:
constructor IdeficsVisionTransformer(config: IdeficsVisionConfig) - added:
field IdeficsVisionModelOutput.attentions: Optional[tuple[torch.FloatTensor, ...]] - added:
field IdeficsVisionModelOutput.hidden_states: Optional[tuple[torch.FloatTensor, ...]] - added:
field IdeficsVisionModelOutput.image_embeds: Optional[torch.FloatTensor] - added:
field IdeficsVisionModelOutput.last_hidden_state: Optional[torch.FloatTensor] - added:
function eager_attention_forward(module: nn.Module, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: Optional[torch.Tensor], scaling: float, dropout: float=…, **kwargs) - added:
method IdeficsVisionAttention.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, causal_attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…) -> tuple[torch.Tensor, Optional[torch.Tensor]] - added:
method IdeficsVisionEmbeddings.forward(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.Tensor - added:
method IdeficsVisionEmbeddings.interpolate_pos_encoding(embeddings: torch.Tensor, height: int, width: int) -> torch.Tensor - added:
method IdeficsVisionEncoder.forward(inputs_embeds, attention_mask: Optional[torch.Tensor]=…, causal_attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method IdeficsVisionEncoderLayer.forward(hidden_states: torch.Tensor, attention_mask: torch.Tensor, causal_attention_mask: torch.Tensor, output_attentions: Optional[bool]=…) -> tuple[torch.FloatTensor] - added:
method IdeficsVisionMLP.forward(hidden_states: torch.Tensor) -> torch.Tensor - added:
method IdeficsVisionTransformer.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
variable logger
- added:
class TFIdeficsVisionAttention(tf.keras.layers.Layer) - added:
class TFIdeficsVisionEmbeddings(tf.keras.layers.Layer) - added:
class TFIdeficsVisionEncoder(tf.keras.layers.Layer) - added:
class TFIdeficsVisionEncoderLayer(tf.keras.layers.Layer) - added:
class TFIdeficsVisionMLP(tf.keras.layers.Layer) - added:
class TFIdeficsVisionModelOutput(ModelOutput) - added:
class TFIdeficsVisionTransformer(TFPreTrainedModel) - added:
constructor TFIdeficsVisionAttention(config, **kwargs) - added:
constructor TFIdeficsVisionEmbeddings(config: IdeficsVisionConfig, **kwargs) - added:
constructor TFIdeficsVisionEncoder(config: IdeficsVisionConfig, **kwargs) - added:
constructor TFIdeficsVisionEncoderLayer(config: IdeficsVisionConfig, **kwargs) - added:
constructor TFIdeficsVisionMLP(config, **kwargs) - added:
constructor TFIdeficsVisionTransformer(config: IdeficsVisionConfig, **kwargs) - added:
field TFIdeficsVisionModelOutput.attentions: Optional[tuple[tf.Tensor]] - added:
field TFIdeficsVisionModelOutput.hidden_states: Optional[tuple[tf.Tensor]] - added:
field TFIdeficsVisionModelOutput.image_embeds: Optional[tf.Tensor] - added:
field TFIdeficsVisionModelOutput.last_hidden_state: Optional[tf.Tensor] - added:
method TFIdeficsVisionAttention.build(input_shape=…) - added:
method TFIdeficsVisionAttention.call(hidden_states: tf.Tensor, attention_mask: Optional[tf.Tensor]=…, causal_attention_mask: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…) -> tuple[tf.Tensor, Optional[tf.Tensor], Optional[tuple[tf.Tensor]]] - added:
method TFIdeficsVisionEmbeddings.build(input_shape=…) - added:
method TFIdeficsVisionEmbeddings.call(pixel_values: tf.Tensor, interpolate_pos_encoding: bool=…) -> tf.Tensor - added:
method TFIdeficsVisionEmbeddings.interpolate_pos_encoding(embeddings: tf.Tensor, height: int, width: int) -> tf.Tensor - added:
method TFIdeficsVisionEncoder.build(input_shape=…) - added:
method TFIdeficsVisionEncoder.call(inputs_embeds, attention_mask: Optional[tf.Tensor]=…, causal_attention_mask: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: Optional[bool]=…) -> Union[tuple, TFBaseModelOutput] - added:
method TFIdeficsVisionEncoderLayer.build(input_shape=…) - added:
method TFIdeficsVisionEncoderLayer.call(hidden_states: tf.Tensor, attention_mask: tf.Tensor, causal_attention_mask: tf.Tensor, output_attentions: Optional[bool]=…) -> tuple[tf.Tensor] - added:
method TFIdeficsVisionMLP.build(input_shape=…) - added:
method TFIdeficsVisionMLP.call(hidden_states: tf.Tensor) -> tf.Tensor - added:
method TFIdeficsVisionTransformer.build(input_shape=…) - added:
method TFIdeficsVisionTransformer.call(pixel_values: Optional[tf.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…, training: Optional[bool]=…) -> Union[tuple, TFBaseModelOutputWithPooling] - added:
variable logger
- added:
export define_import_structure
- added:
class Idefics2Config(PretrainedConfig) - added:
constructor Idefics2Config(use_cache=…, image_token_id=…, tie_word_embeddings=…, vision_config=…, perceiver_config=…, text_config=…, **kwargs) - added:
field Idefics2Config.model_type - added:
field Idefics2Config.sub_configs
- added:
class Idefics2ImageProcessor(BaseImageProcessor) - added:
constructor Idefics2ImageProcessor(do_convert_rgb: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, do_image_splitting: bool=…, **kwargs) -> None - added:
field Idefics2ImageProcessor.model_input_names - added:
method Idefics2ImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method Idefics2ImageProcessor.preprocess(images: ImageInput, do_convert_rgb: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, d… - added:
method Idefics2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Idefics2ImageProcessor.split_image(image: np.ndarray, input_data_format: Optional[Union[str, ChannelDimension]]=…)
- added:
class Idefics2ImageProcessorFast(BaseImageProcessorFast) - added:
field Idefics2ImageProcessorFast.do_convert_rgb - added:
field Idefics2ImageProcessorFast.do_image_splitting - added:
field Idefics2ImageProcessorFast.do_normalize - added:
field Idefics2ImageProcessorFast.do_pad - added:
field Idefics2ImageProcessorFast.do_rescale - added:
field Idefics2ImageProcessorFast.do_resize - added:
field Idefics2ImageProcessorFast.image_mean - added:
field Idefics2ImageProcessorFast.image_std - added:
field Idefics2ImageProcessorFast.model_input_names - added:
field Idefics2ImageProcessorFast.resample - added:
field Idefics2ImageProcessorFast.size - added:
field Idefics2ImageProcessorFast.valid_kwargs - added:
method Idefics2ImageProcessorFast.convert_to_rgb(image: ImageInput) -> ImageInput - added:
method Idefics2ImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], fill: int=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Idefics2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Idefics2FastImageProcessorKwargs]) -> BatchFeature - added:
method Idefics2ImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method Idefics2ImageProcessorFast.split_images(images: torch.Tensor) -> list[torch.Tensor]
- added:
class Idefics2ForConditionalGeneration(Idefics2PreTrainedModel, GenerationMixin) - added:
class Idefics2Model(Idefics2PreTrainedModel) - added:
class Idefics2PreTrainedModel(PreTrainedModel) - added:
constructor Idefics2ForConditionalGeneration(config) - added:
constructor Idefics2Model(config: Idefics2Config) - added:
field Idefics2PreTrainedModel.base_model_prefix - added:
field Idefics2PreTrainedModel.config: Idefics2Config - added:
field Idefics2PreTrainedModel.supports_gradient_checkpointing - added:
method Idefics2ForConditionalGeneration.disable_input_require_grads() - added:
method Idefics2ForConditionalGeneration.enable_input_require_grads() - added:
method Idefics2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTens… - added:
method Idefics2ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method Idefics2ForConditionalGeneration.get_input_embeddings() - added:
method Idefics2ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, pixel_values=…, pixel_attention_mask=…, image_hidden_states=…, logits_to_keep=…, **kwargs) - added:
method Idefics2ForConditionalGeneration.set_input_embeddings(value) - added:
method Idefics2Model.disable_input_require_grads() - added:
method Idefics2Model.enable_input_require_grads() - added:
method Idefics2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTensor]=…, use_cache: O… - added:
method Idefics2Model.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method Idefics2Model.get_input_embeddings() - added:
method Idefics2Model.inputs_merger(input_ids: torch.LongTensor, inputs_embeds: Optional[torch.Tensor], image_hidden_states: Optional[torch.Tensor]) - added:
method Idefics2Model.set_input_embeddings(value)
- added:
class Idefics2Processor(ProcessorMixin) - added:
constructor Idefics2Processor(image_processor, tokenizer=…, image_seq_len: int=…, chat_template: Optional[str]=…, **kwargs) - added:
field Idefics2Processor.attributes - added:
field Idefics2Processor.image_processor_class - added:
field Idefics2Processor.tokenizer_class
- added:
export define_import_structure
- added:
class Idefics3Config(PretrainedConfig) - added:
class Idefics3VisionConfig(PretrainedConfig) - added:
constructor Idefics3Config(use_cache=…, image_token_id=…, tie_word_embeddings=…, vision_config=…, text_config=…, scale_factor=…, pad_token_id=…, **kwargs) - added:
constructor Idefics3VisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, **kwargs) - added:
field Idefics3Config.model_type - added:
field Idefics3Config.sub_configs - added:
field Idefics3VisionConfig.base_config_key - added:
field Idefics3VisionConfig.model_type
- added:
class Idefics3ImageProcessor(BaseImageProcessor) - added:
constructor Idefics3ImageProcessor(do_convert_rgb: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_image_splitting: bool=…, max_image_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…,… - added:
field Idefics3ImageProcessor.model_input_names - added:
method Idefics3ImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Idefics3ImageProcessor.pad(images: list[list[np.ndarray]], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method Idefics3ImageProcessor.preprocess(images: ImageInput, do_convert_rgb: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_image_splitting: Optional[bool]=…, do_rescale: Optional[bool]=…, max_image_size: Optional[dict[str, int]]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional… - added:
method Idefics3ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Idefics3ImageProcessor.resize_for_vision_encoder(image: np.ndarray, vision_encoder_max_size: int, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method Idefics3ImageProcessor.split_image(image, max_image_size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…)
- added:
class Idefics3ImageProcessorFast(BaseImageProcessorFast) - added:
field Idefics3ImageProcessorFast.do_convert_rgb - added:
field Idefics3ImageProcessorFast.do_image_splitting - added:
field Idefics3ImageProcessorFast.do_normalize - added:
field Idefics3ImageProcessorFast.do_pad - added:
field Idefics3ImageProcessorFast.do_rescale - added:
field Idefics3ImageProcessorFast.do_resize - added:
field Idefics3ImageProcessorFast.image_mean - added:
field Idefics3ImageProcessorFast.image_std - added:
field Idefics3ImageProcessorFast.max_image_size - added:
field Idefics3ImageProcessorFast.resample - added:
field Idefics3ImageProcessorFast.return_row_col_info - added:
field Idefics3ImageProcessorFast.size - added:
field Idefics3ImageProcessorFast.valid_kwargs - added:
method Idefics3ImageProcessorFast.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Idefics3ImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], fill: int=…, return_pixel_mask: bool=…) - added:
method Idefics3ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Idefics3FastImageProcessorKwargs]) -> BatchFeature - added:
method Idefics3ImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method Idefics3ImageProcessorFast.resize_for_vision_encoder(image: torch.Tensor, vision_encoder_max_size: int, interpolation: Optional[F.InterpolationMode]=…) - added:
method Idefics3ImageProcessorFast.split_images(images: torch.Tensor, max_image_size: dict[str, int], interpolation: Optional[F.InterpolationMode]=…) - added:
method Idefics3ImageProcessorFast.to_dict()
- added:
class Idefics3ForConditionalGeneration(Idefics3PreTrainedModel, GenerationMixin) - added:
class Idefics3Model(Idefics3PreTrainedModel) - added:
class Idefics3PreTrainedModel(PreTrainedModel) - added:
class Idefics3VisionTransformer(Idefics3PreTrainedModel) - added:
constructor Idefics3ForConditionalGeneration(config) - added:
constructor Idefics3Model(config: Idefics3Config) - added:
constructor Idefics3VisionTransformer(config: Idefics3VisionConfig) - added:
field Idefics3PreTrainedModel.base_model_prefix - added:
field Idefics3PreTrainedModel.config: Idefics3Config - added:
field Idefics3PreTrainedModel.supports_gradient_checkpointing - added:
field Idefics3VisionTransformer.config: Idefics3VisionConfig - added:
method Idefics3ForConditionalGeneration.disable_input_require_grads() - added:
method Idefics3ForConditionalGeneration.enable_input_require_grads() - added:
method Idefics3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTens… - added:
method Idefics3ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method Idefics3ForConditionalGeneration.get_input_embeddings() - added:
method Idefics3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, pixel_values=…, pixel_attention_mask=…, image_hidden_states=…, logits_to_keep=…, **kwargs) - added:
method Idefics3ForConditionalGeneration.set_input_embeddings(value) - added:
method Idefics3Model.disable_input_require_grads() - added:
method Idefics3Model.enable_input_require_grads() - added:
method Idefics3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTensor]=…, use_cache: O… - added:
method Idefics3Model.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method Idefics3Model.get_input_embeddings() - added:
method Idefics3Model.inputs_merger(input_ids: torch.LongTensor, inputs_embeds: Optional[torch.Tensor], image_hidden_states: Optional[torch.Tensor]) - added:
method Idefics3Model.set_input_embeddings(value) - added:
method Idefics3VisionTransformer.forward(pixel_values, patch_attention_mask: Optional[torch.BoolTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutput] - added:
method Idefics3VisionTransformer.get_input_embeddings() - added:
method Idefics3VisionTransformer.set_input_embeddings(value)
- added:
class Idefics3Processor(ProcessorMixin) - added:
constructor Idefics3Processor(image_processor, tokenizer=…, image_seq_len: int=…, chat_template: Optional[str]=…, **kwargs) - added:
field Idefics3Processor.attributes - added:
field Idefics3Processor.image_processor_class - added:
field Idefics3Processor.tokenizer_class
- added:
export define_import_structure
- added:
class IJepaConfig(PretrainedConfig) - added:
constructor IJepaConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, pooler_output_size=…, pooler_act=…, **kwargs) - added:
field IJepaConfig.model_type
- added:
class IJepaForImageClassification(IJepaPreTrainedModel) - added:
class IJepaModel(IJepaPreTrainedModel) - added:
class IJepaPreTrainedModel(PreTrainedModel) - added:
constructor IJepaForImageClassification(config: IJepaConfig) - added:
constructor IJepaModel(config: IJepaConfig, add_pooling_layer: bool=…, use_mask_token: bool=…) - added:
field IJepaPreTrainedModel.base_model_prefix - added:
field IJepaPreTrainedModel.config: IJepaConfig - added:
field IJepaPreTrainedModel.main_input_name - added:
field IJepaPreTrainedModel.supports_gradient_checkpointing - added:
method IJepaForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method IJepaModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method IJepaModel.get_input_embeddings() -> IJepaPatchEmbeddings
- added:
class IJepaForImageClassification(IJepaPreTrainedModel, ViTForImageClassification) - added:
class IJepaModel(IJepaPreTrainedModel, ViTModel) - added:
class IJepaPreTrainedModel(ViTPreTrainedModel) - added:
constructor IJepaForImageClassification(config: IJepaConfig) - added:
constructor IJepaModel(config: IJepaConfig, add_pooling_layer: bool=…, use_mask_token: bool=…) - added:
method IJepaForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput
- added:
export define_import_structure
- added:
class ImageGPTConfig(PretrainedConfig) - added:
class ImageGPTOnnxConfig(OnnxConfig) - added:
constructor ImageGPTConfig(vocab_size=…, n_positions=…, n_embd=…, n_layer=…, n_head=…, n_inner=…, activation_function=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, scale_attn_weights=…, use_cache=…, tie_word_embeddings=…, scale_attn_by_inverse_layer_idx=…, reorder_and_upcast_attn=…, **kwargs) - added:
field ImageGPTConfig.attribute_map - added:
field ImageGPTConfig.keys_to_ignore_at_inference - added:
field ImageGPTConfig.model_type - added:
method ImageGPTOnnxConfig.generate_dummy_inputs(preprocessor: FeatureExtractionMixin, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…) -> Mapping[str, Any] - added:
property ImageGPTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ImageGPTFeatureExtractor(ImageGPTImageProcessor) - added:
constructor ImageGPTFeatureExtractor(*args, **kwargs) -> None
- added:
class ImageGPTImageProcessor(BaseImageProcessor) - added:
constructor ImageGPTImageProcessor(clusters: Optional[Union[list[list[int]], np.ndarray]]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_normalize: bool=…, do_color_quantize: bool=…, **kwargs) -> None - added:
field ImageGPTImageProcessor.model_input_names - added:
method ImageGPTImageProcessor.normalize(image: np.ndarray, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method ImageGPTImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_normalize: Optional[bool]=…, do_color_quantize: Optional[bool]=…, clusters: Optional[Union[list[list[int]], np.ndarray]]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[Union[str, ChannelDimension]]=…,… - added:
method ImageGPTImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method ImageGPTImageProcessor.to_dict()
- added:
class ImageGPTImageProcessorFast(BaseImageProcessorFast) - added:
constructor ImageGPTImageProcessorFast(clusters: Optional[Union[list, np.ndarray, torch.Tensor]]=…, **kwargs: Unpack[ImageGPTFastImageProcessorKwargs]) - added:
field ImageGPTImageProcessorFast.clusters - added:
field ImageGPTImageProcessorFast.do_color_quantize - added:
field ImageGPTImageProcessorFast.do_normalize - added:
field ImageGPTImageProcessorFast.do_rescale - added:
field ImageGPTImageProcessorFast.image_mean - added:
field ImageGPTImageProcessorFast.image_std - added:
field ImageGPTImageProcessorFast.model_input_names - added:
field ImageGPTImageProcessorFast.resample - added:
field ImageGPTImageProcessorFast.valid_kwargs - added:
method ImageGPTImageProcessorFast.to_dict()
- added:
class ImageGPTForCausalImageModeling(ImageGPTPreTrainedModel, GenerationMixin) - added:
class ImageGPTForImageClassification(ImageGPTPreTrainedModel) - added:
class ImageGPTModel(ImageGPTPreTrainedModel) - added:
class ImageGPTPreTrainedModel(PreTrainedModel) - added:
constructor ImageGPTForCausalImageModeling(config: ImageGPTConfig) - added:
constructor ImageGPTForImageClassification(config: ImageGPTConfig) - added:
constructor ImageGPTModel(config: ImageGPTConfig) - added:
constructor ImageGPTPreTrainedModel(*inputs, **kwargs) - added:
field ImageGPTPreTrainedModel.base_model_prefix - added:
field ImageGPTPreTrainedModel.config: ImageGPTConfig - added:
field ImageGPTPreTrainedModel.load_tf_weights - added:
field ImageGPTPreTrainedModel.main_input_name - added:
field ImageGPTPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_imagegpt(model, config, imagegpt_checkpoint_path) - added:
method ImageGPTForCausalImageModeling.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optio… - added:
method ImageGPTForImageClassification.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentio… - added:
method ImageGPTModel.forward(input_ids: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]… - added:
method ImageGPTModel.get_input_embeddings() - added:
method ImageGPTModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class InformerConfig(PretrainedConfig) - added:
constructor InformerConfig(prediction_length: Optional[int]=…, context_length: Optional[int]=…, distribution_output: str=…, loss: str=…, input_size: int=…, lags_sequence: Optional[list[int]]=…, scaling: Optional[Union[str, bool]]=…, num_dynamic_real_features: int=…, num_static_real_features: int=…, num_static_categorical_features: int=…, num_time_features: int=…, cardinality: Optional[list[int]]=… - added:
field InformerConfig.attribute_map - added:
field InformerConfig.model_type
- added:
class InformerForPrediction(InformerPreTrainedModel) - added:
class InformerModel(InformerPreTrainedModel) - added:
class InformerPreTrainedModel(PreTrainedModel) - added:
constructor InformerForPrediction(config: InformerConfig) - added:
constructor InformerModel(config: InformerConfig) - added:
field InformerPreTrainedModel.base_model_prefix - added:
field InformerPreTrainedModel.config: InformerConfig - added:
field InformerPreTrainedModel.main_input_name - added:
field InformerPreTrainedModel.supports_gradient_checkpointing - added:
method InformerForPrediction.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, future_observed_mask: Optional[torch.Tensor]=…, decoder_attention_mask: Opt… - added:
method InformerForPrediction.generate(past_values: torch.Tensor, past_time_features: torch.Tensor, future_time_features: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> SampleTSPredictionOutput - added:
method InformerForPrediction.get_decoder() - added:
method InformerForPrediction.get_encoder() - added:
method InformerForPrediction.output_distribution(params, loc=…, scale=…, trailing_n=…) -> torch.distributions.Distribution - added:
method InformerForPrediction.output_params(dec_output) - added:
method InformerModel.create_network_inputs(past_values: torch.Tensor, past_time_features: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, past_observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…) - added:
method InformerModel.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Ten… - added:
method InformerModel.get_encoder() - added:
method InformerModel.get_lagged_subsequences(sequence: torch.Tensor, subsequences_length: int, shift: int=…) -> torch.Tensor
- added:
class InformerForPrediction(TimeSeriesTransformerForPrediction) - added:
class InformerModel(TimeSeriesTransformerModel) - added:
class InformerPreTrainedModel(PreTrainedModel) - added:
constructor InformerForPrediction(config: InformerConfig) - added:
constructor InformerModel(config: InformerConfig) - added:
field InformerPreTrainedModel.base_model_prefix - added:
field InformerPreTrainedModel.config: InformerConfig - added:
field InformerPreTrainedModel.main_input_name - added:
field InformerPreTrainedModel.supports_gradient_checkpointing - added:
method InformerForPrediction.forward(**super_kwargs) - added:
method InformerModel.forward(**super_kwargs)
- added:
export define_import_structure
- added:
class InstructBlipConfig(PretrainedConfig) - added:
class InstructBlipQFormerConfig(PretrainedConfig) - added:
class InstructBlipVisionConfig(PretrainedConfig) - added:
constructor InstructBlipConfig(vision_config=…, qformer_config=…, text_config=…, num_query_tokens=…, image_token_index=…, **kwargs) - added:
constructor InstructBlipQFormerConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, cross_attention_frequency=…, encoder_hidden_size=…, **kwargs) - added:
constructor InstructBlipVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, **kwargs) - added:
field InstructBlipConfig.attribute_map - added:
field InstructBlipConfig.model_type - added:
field InstructBlipConfig.sub_configs - added:
field InstructBlipQFormerConfig.base_config_key - added:
field InstructBlipQFormerConfig.model_type - added:
field InstructBlipVisionConfig.base_config_key - added:
field InstructBlipVisionConfig.model_type - added:
static method InstructBlipConfig.from_vision_qformer_text_configs(vision_config: InstructBlipVisionConfig, qformer_config: InstructBlipQFormerConfig, text_config: PretrainedConfig, **kwargs)
- added:
class InstructBlipForConditionalGeneration(InstructBlipPreTrainedModel, GenerationMixin) - added:
class InstructBlipModel(InstructBlipPreTrainedModel) - added:
class InstructBlipPreTrainedModel(PreTrainedModel) - added:
class InstructBlipQFormerModel(InstructBlipPreTrainedModel) - added:
class InstructBlipVisionModel(InstructBlipPreTrainedModel) - added:
constructor InstructBlipForConditionalGeneration(config: InstructBlipConfig) - added:
constructor InstructBlipModel(config: InstructBlipConfig) - added:
constructor InstructBlipQFormerModel(config: InstructBlipQFormerConfig) - added:
constructor InstructBlipVisionModel(config: InstructBlipVisionConfig) - added:
field InstructBlipForConditionalGeneration.config: InstructBlipConfig - added:
field InstructBlipForConditionalGeneration.main_input_name - added:
field InstructBlipModel.main_input_name - added:
field InstructBlipPreTrainedModel.base_model_prefix - added:
field InstructBlipPreTrainedModel.config: InstructBlipConfig - added:
field InstructBlipPreTrainedModel.supports_gradient_checkpointing - added:
field InstructBlipVisionModel.config: InstructBlipVisionConfig - added:
field InstructBlipVisionModel.main_input_name - added:
method InstructBlipForConditionalGeneration.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.F… - added:
method InstructBlipForConditionalGeneration.generate(pixel_values: torch.FloatTensor, qformer_input_ids: Optional[torch.LongTensor]=…, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method InstructBlipForConditionalGeneration.get_decoder() - added:
method InstructBlipForConditionalGeneration.get_encoder() - added:
method InstructBlipForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, qformer_input_ids: torch.LongTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method InstructBlipForConditionalGeneration.get_input_embeddings() - added:
method InstructBlipForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method InstructBlipForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method InstructBlipForConditionalGeneration.set_input_embeddings(value) - added:
method InstructBlipForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method InstructBlipModel.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, interpola… - added:
method InstructBlipModel.get_input_embeddings() - added:
method InstructBlipModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method InstructBlipModel.set_input_embeddings(value) - added:
method InstructBlipQFormerModel.forward(input_ids: torch.LongTensor, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, query_embeds: Optional[torch.Tensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[t… - added:
method InstructBlipQFormerModel.get_extended_attention_mask(attention_mask: torch.Tensor, input_shape: tuple[int], device: torch.device, has_query: bool=…) -> torch.Tensor - added:
method InstructBlipQFormerModel.get_input_embeddings() - added:
method InstructBlipQFormerModel.set_input_embeddings(value) - added:
method InstructBlipVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPooling] - added:
method InstructBlipVisionModel.get_input_embeddings()
- added:
class InstructBlipProcessor(ProcessorMixin) - added:
constructor InstructBlipProcessor(image_processor, tokenizer, qformer_tokenizer, num_query_tokens=…, **kwargs) - added:
field InstructBlipProcessor.attributes - added:
field InstructBlipProcessor.image_processor_class - added:
field InstructBlipProcessor.qformer_tokenizer_class - added:
field InstructBlipProcessor.tokenizer_class - added:
method InstructBlipProcessor.save_pretrained(save_directory, **kwargs) - added:
property InstructBlipProcessor.model_input_names - added:
static method InstructBlipProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
export define_import_structure
- added:
class InstructBlipVideoConfig(PretrainedConfig) - added:
class InstructBlipVideoQFormerConfig(PretrainedConfig) - added:
class InstructBlipVideoVisionConfig(PretrainedConfig) - added:
constructor InstructBlipVideoConfig(vision_config=…, qformer_config=…, text_config=…, num_query_tokens=…, video_token_index=…, **kwargs) - added:
constructor InstructBlipVideoQFormerConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, cross_attention_frequency=…, encoder_hidden_size=…, **kwargs) - added:
constructor InstructBlipVideoVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, **kwargs) - added:
field InstructBlipVideoConfig.attribute_map - added:
field InstructBlipVideoConfig.model_type - added:
field InstructBlipVideoConfig.sub_configs - added:
field InstructBlipVideoQFormerConfig.base_config_key - added:
field InstructBlipVideoQFormerConfig.model_type - added:
field InstructBlipVideoVisionConfig.base_config_key - added:
field InstructBlipVideoVisionConfig.model_type - added:
static method InstructBlipVideoConfig.from_vision_qformer_text_configs(vision_config: InstructBlipVideoVisionConfig, qformer_config: InstructBlipVideoQFormerConfig, text_config: PretrainedConfig, **kwargs)
- added:
class InstructBlipVideoImageProcessor(BaseImageProcessor) - added:
constructor InstructBlipVideoImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field InstructBlipVideoImageProcessor.model_input_names - added:
method InstructBlipVideoImageProcessor.preprocess(images: Optional[VideoInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors… - added:
method InstructBlipVideoImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class InstructBlipVideoForConditionalGeneration(InstructBlipVideoPreTrainedModel, GenerationMixin) - added:
class InstructBlipVideoModel(InstructBlipVideoPreTrainedModel) - added:
class InstructBlipVideoPreTrainedModel(PreTrainedModel) - added:
class InstructBlipVideoQFormerModel(InstructBlipVideoPreTrainedModel) - added:
class InstructBlipVideoVisionModel(InstructBlipVideoPreTrainedModel) - added:
constructor InstructBlipVideoForConditionalGeneration(config: InstructBlipVideoConfig) - added:
constructor InstructBlipVideoModel(config: InstructBlipVideoConfig) - added:
constructor InstructBlipVideoQFormerModel(config: InstructBlipVideoQFormerConfig) - added:
constructor InstructBlipVideoVisionModel(config: InstructBlipVideoVisionConfig) - added:
field InstructBlipVideoForConditionalGeneration.config: InstructBlipVideoConfig - added:
field InstructBlipVideoForConditionalGeneration.main_input_name - added:
field InstructBlipVideoModel.main_input_name - added:
field InstructBlipVideoPreTrainedModel.base_model_prefix - added:
field InstructBlipVideoPreTrainedModel.config: InstructBlipVideoConfig - added:
field InstructBlipVideoPreTrainedModel.supports_gradient_checkpointing - added:
field InstructBlipVideoVisionModel.config: InstructBlipVideoVisionConfig - added:
field InstructBlipVideoVisionModel.main_input_name - added:
method InstructBlipVideoForConditionalGeneration.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[to… - added:
method InstructBlipVideoForConditionalGeneration.generate(pixel_values: torch.FloatTensor, qformer_input_ids: Optional[torch.LongTensor]=…, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method InstructBlipVideoForConditionalGeneration.get_decoder() - added:
method InstructBlipVideoForConditionalGeneration.get_encoder() - added:
method InstructBlipVideoForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, qformer_input_ids: torch.LongTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method InstructBlipVideoForConditionalGeneration.get_input_embeddings() - added:
method InstructBlipVideoForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method InstructBlipVideoForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method InstructBlipVideoForConditionalGeneration.get_video_features(pixel_values: torch.FloatTensor, qformer_input_ids: torch.LongTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method InstructBlipVideoForConditionalGeneration.set_input_embeddings(value) - added:
method InstructBlipVideoForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method InstructBlipVideoModel.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, outp… - added:
method InstructBlipVideoModel.get_input_embeddings() - added:
method InstructBlipVideoModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method InstructBlipVideoModel.set_input_embeddings(value) - added:
method InstructBlipVideoQFormerModel.forward(input_ids: torch.LongTensor, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, query_embeds: Optional[torch.Tensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Un… - added:
method InstructBlipVideoQFormerModel.get_extended_attention_mask(attention_mask: torch.Tensor, input_shape: tuple[int], device: torch.device, has_query: bool=…) -> torch.Tensor - added:
method InstructBlipVideoQFormerModel.get_input_embeddings() - added:
method InstructBlipVideoQFormerModel.set_input_embeddings(value) - added:
method InstructBlipVideoVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPooling] - added:
method InstructBlipVideoVisionModel.get_input_embeddings()
- added:
class InstructBlipVideoConfig(PretrainedConfig) - added:
class InstructBlipVideoForConditionalGeneration(InstructBlipForConditionalGeneration) - added:
class InstructBlipVideoModel(InstructBlipModel) - added:
class InstructBlipVideoPreTrainedModel(InstructBlipPreTrainedModel) - added:
class InstructBlipVideoQFormerConfig(InstructBlipQFormerConfig) - added:
class InstructBlipVideoQFormerModel(InstructBlipQFormerModel) - added:
class InstructBlipVideoVisionConfig(InstructBlipVisionConfig) - added:
class InstructBlipVideoVisionModel(InstructBlipVisionModel) - added:
constructor InstructBlipVideoConfig(vision_config=…, qformer_config=…, text_config=…, num_query_tokens=…, video_token_index=…, **kwargs) - added:
field InstructBlipVideoConfig.attribute_map - added:
field InstructBlipVideoConfig.model_type - added:
field InstructBlipVideoConfig.sub_configs - added:
method InstructBlipVideoForConditionalGeneration.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[to… - added:
method InstructBlipVideoForConditionalGeneration.generate(pixel_values: torch.FloatTensor, qformer_input_ids: Optional[torch.LongTensor]=…, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **generate_kwargs) -> torch.LongTensor - added:
method InstructBlipVideoForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, qformer_input_ids: torch.LongTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method InstructBlipVideoForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor) - added:
method InstructBlipVideoForConditionalGeneration.get_video_features(pixel_values: torch.FloatTensor, qformer_input_ids: torch.LongTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, interpolate_pos_encoding: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method InstructBlipVideoModel.forward(pixel_values: torch.FloatTensor, qformer_input_ids: torch.FloatTensor, qformer_attention_mask: Optional[torch.LongTensor]=…, input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, outp… - added:
static method InstructBlipVideoConfig.from_vision_qformer_text_configs(vision_config: InstructBlipVideoVisionConfig, qformer_config: InstructBlipVideoQFormerConfig, text_config: PretrainedConfig, **kwargs)
- added:
class InstructBlipVideoProcessor(ProcessorMixin) - added:
constructor InstructBlipVideoProcessor(video_processor, tokenizer, qformer_tokenizer, num_query_tokens=…, **kwargs) - added:
field InstructBlipVideoProcessor.attributes - added:
field InstructBlipVideoProcessor.qformer_tokenizer_class - added:
field InstructBlipVideoProcessor.tokenizer_class - added:
field InstructBlipVideoProcessor.video_processor_class - added:
method InstructBlipVideoProcessor.save_pretrained(save_directory, **kwargs) - added:
property InstructBlipVideoProcessor.model_input_names - added:
static method InstructBlipVideoProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
class InstructBlipVideoVideoProcessor(BaseVideoProcessor) - added:
constructor InstructBlipVideoVideoProcessor(**kwargs: Unpack[InstructBlipVideoVideoProcessorInitKwargs]) - added:
field InstructBlipVideoVideoProcessor.default_to_square - added:
field InstructBlipVideoVideoProcessor.do_convert_rgb - added:
field InstructBlipVideoVideoProcessor.do_normalize - added:
field InstructBlipVideoVideoProcessor.do_rescale - added:
field InstructBlipVideoVideoProcessor.do_resize - added:
field InstructBlipVideoVideoProcessor.do_sample_frames - added:
field InstructBlipVideoVideoProcessor.image_mean - added:
field InstructBlipVideoVideoProcessor.image_std - added:
field InstructBlipVideoVideoProcessor.model_input_names - added:
field InstructBlipVideoVideoProcessor.resample - added:
field InstructBlipVideoVideoProcessor.size - added:
field InstructBlipVideoVideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class InternVLConfig(PretrainedConfig) - added:
class InternVLVisionConfig(PretrainedConfig) - added:
constructor InternVLConfig(vision_config=…, text_config=…, image_token_id=…, image_seq_length=…, downsample_ratio=…, projector_hidden_act=…, vision_feature_layer=…, vision_feature_select_strategy=…, **kwargs) - added:
constructor InternVLVisionConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, attention_bias=…, use_qk_norm=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_dropout=…, projection_dropout=…, initializer_range=…, norm_type=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, use_mask_token=…, use_absolute_position_embeddings=…, layer_scale_init_value=… - added:
field InternVLConfig.model_type - added:
field InternVLConfig.sub_configs - added:
field InternVLVisionConfig.base_config_key - added:
field InternVLVisionConfig.model_type
- added:
class InternVLForConditionalGeneration(InternVLPreTrainedModel, GenerationMixin) - added:
class InternVLModel(InternVLPreTrainedModel) - added:
class InternVLPreTrainedModel(PreTrainedModel) - added:
class InternVLVisionModel(InternVLVisionPreTrainedModel) - added:
class InternVLVisionPreTrainedModel(PreTrainedModel) - added:
constructor InternVLForConditionalGeneration(config: InternVLConfig) - added:
constructor InternVLModel(config: InternVLConfig) - added:
constructor InternVLVisionModel(config: InternVLVisionConfig) -> None - added:
field InternVLPreTrainedModel.base_model_prefix - added:
field InternVLPreTrainedModel.config: InternVLConfig - added:
field InternVLPreTrainedModel.supports_gradient_checkpointing - added:
field InternVLVisionPreTrainedModel.base_model_prefix - added:
field InternVLVisionPreTrainedModel.config: InternVLVisionConfig - added:
field InternVLVisionPreTrainedModel.main_input_name - added:
field InternVLVisionPreTrainedModel.supports_gradient_checkpointing - added:
method InternVLForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional… - added:
method InternVLForConditionalGeneration.get_decoder() - added:
method InternVLForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method InternVLForConditionalGeneration.get_input_embeddings() - added:
method InternVLForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method InternVLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method InternVLForConditionalGeneration.set_decoder(decoder) - added:
method InternVLForConditionalGeneration.set_input_embeddings(value) - added:
method InternVLModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, cache_posi… - added:
method InternVLModel.get_decoder() - added:
method InternVLModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method InternVLModel.get_input_embeddings() - added:
method InternVLModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method InternVLModel.pixel_shuffle(vision_features: torch.Tensor, scale_factor: float=…) - added:
method InternVLModel.set_decoder(decoder) - added:
method InternVLModel.set_input_embeddings(value) - added:
method InternVLVisionModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…) -> Union[tuple, InternVLVisionModelOutputWithPooling] - added:
method InternVLVisionModel.get_input_embeddings() - added:
property InternVLForConditionalGeneration.language_model - added:
property InternVLForConditionalGeneration.multi_modal_projector - added:
property InternVLForConditionalGeneration.vision_tower
- added:
class InternVLForConditionalGeneration(LlavaForConditionalGeneration) - added:
class InternVLModel(LlavaModel) - added:
class InternVLPreTrainedModel(LlavaPreTrainedModel) - added:
class InternVLVisionModel(InternVLVisionPreTrainedModel) - added:
class InternVLVisionPreTrainedModel(PreTrainedModel) - added:
constructor InternVLVisionModel(config: InternVLVisionConfig) -> None - added:
field InternVLVisionPreTrainedModel.base_model_prefix - added:
field InternVLVisionPreTrainedModel.config: InternVLVisionConfig - added:
field InternVLVisionPreTrainedModel.main_input_name - added:
field InternVLVisionPreTrainedModel.supports_gradient_checkpointing - added:
method InternVLForConditionalGeneration.forward(**super_kwargs) - added:
method InternVLModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, cache_posi… - added:
method InternVLModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method InternVLModel.pixel_shuffle(vision_features: torch.Tensor, scale_factor: float=…) - added:
method InternVLVisionModel.forward(pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…) -> Union[tuple, InternVLVisionModelOutputWithPooling] - added:
method InternVLVisionModel.get_input_embeddings()
- added:
class InternVLProcessor(ProcessorMixin) - added:
constructor InternVLProcessor(image_processor=…, tokenizer=…, video_processor=…, image_seq_length: int=…, chat_template=…, **kwargs) - added:
field InternVLProcessor.attributes - added:
field InternVLProcessor.image_processor_class - added:
field InternVLProcessor.tokenizer_class - added:
field InternVLProcessor.video_processor_class - added:
property InternVLProcessor.model_input_names
- added:
class InternVLVideoProcessor(BaseVideoProcessor) - added:
constructor InternVLVideoProcessor(**kwargs: Unpack[InternVLVideoProcessorInitKwargs]) - added:
field InternVLVideoProcessor.do_convert_rgb - added:
field InternVLVideoProcessor.do_normalize - added:
field InternVLVideoProcessor.do_rescale - added:
field InternVLVideoProcessor.do_resize - added:
field InternVLVideoProcessor.do_sample_frames - added:
field InternVLVideoProcessor.image_mean - added:
field InternVLVideoProcessor.image_std - added:
field InternVLVideoProcessor.initial_shift - added:
field InternVLVideoProcessor.model_input_names - added:
field InternVLVideoProcessor.resample - added:
field InternVLVideoProcessor.size - added:
field InternVLVideoProcessor.valid_kwargs - added:
method InternVLVideoProcessor.sample_frames(metadata: VideoMetadata, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, initial_shift: Optional[Union[bool, float, int]]=…, **kwargs)
- added:
export define_import_structure
- added:
class JambaConfig(PretrainedConfig) - added:
constructor JambaConfig(vocab_size=…, tie_word_embeddings=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, num_logits_to_keep=…, output_router_logits=…, router_aux_loss_coef=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, sliding_window=…, max_position_embeddings=…, attentio… - added:
field JambaConfig.keys_to_ignore_at_inference - added:
field JambaConfig.model_type - added:
property JambaConfig.layers_block_type - added:
property JambaConfig.layers_num_experts
- added:
class JambaForCausalLM(JambaPreTrainedModel, GenerationMixin) - added:
class JambaForSequenceClassification(GenericForSequenceClassification, JambaPreTrainedModel) - added:
class JambaModel(JambaPreTrainedModel) - added:
class JambaPreTrainedModel(PreTrainedModel) - added:
constructor JambaForCausalLM(config: JambaConfig) - added:
constructor JambaModel(config: JambaConfig) - added:
field JambaPreTrainedModel.base_model_prefix - added:
field JambaPreTrainedModel.config: JambaConfig - added:
field JambaPreTrainedModel.supports_gradient_checkpointing - added:
method JambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optio… - added:
method JambaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, output_router_logits=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method JambaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[HybridMambaAttentionDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Optional[…
- added:
export define_import_structure
- added:
class JanusConfig(PretrainedConfig) - added:
class JanusVQVAEConfig(PretrainedConfig) - added:
class JanusVisionConfig(PretrainedConfig) - added:
constructor JanusConfig(text_config=…, vision_config=…, vq_config=…, image_token_id=…, **kwargs) - added:
constructor JanusVQVAEConfig(embed_dim: int=…, num_embeddings: int=…, double_latent: bool=…, latent_channels: int=…, num_patches: int=…, in_channels: int=…, out_channels: int=…, base_channels: int=…, channel_multiplier: list[int]=…, num_res_blocks: int=…, dropout: float=…, initializer_range=…, projection_dim=…, num_hidden_layers=…, hidden_act=…, image_token_embed_dim=…, **kwargs) - added:
constructor JanusVisionConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, patch_size=…, image_size=…, attention_dropout=…, layer_norm_eps=…, hidden_act=…, mlp_ratio=…, attention_bias=…, hidden_dropout_rate=…, projection_dim=…, projection_dropout=…, use_qk_norm=…, initializer_range=…, depth=…, num_image_tokens=…, **kwargs) - added:
field JanusConfig.model_type - added:
field JanusConfig.sub_configs - added:
field JanusVQVAEConfig.base_config_key - added:
field JanusVQVAEConfig.model_type - added:
field JanusVisionConfig.base_config_key - added:
field JanusVisionConfig.model_type
- added:
class JanusImageProcessor(BaseImageProcessor) - added:
constructor JanusImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, min_size: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, do_pad: Optional[bool]=…, **kwargs) -> None - added:
field JanusImageProcessor.model_input_names - added:
method JanusImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method JanusImageProcessor.postprocess(images: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[list[float]]=…, image_std: Optional[list[float]]=…, input_data_format: Optional[str]=…, return_tensors: Optional[str]=…) - added:
method JanusImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, Te… - added:
method JanusImageProcessor.resize(image: np.ndarray, size: Union[dict[str, int], int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method JanusImageProcessor.unnormalize(image: np.ndarray, image_mean: Union[float, Iterable[float]], image_std: Union[float, Iterable[float]], input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray
- added:
class JanusImageProcessorFast(BaseImageProcessorFast) - added:
constructor JanusImageProcessorFast(**kwargs: Unpack[JanusFastImageProcessorKwargs]) - added:
field JanusImageProcessorFast.do_normalize - added:
field JanusImageProcessorFast.do_pad - added:
field JanusImageProcessorFast.do_rescale - added:
field JanusImageProcessorFast.do_resize - added:
field JanusImageProcessorFast.image_mean - added:
field JanusImageProcessorFast.image_std - added:
field JanusImageProcessorFast.min_size - added:
field JanusImageProcessorFast.resample - added:
field JanusImageProcessorFast.size - added:
field JanusImageProcessorFast.valid_kwargs - added:
method JanusImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method JanusImageProcessorFast.postprocess(images: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[list[float]]=…, image_std: Optional[list[float]]=…, return_tensors: Optional[str]=…) -> torch.Tensor - added:
method JanusImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, min_size: int, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor
- added:
class JanusForConditionalGeneration(JanusPreTrainedModel, GenerationMixin) - added:
class JanusModel(JanusPreTrainedModel) - added:
class JanusPreTrainedModel(PreTrainedModel) - added:
class JanusVQVAE(JanusPreTrainedModel) - added:
class JanusVisionModel(JanusPreTrainedModel) - added:
constructor JanusForConditionalGeneration(config: JanusConfig) - added:
constructor JanusModel(config: JanusConfig) - added:
constructor JanusVQVAE(config: JanusVQVAEConfig) - added:
constructor JanusVisionModel(config: JanusVisionConfig) - added:
field JanusPreTrainedModel.base_model_prefix - added:
field JanusPreTrainedModel.config: JanusConfig - added:
field JanusPreTrainedModel.supports_gradient_checkpointing - added:
field JanusVQVAE.config: JanusVQVAEConfig - added:
field JanusVQVAE.main_input_name - added:
field JanusVisionModel.config: JanusVisionConfig - added:
field JanusVisionModel.main_input_name - added:
method JanusForConditionalGeneration.decode_image_tokens(image_tokens: torch.Tensor) - added:
method JanusForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optio… - added:
method JanusForConditionalGeneration.generate(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.LongTensor]=…, logits_processor: Optional[LogitsProcessorList]=…, **kwargs) - added:
method JanusForConditionalGeneration.get_input_embeddings() - added:
method JanusForConditionalGeneration.prepare_embeddings_for_image_generation(inputs: torch.Tensor) -> torch.Tensor - added:
method JanusForConditionalGeneration.prepare_inputs_for_generation(input_ids, pixel_values=…, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method JanusForConditionalGeneration.set_input_embeddings(value) - added:
method JanusModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method JanusModel.get_image_features(pixel_values) - added:
method JanusModel.get_input_embeddings() - added:
method JanusModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method JanusModel.set_input_embeddings(value) - added:
method JanusVQVAE.decode(image_tokens: torch.LongTensor) -> torch.FloatTensor - added:
method JanusVQVAE.encode(pixel_values: torch.LongTensor) - added:
method JanusVQVAE.forward(pixel_values: torch.FloatTensor) -> tuple[torch.FloatTensor, torch.FloatTensor] - added:
method JanusVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPooling] - added:
method JanusVisionModel.get_input_embeddings()
- added:
class JanusConfig(PretrainedConfig) - added:
class JanusForConditionalGeneration(JanusPreTrainedModel, GenerationMixin) - added:
class JanusImageProcessor(BlipImageProcessor) - added:
class JanusModel(JanusPreTrainedModel) - added:
class JanusPreTrainedModel(PreTrainedModel) - added:
class JanusVQVAE(ChameleonVQVAE) - added:
class JanusVQVAEConfig(ChameleonVQVAEConfig) - added:
class JanusVisionConfig(SiglipVisionConfig) - added:
class JanusVisionModel(Blip2VisionModel) - added:
constructor JanusConfig(text_config=…, vision_config=…, vq_config=…, image_token_id=…, **kwargs) - added:
constructor JanusForConditionalGeneration(config: JanusConfig) - added:
constructor JanusImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, min_size: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, do_pad: Optional[bool]=…, **kwargs) - added:
constructor JanusModel(config: JanusConfig) - added:
constructor JanusVQVAE(config: JanusVQVAEConfig) - added:
constructor JanusVQVAEConfig(embed_dim: int=…, num_embeddings: int=…, double_latent: bool=…, latent_channels: int=…, num_patches: int=…, in_channels: int=…, out_channels: int=…, base_channels: int=…, channel_multiplier: list[int]=…, num_res_blocks: int=…, dropout: float=…, initializer_range=…, projection_dim=…, num_hidden_layers=…, hidden_act=…, image_token_embed_dim=…, **kwargs) - added:
constructor JanusVisionConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, patch_size=…, image_size=…, attention_dropout=…, layer_norm_eps=…, hidden_act=…, mlp_ratio=…, attention_bias=…, hidden_dropout_rate=…, projection_dim=…, projection_dropout=…, use_qk_norm=…, initializer_range=…, depth=…, num_image_tokens=…, **kwargs) - added:
constructor JanusVisionModel(config: JanusVisionConfig) - added:
field JanusConfig.model_type - added:
field JanusConfig.sub_configs - added:
field JanusPreTrainedModel.base_model_prefix - added:
field JanusPreTrainedModel.config: JanusConfig - added:
field JanusPreTrainedModel.supports_gradient_checkpointing - added:
field JanusVQVAE.main_input_name - added:
field JanusVisionConfig.base_config_key - added:
field JanusVisionConfig.model_type - added:
method JanusForConditionalGeneration.decode_image_tokens(image_tokens: torch.Tensor) - added:
method JanusForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optio… - added:
method JanusForConditionalGeneration.generate(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.LongTensor]=…, logits_processor: Optional[LogitsProcessorList]=…, **kwargs) - added:
method JanusForConditionalGeneration.get_input_embeddings() - added:
method JanusForConditionalGeneration.prepare_embeddings_for_image_generation(inputs: torch.Tensor) -> torch.Tensor - added:
method JanusForConditionalGeneration.prepare_inputs_for_generation(input_ids, pixel_values=…, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method JanusForConditionalGeneration.set_input_embeddings(value) - added:
method JanusImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method JanusImageProcessor.postprocess(images: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[list[float]]=…, image_std: Optional[list[float]]=…, input_data_format: Optional[str]=…, return_tensors: Optional[str]=…) - added:
method JanusImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, Te… - added:
method JanusImageProcessor.resize(image: np.ndarray, size: Union[dict[str, int], int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method JanusImageProcessor.unnormalize(image: np.ndarray, image_mean: Union[float, Iterable[float]], image_std: Union[float, Iterable[float]], input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method JanusModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method JanusModel.get_image_features(pixel_values) - added:
method JanusModel.get_input_embeddings() - added:
method JanusModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method JanusModel.set_input_embeddings(value) - added:
method JanusVQVAE.decode(image_tokens: torch.LongTensor) -> torch.FloatTensor - added:
method JanusVQVAE.forward(pixel_values: torch.FloatTensor) -> tuple[torch.FloatTensor, torch.FloatTensor]
- added:
class JanusProcessor(ProcessorMixin) - added:
constructor JanusProcessor(image_processor, tokenizer, chat_template=…, use_default_system_prompt=…, **kwargs) - added:
field JanusProcessor.attributes - added:
field JanusProcessor.image_processor_class - added:
field JanusProcessor.tokenizer_class - added:
method JanusProcessor.postprocess(images: ImageInput, **kwargs)
- added:
export define_import_structure
- added:
class JetMoeConfig(PretrainedConfig) - added:
constructor JetMoeConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_key_value_heads=…, kv_channels=…, intermediate_size=…, max_position_embeddings=…, activation_function=…, num_local_experts=…, num_experts_per_tok=…, output_router_logits=…, aux_loss_coef=…, use_cache=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rms_norm_eps=…, initializer_range=…, attention_d… - added:
field JetMoeConfig.attribute_map - added:
field JetMoeConfig.keys_to_ignore_at_inference - added:
field JetMoeConfig.model_type
- added:
class JetMoeForCausalLM(JetMoePreTrainedModel, GenerationMixin) - added:
class JetMoeForSequenceClassification(GenericForSequenceClassification, JetMoePreTrainedModel) - added:
class JetMoeModel(JetMoePreTrainedModel) - added:
class JetMoePreTrainedModel(PreTrainedModel) - added:
constructor JetMoeForCausalLM(config) - added:
constructor JetMoeModel(config: JetMoeConfig) - added:
field JetMoePreTrainedModel.base_model_prefix - added:
field JetMoePreTrainedModel.config: JetMoeConfig - added:
field JetMoePreTrainedModel.supports_gradient_checkpointing - added:
method JetMoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router… - added:
method JetMoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Opt…
- added:
export define_import_structure
- added:
class Kosmos2Config(PretrainedConfig) - added:
constructor Kosmos2Config(text_config=…, vision_config=…, latent_query_num=…, **kwargs) - added:
field Kosmos2Config.model_type - added:
field Kosmos2Config.sub_configs
- added:
class Kosmos2ForConditionalGeneration(Kosmos2PreTrainedModel, GenerationMixin) - added:
class Kosmos2Model(Kosmos2PreTrainedModel) - added:
class Kosmos2PreTrainedModel(PreTrainedModel) - added:
constructor Kosmos2ForConditionalGeneration(config: Kosmos2Config) - added:
constructor Kosmos2Model(config: Kosmos2Config) - added:
field Kosmos2ForConditionalGeneration.config: Kosmos2Config - added:
field Kosmos2ForConditionalGeneration.main_input_name - added:
field Kosmos2Model.config: Kosmos2Config - added:
field Kosmos2Model.main_input_name - added:
field Kosmos2PreTrainedModel.config: Kosmos2Config - added:
field Kosmos2PreTrainedModel.supports_gradient_checkpointing - added:
method Kosmos2ForConditionalGeneration.forward(pixel_values: Optional[torch.Tensor]=…, input_ids: Optional[torch.Tensor]=…, image_embeds_position_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, image_embeds: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, position_ids: Optional[to… - added:
method Kosmos2ForConditionalGeneration.generate(pixel_values: Optional[torch.Tensor]=…, image_embeds_position_mask: Optional[torch.Tensor]=…, input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, image_embeds: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, **kwargs) - added:
method Kosmos2ForConditionalGeneration.get_input_embeddings() -> nn.Module - added:
method Kosmos2ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Kosmos2ForConditionalGeneration.set_input_embeddings(value) - added:
method Kosmos2ForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method Kosmos2Model.forward(pixel_values: Optional[torch.Tensor]=…, input_ids: Optional[torch.Tensor]=…, image_embeds_position_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, image_embeds: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, use_… - added:
method Kosmos2Model.get_image_features(pixel_values: torch.FloatTensor, return_attentions: Optional[bool]=…, interpolate_pos_encoding: Optional[bool]=…) - added:
method Kosmos2Model.get_input_embeddings() -> nn.Module - added:
method Kosmos2Model.set_input_embeddings(value)
- added:
class Kosmos2Processor(ProcessorMixin) - added:
constructor Kosmos2Processor(image_processor, tokenizer, num_patch_index_tokens=…, *kwargs) - added:
field Kosmos2Processor.attributes - added:
field Kosmos2Processor.image_processor_class - added:
field Kosmos2Processor.tokenizer_class - added:
method Kosmos2Processor.post_process_generation(text, cleanup_and_extract=…) - added:
method Kosmos2Processor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, **kwargs) - added:
method Kosmos2Processor.preprocess_examples(texts: Union[TextInput, list[TextInput]], images: Optional[ImageInput]=…, bboxes: BboxInput=…, num_image_tokens: Optional[int]=…) -> Union[str, list[str]] - added:
property Kosmos2Processor.model_input_names
- added:
export define_import_structure
- added:
class Kosmos2_5Config(PretrainedConfig) - added:
constructor Kosmos2_5Config(text_config=…, vision_config=…, latent_query_num=…, **kwargs) - added:
field Kosmos2_5Config.model_type - added:
field Kosmos2_5Config.sub_configs
- added:
class Kosmos2_5ImageProcessor(BaseImageProcessor) - added:
constructor Kosmos2_5ImageProcessor(do_convert_rgb: bool=…, do_normalize: bool=…, patch_size: Optional[dict[str, int]]=…, max_patches: int=…, **kwargs) -> None - added:
field Kosmos2_5ImageProcessor.model_input_names - added:
method Kosmos2_5ImageProcessor.extract_flattened_patches(image: np.ndarray, max_patches: int, patch_size: dict, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Kosmos2_5ImageProcessor.normalize(image: np.ndarray, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Kosmos2_5ImageProcessor.preprocess(images: ImageInput, do_convert_rgb: Optional[bool]=…, do_normalize: Optional[bool]=…, max_patches: Optional[int]=…, patch_size: Optional[dict[str, int]]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> ImageInput
- added:
class Kosmos2_5ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Kosmos2_5ImageProcessorFast(**kwargs: Unpack[Kosmos2_5FastImageProcessorKwargs]) - added:
field Kosmos2_5ImageProcessorFast.do_convert_rgb - added:
field Kosmos2_5ImageProcessorFast.do_normalize - added:
field Kosmos2_5ImageProcessorFast.max_patches - added:
field Kosmos2_5ImageProcessorFast.patch_size - added:
field Kosmos2_5ImageProcessorFast.rescale_factor - added:
field Kosmos2_5ImageProcessorFast.valid_kwargs - added:
method Kosmos2_5ImageProcessorFast.extract_flattened_patches(image: torch.Tensor, max_patches: int, patch_size: dict) -> torch.Tensor - added:
method Kosmos2_5ImageProcessorFast.normalize(image: torch.Tensor, **kwargs) -> torch.Tensor - added:
method Kosmos2_5ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Kosmos2_5FastImageProcessorKwargs]) -> BatchFeature
- added:
class Kosmos2_5ForConditionalGeneration(Kosmos2_5PreTrainedModel, GenerationMixin) - added:
class Kosmos2_5Model(Kosmos2_5PreTrainedModel) - added:
class Kosmos2_5PreTrainedModel(PreTrainedModel) - added:
constructor Kosmos2_5ForConditionalGeneration(config: Kosmos2_5Config) - added:
constructor Kosmos2_5Model(config: Kosmos2_5Config) - added:
field Kosmos2_5ForConditionalGeneration.config_class - added:
field Kosmos2_5Model.config_class - added:
field Kosmos2_5PreTrainedModel.config_class - added:
field Kosmos2_5PreTrainedModel.supports_gradient_checkpointing - added:
method Kosmos2_5ForConditionalGeneration.forward(input_ids: Optional[torch.Tensor]=…, flattened_patches: Optional[torch.Tensor]=…, width: Optional[torch.Tensor]=…, height: Optional[torch.Tensor]=…, image_embeds_position_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, image_embeds: Optional[torch.Tensor]=… - added:
method Kosmos2_5ForConditionalGeneration.get_input_embeddings() -> nn.Module - added:
method Kosmos2_5ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Kosmos2_5ForConditionalGeneration.prepare_inputs_for_generation(input_ids, flattened_patches=…, image_embeds=…, image_embeds_position_mask=…, past_key_values=…, attention_mask=…, use_cache=…, cache_position=…, position_ids=…, **model_kwargs) - added:
method Kosmos2_5ForConditionalGeneration.set_input_embeddings(value) - added:
method Kosmos2_5ForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method Kosmos2_5Model.forward(input_ids: Optional[torch.Tensor]=…, flattened_patches: Optional[torch.Tensor]=…, width: Optional[torch.Tensor]=…, height: Optional[torch.Tensor]=…, image_embeds_position_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, image_embeds: Optional[torch.Tensor]=…, inputs_embeds: O… - added:
method Kosmos2_5Model.get_input_embeddings() -> nn.Module - added:
method Kosmos2_5Model.set_input_embeddings(value)
- added:
class Kosmos2_5Processor(ProcessorMixin) - added:
constructor Kosmos2_5Processor(image_processor, tokenizer) - added:
field Kosmos2_5Processor.attributes - added:
field Kosmos2_5Processor.image_processor_class - added:
field Kosmos2_5Processor.tokenizer_class - added:
method Kosmos2_5Processor.batch_decode(*args, **kwargs) - added:
method Kosmos2_5Processor.decode(*args, **kwargs) - added:
property Kosmos2_5Processor.model_input_names
- added:
export define_import_structure
- added:
class KyutaiSpeechToTextConfig(PretrainedConfig) - added:
constructor KyutaiSpeechToTextConfig(codebook_vocab_size=…, vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, max_position_embeddings=…, rope_theta=…, hidden_act=…, head_dim=…, initializer_range=…, use_cache=…, sliding_window=…, attention_dropout=…, ffn_dim=…, rms_norm_eps=…, num_codebooks=…, audio_bos_token_id=…, audio_pad_token_id=…, tie_word_embedd… - added:
field KyutaiSpeechToTextConfig.keys_to_ignore_at_inference - added:
field KyutaiSpeechToTextConfig.model_type - added:
field KyutaiSpeechToTextConfig.sub_configs
- added:
class KyutaiSpeechToTextFeatureExtractor(SequenceFeatureExtractor) - added:
constructor KyutaiSpeechToTextFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, chunk_length_s: Optional[float]=…, overlap: Optional[float]=…, audio_delay_seconds: Optional[float]=…, audio_silence_prefix_seconds: Optional[float]=…, **kwargs) - added:
field KyutaiSpeechToTextFeatureExtractor.model_input_names - added:
property KyutaiSpeechToTextFeatureExtractor.chunk_length -> Optional[int] - added:
property KyutaiSpeechToTextFeatureExtractor.chunk_stride -> Optional[int]
- added:
class KyutaiSpeechToTextForConditionalGeneration(KyutaiSpeechToTextPreTrainedModel, GenerationMixin) - added:
class KyutaiSpeechToTextModel(KyutaiSpeechToTextPreTrainedModel) - added:
class KyutaiSpeechToTextPreTrainedModel(PreTrainedModel) - added:
constructor KyutaiSpeechToTextForConditionalGeneration(config) - added:
constructor KyutaiSpeechToTextModel(config) - added:
field KyutaiSpeechToTextPreTrainedModel.base_model_prefix - added:
field KyutaiSpeechToTextPreTrainedModel.config: KyutaiSpeechToTextConfig - added:
field KyutaiSpeechToTextPreTrainedModel.main_input_name - added:
field KyutaiSpeechToTextPreTrainedModel.supports_gradient_checkpointing - added:
method KyutaiSpeechToTextForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Uni… - added:
method KyutaiSpeechToTextForConditionalGeneration.generate(*args, **kwargs) - added:
method KyutaiSpeechToTextForConditionalGeneration.prepare_inputs_for_generation(*args, audio_tokens: Optional[torch.LongTensor]=…, input_values: Optional[torch.FloatTensor]=…, padding_mask: Optional[torch.Tensor]=…, audio_window_size: Optional[int]=…, current_window: Optional[tuple[int, int]]=…, encoder_past_key_values: Optional[Cache]=…, padding_cache: Optional[KyutaiSpeechToTextConv1dPaddingCac… - added:
method KyutaiSpeechToTextForConditionalGeneration.save_pretrained(*args, **kwargs) - added:
method KyutaiSpeechToTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
static method KyutaiSpeechToTextForConditionalGeneration.from_pretrained(*args, **kwargs)
- added:
class KyutaiSpeechToTextFeatureExtractor(EncodecFeatureExtractor) - added:
class KyutaiSpeechToTextForConditionalGeneration(LlamaForCausalLM, GenerationMixin) - added:
class KyutaiSpeechToTextModel(MoshiModel) - added:
class KyutaiSpeechToTextPreTrainedModel(MoshiPreTrainedModel) - added:
constructor KyutaiSpeechToTextFeatureExtractor(audio_delay_seconds: Optional[float]=…, audio_silence_prefix_seconds: Optional[float]=…, **super_kwargs) - added:
constructor KyutaiSpeechToTextForConditionalGeneration(config) - added:
constructor KyutaiSpeechToTextModel(config) - added:
method KyutaiSpeechToTextForConditionalGeneration.forward(**super_kwargs) - added:
method KyutaiSpeechToTextForConditionalGeneration.generate(*args, **kwargs) - added:
method KyutaiSpeechToTextForConditionalGeneration.prepare_inputs_for_generation(*args, audio_tokens: Optional[torch.LongTensor]=…, input_values: Optional[torch.FloatTensor]=…, padding_mask: Optional[torch.Tensor]=…, audio_window_size: Optional[int]=…, current_window: Optional[tuple[int, int]]=…, encoder_past_key_values: Optional[Cache]=…, padding_cache: Optional[KyutaiSpeechToTextConv1dPaddingCac… - added:
method KyutaiSpeechToTextForConditionalGeneration.save_pretrained(*args, **kwargs) - added:
static method KyutaiSpeechToTextForConditionalGeneration.from_pretrained(*args, **kwargs)
- added:
class KyutaiSpeechToTextProcessor(ProcessorMixin) - added:
field KyutaiSpeechToTextProcessor.feature_extractor_class - added:
field KyutaiSpeechToTextProcessor.tokenizer_class - added:
field KyutaiSpeechToTextProcessor.valid_processor_kwargs
- added:
export define_import_structure
- added:
class LayoutLMConfig(PretrainedConfig) - added:
class LayoutLMOnnxConfig(OnnxConfig) - added:
constructor LayoutLMConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, use_cache=…, max_2d_position_embeddings=…, **kwargs) - added:
constructor LayoutLMOnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…) - added:
field LayoutLMConfig.model_type - added:
method LayoutLMOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property LayoutLMOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class LayoutLMForMaskedLM(LayoutLMPreTrainedModel) - added:
class LayoutLMForQuestionAnswering(LayoutLMPreTrainedModel) - added:
class LayoutLMForSequenceClassification(LayoutLMPreTrainedModel) - added:
class LayoutLMForTokenClassification(LayoutLMPreTrainedModel) - added:
class LayoutLMModel(LayoutLMPreTrainedModel) - added:
class LayoutLMPreTrainedModel(PreTrainedModel) - added:
constructor LayoutLMForMaskedLM(config) - added:
constructor LayoutLMForQuestionAnswering(config, has_visual_segment_embedding=…) - added:
constructor LayoutLMForSequenceClassification(config) - added:
constructor LayoutLMForTokenClassification(config) - added:
constructor LayoutLMModel(config) - added:
field LayoutLMPreTrainedModel.base_model_prefix - added:
field LayoutLMPreTrainedModel.config: LayoutLMConfig - added:
field LayoutLMPreTrainedModel.supports_gradient_checkpointing - added:
method LayoutLMForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optio… - added:
method LayoutLMForMaskedLM.get_input_embeddings() - added:
method LayoutLMForMaskedLM.get_output_embeddings() - added:
method LayoutLMForMaskedLM.set_output_embeddings(new_embeddings) - added:
method LayoutLMForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_po… - added:
method LayoutLMForQuestionAnswering.get_input_embeddings() - added:
method LayoutLMForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_att… - added:
method LayoutLMForSequenceClassification.get_input_embeddings() - added:
method LayoutLMForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attent… - added:
method LayoutLMForTokenClassification.get_input_embeddings() - added:
method LayoutLMModel.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[… - added:
method LayoutLMModel.get_input_embeddings() - added:
method LayoutLMModel.set_input_embeddings(value)
- added:
class TFLayoutLMForMaskedLM(TFLayoutLMPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFLayoutLMForQuestionAnswering(TFLayoutLMPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFLayoutLMForSequenceClassification(TFLayoutLMPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFLayoutLMForTokenClassification(TFLayoutLMPreTrainedModel, TFTokenClassificationLoss) - added:
class TFLayoutLMMainLayer(keras.layers.Layer) - added:
class TFLayoutLMModel(TFLayoutLMPreTrainedModel) - added:
class TFLayoutLMPreTrainedModel(TFPreTrainedModel) - added:
constructor TFLayoutLMForMaskedLM(config: LayoutLMConfig, *inputs, **kwargs) - added:
constructor TFLayoutLMForQuestionAnswering(config: LayoutLMConfig, *inputs, **kwargs) - added:
constructor TFLayoutLMForSequenceClassification(config: LayoutLMConfig, *inputs, **kwargs) - added:
constructor TFLayoutLMForTokenClassification(config: LayoutLMConfig, *inputs, **kwargs) - added:
constructor TFLayoutLMMainLayer(config: LayoutLMConfig, add_pooling_layer: bool=…, **kwargs) - added:
constructor TFLayoutLMModel(config: LayoutLMConfig, *inputs, **kwargs) - added:
field TFLayoutLMMainLayer.config_class - added:
field TFLayoutLMPreTrainedModel.base_model_prefix - added:
field TFLayoutLMPreTrainedModel.config_class - added:
method TFLayoutLMForMaskedLM.build(input_shape=…) - added:
method TFLayoutLMForMaskedLM.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_sta… - added:
method TFLayoutLMForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFLayoutLMForMaskedLM.get_prefix_bias_name() -> str - added:
method TFLayoutLMForQuestionAnswering.build(input_shape=…) - added:
method TFLayoutLMForQuestionAnswering.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_h… - added:
method TFLayoutLMForSequenceClassification.build(input_shape=…) - added:
method TFLayoutLMForSequenceClassification.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, out… - added:
method TFLayoutLMForTokenClassification.build(input_shape=…) - added:
method TFLayoutLMForTokenClassification.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output… - added:
method TFLayoutLMMainLayer.build(input_shape=…) - added:
method TFLayoutLMMainLayer.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=… - added:
method TFLayoutLMMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFLayoutLMMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFLayoutLMModel.build(input_shape=…) - added:
method TFLayoutLMModel.call(input_ids: TFModelInputType | None=…, bbox: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, e… - added:
property TFLayoutLMPreTrainedModel.input_signature
- added:
class LayoutLMTokenizer(PreTrainedTokenizer) - added:
constructor LayoutLMTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field LayoutLMTokenizer.vocab_files_names - added:
method LayoutLMTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutLMTokenizer.convert_tokens_to_string(tokens) - added:
method LayoutLMTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LayoutLMTokenizer.get_vocab() - added:
method LayoutLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LayoutLMTokenizer.do_lower_case - added:
property LayoutLMTokenizer.vocab_size
- added:
class LayoutLMTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LayoutLMTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field LayoutLMTokenizerFast.slow_tokenizer_class - added:
field LayoutLMTokenizerFast.vocab_files_names - added:
method LayoutLMTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LayoutLMTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class LayoutLMv2Config(PretrainedConfig) - added:
constructor LayoutLMv2Config(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, max_2d_position_embeddings=…, max_rel_pos=…, rel_pos_bins=…, fast_qkv=…, max_rel_2d_pos=…, rel_2d_pos_bins… - added:
field LayoutLMv2Config.model_type - added:
method LayoutLMv2Config.get_detectron2_config() - added:
static method LayoutLMv2Config.get_default_detectron2_config()
- added:
class LayoutLMv2FeatureExtractor(LayoutLMv2ImageProcessor) - added:
constructor LayoutLMv2FeatureExtractor(*args, **kwargs) -> None
- added:
class LayoutLMv2ImageProcessor(BaseImageProcessor) - added:
constructor LayoutLMv2ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, apply_ocr: bool=…, ocr_lang: Optional[str]=…, tesseract_config: Optional[str]=…, **kwargs) -> None - added:
field LayoutLMv2ImageProcessor.model_input_names - added:
method LayoutLMv2ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, apply_ocr: Optional[bool]=…, ocr_lang: Optional[str]=…, tesseract_config: Optional[str]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension… - added:
method LayoutLMv2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LayoutLMv2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor LayoutLMv2ImageProcessorFast(**kwargs: Unpack[LayoutLMv2FastImageProcessorKwargs]) - added:
field LayoutLMv2ImageProcessorFast.apply_ocr - added:
field LayoutLMv2ImageProcessorFast.do_resize - added:
field LayoutLMv2ImageProcessorFast.ocr_lang - added:
field LayoutLMv2ImageProcessorFast.resample - added:
field LayoutLMv2ImageProcessorFast.rescale_factor - added:
field LayoutLMv2ImageProcessorFast.size - added:
field LayoutLMv2ImageProcessorFast.tesseract_config - added:
field LayoutLMv2ImageProcessorFast.valid_kwargs - added:
method LayoutLMv2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LayoutLMv2FastImageProcessorKwargs]) -> BatchFeature
- added:
class LayoutLMv2ForQuestionAnswering(LayoutLMv2PreTrainedModel) - added:
class LayoutLMv2ForSequenceClassification(LayoutLMv2PreTrainedModel) - added:
class LayoutLMv2ForTokenClassification(LayoutLMv2PreTrainedModel) - added:
class LayoutLMv2Layer(GradientCheckpointingLayer) - added:
class LayoutLMv2Model(LayoutLMv2PreTrainedModel) - added:
class LayoutLMv2PreTrainedModel(PreTrainedModel) - added:
constructor LayoutLMv2ForQuestionAnswering(config, has_visual_segment_embedding=…) - added:
constructor LayoutLMv2ForSequenceClassification(config) - added:
constructor LayoutLMv2ForTokenClassification(config) - added:
constructor LayoutLMv2Layer(config) - added:
constructor LayoutLMv2Model(config) - added:
field LayoutLMv2PreTrainedModel.base_model_prefix - added:
field LayoutLMv2PreTrainedModel.config: LayoutLMv2Config - added:
method LayoutLMv2ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, image: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positio… - added:
method LayoutLMv2ForQuestionAnswering.get_input_embeddings() - added:
method LayoutLMv2ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, image: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method LayoutLMv2ForSequenceClassification.get_input_embeddings() - added:
method LayoutLMv2ForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, image: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Opt… - added:
method LayoutLMv2ForTokenClassification.get_input_embeddings() - added:
method LayoutLMv2Layer.feed_forward_chunk(attention_output) - added:
method LayoutLMv2Layer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, rel_pos=…, rel_2d_pos=…) - added:
method LayoutLMv2Model.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, image: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[… - added:
method LayoutLMv2Model.get_input_embeddings() - added:
method LayoutLMv2Model.set_input_embeddings(value)
- added:
class LayoutLMv2Processor(ProcessorMixin) - added:
constructor LayoutLMv2Processor(image_processor=…, tokenizer=…, **kwargs) - added:
field LayoutLMv2Processor.attributes - added:
field LayoutLMv2Processor.image_processor_class - added:
field LayoutLMv2Processor.tokenizer_class - added:
method LayoutLMv2Processor.get_overflowing_images(images, overflow_to_sample_mapping) - added:
property LayoutLMv2Processor.feature_extractor - added:
property LayoutLMv2Processor.feature_extractor_class - added:
property LayoutLMv2Processor.model_input_names
- added:
class LayoutLMv2Tokenizer(PreTrainedTokenizer) - added:
constructor LayoutLMv2Tokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, tokenize_chinese_chars=…, strip_accents=…, model_max_length: int=…, additional_special_tokens: Optional[list[str]]=…, **kwargs) - added:
field LayoutLMv2Tokenizer.vocab_files_names - added:
method LayoutLMv2Tokenizer.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max… - added:
method LayoutLMv2Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutLMv2Tokenizer.convert_tokens_to_string(tokens) - added:
method LayoutLMv2Tokenizer.encode(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, p… - added:
method LayoutLMv2Tokenizer.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]… - added:
method LayoutLMv2Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LayoutLMv2Tokenizer.get_vocab() - added:
method LayoutLMv2Tokenizer.prepare_for_model(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optiona… - added:
method LayoutLMv2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutLMv2Tokenizer.truncate_sequences(ids: list[int], token_boxes: list[list[int]], pair_ids: Optional[list[int]]=…, pair_token_boxes: Optional[list[list[int]]]=…, labels: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[list[int], list[int], list[int]] - added:
property LayoutLMv2Tokenizer.do_lower_case - added:
property LayoutLMv2Tokenizer.vocab_size
- added:
class LayoutLMv2TokenizerFast(PreTrainedTokenizerFast) - added:
constructor LayoutLMv2TokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field LayoutLMv2TokenizerFast.slow_tokenizer_class - added:
field LayoutLMv2TokenizerFast.vocab_files_names - added:
method LayoutLMv2TokenizerFast.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…,… - added:
method LayoutLMv2TokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LayoutLMv2TokenizerFast.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[… - added:
method LayoutLMv2TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutLMv2TokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str]
- added:
export define_import_structure
- added:
class LayoutLMv3Config(PretrainedConfig) - added:
class LayoutLMv3OnnxConfig(OnnxConfig) - added:
constructor LayoutLMv3Config(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_2d_position_embeddings=…, coordinate_size=…, shape_size=…, has_relativ… - added:
field LayoutLMv3Config.model_type - added:
field LayoutLMv3OnnxConfig.torch_onnx_minimum_version - added:
method LayoutLMv3OnnxConfig.generate_dummy_inputs(processor: ProcessorMixin, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…) -> Mapping[str, Any] - added:
property LayoutLMv3OnnxConfig.atol_for_validation -> float - added:
property LayoutLMv3OnnxConfig.default_onnx_opset -> int - added:
property LayoutLMv3OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class LayoutLMv3FeatureExtractor(LayoutLMv3ImageProcessor) - added:
constructor LayoutLMv3FeatureExtractor(*args, **kwargs) -> None
- added:
class LayoutLMv3ImageProcessor(BaseImageProcessor) - added:
constructor LayoutLMv3ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_value: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, Iterable[float]]]=…, image_std: Optional[Union[float, Iterable[float]]]=…, apply_ocr: bool=…, ocr_lang: Optional[str]=…, tesseract_config: Optional[str]=…, **kwargs) -> None - added:
field LayoutLMv3ImageProcessor.model_input_names - added:
method LayoutLMv3ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, Iterable[float]]]=…, image_std: Optional[Union[float, Iterable[float]]]=…, apply_ocr: Optional[bool]=…, ocr_lang: Optional[str]=…,… - added:
method LayoutLMv3ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LayoutLMv3ImageProcessorFast(BaseImageProcessorFast) - added:
constructor LayoutLMv3ImageProcessorFast(**kwargs: Unpack[LayoutLMv3FastImageProcessorKwargs]) - added:
field LayoutLMv3ImageProcessorFast.apply_ocr - added:
field LayoutLMv3ImageProcessorFast.do_normalize - added:
field LayoutLMv3ImageProcessorFast.do_rescale - added:
field LayoutLMv3ImageProcessorFast.do_resize - added:
field LayoutLMv3ImageProcessorFast.image_mean - added:
field LayoutLMv3ImageProcessorFast.image_std - added:
field LayoutLMv3ImageProcessorFast.ocr_lang - added:
field LayoutLMv3ImageProcessorFast.resample - added:
field LayoutLMv3ImageProcessorFast.size - added:
field LayoutLMv3ImageProcessorFast.tesseract_config - added:
field LayoutLMv3ImageProcessorFast.valid_kwargs - added:
method LayoutLMv3ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LayoutLMv3FastImageProcessorKwargs]) -> BatchFeature
- added:
class LayoutLMv3ForQuestionAnswering(LayoutLMv3PreTrainedModel) - added:
class LayoutLMv3ForSequenceClassification(LayoutLMv3PreTrainedModel) - added:
class LayoutLMv3ForTokenClassification(LayoutLMv3PreTrainedModel) - added:
class LayoutLMv3Model(LayoutLMv3PreTrainedModel) - added:
class LayoutLMv3PreTrainedModel(PreTrainedModel) - added:
constructor LayoutLMv3ForQuestionAnswering(config) - added:
constructor LayoutLMv3ForSequenceClassification(config) - added:
constructor LayoutLMv3ForTokenClassification(config) - added:
constructor LayoutLMv3Model(config) - added:
field LayoutLMv3PreTrainedModel.base_model_prefix - added:
field LayoutLMv3PreTrainedModel.config: LayoutLMv3Config - added:
method LayoutLMv3ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor… - added:
method LayoutLMv3ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_… - added:
method LayoutLMv3ForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_atte… - added:
method LayoutLMv3Model.calculate_visual_bbox(device, dtype, batch_size) - added:
method LayoutLMv3Model.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Op… - added:
method LayoutLMv3Model.forward_image(pixel_values) - added:
method LayoutLMv3Model.get_input_embeddings() - added:
method LayoutLMv3Model.init_visual_bbox(image_size=…, max_len=…) - added:
method LayoutLMv3Model.set_input_embeddings(value)
- added:
class TFLayoutLMv3ForQuestionAnswering(TFLayoutLMv3PreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFLayoutLMv3ForSequenceClassification(TFLayoutLMv3PreTrainedModel, TFSequenceClassificationLoss) - added:
class TFLayoutLMv3ForTokenClassification(TFLayoutLMv3PreTrainedModel, TFTokenClassificationLoss) - added:
class TFLayoutLMv3Model(TFLayoutLMv3PreTrainedModel) - added:
class TFLayoutLMv3PreTrainedModel(TFPreTrainedModel) - added:
constructor TFLayoutLMv3ForQuestionAnswering(config: LayoutLMv3Config, **kwargs) - added:
constructor TFLayoutLMv3ForSequenceClassification(config: LayoutLMv3Config, **kwargs) - added:
constructor TFLayoutLMv3ForTokenClassification(config: LayoutLMv3Config, **kwargs) - added:
constructor TFLayoutLMv3Model(config, *inputs, **kwargs) - added:
field TFLayoutLMv3PreTrainedModel.base_model_prefix - added:
field TFLayoutLMv3PreTrainedModel.config_class - added:
method TFLayoutLMv3ForQuestionAnswering.build(input_shape=…) - added:
method TFLayoutLMv3ForQuestionAnswering.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, start_positions: tf.Tensor | None=…, end_positions: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, bbox: tf… - added:
method TFLayoutLMv3ForSequenceClassification.build(input_shape=…) - added:
method TFLayoutLMv3ForSequenceClassification.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, bbox: tf.Tensor | N… - added:
method TFLayoutLMv3ForTokenClassification.build(input_shape=…) - added:
method TFLayoutLMv3ForTokenClassification.call(input_ids: tf.Tensor | None=…, bbox: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None… - added:
method TFLayoutLMv3Model.build(input_shape=…) - added:
method TFLayoutLMv3Model.call(input_ids: tf.Tensor | None=…, bbox: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, pixel_values: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, trainin… - added:
property TFLayoutLMv3PreTrainedModel.input_signature
- added:
class LayoutLMv3Processor(ProcessorMixin) - added:
constructor LayoutLMv3Processor(image_processor=…, tokenizer=…, **kwargs) - added:
field LayoutLMv3Processor.attributes - added:
field LayoutLMv3Processor.image_processor_class - added:
field LayoutLMv3Processor.tokenizer_class - added:
method LayoutLMv3Processor.get_overflowing_images(images, overflow_to_sample_mapping) - added:
property LayoutLMv3Processor.feature_extractor - added:
property LayoutLMv3Processor.feature_extractor_class - added:
property LayoutLMv3Processor.model_input_names
- added:
class LayoutLMv3Tokenizer(PreTrainedTokenizer) - added:
constructor LayoutLMv3Tokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, **kwargs) - added:
field LayoutLMv3Tokenizer.model_input_names - added:
field LayoutLMv3Tokenizer.vocab_files_names - added:
method LayoutLMv3Tokenizer.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max… - added:
method LayoutLMv3Tokenizer.bpe(token) - added:
method LayoutLMv3Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutLMv3Tokenizer.convert_tokens_to_string(tokens) - added:
method LayoutLMv3Tokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutLMv3Tokenizer.encode(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, p… - added:
method LayoutLMv3Tokenizer.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]… - added:
method LayoutLMv3Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LayoutLMv3Tokenizer.get_vocab() - added:
method LayoutLMv3Tokenizer.prepare_for_model(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optiona… - added:
method LayoutLMv3Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method LayoutLMv3Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutLMv3Tokenizer.truncate_sequences(ids: list[int], token_boxes: list[list[int]], pair_ids: Optional[list[int]]=…, pair_token_boxes: Optional[list[list[int]]]=…, labels: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[list[int], list[int], list[int]] - added:
property LayoutLMv3Tokenizer.vocab_size
- added:
class LayoutLMv3TokenizerFast(PreTrainedTokenizerFast) - added:
constructor LayoutLMv3TokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, **kwargs) - added:
field LayoutLMv3TokenizerFast.model_input_names - added:
field LayoutLMv3TokenizerFast.slow_tokenizer_class - added:
field LayoutLMv3TokenizerFast.vocab_files_names - added:
method LayoutLMv3TokenizerFast.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…,… - added:
method LayoutLMv3TokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LayoutLMv3TokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutLMv3TokenizerFast.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[… - added:
method LayoutLMv3TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutLMv3TokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str]
- added:
export define_import_structure
- added:
class LayoutXLMProcessor(ProcessorMixin) - added:
constructor LayoutXLMProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field LayoutXLMProcessor.attributes - added:
field LayoutXLMProcessor.image_processor_class - added:
field LayoutXLMProcessor.tokenizer_class - added:
method LayoutXLMProcessor.get_overflowing_images(images, overflow_to_sample_mapping) - added:
property LayoutXLMProcessor.feature_extractor - added:
property LayoutXLMProcessor.feature_extractor_class - added:
property LayoutXLMProcessor.model_input_names
- added:
class LayoutXLMTokenizer(PreTrainedTokenizer) - added:
constructor LayoutXLMTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field LayoutXLMTokenizer.model_input_names - added:
field LayoutXLMTokenizer.vocab_files_names - added:
method LayoutXLMTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutXLMTokenizer.convert_tokens_to_string(tokens) - added:
method LayoutXLMTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutXLMTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LayoutXLMTokenizer.get_vocab() - added:
method LayoutXLMTokenizer.prepare_for_model(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional… - added:
method LayoutXLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutXLMTokenizer.truncate_sequences(ids: list[int], token_boxes: list[list[int]], pair_ids: Optional[list[int]]=…, pair_token_boxes: Optional[list[list[int]]]=…, labels: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[list[int], list[int], list[int]] - added:
property LayoutXLMTokenizer.vocab_size
- added:
class LayoutXLMTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LayoutXLMTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, cls_token_box=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, **kwargs) - added:
field LayoutXLMTokenizerFast.model_input_names - added:
field LayoutXLMTokenizerFast.slow_tokenizer_class - added:
field LayoutXLMTokenizerFast.vocab_files_names - added:
method LayoutXLMTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutXLMTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LayoutXLMTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LayoutXLMTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str]
- added:
export define_import_structure
- added:
class LEDConfig(PretrainedConfig) - added:
constructor LEDConfig(vocab_size=…, max_encoder_position_embeddings=…, max_decoder_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…,… - added:
field LEDConfig.attribute_map - added:
field LEDConfig.model_type
- added:
class LEDForConditionalGeneration(LEDPreTrainedModel, GenerationMixin) - added:
class LEDForQuestionAnswering(LEDPreTrainedModel) - added:
class LEDForSequenceClassification(LEDPreTrainedModel) - added:
class LEDModel(LEDPreTrainedModel) - added:
class LEDPreTrainedModel(PreTrainedModel) - added:
constructor LEDForConditionalGeneration(config: LEDConfig) - added:
constructor LEDForQuestionAnswering(config) - added:
constructor LEDForSequenceClassification(config: LEDConfig, **kwargs) - added:
constructor LEDModel(config: LEDConfig) - added:
field LEDForConditionalGeneration.base_model_prefix - added:
field LEDPreTrainedModel.base_model_prefix - added:
field LEDPreTrainedModel.config: LEDConfig - added:
field LEDPreTrainedModel.supports_gradient_checkpointing - added:
method LEDForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[to… - added:
method LEDForConditionalGeneration.get_decoder() - added:
method LEDForConditionalGeneration.get_encoder() - added:
method LEDForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method LEDForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method LEDForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.… - added:
method LEDForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[t… - added:
method LEDModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=… - added:
method LEDModel.get_encoder() - added:
method LEDModel.get_input_embeddings() - added:
method LEDModel.set_input_embeddings(value) - added:
property LEDPreTrainedModel.dummy_inputs
- added:
class TFLEDForConditionalGeneration(TFLEDPreTrainedModel) - added:
class TFLEDModel(TFLEDPreTrainedModel) - added:
class TFLEDPreTrainedModel(TFPreTrainedModel) - added:
constructor TFLEDForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFLEDModel(config, *inputs, **kwargs) - added:
field TFLEDPreTrainedModel.base_model_prefix - added:
field TFLEDPreTrainedModel.config_class - added:
method TFLEDForConditionalGeneration.build(input_shape=…) - added:
method TFLEDForConditionalGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: TFLEDEncoderBaseModelOutput | None=…, global_att… - added:
method TFLEDForConditionalGeneration.get_bias() - added:
method TFLEDForConditionalGeneration.get_decoder() - added:
method TFLEDForConditionalGeneration.get_encoder() - added:
method TFLEDForConditionalGeneration.get_output_embeddings() - added:
method TFLEDForConditionalGeneration.hf_compute_loss(labels, logits) - added:
method TFLEDForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFLEDForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, head_mask=…, decoder_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFLEDForConditionalGeneration.serving_output(output) - added:
method TFLEDForConditionalGeneration.set_bias(value) - added:
method TFLEDForConditionalGeneration.set_output_embeddings(value) - added:
method TFLEDModel.build(input_shape=…) - added:
method TFLEDModel.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, encoder_outputs: tf.Tensor | None=…, global_attention_mask: tf.Tensor | None=…, past_key_values: tuple[tuple[tf.Tensor]] | None=…, inputs_embeds: tf.… - added:
method TFLEDModel.get_decoder() - added:
method TFLEDModel.get_encoder() - added:
method TFLEDModel.serving_output(output) - added:
property TFLEDPreTrainedModel.input_signature
- added:
class LEDTokenizer(PreTrainedTokenizer) - added:
constructor LEDTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field LEDTokenizer.model_input_names - added:
field LEDTokenizer.vocab_files_names - added:
method LEDTokenizer.bpe(token) - added:
method LEDTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LEDTokenizer.convert_tokens_to_string(tokens) - added:
method LEDTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LEDTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LEDTokenizer.get_vocab() - added:
method LEDTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method LEDTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LEDTokenizer.vocab_size
- added:
class LEDTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LEDTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field LEDTokenizerFast.model_input_names - added:
field LEDTokenizerFast.slow_tokenizer_class - added:
field LEDTokenizerFast.vocab_files_names - added:
method LEDTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LEDTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LEDTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LEDTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class LevitConfig(PretrainedConfig) - added:
class LevitOnnxConfig(OnnxConfig) - added:
constructor LevitConfig(image_size=…, num_channels=…, kernel_size=…, stride=…, padding=…, patch_size=…, hidden_sizes=…, num_attention_heads=…, depths=…, key_dim=…, drop_path_rate=…, mlp_ratio=…, attention_ratio=…, initializer_range=…, **kwargs) - added:
field LevitConfig.model_type - added:
field LevitOnnxConfig.torch_onnx_minimum_version - added:
property LevitOnnxConfig.atol_for_validation -> float - added:
property LevitOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class LevitFeatureExtractor(LevitImageProcessor) - added:
constructor LevitFeatureExtractor(*args, **kwargs) -> None
- added:
class LevitImageProcessor(BaseImageProcessor) - added:
constructor LevitImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, Iterable[float]]]=…, image_std: Optional[Union[float, Iterable[float]]]=…, **kwargs) -> None - added:
field LevitImageProcessor.model_input_names - added:
method LevitImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, Iterable[float]]]=…, image_std:… - added:
method LevitImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LevitImageProcessorFast(BaseImageProcessorFast) - added:
field LevitImageProcessorFast.crop_size - added:
field LevitImageProcessorFast.default_to_square - added:
field LevitImageProcessorFast.do_center_crop - added:
field LevitImageProcessorFast.do_convert_rgb - added:
field LevitImageProcessorFast.do_normalize - added:
field LevitImageProcessorFast.do_rescale - added:
field LevitImageProcessorFast.do_resize - added:
field LevitImageProcessorFast.image_mean - added:
field LevitImageProcessorFast.image_std - added:
field LevitImageProcessorFast.resample - added:
field LevitImageProcessorFast.size - added:
method LevitImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor
- added:
class LevitForImageClassification(LevitPreTrainedModel) - added:
class LevitForImageClassificationWithTeacher(LevitPreTrainedModel) - added:
class LevitModel(LevitPreTrainedModel) - added:
class LevitPreTrainedModel(PreTrainedModel) - added:
constructor LevitForImageClassification(config) - added:
constructor LevitForImageClassificationWithTeacher(config) - added:
constructor LevitModel(config) - added:
field LevitPreTrainedModel.base_model_prefix - added:
field LevitPreTrainedModel.config: LevitConfig - added:
field LevitPreTrainedModel.main_input_name - added:
method LevitForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method LevitForImageClassificationWithTeacher.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, LevitForImageClassificationWithTeacherOutput] - added:
method LevitModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class Lfm2Config(PretrainedConfig) - added:
constructor Lfm2Config(vocab_size: int=…, hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads: int=…, max_position_embeddings: int=…, initializer_range: float=…, norm_eps: float=…, use_cache: bool=…, pad_token_id: int=…, bos_token_id: int=…, eos_token_id: int=…, tie_word_embeddings: bool=…, rope_theta: float=…, conv_bias: bool=…,… - added:
field Lfm2Config.keys_to_ignore_at_inference - added:
field Lfm2Config.model_type
- added:
class Lfm2ForCausalLM(Lfm2PreTrainedModel, GenerationMixin) - added:
class Lfm2Model(Lfm2PreTrainedModel) - added:
class Lfm2PreTrainedModel(PreTrainedModel) - added:
constructor Lfm2ForCausalLM(config) - added:
constructor Lfm2Model(config: Lfm2Config) - added:
field Lfm2PreTrainedModel.base_model_prefix - added:
field Lfm2PreTrainedModel.config: Lfm2Config - added:
field Lfm2PreTrainedModel.supports_gradient_checkpointing - added:
method Lfm2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **… - added:
method Lfm2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Lfm2HybridConvCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Lfm2ForCausalLM(LlamaForCausalLM) - added:
class Lfm2Model(LlamaModel) - added:
class Lfm2PreTrainedModel(LlamaPreTrainedModel) - added:
constructor Lfm2Model(config: Lfm2Config) - added:
method Lfm2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Lfm2HybridConvCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
export define_import_structure
- added:
class Lfm2VlConfig(PretrainedConfig) - added:
constructor Lfm2VlConfig(vision_config=…, text_config=…, image_token_id=…, projector_hidden_act=…, projector_hidden_size=…, projector_bias=…, downsample_factor=…, **kwargs) - added:
field Lfm2VlConfig.model_type - added:
field Lfm2VlConfig.sub_configs
- added:
class Lfm2VlImageProcessorFast(BaseImageProcessorFast) - added:
constructor Lfm2VlImageProcessorFast(**kwargs: Unpack[Lfm2VlFastImageProcessorKwargs]) - added:
field Lfm2VlImageProcessorFast.do_image_splitting - added:
field Lfm2VlImageProcessorFast.do_normalize - added:
field Lfm2VlImageProcessorFast.do_pad - added:
field Lfm2VlImageProcessorFast.do_rescale - added:
field Lfm2VlImageProcessorFast.do_resize - added:
field Lfm2VlImageProcessorFast.downsample_factor - added:
field Lfm2VlImageProcessorFast.encoder_patch_size - added:
field Lfm2VlImageProcessorFast.image_mean - added:
field Lfm2VlImageProcessorFast.image_std - added:
field Lfm2VlImageProcessorFast.max_image_tokens - added:
field Lfm2VlImageProcessorFast.max_pixels_tolerance - added:
field Lfm2VlImageProcessorFast.max_tiles - added:
field Lfm2VlImageProcessorFast.min_image_tokens - added:
field Lfm2VlImageProcessorFast.min_tiles - added:
field Lfm2VlImageProcessorFast.model_input_names - added:
field Lfm2VlImageProcessorFast.resample - added:
field Lfm2VlImageProcessorFast.rescale_factor - added:
field Lfm2VlImageProcessorFast.return_row_col_info - added:
field Lfm2VlImageProcessorFast.size - added:
field Lfm2VlImageProcessorFast.tile_size - added:
field Lfm2VlImageProcessorFast.use_thumbnail - added:
field Lfm2VlImageProcessorFast.valid_kwargs - added:
method Lfm2VlImageProcessorFast.crop_image_to_patches(image: torch.Tensor, min_tiles: int, max_tiles: int, tile_size: int, use_thumbnail: bool, thumbnail_size: tuple[int], interpolation: F.InterpolationMode=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method Lfm2VlImageProcessorFast.resize_and_split(images: torch.Tensor, downsample_factor: int, min_tiles: int, max_tiles: int, use_thumbnail: bool, min_image_tokens: int, max_image_tokens: int, encoder_patch_size: int, tile_size: int, max_pixels_tolerance: float, interpolation: F.InterpolationMode) -> torch.Tensor - added:
method Lfm2VlImageProcessorFast.smart_resize(height: int, width: int, downsample_factor: int, min_image_tokens: int, max_image_tokens: int, encoder_patch_size: int) -> tuple[int, int]
- added:
class Lfm2VlForConditionalGeneration(Lfm2VlPreTrainedModel, GenerationMixin) - added:
class Lfm2VlModel(Lfm2VlPreTrainedModel) - added:
class Lfm2VlPreTrainedModel(PreTrainedModel) - added:
constructor Lfm2VlForConditionalGeneration(config: Lfm2VlConfig) - added:
constructor Lfm2VlModel(config: Lfm2VlConfig) - added:
field Lfm2VlPreTrainedModel.base_model_prefix - added:
field Lfm2VlPreTrainedModel.config: Lfm2VlConfig - added:
field Lfm2VlPreTrainedModel.supports_gradient_checkpointing - added:
method Lfm2VlForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, spatial_shapes: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: O… - added:
method Lfm2VlForConditionalGeneration.get_decoder() - added:
method Lfm2VlForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, spatial_shapes: torch.Tensor, pixel_attention_mask: torch.Tensor, **kwargs) - added:
method Lfm2VlForConditionalGeneration.get_input_embeddings() - added:
method Lfm2VlForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Lfm2VlForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Lfm2VlForConditionalGeneration.set_decoder(decoder) - added:
method Lfm2VlForConditionalGeneration.set_input_embeddings(value) - added:
method Lfm2VlModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, spatial_shapes: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…,… - added:
method Lfm2VlModel.get_decoder() - added:
method Lfm2VlModel.get_image_features(pixel_values: torch.FloatTensor, spatial_shapes: torch.Tensor, pixel_attention_mask: torch.Tensor, **kwargs) -> list[torch.Tensor] - added:
method Lfm2VlModel.get_input_embeddings() - added:
method Lfm2VlModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Lfm2VlModel.set_decoder(decoder) - added:
method Lfm2VlModel.set_input_embeddings(value) - added:
property Lfm2VlForConditionalGeneration.language_model - added:
property Lfm2VlForConditionalGeneration.multi_modal_projector - added:
property Lfm2VlForConditionalGeneration.vision_tower
- added:
class Lfm2VlForConditionalGeneration(LlavaForConditionalGeneration) - added:
class Lfm2VlModel(LlavaModel) - added:
class Lfm2VlPreTrainedModel(LlavaPreTrainedModel) - added:
constructor Lfm2VlModel(config: Lfm2VlConfig) - added:
method Lfm2VlForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, spatial_shapes: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: O… - added:
method Lfm2VlForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, spatial_shapes: torch.Tensor, pixel_attention_mask: torch.Tensor, **kwargs) - added:
method Lfm2VlModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, spatial_shapes: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…,… - added:
method Lfm2VlModel.get_image_features(pixel_values: torch.FloatTensor, spatial_shapes: torch.Tensor, pixel_attention_mask: torch.Tensor, **kwargs) -> list[torch.Tensor] - added:
method Lfm2VlModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor)
- added:
class Lfm2VlProcessor(ProcessorMixin) - added:
constructor Lfm2VlProcessor(image_processor, tokenizer, chat_template: Optional[str]=…, use_image_special_tokens: Optional[bool]=…, **kwargs) - added:
field Lfm2VlProcessor.attributes - added:
field Lfm2VlProcessor.image_processor_class - added:
field Lfm2VlProcessor.tokenizer_class - added:
method Lfm2VlProcessor.batch_decode(*args, **kwargs) - added:
method Lfm2VlProcessor.decode(*args, **kwargs) - added:
method Lfm2VlProcessor.expand_text_with_placeholders(text: list[str], images: list[list[ImageInput]], image_rows: list[list[int]], image_cols: list[list[int]], image_sizes: list[list[int]], use_image_special_tokens: bool, **images_kwargs) - added:
property Lfm2VlProcessor.model_input_names
- added:
export define_import_structure
- added:
class LightGlueConfig(PretrainedConfig) - added:
constructor LightGlueConfig(keypoint_detector_config: SuperPointConfig=…, descriptor_dim: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads=…, depth_confidence: float=…, width_confidence: float=…, filter_threshold: float=…, initializer_range: float=…, hidden_act: str=…, attention_dropout=…, attention_bias=…, trust_remote_code: bool=…, **kwargs) - added:
field LightGlueConfig.model_type - added:
field LightGlueConfig.sub_configs
- added:
class LightGlueImageProcessor(BaseImageProcessor) - added:
constructor LightGlueImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_grayscale: bool=…, **kwargs) -> None - added:
field LightGlueImageProcessor.model_input_names - added:
method LightGlueImageProcessor.plot_keypoint_matching(images: ImageInput, keypoint_matching_output: LightGlueKeypointMatchingOutput) - added:
method LightGlueImageProcessor.post_process_keypoint_matching(outputs: LightGlueKeypointMatchingOutput, target_sizes: Union[TensorType, list[tuple]], threshold: float=…) -> list[dict[str, torch.Tensor]] - added:
method LightGlueImageProcessor.preprocess(images, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_grayscale: Optional[bool]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, *… - added:
method LightGlueImageProcessor.resize(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method LightGlueImageProcessor.visualize_keypoint_matching(images: ImageInput, keypoint_matching_output: list[dict[str, torch.Tensor]]) -> list[Image.Image]
- added:
class LightGlueForKeypointMatching(LightGluePreTrainedModel) - added:
class LightGluePreTrainedModel(PreTrainedModel) - added:
constructor LightGlueForKeypointMatching(config: LightGlueConfig) - added:
field LightGluePreTrainedModel.base_model_prefix - added:
field LightGluePreTrainedModel.config: LightGlueConfig - added:
field LightGluePreTrainedModel.main_input_name - added:
field LightGluePreTrainedModel.supports_gradient_checkpointing - added:
method LightGlueForKeypointMatching.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> Union[tuple, LightGlueKeypointMatchingOutput]
- added:
class LightGlueConfig(PretrainedConfig) - added:
class LightGlueForKeypointMatching(LightGluePreTrainedModel) - added:
class LightGlueImageProcessor(SuperGlueImageProcessor) - added:
class LightGluePreTrainedModel(PreTrainedModel) - added:
constructor LightGlueConfig(keypoint_detector_config: SuperPointConfig=…, descriptor_dim: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_key_value_heads=…, depth_confidence: float=…, width_confidence: float=…, filter_threshold: float=…, initializer_range: float=…, hidden_act: str=…, attention_dropout=…, attention_bias=…, trust_remote_code: bool=…, **kwargs) - added:
constructor LightGlueForKeypointMatching(config: LightGlueConfig) - added:
field LightGlueConfig.model_type - added:
field LightGlueConfig.sub_configs - added:
field LightGluePreTrainedModel.base_model_prefix - added:
field LightGluePreTrainedModel.config: LightGlueConfig - added:
field LightGluePreTrainedModel.main_input_name - added:
field LightGluePreTrainedModel.supports_gradient_checkpointing - added:
method LightGlueForKeypointMatching.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> Union[tuple, LightGlueKeypointMatchingOutput] - added:
method LightGlueImageProcessor.plot_keypoint_matching(images: ImageInput, keypoint_matching_output: LightGlueKeypointMatchingOutput) - added:
method LightGlueImageProcessor.post_process_keypoint_matching(outputs: LightGlueKeypointMatchingOutput, target_sizes: Union[TensorType, list[tuple]], threshold: float=…) -> list[dict[str, torch.Tensor]] - added:
method LightGlueImageProcessor.visualize_keypoint_matching(images: ImageInput, keypoint_matching_output: list[dict[str, torch.Tensor]]) -> list[Image.Image]
- added:
export define_import_structure
- added:
class LiltConfig(PretrainedConfig) - added:
constructor LiltConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, classifier_dropout=…, channel_shrink_ratio=…, max_2d_position_embeddings=…, **kwargs) - added:
field LiltConfig.model_type
- added:
class LiltForQuestionAnswering(LiltPreTrainedModel) - added:
class LiltForSequenceClassification(LiltPreTrainedModel) - added:
class LiltForTokenClassification(LiltPreTrainedModel) - added:
class LiltModel(LiltPreTrainedModel) - added:
class LiltPreTrainedModel(PreTrainedModel) - added:
constructor LiltForQuestionAnswering(config) - added:
constructor LiltForSequenceClassification(config) - added:
constructor LiltForTokenClassification(config) - added:
constructor LiltModel(config, add_pooling_layer=…) - added:
field LiltPreTrainedModel.base_model_prefix - added:
field LiltPreTrainedModel.config: LiltConfig - added:
field LiltPreTrainedModel.supports_gradient_checkpointing - added:
method LiltForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positi… - added:
method LiltForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions:… - added:
method LiltForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, bbox: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions… - added:
method LiltModel.forward(input_ids: Optional[torch.Tensor]=…, bbox: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool… - added:
method LiltModel.get_input_embeddings() - added:
method LiltModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class LlamaConfig(PretrainedConfig) - added:
constructor LlamaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, pretraining_tp=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, mlp_bias=… - added:
field LlamaConfig.base_model_pp_plan - added:
field LlamaConfig.base_model_tp_plan - added:
field LlamaConfig.keys_to_ignore_at_inference - added:
field LlamaConfig.model_type
- added:
class FlaxLlamaForCausalLM(FlaxLlamaPreTrainedModel) - added:
class FlaxLlamaModel(FlaxLlamaPreTrainedModel) - added:
class FlaxLlamaPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxLlamaPreTrainedModel(config: LlamaConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxLlamaForCausalLM.module_class - added:
field FlaxLlamaModel.module_class - added:
field FlaxLlamaPreTrainedModel.base_model_prefix - added:
field FlaxLlamaPreTrainedModel.config_class - added:
field FlaxLlamaPreTrainedModel.module_class: nn.Module - added:
method FlaxLlamaForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxLlamaForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxLlamaPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxLlamaPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class LlamaForCausalLM(LlamaPreTrainedModel, GenerationMixin) - added:
class LlamaForQuestionAnswering(GenericForQuestionAnswering, LlamaPreTrainedModel) - added:
class LlamaForSequenceClassification(GenericForSequenceClassification, LlamaPreTrainedModel) - added:
class LlamaForTokenClassification(GenericForTokenClassification, LlamaPreTrainedModel) - added:
class LlamaModel(LlamaPreTrainedModel) - added:
class LlamaPreTrainedModel(PreTrainedModel) - added:
constructor LlamaForCausalLM(config) - added:
constructor LlamaModel(config: LlamaConfig) - added:
field LlamaForQuestionAnswering.base_model_prefix - added:
field LlamaPreTrainedModel.base_model_prefix - added:
field LlamaPreTrainedModel.config: LlamaConfig - added:
field LlamaPreTrainedModel.supports_gradient_checkpointing - added:
method LlamaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method LlamaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class LlamaTokenizer(PreTrainedTokenizer) - added:
constructor LlamaTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, pad_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, add_bos_token=…, add_eos_token=…, clean_up_tokenization_spaces=…, use_default_system_prompt=…, spaces_between_special_tokens=…, legacy=…, add_prefix_space=…, **kwargs) - added:
field LlamaTokenizer.model_input_names - added:
field LlamaTokenizer.vocab_files_names - added:
method LlamaTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LlamaTokenizer.convert_tokens_to_string(tokens) - added:
method LlamaTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LlamaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LlamaTokenizer.get_spm_processor(from_slow=…) - added:
method LlamaTokenizer.get_vocab() - added:
method LlamaTokenizer.save_vocabulary(save_directory, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LlamaTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
property LlamaTokenizer.unk_token_length - added:
property LlamaTokenizer.vocab_size
- added:
class LlamaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LlamaTokenizerFast(vocab_file=…, tokenizer_file=…, clean_up_tokenization_spaces=…, unk_token=…, bos_token=…, eos_token=…, add_bos_token=…, add_eos_token=…, use_default_system_prompt=…, legacy=…, add_prefix_space=…, **kwargs) - added:
field LlamaTokenizerFast.model_input_names - added:
field LlamaTokenizerFast.padding_side - added:
field LlamaTokenizerFast.slow_tokenizer_class - added:
field LlamaTokenizerFast.vocab_files_names - added:
method LlamaTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LlamaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method LlamaTokenizerFast.update_post_processor() - added:
property LlamaTokenizerFast.add_bos_token - added:
property LlamaTokenizerFast.add_eos_token
- added:
export define_import_structure
- added:
class Llama4Config(PretrainedConfig) - added:
class Llama4TextConfig(PretrainedConfig) - added:
class Llama4VisionConfig(PretrainedConfig) - added:
constructor Llama4Config(vision_config=…, text_config=…, boi_token_index=…, eoi_token_index=…, image_token_index=…, tie_word_embeddings=…, **kwargs) - added:
constructor Llama4TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, intermediate_size_mlp=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_dropout=…, num_experts_per_tok… - added:
constructor Llama4VisionConfig(hidden_size: int=…, hidden_act: str=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_channels: int=…, intermediate_size: int=…, vision_output_dim: int=…, image_size: int=…, patch_size: int=…, norm_eps: float=…, vision_feature_select_strategy=…, initializer_range: float=…, pixel_shuffle_ratio=…, projector_input_dim=…, projector_output_dim=…, multi_modal_p… - added:
field Llama4Config.attribute_map - added:
field Llama4Config.base_model_tp_plan - added:
field Llama4Config.model_type - added:
field Llama4Config.sub_configs - added:
field Llama4TextConfig.base_model_ep_plan - added:
field Llama4TextConfig.base_model_tp_plan - added:
field Llama4TextConfig.keys_to_ignore_at_inference - added:
field Llama4TextConfig.model_type - added:
field Llama4VisionConfig.base_config_key - added:
field Llama4VisionConfig.base_model_tp_plan - added:
field Llama4VisionConfig.model_type
- added:
class Llama4ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Llama4ImageProcessorFast(**kwargs: Unpack[Llama4ImageProcessorKwargs]) - added:
field Llama4ImageProcessorFast.do_convert_rgb - added:
field Llama4ImageProcessorFast.do_normalize - added:
field Llama4ImageProcessorFast.do_rescale - added:
field Llama4ImageProcessorFast.do_resize - added:
field Llama4ImageProcessorFast.image_mean - added:
field Llama4ImageProcessorFast.image_std - added:
field Llama4ImageProcessorFast.max_patches - added:
field Llama4ImageProcessorFast.resample - added:
field Llama4ImageProcessorFast.resize_to_max_canvas - added:
field Llama4ImageProcessorFast.size - added:
field Llama4ImageProcessorFast.valid_kwargs - added:
method Llama4ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Llama4ImageProcessorKwargs]) -> BatchFeature - added:
method Llama4ImageProcessorFast.rescale_and_normalize(images: torch.Tensor, do_rescale: bool, rescale_factor: float, do_normalize: bool, image_mean: Union[float, list[float]], image_std: Union[float, list[float]]) -> torch.Tensor
- added:
class Llama4ForCausalLM(Llama4PreTrainedModel, GenerationMixin) - added:
class Llama4ForConditionalGeneration(Llama4PreTrainedModel, GenerationMixin) - added:
class Llama4PreTrainedModel(PreTrainedModel) - added:
class Llama4TextModel(Llama4PreTrainedModel) - added:
class Llama4VisionModel(Llama4PreTrainedModel) - added:
constructor Llama4ForCausalLM(config: Llama4TextConfig) - added:
constructor Llama4ForConditionalGeneration(config: Llama4Config) - added:
constructor Llama4TextModel(config: Llama4TextConfig) - added:
constructor Llama4VisionModel(config: Llama4VisionConfig) - added:
field Llama4ForCausalLM.base_model_prefix - added:
field Llama4ForCausalLM.config: Llama4TextConfig - added:
field Llama4ForConditionalGeneration.base_model_prefix - added:
field Llama4ForConditionalGeneration.config: Llama4Config - added:
field Llama4PreTrainedModel.config: Llama4Config - added:
field Llama4PreTrainedModel.supports_gradient_checkpointing - added:
field Llama4TextModel.base_model_prefix - added:
field Llama4TextModel.config: Llama4TextConfig - added:
field Llama4VisionModel.base_model_prefix - added:
field Llama4VisionModel.config: Llama4VisionConfig - added:
method Llama4ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_ke… - added:
method Llama4ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[s… - added:
method Llama4ForConditionalGeneration.get_decoder() - added:
method Llama4ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_select_strategy: str, **kwargs) - added:
method Llama4ForConditionalGeneration.get_input_embeddings() - added:
method Llama4ForConditionalGeneration.get_output_embeddings() - added:
method Llama4ForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Llama4ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Llama4ForConditionalGeneration.set_decoder(decoder) - added:
method Llama4ForConditionalGeneration.set_input_embeddings(value) - added:
method Llama4ForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method Llama4TextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutputWithPast] - added:
method Llama4VisionModel.forward(pixel_values: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[BaseModelOutput, tuple[torch.Tensor, ...]] - added:
method Llama4VisionModel.get_input_embeddings()
- added:
class Llama4Processor(ProcessorMixin) - added:
constructor Llama4Processor(image_processor=…, tokenizer=…, patch_size: int=…, pixel_shuffle_ratio: float=…, fake_image_token=…, image_token=…, start_of_image_token=…, end_of_image_token=…, patch_token=…, tile_x_separator_token=…, tile_y_separator_token=…, chat_template=…, **kwargs) - added:
field Llama4Processor.attributes - added:
field Llama4Processor.image_processor_class - added:
field Llama4Processor.tokenizer_class
- added:
export define_import_structure
- added:
class LlavaConfig(PretrainedConfig) - added:
constructor LlavaConfig(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, vision_feature_select_strategy=…, vision_feature_layer=…, image_seq_length=…, multimodal_projector_bias=…, **kwargs) - added:
field LlavaConfig.attribute_map - added:
field LlavaConfig.model_type - added:
field LlavaConfig.sub_configs
- added:
class LlavaImageProcessor(BaseImageProcessor) - added:
constructor LlavaImageProcessor(do_pad: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…,… - added:
field LlavaImageProcessor.model_input_names - added:
method LlavaImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method LlavaImageProcessor.preprocess(images: ImageInput, do_pad: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…,… - added:
method LlavaImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LlavaImageProcessorFast(BaseImageProcessorFast) - added:
constructor LlavaImageProcessorFast(**kwargs: Unpack[LlavaFastImageProcessorKwargs]) -> None - added:
field LlavaImageProcessorFast.crop_size - added:
field LlavaImageProcessorFast.default_to_square - added:
field LlavaImageProcessorFast.do_center_crop - added:
field LlavaImageProcessorFast.do_convert_rgb - added:
field LlavaImageProcessorFast.do_normalize - added:
field LlavaImageProcessorFast.do_pad - added:
field LlavaImageProcessorFast.do_rescale - added:
field LlavaImageProcessorFast.do_resize - added:
field LlavaImageProcessorFast.image_mean - added:
field LlavaImageProcessorFast.image_std - added:
field LlavaImageProcessorFast.resample - added:
field LlavaImageProcessorFast.size - added:
field LlavaImageProcessorFast.valid_kwargs - added:
method LlavaImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method LlavaImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LlavaFastImageProcessorKwargs]) -> BatchFeature
- added:
class LlavaForConditionalGeneration(LlavaPreTrainedModel, GenerationMixin) - added:
class LlavaModel(LlavaPreTrainedModel) - added:
class LlavaPreTrainedModel(PreTrainedModel) - added:
constructor LlavaForConditionalGeneration(config: LlavaConfig) - added:
constructor LlavaModel(config: LlavaConfig) - added:
field LlavaPreTrainedModel.base_model_prefix - added:
field LlavaPreTrainedModel.config: LlavaConfig - added:
field LlavaPreTrainedModel.supports_gradient_checkpointing - added:
method LlavaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[st… - added:
method LlavaForConditionalGeneration.get_decoder() - added:
method LlavaForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method LlavaForConditionalGeneration.get_input_embeddings() - added:
method LlavaForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method LlavaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaForConditionalGeneration.set_decoder(decoder) - added:
method LlavaForConditionalGeneration.set_input_embeddings(value) - added:
method LlavaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, cache_positio… - added:
method LlavaModel.get_decoder() - added:
method LlavaModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, **kwargs) - added:
method LlavaModel.get_input_embeddings() - added:
method LlavaModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method LlavaModel.set_decoder(decoder) - added:
method LlavaModel.set_input_embeddings(value) - added:
property LlavaForConditionalGeneration.language_model - added:
property LlavaForConditionalGeneration.multi_modal_projector - added:
property LlavaForConditionalGeneration.vision_tower
- added:
class LlavaProcessor(ProcessorMixin) - added:
constructor LlavaProcessor(image_processor=…, tokenizer=…, patch_size=…, vision_feature_select_strategy=…, chat_template=…, image_token=…, num_additional_image_tokens=…, **kwargs) - added:
field LlavaProcessor.attributes - added:
field LlavaProcessor.image_processor_class - added:
field LlavaProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class LlavaNextConfig(PretrainedConfig) - added:
constructor LlavaNextConfig(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, vision_feature_select_strategy=…, vision_feature_layer=…, image_grid_pinpoints=…, tie_word_embeddings=…, image_seq_length=…, multimodal_projector_bias=…, **kwargs) - added:
field LlavaNextConfig.attribute_map - added:
field LlavaNextConfig.model_type - added:
field LlavaNextConfig.sub_configs
- added:
class LlavaNextImageProcessor(BaseImageProcessor) - added:
constructor LlavaNextImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, image_grid_pinpoints: Optional[list]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]… - added:
field LlavaNextImageProcessor.model_input_names - added:
method LlavaNextImageProcessor.get_image_patches(image: np.ndarray, grid_pinpoints, size: tuple, patch_size: int, resample: PILImageResampling, data_format: ChannelDimension, input_data_format: ChannelDimension) -> list[np.ndarray] - added:
method LlavaNextImageProcessor.pad(image: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method LlavaNextImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, image_grid_pinpoints: Optional[list]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float,… - added:
method LlavaNextImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LlavaNextImageProcessorFast(BaseImageProcessorFast) - added:
constructor LlavaNextImageProcessorFast(**kwargs: Unpack[LlavaNextFastImageProcessorKwargs]) - added:
field LlavaNextImageProcessorFast.crop_size - added:
field LlavaNextImageProcessorFast.default_to_square - added:
field LlavaNextImageProcessorFast.do_center_crop - added:
field LlavaNextImageProcessorFast.do_convert_rgb - added:
field LlavaNextImageProcessorFast.do_normalize - added:
field LlavaNextImageProcessorFast.do_pad - added:
field LlavaNextImageProcessorFast.do_rescale - added:
field LlavaNextImageProcessorFast.do_resize - added:
field LlavaNextImageProcessorFast.image_grid_pinpoints - added:
field LlavaNextImageProcessorFast.image_mean - added:
field LlavaNextImageProcessorFast.image_std - added:
field LlavaNextImageProcessorFast.resample - added:
field LlavaNextImageProcessorFast.size - added:
field LlavaNextImageProcessorFast.valid_kwargs - added:
method LlavaNextImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LlavaNextFastImageProcessorKwargs]) -> BatchFeature
- added:
class LlavaNextForConditionalGeneration(LlavaNextPreTrainedModel, GenerationMixin) - added:
class LlavaNextModel(LlavaNextPreTrainedModel) - added:
class LlavaNextPreTrainedModel(PreTrainedModel) - added:
constructor LlavaNextForConditionalGeneration(config: LlavaNextConfig) - added:
constructor LlavaNextModel(config: LlavaNextConfig) - added:
field LlavaNextPreTrainedModel.base_model_prefix - added:
field LlavaNextPreTrainedModel.config: LlavaNextConfig - added:
field LlavaNextPreTrainedModel.supports_gradient_checkpointing - added:
method LlavaNextForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]… - added:
method LlavaNextForConditionalGeneration.get_decoder() - added:
method LlavaNextForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextForConditionalGeneration.get_input_embeddings() - added:
method LlavaNextForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method LlavaNextForConditionalGeneration.pack_image_features(image_features, image_sizes, vision_feature_select_strategy, image_newline=…) - added:
method LlavaNextForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, image_sizes=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaNextForConditionalGeneration.set_decoder(decoder) - added:
method LlavaNextForConditionalGeneration.set_input_embeddings(value) - added:
method LlavaNextModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_… - added:
method LlavaNextModel.get_decoder() - added:
method LlavaNextModel.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextModel.get_input_embeddings() - added:
method LlavaNextModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method LlavaNextModel.pack_image_features(image_features, image_sizes, vision_feature_select_strategy, image_newline=…) - added:
method LlavaNextModel.set_decoder(decoder) - added:
method LlavaNextModel.set_input_embeddings(value) - added:
property LlavaNextForConditionalGeneration.language_model - added:
property LlavaNextForConditionalGeneration.multi_modal_projector - added:
property LlavaNextForConditionalGeneration.vision_tower
- added:
class LlavaNextProcessor(ProcessorMixin) - added:
constructor LlavaNextProcessor(image_processor=…, tokenizer=…, patch_size=…, vision_feature_select_strategy=…, chat_template=…, image_token=…, num_additional_image_tokens=…, **kwargs) - added:
field LlavaNextProcessor.attributes - added:
field LlavaNextProcessor.image_processor_class - added:
field LlavaNextProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class LlavaNextVideoConfig(PretrainedConfig) - added:
constructor LlavaNextVideoConfig(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, multimodal_projector_bias=…, vision_feature_select_strategy=…, vision_feature_layer=…, image_grid_pinpoints=…, tie_word_embeddings=…, video_token_index=…, spatial_pool_mode=…, spatial_pool_stride=…, image_seq_length=…, video_seq_length=…, **kwargs) - added:
field LlavaNextVideoConfig.attribute_map - added:
field LlavaNextVideoConfig.model_type - added:
field LlavaNextVideoConfig.sub_configs
- added:
class LlavaNextVideoImageProcessor(BaseImageProcessor) - added:
constructor LlavaNextVideoImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, image_grid_pinpoints: Optional[list]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[fl… - added:
field LlavaNextVideoImageProcessor.model_input_names - added:
method LlavaNextVideoImageProcessor.preprocess(images: VideoInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Option… - added:
method LlavaNextVideoImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class LlavaNextVideoForConditionalGeneration(LlavaNextVideoPreTrainedModel, GenerationMixin) - added:
class LlavaNextVideoModel(LlavaNextVideoPreTrainedModel) - added:
class LlavaNextVideoPreTrainedModel(PreTrainedModel) - added:
constructor LlavaNextVideoForConditionalGeneration(config: LlavaNextVideoConfig) - added:
constructor LlavaNextVideoModel(config: LlavaNextVideoConfig) - added:
field LlavaNextVideoPreTrainedModel.base_model_prefix - added:
field LlavaNextVideoPreTrainedModel.config: LlavaNextVideoConfig - added:
field LlavaNextVideoPreTrainedModel.supports_gradient_checkpointing - added:
method LlavaNextVideoForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]… - added:
method LlavaNextVideoForConditionalGeneration.get_decoder() - added:
method LlavaNextVideoForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoForConditionalGeneration.get_input_embeddings() - added:
method LlavaNextVideoForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method LlavaNextVideoForConditionalGeneration.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoForConditionalGeneration.pack_image_features(image_features, image_sizes, vision_feature_select_strategy, image_newline=…) - added:
method LlavaNextVideoForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_values_videos=…, image_sizes=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaNextVideoForConditionalGeneration.set_decoder(decoder) - added:
method LlavaNextVideoForConditionalGeneration.set_input_embeddings(value) - added:
method LlavaNextVideoModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_… - added:
method LlavaNextVideoModel.get_decoder() - added:
method LlavaNextVideoModel.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoModel.get_input_embeddings() - added:
method LlavaNextVideoModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method LlavaNextVideoModel.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoModel.pack_image_features(image_features, image_sizes, vision_feature_select_strategy, image_newline=…) - added:
method LlavaNextVideoModel.set_decoder(decoder) - added:
method LlavaNextVideoModel.set_input_embeddings(value) - added:
property LlavaNextVideoForConditionalGeneration.language_model - added:
property LlavaNextVideoForConditionalGeneration.multi_modal_projector - added:
property LlavaNextVideoForConditionalGeneration.vision_tower
- added:
class LlavaNextVideoConfig(PretrainedConfig) - added:
class LlavaNextVideoForConditionalGeneration(LlavaNextForConditionalGeneration) - added:
class LlavaNextVideoModel(LlavaNextModel) - added:
constructor LlavaNextVideoConfig(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, multimodal_projector_bias=…, vision_feature_select_strategy=…, vision_feature_layer=…, image_grid_pinpoints=…, tie_word_embeddings=…, video_token_index=…, spatial_pool_mode=…, spatial_pool_stride=…, image_seq_length=…, video_seq_length=…, **kwargs) - added:
constructor LlavaNextVideoModel(config: LlavaNextVideoConfig, **super_kwargs) - added:
field LlavaNextVideoConfig.attribute_map - added:
field LlavaNextVideoConfig.model_type - added:
field LlavaNextVideoConfig.sub_configs - added:
method LlavaNextVideoForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]… - added:
method LlavaNextVideoForConditionalGeneration.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_values_videos=…, image_sizes=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaNextVideoModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_… - added:
method LlavaNextVideoModel.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaNextVideoModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method LlavaNextVideoModel.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…)
- added:
class LlavaNextVideoProcessor(ProcessorMixin) - added:
constructor LlavaNextVideoProcessor(video_processor=…, image_processor=…, tokenizer=…, chat_template=…, patch_size=…, vision_feature_select_strategy=…, video_token=…, image_token=…, num_additional_image_tokens=…, **kwargs) - added:
field LlavaNextVideoProcessor.attributes - added:
field LlavaNextVideoProcessor.image_processor_class - added:
field LlavaNextVideoProcessor.tokenizer_class - added:
field LlavaNextVideoProcessor.video_processor_class
- added:
class LlavaNextVideoVideoProcessor(BaseVideoProcessor) - added:
constructor LlavaNextVideoVideoProcessor(**kwargs: Unpack[LlavaNextVideoFastVideoProcessorInitKwargs]) - added:
field LlavaNextVideoVideoProcessor.crop_size - added:
field LlavaNextVideoVideoProcessor.default_to_square - added:
field LlavaNextVideoVideoProcessor.do_center_crop - added:
field LlavaNextVideoVideoProcessor.do_convert_rgb - added:
field LlavaNextVideoVideoProcessor.do_normalize - added:
field LlavaNextVideoVideoProcessor.do_rescale - added:
field LlavaNextVideoVideoProcessor.do_resize - added:
field LlavaNextVideoVideoProcessor.do_sample_frames - added:
field LlavaNextVideoVideoProcessor.image_mean - added:
field LlavaNextVideoVideoProcessor.image_std - added:
field LlavaNextVideoVideoProcessor.model_input_names - added:
field LlavaNextVideoVideoProcessor.resample - added:
field LlavaNextVideoVideoProcessor.size - added:
field LlavaNextVideoVideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class LlavaOnevisionConfig(PretrainedConfig) - added:
constructor LlavaOnevisionConfig(vision_config=…, text_config=…, image_token_index=…, video_token_index=…, projector_hidden_act=…, vision_feature_select_strategy=…, vision_feature_layer=…, vision_aspect_ratio=…, image_grid_pinpoints=…, tie_word_embeddings=…, multimodal_projector_bias=…, **kwargs) - added:
field LlavaOnevisionConfig.attribute_map - added:
field LlavaOnevisionConfig.model_type - added:
field LlavaOnevisionConfig.sub_configs
- added:
class LlavaOnevisionImageProcessor(BaseImageProcessor) - added:
constructor LlavaOnevisionImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, image_grid_pinpoints: Optional[list]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: Optional[bool]=…, do_convert_rgb: bool=…, **k… - added:
field LlavaOnevisionImageProcessor.model_input_names - added:
method LlavaOnevisionImageProcessor.get_image_patches(image: np.ndarray, grid_pinpoints, size: tuple, patch_size: int, resample: PILImageResampling, data_format: ChannelDimension, input_data_format: ChannelDimension) -> list[np.ndarray] - added:
method LlavaOnevisionImageProcessor.pad(image: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method LlavaOnevisionImageProcessor.pad_to_square(image: np.ndarray, background_color: Union[int, tuple[int, int, int]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method LlavaOnevisionImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, image_grid_pinpoints: Optional[list]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[f…
- added:
class LlavaOnevisionImageProcessorFast(BaseImageProcessorFast) - added:
constructor LlavaOnevisionImageProcessorFast(**kwargs: Unpack[LlavaOnevisionFastImageProcessorKwargs]) - added:
field LlavaOnevisionImageProcessorFast.crop_size - added:
field LlavaOnevisionImageProcessorFast.default_to_square - added:
field LlavaOnevisionImageProcessorFast.do_center_crop - added:
field LlavaOnevisionImageProcessorFast.do_convert_rgb - added:
field LlavaOnevisionImageProcessorFast.do_normalize - added:
field LlavaOnevisionImageProcessorFast.do_pad - added:
field LlavaOnevisionImageProcessorFast.do_rescale - added:
field LlavaOnevisionImageProcessorFast.do_resize - added:
field LlavaOnevisionImageProcessorFast.image_grid_pinpoints - added:
field LlavaOnevisionImageProcessorFast.image_mean - added:
field LlavaOnevisionImageProcessorFast.image_std - added:
field LlavaOnevisionImageProcessorFast.model_input_names - added:
field LlavaOnevisionImageProcessorFast.resample - added:
field LlavaOnevisionImageProcessorFast.size - added:
field LlavaOnevisionImageProcessorFast.valid_kwargs - added:
method LlavaOnevisionImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method LlavaOnevisionImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LlavaOnevisionFastImageProcessorKwargs]) -> BatchFeature
- added:
class LlavaOnevisionForConditionalGeneration(LlavaOnevisionPreTrainedModel, GenerationMixin) - added:
class LlavaOnevisionModel(LlavaOnevisionPreTrainedModel) - added:
class LlavaOnevisionPreTrainedModel(PreTrainedModel) - added:
constructor LlavaOnevisionForConditionalGeneration(config: LlavaOnevisionConfig) - added:
constructor LlavaOnevisionModel(config) - added:
field LlavaOnevisionPreTrainedModel.base_model_prefix - added:
field LlavaOnevisionPreTrainedModel.config: LlavaOnevisionConfig - added:
field LlavaOnevisionPreTrainedModel.supports_gradient_checkpointing - added:
method LlavaOnevisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes_videos: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Ca… - added:
method LlavaOnevisionForConditionalGeneration.get_decoder() - added:
method LlavaOnevisionForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaOnevisionForConditionalGeneration.get_input_embeddings() - added:
method LlavaOnevisionForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method LlavaOnevisionForConditionalGeneration.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method LlavaOnevisionForConditionalGeneration.pack_image_features(image_features, image_sizes, vision_feature_select_strategy, image_newline=…) - added:
method LlavaOnevisionForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, image_sizes=…, pixel_values_videos=…, image_sizes_videos=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaOnevisionForConditionalGeneration.set_decoder(decoder) - added:
method LlavaOnevisionForConditionalGeneration.set_input_embeddings(value) - added:
method LlavaOnevisionModel.apply_pooling(image_features) - added:
method LlavaOnevisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes_videos: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embe… - added:
method LlavaOnevisionModel.get_decoder() - added:
method LlavaOnevisionModel.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, vision_aspect_ratio: Optional[str]=…, batch_num_images: Optional[torch.LongTensor]=…) - added:
method LlavaOnevisionModel.get_input_embeddings() - added:
method LlavaOnevisionModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method LlavaOnevisionModel.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Union[int, list[int]], vision_feature_select_strategy: str) - added:
method LlavaOnevisionModel.pack_image_features(image_features, image_sizes, image_newline=…, vision_aspect_ratio=…) - added:
method LlavaOnevisionModel.set_decoder(decoder) - added:
method LlavaOnevisionModel.set_input_embeddings(value) - added:
property LlavaOnevisionForConditionalGeneration.language_model - added:
property LlavaOnevisionForConditionalGeneration.multi_modal_projector - added:
property LlavaOnevisionForConditionalGeneration.vision_tower
- added:
class LlavaOnevisionForConditionalGeneration(LlavaNextVideoForConditionalGeneration) - added:
class LlavaOnevisionImageProcessorFast(LlavaNextImageProcessorFast) - added:
class LlavaOnevisionModel(LlavaNextVideoModel) - added:
class LlavaOnevisionPreTrainedModel(LlavaNextVideoPreTrainedModel) - added:
constructor LlavaOnevisionModel(config) - added:
field LlavaOnevisionImageProcessorFast.crop_size - added:
field LlavaOnevisionImageProcessorFast.default_to_square - added:
field LlavaOnevisionImageProcessorFast.do_center_crop - added:
field LlavaOnevisionImageProcessorFast.do_convert_rgb - added:
field LlavaOnevisionImageProcessorFast.do_normalize - added:
field LlavaOnevisionImageProcessorFast.do_pad - added:
field LlavaOnevisionImageProcessorFast.do_rescale - added:
field LlavaOnevisionImageProcessorFast.do_resize - added:
field LlavaOnevisionImageProcessorFast.image_grid_pinpoints - added:
field LlavaOnevisionImageProcessorFast.image_mean - added:
field LlavaOnevisionImageProcessorFast.image_std - added:
field LlavaOnevisionImageProcessorFast.model_input_names - added:
field LlavaOnevisionImageProcessorFast.resample - added:
field LlavaOnevisionImageProcessorFast.size - added:
method LlavaOnevisionForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes_videos: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Ca… - added:
method LlavaOnevisionForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, image_sizes=…, pixel_values_videos=…, image_sizes_videos=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method LlavaOnevisionImageProcessorFast.pad_to_square(images: torch.Tensor, background_color: Union[int, tuple[int, int, int]]=…) -> torch.Tensor - added:
method LlavaOnevisionImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[LlavaOnevisionFastImageProcessorKwargs]) -> BatchFeature - added:
method LlavaOnevisionModel.apply_pooling(image_features) - added:
method LlavaOnevisionModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_sizes_videos: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embe… - added:
method LlavaOnevisionModel.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…, vision_aspect_ratio: Optional[str]=…, batch_num_images: Optional[torch.LongTensor]=…) - added:
method LlavaOnevisionModel.get_video_features(pixel_values: torch.FloatTensor, vision_feature_layer: Union[int, list[int]], vision_feature_select_strategy: str) - added:
method LlavaOnevisionModel.pack_image_features(image_features, image_sizes, image_newline=…, vision_aspect_ratio=…)
- added:
class LlavaOnevisionProcessor(ProcessorMixin) - added:
constructor LlavaOnevisionProcessor(image_processor=…, tokenizer=…, video_processor=…, num_image_tokens=…, vision_feature_select_strategy=…, chat_template=…, image_token=…, video_token=…, vision_aspect_ratio=…, **kwargs) - added:
field LlavaOnevisionProcessor.attributes - added:
field LlavaOnevisionProcessor.image_processor_class - added:
field LlavaOnevisionProcessor.tokenizer_class - added:
field LlavaOnevisionProcessor.video_processor_class
- added:
class LlavaOnevisionVideoProcessor(BaseVideoProcessor) - added:
constructor LlavaOnevisionVideoProcessor(**kwargs: Unpack[LlavaOnevisionFastVideoProcessorInitKwargs]) - added:
field LlavaOnevisionVideoProcessor.crop_size - added:
field LlavaOnevisionVideoProcessor.default_to_square - added:
field LlavaOnevisionVideoProcessor.do_center_crop - added:
field LlavaOnevisionVideoProcessor.do_convert_rgb - added:
field LlavaOnevisionVideoProcessor.do_normalize - added:
field LlavaOnevisionVideoProcessor.do_rescale - added:
field LlavaOnevisionVideoProcessor.do_resize - added:
field LlavaOnevisionVideoProcessor.do_sample_frames - added:
field LlavaOnevisionVideoProcessor.image_mean - added:
field LlavaOnevisionVideoProcessor.image_std - added:
field LlavaOnevisionVideoProcessor.model_input_names - added:
field LlavaOnevisionVideoProcessor.resample - added:
field LlavaOnevisionVideoProcessor.rescale_factor - added:
field LlavaOnevisionVideoProcessor.size - added:
field LlavaOnevisionVideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class LongcatFlashConfig(PretrainedConfig) - added:
constructor LongcatFlashConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, ffn_hidden_size=…, q_lora_r… - added:
field LongcatFlashConfig.base_model_pp_plan - added:
field LongcatFlashConfig.base_model_tp_plan - added:
field LongcatFlashConfig.keys_to_ignore_at_inference - added:
field LongcatFlashConfig.model_type
- added:
class LongcatFlashForCausalLM(LongcatFlashPreTrainedModel, GenerationMixin) - added:
class LongcatFlashModel(LongcatFlashPreTrainedModel) - added:
class LongcatFlashPreTrainedModel(PreTrainedModel) - added:
constructor LongcatFlashForCausalLM(config) - added:
constructor LongcatFlashModel(config) - added:
field LongcatFlashPreTrainedModel.base_model_prefix - added:
field LongcatFlashPreTrainedModel.config: LongcatFlashConfig - added:
field LongcatFlashPreTrainedModel.supports_gradient_checkpointing - added:
method LongcatFlashForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tenso… - added:
method LongcatFlashModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class LongcatFlashForCausalLM(DeepseekV3ForCausalLM) - added:
class LongcatFlashModel(DeepseekV3Model) - added:
class LongcatFlashPreTrainedModel(DeepseekV3PreTrainedModel) - added:
constructor LongcatFlashForCausalLM(config) - added:
constructor LongcatFlashModel(config) - added:
method LongcatFlashModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs])
- added:
export define_import_structure
- added:
class LongformerConfig(PretrainedConfig) - added:
class LongformerOnnxConfig(OnnxConfig) - added:
constructor LongformerConfig(attention_window: Union[list[int], int]=…, sep_token_id: int=…, pad_token_id: int=…, bos_token_id: int=…, eos_token_id: int=…, vocab_size: int=…, hidden_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, intermediate_size: int=…, hidden_act: str=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…, max_position_embeddings: int=…, type… - added:
constructor LongformerOnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…) - added:
field LongformerConfig.model_type - added:
method LongformerOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizerBase, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property LongformerOnnxConfig.atol_for_validation -> float - added:
property LongformerOnnxConfig.default_onnx_opset -> int - added:
property LongformerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property LongformerOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class LongformerForMaskedLM(LongformerPreTrainedModel) - added:
class LongformerForMultipleChoice(LongformerPreTrainedModel) - added:
class LongformerForQuestionAnswering(LongformerPreTrainedModel) - added:
class LongformerForSequenceClassification(LongformerPreTrainedModel) - added:
class LongformerForTokenClassification(LongformerPreTrainedModel) - added:
class LongformerModel(LongformerPreTrainedModel) - added:
class LongformerPreTrainedModel(PreTrainedModel) - added:
class LongformerSelfAttention(nn.Module) - added:
constructor LongformerForMaskedLM(config) - added:
constructor LongformerForMultipleChoice(config) - added:
constructor LongformerForQuestionAnswering(config) - added:
constructor LongformerForSequenceClassification(config) - added:
constructor LongformerForTokenClassification(config) - added:
constructor LongformerModel(config, add_pooling_layer=…) - added:
constructor LongformerSelfAttention(config, layer_id) - added:
field LongformerPreTrainedModel.base_model_prefix - added:
field LongformerPreTrainedModel.config: LongformerConfig - added:
field LongformerPreTrainedModel.supports_gradient_checkpointing - added:
method LongformerForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, out… - added:
method LongformerForMaskedLM.get_output_embeddings() - added:
method LongformerForMaskedLM.set_output_embeddings(new_embeddings) - added:
method LongformerForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=… - added:
method LongformerForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optiona… - added:
method LongformerForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optiona… - added:
method LongformerForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[b… - added:
method LongformerModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, global_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, ret… - added:
method LongformerModel.get_input_embeddings() - added:
method LongformerModel.set_input_embeddings(value) - added:
method LongformerSelfAttention.forward(hidden_states, attention_mask=…, layer_head_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…, output_attentions=…)
- added:
class TFLongformerForMaskedLM(TFLongformerPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFLongformerForMultipleChoice(TFLongformerPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFLongformerForQuestionAnswering(TFLongformerPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFLongformerForSequenceClassification(TFLongformerPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFLongformerForTokenClassification(TFLongformerPreTrainedModel, TFTokenClassificationLoss) - added:
class TFLongformerModel(TFLongformerPreTrainedModel) - added:
class TFLongformerPreTrainedModel(TFPreTrainedModel) - added:
class TFLongformerSelfAttention(keras.layers.Layer) - added:
constructor TFLongformerForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFLongformerForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFLongformerForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFLongformerForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFLongformerForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFLongformerModel(config, *inputs, **kwargs) - added:
constructor TFLongformerSelfAttention(config, layer_id, **kwargs) - added:
field TFLongformerPreTrainedModel.base_model_prefix - added:
field TFLongformerPreTrainedModel.config_class - added:
method TFLongformerForMaskedLM.build(input_shape=…) - added:
method TFLongformerForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…… - added:
method TFLongformerForMaskedLM.get_lm_head() - added:
method TFLongformerForMaskedLM.get_prefix_bias_name() - added:
method TFLongformerForMultipleChoice.build(input_shape=…) - added:
method TFLongformerForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool |… - added:
method TFLongformerForQuestionAnswering.build(input_shape=…) - added:
method TFLongformerForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool… - added:
method TFLongformerForSequenceClassification.build(input_shape=…) - added:
method TFLongformerForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions:… - added:
method TFLongformerForTokenClassification.build(input_shape=…) - added:
method TFLongformerForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bo… - added:
method TFLongformerModel.build(input_shape=…) - added:
method TFLongformerModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, global_attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, outp… - added:
method TFLongformerSelfAttention.build(input_shape=…) - added:
method TFLongformerSelfAttention.call(inputs, training=…) - added:
method TFLongformerSelfAttention.reshape_and_transpose(vector, batch_size) - added:
property TFLongformerForMultipleChoice.input_signature - added:
property TFLongformerPreTrainedModel.input_signature
- added:
class LongformerTokenizer(PreTrainedTokenizer) - added:
constructor LongformerTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field LongformerTokenizer.model_input_names - added:
field LongformerTokenizer.vocab_files_names - added:
method LongformerTokenizer.bpe(token) - added:
method LongformerTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LongformerTokenizer.convert_tokens_to_string(tokens) - added:
method LongformerTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LongformerTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LongformerTokenizer.get_vocab() - added:
method LongformerTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method LongformerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LongformerTokenizer.vocab_size
- added:
class LongformerTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LongformerTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field LongformerTokenizerFast.model_input_names - added:
field LongformerTokenizerFast.slow_tokenizer_class - added:
field LongformerTokenizerFast.vocab_files_names - added:
method LongformerTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LongformerTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LongformerTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LongformerTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class LongT5Config(PretrainedConfig) - added:
class LongT5OnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor LongT5Config(vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, local_radius=…, global_block_size=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, encoder_attention_type=…, use_cache=…, pad_token_id=…, eos_token_id=…, **kw… - added:
field LongT5Config.attribute_map - added:
field LongT5Config.keys_to_ignore_at_inference - added:
field LongT5Config.model_type - added:
property LongT5OnnxConfig.default_onnx_opset -> int - added:
property LongT5OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxLongT5ForConditionalGeneration(FlaxLongT5PreTrainedModel) - added:
class FlaxLongT5Model(FlaxLongT5PreTrainedModel) - added:
class FlaxLongT5PreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxLongT5PreTrainedModel(config: LongT5Config, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxLongT5ForConditionalGeneration.module_class - added:
field FlaxLongT5Model.module_class - added:
field FlaxLongT5PreTrainedModel.base_model_prefix - added:
field FlaxLongT5PreTrainedModel.config_class - added:
field FlaxLongT5PreTrainedModel.module_class: nn.Module - added:
method FlaxLongT5ForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxLongT5ForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxLongT5ForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxLongT5PreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxLongT5PreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxLongT5PreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxLongT5PreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxLongT5PreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class LongT5EncoderModel(LongT5PreTrainedModel) - added:
class LongT5ForConditionalGeneration(LongT5PreTrainedModel, GenerationMixin) - added:
class LongT5Model(LongT5PreTrainedModel) - added:
class LongT5PreTrainedModel(PreTrainedModel) - added:
constructor LongT5EncoderModel(config: LongT5Config) - added:
constructor LongT5ForConditionalGeneration(config: LongT5Config) - added:
constructor LongT5Model(config: LongT5Config) - added:
field LongT5PreTrainedModel.base_model_prefix - added:
field LongT5PreTrainedModel.config: LongT5Config - added:
field LongT5PreTrainedModel.supports_gradient_checkpointing - added:
method LongT5EncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method LongT5EncoderModel.get_encoder() - added:
method LongT5EncoderModel.get_input_embeddings() - added:
method LongT5EncoderModel.set_input_embeddings(new_embeddings) - added:
method LongT5ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optio… - added:
method LongT5ForConditionalGeneration.get_encoder() - added:
method LongT5ForConditionalGeneration.get_input_embeddings() - added:
method LongT5ForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method LongT5ForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method LongT5Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[tor… - added:
method LongT5Model.get_encoder() - added:
method LongT5Model.get_input_embeddings() - added:
method LongT5Model.set_input_embeddings(new_embeddings) - added:
property LongT5PreTrainedModel.dummy_inputs - added:
static method LongT5PreTrainedModel.from_pretrained(*args, **kwargs)
- added:
export define_import_structure
- added:
class LukeConfig(PretrainedConfig) - added:
constructor LukeConfig(vocab_size=…, entity_vocab_size=…, hidden_size=…, entity_emb_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_entity_aware_attention=…, classifier_dropout=…, pad_token_id=…, bos_token_id=…, eos_… - added:
field LukeConfig.model_type
- added:
class LukeForEntityClassification(LukePreTrainedModel) - added:
class LukeForEntityPairClassification(LukePreTrainedModel) - added:
class LukeForEntitySpanClassification(LukePreTrainedModel) - added:
class LukeForMaskedLM(LukePreTrainedModel) - added:
class LukeForMultipleChoice(LukePreTrainedModel) - added:
class LukeForQuestionAnswering(LukePreTrainedModel) - added:
class LukeForSequenceClassification(LukePreTrainedModel) - added:
class LukeForTokenClassification(LukePreTrainedModel) - added:
class LukeModel(LukePreTrainedModel) - added:
class LukePreTrainedModel(PreTrainedModel) - added:
constructor LukeForEntityClassification(config) - added:
constructor LukeForEntityPairClassification(config) - added:
constructor LukeForEntitySpanClassification(config) - added:
constructor LukeForMaskedLM(config) - added:
constructor LukeForMultipleChoice(config) - added:
constructor LukeForQuestionAnswering(config) - added:
constructor LukeForSequenceClassification(config) - added:
constructor LukeForTokenClassification(config) - added:
constructor LukeModel(config: LukeConfig, add_pooling_layer: bool=…) - added:
field LukePreTrainedModel.base_model_prefix - added:
field LukePreTrainedModel.config: LukeConfig - added:
field LukePreTrainedModel.supports_gradient_checkpointing - added:
method LukeForEntityClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional… - added:
method LukeForEntityPairClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Opti… - added:
method LukeForEntitySpanClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.LongTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optio… - added:
method LukeForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.LongTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional[torch.LongTe… - added:
method LukeForMaskedLM.get_output_embeddings() - added:
method LukeForMaskedLM.set_output_embeddings(new_embeddings) - added:
method LukeForMaskedLM.tie_weights() - added:
method LukeForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional[torch… - added:
method LukeForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.FloatTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional[t… - added:
method LukeForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Option… - added:
method LukeForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional[… - added:
method LukeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, entity_ids: Optional[torch.LongTensor]=…, entity_attention_mask: Optional[torch.FloatTensor]=…, entity_token_type_ids: Optional[torch.LongTensor]=…, entity_position_ids: Optional[torch.LongTensor]… - added:
method LukeModel.get_entity_embeddings() - added:
method LukeModel.get_extended_attention_mask(word_attention_mask: torch.LongTensor, entity_attention_mask: Optional[torch.LongTensor]) - added:
method LukeModel.get_input_embeddings() - added:
method LukeModel.set_entity_embeddings(value) - added:
method LukeModel.set_input_embeddings(value)
- added:
class LukeTokenizer(PreTrainedTokenizer) - added:
constructor LukeTokenizer(vocab_file, merges_file, entity_vocab_file, task=…, max_entity_length=…, max_mention_length=…, entity_token_1=…, entity_token_2=…, entity_unk_token=…, entity_pad_token=…, entity_mask_token=…, entity_mask2_token=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field LukeTokenizer.model_input_names - added:
field LukeTokenizer.vocab_files_names - added:
method LukeTokenizer.bpe(token) - added:
method LukeTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LukeTokenizer.convert_tokens_to_string(tokens) - added:
method LukeTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LukeTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LukeTokenizer.get_vocab() - added:
method LukeTokenizer.pad(encoded_inputs: Union[BatchEncoding, list[BatchEncoding], dict[str, EncodedInput], dict[str, list[EncodedInput]], list[dict[str, EncodedInput]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, max_entity_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_attention_mask: Optional[bool]=…, return_t… - added:
method LukeTokenizer.prepare_for_model(ids: list[int], pair_ids: Optional[list[int]]=…, entity_ids: Optional[list[int]]=…, pair_entity_ids: Optional[list[int]]=…, entity_token_spans: Optional[list[tuple[int, int]]]=…, pair_entity_token_spans: Optional[list[tuple[int, int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy… - added:
method LukeTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method LukeTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LukeTokenizer.vocab_size
- added:
export define_import_structure
- added:
class LxmertConfig(PretrainedConfig) - added:
constructor LxmertConfig(vocab_size=…, hidden_size=…, num_attention_heads=…, num_qa_labels=…, num_object_labels=…, num_attr_labels=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, l_layers=…, x_layers=…, r_layers=…, visual_feat_dim=…, visual_pos_dim=…, visual_loss_norma… - added:
field LxmertConfig.attribute_map - added:
field LxmertConfig.model_type
- added:
class LxmertEncoder(nn.Module) - added:
class LxmertForPreTraining(LxmertPreTrainedModel) - added:
class LxmertForQuestionAnswering(LxmertPreTrainedModel) - added:
class LxmertModel(LxmertPreTrainedModel) - added:
class LxmertPreTrainedModel(PreTrainedModel) - added:
class LxmertVisualFeatureEncoder(nn.Module) - added:
class LxmertXLayer(nn.Module) - added:
constructor LxmertEncoder(config) - added:
constructor LxmertForPreTraining(config) - added:
constructor LxmertForQuestionAnswering(config) - added:
constructor LxmertModel(config) - added:
constructor LxmertVisualFeatureEncoder(config) - added:
constructor LxmertXLayer(config) - added:
field LxmertPreTrainedModel.base_model_prefix - added:
field LxmertPreTrainedModel.config: LxmertConfig - added:
field LxmertPreTrainedModel.load_tf_weights - added:
method LxmertEncoder.forward(lang_feats, lang_attention_mask, visual_feats, visual_pos, visual_attention_mask=…, output_attentions=…) - added:
method LxmertForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, visual_feats: Optional[torch.FloatTensor]=…, visual_pos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, obj… - added:
method LxmertForPreTraining.get_qa_logit_layer() -> nn.Module - added:
method LxmertForPreTraining.resize_num_qa_labels(num_labels) - added:
method LxmertForPreTraining.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method LxmertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, visual_feats: Optional[torch.FloatTensor]=…, visual_pos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.Tensor]=…, o… - added:
method LxmertForQuestionAnswering.get_qa_logit_layer() -> nn.Module - added:
method LxmertForQuestionAnswering.resize_num_qa_labels(num_labels) - added:
method LxmertModel.forward(input_ids: Optional[torch.LongTensor]=…, visual_feats: Optional[torch.FloatTensor]=…, visual_pos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method LxmertModel.get_input_embeddings() - added:
method LxmertModel.set_input_embeddings(new_embeddings) - added:
method LxmertVisualFeatureEncoder.forward(visual_feats, visual_pos) - added:
method LxmertXLayer.cross_att(lang_input, lang_attention_mask, visual_input, visual_attention_mask, output_x_attentions=…) - added:
method LxmertXLayer.forward(lang_feats, lang_attention_mask, visual_feats, visual_attention_mask, output_attentions=…) - added:
method LxmertXLayer.output_fc(lang_input, visual_input) - added:
method LxmertXLayer.self_att(lang_input, lang_attention_mask, visual_input, visual_attention_mask)
- added:
class TFLxmertForPreTraining(TFLxmertPreTrainedModel) - added:
class TFLxmertMainLayer(keras.layers.Layer) - added:
class TFLxmertModel(TFLxmertPreTrainedModel) - added:
class TFLxmertPreTrainedModel(TFPreTrainedModel) - added:
class TFLxmertVisualFeatureEncoder(keras.layers.Layer) - added:
constructor TFLxmertForPreTraining(config, *inputs, **kwargs) - added:
constructor TFLxmertMainLayer(config, **kwargs) - added:
constructor TFLxmertModel(config, *inputs, **kwargs) - added:
constructor TFLxmertVisualFeatureEncoder(config, **kwargs) - added:
field TFLxmertMainLayer.config_class - added:
field TFLxmertPreTrainedModel.base_model_prefix - added:
field TFLxmertPreTrainedModel.config_class - added:
method TFLxmertForPreTraining.build(input_shape=…) - added:
method TFLxmertForPreTraining.call(input_ids: TFModelInputType | None=…, visual_feats: tf.Tensor | None=…, visual_pos: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, visual_attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, masked_lm_labels: tf.Tensor | None=…, obj_labels: dict[str, tuple[tf.Tensor, tf.Tensor]] | None=…, matched_lab… - added:
method TFLxmertForPreTraining.get_lm_head() - added:
method TFLxmertForPreTraining.get_prefix_bias_name() - added:
method TFLxmertMainLayer.build(input_shape=…) - added:
method TFLxmertMainLayer.call(input_ids=…, visual_feats=…, visual_pos=…, attention_mask=…, visual_attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFLxmertMainLayer.get_input_embeddings() - added:
method TFLxmertMainLayer.set_input_embeddings(value) - added:
method TFLxmertModel.build(input_shape=…) - added:
method TFLxmertModel.call(input_ids: TFModelInputType | None=…, visual_feats: tf.Tensor | None=…, visual_pos: tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, visual_attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None… - added:
method TFLxmertVisualFeatureEncoder.build(input_shape=…) - added:
method TFLxmertVisualFeatureEncoder.call(visn_input, training=…) - added:
property TFLxmertForPreTraining.dummy_inputs - added:
property TFLxmertPreTrainedModel.dummy_inputs - added:
property TFLxmertPreTrainedModel.input_signature
- added:
class LxmertTokenizer(PreTrainedTokenizer) - added:
constructor LxmertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field LxmertTokenizer.vocab_files_names - added:
method LxmertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method LxmertTokenizer.convert_tokens_to_string(tokens) - added:
method LxmertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method LxmertTokenizer.get_vocab() - added:
method LxmertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property LxmertTokenizer.do_lower_case - added:
property LxmertTokenizer.vocab_size
- added:
class LxmertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor LxmertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field LxmertTokenizerFast.slow_tokenizer_class - added:
field LxmertTokenizerFast.vocab_files_names - added:
method LxmertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method LxmertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class M2M100Config(PretrainedConfig) - added:
class M2M100OnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor M2M100Config(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…, sc… - added:
field M2M100Config.attribute_map - added:
field M2M100Config.keys_to_ignore_at_inference - added:
field M2M100Config.model_type - added:
field M2M100OnnxConfig.generate_dummy_inputs - added:
property M2M100OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class M2M100ForConditionalGeneration(M2M100PreTrainedModel, GenerationMixin) - added:
class M2M100Model(M2M100PreTrainedModel) - added:
class M2M100PreTrainedModel(PreTrainedModel) - added:
constructor M2M100ForConditionalGeneration(config: M2M100Config) - added:
constructor M2M100Model(config: M2M100Config) - added:
field M2M100ForConditionalGeneration.base_model_prefix - added:
field M2M100PreTrainedModel.base_model_prefix - added:
field M2M100PreTrainedModel.config: M2M100Config - added:
field M2M100PreTrainedModel.supports_gradient_checkpointing - added:
method M2M100ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple… - added:
method M2M100ForConditionalGeneration.get_decoder() - added:
method M2M100ForConditionalGeneration.get_encoder() - added:
method M2M100Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]… - added:
method M2M100Model.get_encoder() - added:
method M2M100Model.get_input_embeddings() - added:
method M2M100Model.set_input_embeddings(value)
- added:
class M2M100Tokenizer(PreTrainedTokenizer) - added:
constructor M2M100Tokenizer(vocab_file, spm_file, src_lang=…, tgt_lang=…, bos_token=…, eos_token=…, sep_token=…, pad_token=…, unk_token=…, language_codes=…, sp_model_kwargs: Optional[dict[str, Any]]=…, num_madeup_words=…, **kwargs) -> None - added:
field M2M100Tokenizer.model_input_names - added:
field M2M100Tokenizer.prefix_tokens: list[int] - added:
field M2M100Tokenizer.suffix_tokens: list[int] - added:
field M2M100Tokenizer.vocab_files_names - added:
method M2M100Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method M2M100Tokenizer.convert_tokens_to_string(tokens) - added:
method M2M100Tokenizer.get_lang_id(lang: str) -> int - added:
method M2M100Tokenizer.get_lang_token(lang: str) -> str - added:
method M2M100Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method M2M100Tokenizer.get_vocab() -> dict - added:
method M2M100Tokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method M2M100Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method M2M100Tokenizer.set_src_lang_special_tokens(src_lang: str) -> None - added:
method M2M100Tokenizer.set_tgt_lang_special_tokens(tgt_lang: str) -> None - added:
property M2M100Tokenizer.src_lang -> str - added:
property M2M100Tokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class MambaConfig(PretrainedConfig) - added:
constructor MambaConfig(vocab_size=…, hidden_size=…, state_size=…, num_hidden_layers=…, layer_norm_epsilon=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, expand=…, conv_kernel=…, use_bias=…, use_conv_bias=…, hidden_act=…, initializer_range=…, residual_in_fp32=…, time_step_rank=…, time_step_scale=…, time_step_min=…, time_step_max=…, time_step_init_scheme=…, time_step_floor=…, rescale_prenorm_r… - added:
field MambaConfig.model_type
- added:
class MambaCache - added:
class MambaForCausalLM(MambaPreTrainedModel, GenerationMixin) - added:
class MambaModel(MambaPreTrainedModel) - added:
class MambaPreTrainedModel(PreTrainedModel) - added:
constructor MambaCache(config: PretrainedConfig, max_batch_size: int, dtype: torch.dtype=…, device: Union[torch.device, str, None]=…) - added:
constructor MambaForCausalLM(config) - added:
constructor MambaModel(config) - added:
field MambaCache.is_compileable - added:
field MambaPreTrainedModel.base_model_prefix - added:
field MambaPreTrainedModel.config: MambaConfig - added:
field MambaPreTrainedModel.supports_gradient_checkpointing - added:
method MambaCache.reset() - added:
method MambaCache.update_conv_state(layer_idx: int, new_conv_state: torch.Tensor, cache_position: torch.LongTensor) -> torch.Tensor - added:
method MambaCache.update_ssm_state(layer_idx: int, new_ssm_state: torch.Tensor) - added:
method MambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_params: Optional[MambaCache]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…, **kwargs) -> Uni… - added:
method MambaForCausalLM.get_input_embeddings() - added:
method MambaForCausalLM.prepare_inputs_for_generation(input_ids, inputs_embeds=…, use_cache=…, cache_params: Optional[MambaCache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, **kwargs) - added:
method MambaForCausalLM.set_input_embeddings(new_embeddings) - added:
method MambaModel.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, cache_params: Optional[MambaCache]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…) -> Union[tuple, MambaOutput] - added:
method MambaModel.get_input_embeddings() - added:
method MambaModel.load_hook(state_dict, prefix, *args) - added:
method MambaModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class Mamba2Config(PretrainedConfig) - added:
constructor Mamba2Config(num_heads=…, head_dim=…, vocab_size=…, hidden_size=…, state_size=…, num_hidden_layers=…, layer_norm_epsilon=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, expand=…, conv_kernel=…, n_groups=…, use_bias=…, use_conv_bias=…, hidden_act=…, initializer_range=…, residual_in_fp32=…, time_step_rank=…, time_step_min=…, time_step_max=…, time_step_floor=…, time_step_limit=…, resc… - added:
field Mamba2Config.model_type
- added:
class Mamba2ForCausalLM(Mamba2PreTrainedModel, GenerationMixin) - added:
class Mamba2Model(Mamba2PreTrainedModel) - added:
class Mamba2PreTrainedModel(PreTrainedModel) - added:
constructor Mamba2ForCausalLM(config) - added:
constructor Mamba2Model(config) - added:
field Mamba2PreTrainedModel.base_model_prefix - added:
field Mamba2PreTrainedModel.config: Mamba2Config - added:
field Mamba2PreTrainedModel.supports_gradient_checkpointing - added:
method Mamba2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_params: Optional[Mamba2Cache]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) -> Union… - added:
method Mamba2ForCausalLM.get_input_embeddings() - added:
method Mamba2ForCausalLM.prepare_inputs_for_generation(input_ids, inputs_embeds=…, use_cache=…, cache_params: Optional[Mamba2Cache]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) - added:
method Mamba2ForCausalLM.set_input_embeddings(new_embeddings) - added:
method Mamba2Model.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, cache_params: Optional[Mamba2Cache]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, **kwargs) -> Union[tuple, Mamba2Output] - added:
method Mamba2Model.get_input_embeddings() - added:
method Mamba2Model.load_hook(state_dict, prefix, *args) - added:
method Mamba2Model.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class MarianConfig(PretrainedConfig) - added:
class MarianOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor MarianConfig(vocab_size=…, decoder_vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decode… - added:
field MarianConfig.attribute_map - added:
field MarianConfig.keys_to_ignore_at_inference - added:
field MarianConfig.model_type - added:
method MarianOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property MarianOnnxConfig.atol_for_validation -> float - added:
property MarianOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MarianOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxMarianMTModel(FlaxMarianPreTrainedModel) - added:
class FlaxMarianModel(FlaxMarianPreTrainedModel) - added:
class FlaxMarianPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxMarianPreTrainedModel(config: MarianConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxMarianMTModel.dtype: jnp.dtype - added:
field FlaxMarianMTModel.module_class - added:
field FlaxMarianModel.config: MarianConfig - added:
field FlaxMarianModel.dtype: jnp.dtype - added:
field FlaxMarianModel.module_class - added:
field FlaxMarianPreTrainedModel.base_model_prefix: str - added:
field FlaxMarianPreTrainedModel.config_class - added:
field FlaxMarianPreTrainedModel.module_class: nn.Module - added:
method FlaxMarianMTModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dr… - added:
method FlaxMarianMTModel.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxMarianMTModel.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxMarianPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dic… - added:
method FlaxMarianPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxMarianPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxMarianPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class MarianForCausalLM(MarianPreTrainedModel, GenerationMixin) - added:
class MarianMTModel(MarianPreTrainedModel, GenerationMixin) - added:
class MarianModel(MarianPreTrainedModel) - added:
class MarianPreTrainedModel(PreTrainedModel) - added:
constructor MarianForCausalLM(config) - added:
constructor MarianMTModel(config: MarianConfig) - added:
constructor MarianModel(config: MarianConfig) - added:
field MarianMTModel.base_model_prefix - added:
field MarianPreTrainedModel.base_model_prefix - added:
field MarianPreTrainedModel.config: MarianConfig - added:
field MarianPreTrainedModel.supports_gradient_checkpointing - added:
method MarianForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method MarianForCausalLM.get_decoder() - added:
method MarianForCausalLM.get_input_embeddings() - added:
method MarianForCausalLM.set_decoder(decoder) - added:
method MarianForCausalLM.set_input_embeddings(value) - added:
method MarianMTModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Union[tuple[torch.Tensor], BaseM… - added:
method MarianMTModel.get_decoder() - added:
method MarianMTModel.get_encoder() - added:
method MarianMTModel.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MarianMTModel.resize_decoder_token_embeddings(new_num_tokens) - added:
method MarianMTModel.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method MarianMTModel.set_output_embeddings(new_embeddings: nn.Embedding) - added:
method MarianMTModel.tie_weights() - added:
method MarianModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[Union[tuple[torch.Tensor], BaseMod… - added:
method MarianModel.get_decoder_input_embeddings() - added:
method MarianModel.get_encoder() - added:
method MarianModel.get_input_embeddings() - added:
method MarianModel.resize_decoder_token_embeddings(new_num_tokens: int) -> nn.Embedding - added:
method MarianModel.set_decoder_input_embeddings(value) - added:
method MarianModel.set_input_embeddings(value) - added:
property MarianPreTrainedModel.dummy_inputs
- added:
class TFMarianMTModel(TFMarianPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFMarianModel(TFMarianPreTrainedModel) - added:
class TFMarianPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMarianMTModel(config, *inputs, **kwargs) - added:
constructor TFMarianModel(config: MarianConfig, *inputs, **kwargs) - added:
field TFMarianPreTrainedModel.base_model_prefix - added:
field TFMarianPreTrainedModel.config_class - added:
method TFMarianMTModel.build(input_shape=…) - added:
method TFMarianMTModel.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: TFBaseModelOutput | None=…, past_key_values: tuple… - added:
method TFMarianMTModel.get_bias() - added:
method TFMarianMTModel.get_decoder() - added:
method TFMarianMTModel.get_encoder() - added:
method TFMarianMTModel.get_output_embeddings() - added:
method TFMarianMTModel.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFMarianMTModel.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFMarianMTModel.serving_output(output) - added:
method TFMarianMTModel.set_bias(value) - added:
method TFMarianMTModel.set_output_embeddings(value) - added:
method TFMarianModel.build(input_shape=…) - added:
method TFMarianModel.call(input_ids: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: tf.Tensor | None=…, past_key_values: tuple[tuple[tf.… - added:
method TFMarianModel.get_decoder() - added:
method TFMarianModel.get_encoder() - added:
method TFMarianModel.serving_output(output)
- added:
class MarianTokenizer(PreTrainedTokenizer) - added:
constructor MarianTokenizer(source_spm, target_spm, vocab, target_vocab_file=…, source_lang=…, target_lang=…, unk_token=…, eos_token=…, pad_token=…, model_max_length=…, sp_model_kwargs: Optional[dict[str, Any]]=…, separate_vocabs=…, **kwargs) -> None - added:
field MarianTokenizer.model_input_names - added:
field MarianTokenizer.vocab_files_names - added:
method MarianTokenizer.batch_decode(sequences, **kwargs) - added:
method MarianTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method MarianTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method MarianTokenizer.decode(token_ids, **kwargs) - added:
method MarianTokenizer.get_special_tokens_mask(token_ids_0: list, token_ids_1: Optional[list]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MarianTokenizer.get_src_vocab() - added:
method MarianTokenizer.get_tgt_vocab() - added:
method MarianTokenizer.get_vocab() -> dict - added:
method MarianTokenizer.normalize(x: str) -> str - added:
method MarianTokenizer.num_special_tokens_to_add(*args, **kwargs) - added:
method MarianTokenizer.remove_language_code(text: str) - added:
method MarianTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MarianTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class MarkupLMConfig(PretrainedConfig) - added:
constructor MarkupLMConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_xpath_tag_unit_embeddings=…, max_xpath_subs_unit_embeddings=…, tag_pad_i… - added:
field MarkupLMConfig.model_type
- added:
class MarkupLMFeatureExtractor(FeatureExtractionMixin) - added:
constructor MarkupLMFeatureExtractor(**kwargs) - added:
method MarkupLMFeatureExtractor.construct_xpath(xpath_tags, xpath_subscripts) - added:
method MarkupLMFeatureExtractor.get_three_from_single(html_string) - added:
method MarkupLMFeatureExtractor.xpath_soup(element)
- added:
class MarkupLMForQuestionAnswering(MarkupLMPreTrainedModel) - added:
class MarkupLMForSequenceClassification(MarkupLMPreTrainedModel) - added:
class MarkupLMForTokenClassification(MarkupLMPreTrainedModel) - added:
class MarkupLMModel(MarkupLMPreTrainedModel) - added:
class MarkupLMPreTrainedModel(PreTrainedModel) - added:
constructor MarkupLMForQuestionAnswering(config) - added:
constructor MarkupLMForSequenceClassification(config) - added:
constructor MarkupLMForTokenClassification(config) - added:
constructor MarkupLMModel(config, add_pooling_layer=…) - added:
field MarkupLMPreTrainedModel.base_model_prefix - added:
field MarkupLMPreTrainedModel.config: MarkupLMConfig - added:
method MarkupLMForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, xpath_tags_seq: Optional[torch.Tensor]=…, xpath_subs_seq: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.… - added:
method MarkupLMForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, xpath_tags_seq: Optional[torch.Tensor]=…, xpath_subs_seq: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tens… - added:
method MarkupLMForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, xpath_tags_seq: Optional[torch.Tensor]=…, xpath_subs_seq: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]… - added:
method MarkupLMModel.forward(input_ids: Optional[torch.LongTensor]=…, xpath_tags_seq: Optional[torch.LongTensor]=…, xpath_subs_seq: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_atten… - added:
method MarkupLMModel.get_input_embeddings() - added:
method MarkupLMModel.set_input_embeddings(value) - added:
static method MarkupLMPreTrainedModel.from_pretrained(pretrained_model_name_or_path: Optional[Union[str, os.PathLike]], *model_args, **kwargs)
- added:
class MarkupLMProcessor(ProcessorMixin) - added:
field MarkupLMProcessor.feature_extractor_class - added:
field MarkupLMProcessor.parse_html - added:
field MarkupLMProcessor.tokenizer_class
- added:
class MarkupLMTokenizer(PreTrainedTokenizer) - added:
constructor MarkupLMTokenizer(vocab_file, merges_file, tags_dict, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, max_depth=…, max_width=…, pad_width=…, pad_token_label=…, only_label_first_subword=…, **kwargs) - added:
field MarkupLMTokenizer.vocab_files_names - added:
method MarkupLMTokenizer.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, xpaths: Optional[list[list[list[int]]]]=…, node_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_… - added:
method MarkupLMTokenizer.bpe(token) - added:
method MarkupLMTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizer.build_xpath_subs_with_special_tokens(xpath_subs_0: list[int], xpath_subs_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizer.build_xpath_tags_with_special_tokens(xpath_tags_0: list[int], xpath_tags_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizer.convert_tokens_to_string(tokens) - added:
method MarkupLMTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizer.encode(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, xpaths: Optional[list[list[int]]]=…, node_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, pa… - added:
method MarkupLMTokenizer.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, xpaths: Optional[list[list[int]]]=…, node_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=… - added:
method MarkupLMTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MarkupLMTokenizer.get_vocab() - added:
method MarkupLMTokenizer.get_xpath_seq(xpath) - added:
method MarkupLMTokenizer.prepare_for_model(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, xpaths: Optional[list[list[int]]]=…, node_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional… - added:
method MarkupLMTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method MarkupLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MarkupLMTokenizer.truncate_sequences(ids: list[int], xpath_tags_seq: list[list[int]], xpath_subs_seq: list[list[int]], pair_ids: Optional[list[int]]=…, pair_xpath_tags_seq: Optional[list[list[int]]]=…, pair_xpath_subs_seq: Optional[list[list[int]]]=…, labels: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[l… - added:
property MarkupLMTokenizer.vocab_size
- added:
class MarkupLMTokenizerFast(PreTrainedTokenizerFast) - added:
constructor MarkupLMTokenizerFast(vocab_file, merges_file, tags_dict, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, max_depth=…, max_width=…, pad_width=…, pad_token_label=…, only_label_first_subword=…, trim_offsets=…, **kwargs) - added:
field MarkupLMTokenizerFast.slow_tokenizer_class - added:
field MarkupLMTokenizerFast.vocab_files_names - added:
method MarkupLMTokenizerFast.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, xpaths: Optional[list[list[list[int]]]]=…, node_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…,… - added:
method MarkupLMTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MarkupLMTokenizerFast.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, xpaths: Optional[list[list[int]]]=…, node_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[i… - added:
method MarkupLMTokenizerFast.get_xpath_seq(xpath) - added:
method MarkupLMTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MarkupLMTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str]
- added:
export define_import_structure
- added:
class Mask2FormerConfig(PretrainedConfig) - added:
constructor Mask2FormerConfig(backbone_config: Optional[dict]=…, feature_size: int=…, mask_feature_size: int=…, hidden_dim: int=…, encoder_feedforward_dim: int=…, activation_function: str=…, encoder_layers: int=…, decoder_layers: int=…, num_attention_heads: int=…, dropout: float=…, dim_feedforward: int=…, pre_norm: bool=…, enforce_input_projection: bool=…, common_stride: int=…, ignore_value: int=… - added:
field Mask2FormerConfig.attribute_map - added:
field Mask2FormerConfig.backbones_supported - added:
field Mask2FormerConfig.model_type - added:
property Mask2FormerConfig.sub_configs - added:
static method Mask2FormerConfig.from_backbone_config(backbone_config: PretrainedConfig, **kwargs)
- added:
class Mask2FormerImageProcessor(BaseImageProcessor) - added:
constructor Mask2FormerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, size_divisor: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, num_labels: Optional[int]=…, p… - added:
field Mask2FormerImageProcessor.model_input_names - added:
method Mask2FormerImageProcessor.convert_segmentation_map_to_binary_masks(segmentation_map: np.ndarray, instance_id_to_semantic_id: Optional[dict[int, int]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…) - added:
method Mask2FormerImageProcessor.encode_inputs(pixel_values_list: list[ImageInput], segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, pad_size: Optiona… - added:
method Mask2FormerImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, pad_size: Optional[dict[str, int]]=…) -> BatchFeature - added:
method Mask2FormerImageProcessor.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…, return_binary_maps: Optional[bool]=…) -> list[dict] - added:
method Mask2FormerImageProcessor.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method Mask2FormerImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method Mask2FormerImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[dict[int, int]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, size_divisor: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, ima… - added:
method Mask2FormerImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method Mask2FormerImageProcessor.resize(image: np.ndarray, size: dict[str, int], size_divisor: int=…, resample: PILImageResampling=…, data_format=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Mask2FormerImageProcessor.to_dict() -> dict[str, Any]
- added:
class Mask2FormerImageProcessorFast(BaseImageProcessorFast) - added:
constructor Mask2FormerImageProcessorFast(**kwargs: Unpack[Mask2FormerFastImageProcessorKwargs]) -> None - added:
field Mask2FormerImageProcessorFast.default_to_square - added:
field Mask2FormerImageProcessorFast.do_normalize - added:
field Mask2FormerImageProcessorFast.do_pad - added:
field Mask2FormerImageProcessorFast.do_reduce_labels - added:
field Mask2FormerImageProcessorFast.do_rescale - added:
field Mask2FormerImageProcessorFast.do_resize - added:
field Mask2FormerImageProcessorFast.image_mean - added:
field Mask2FormerImageProcessorFast.image_std - added:
field Mask2FormerImageProcessorFast.model_input_names - added:
field Mask2FormerImageProcessorFast.resample - added:
field Mask2FormerImageProcessorFast.rescale_factor - added:
field Mask2FormerImageProcessorFast.size - added:
field Mask2FormerImageProcessorFast.size_divisor - added:
field Mask2FormerImageProcessorFast.valid_kwargs - added:
method Mask2FormerImageProcessorFast.pad(images: torch.Tensor, padded_size: tuple[int, int], segmentation_maps: Optional[torch.Tensor]=…, fill: int=…, ignore_index: int=…) -> BatchFeature - added:
method Mask2FormerImageProcessorFast.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…, return_binary_maps: Optional[bool]=…) -> list[dict] - added:
method Mask2FormerImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method Mask2FormerImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method Mask2FormerImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, **kwargs: Unpack[Mask2FormerFastImageProcessorKwargs]) -> BatchFeature - added:
method Mask2FormerImageProcessorFast.reduce_label(labels: list[torch.Tensor]) - added:
method Mask2FormerImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, size_divisor: int=…, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method Mask2FormerImageProcessorFast.to_dict() -> dict[str, Any]
- added:
class Mask2FormerForUniversalSegmentation(Mask2FormerPreTrainedModel) - added:
class Mask2FormerModel(Mask2FormerPreTrainedModel) - added:
class Mask2FormerPreTrainedModel(PreTrainedModel) - added:
constructor Mask2FormerForUniversalSegmentation(config: Mask2FormerConfig) - added:
constructor Mask2FormerModel(config: Mask2FormerConfig) - added:
field Mask2FormerForUniversalSegmentation.main_input_name - added:
field Mask2FormerModel.main_input_name - added:
field Mask2FormerPreTrainedModel.base_model_prefix - added:
field Mask2FormerPreTrainedModel.config: Mask2FormerConfig - added:
field Mask2FormerPreTrainedModel.main_input_name - added:
method Mask2FormerForUniversalSegmentation.forward(pixel_values: Tensor, mask_labels: Optional[list[Tensor]]=…, class_labels: Optional[list[Tensor]]=…, pixel_mask: Optional[Tensor]=…, output_hidden_states: Optional[bool]=…, output_auxiliary_logits: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Mask2FormerForUniversalSegmentationOutput - added:
method Mask2FormerForUniversalSegmentation.get_auxiliary_logits(classes: torch.Tensor, output_masks: torch.Tensor) - added:
method Mask2FormerForUniversalSegmentation.get_loss(loss_dict: dict[str, Tensor]) -> Tensor - added:
method Mask2FormerForUniversalSegmentation.get_loss_dict(masks_queries_logits: Tensor, class_queries_logits: Tensor, mask_labels: Tensor, class_labels: Tensor, auxiliary_predictions: dict[str, Tensor]) -> dict[str, Tensor] - added:
method Mask2FormerModel.forward(pixel_values: Tensor, pixel_mask: Optional[Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Mask2FormerModelOutput
- added:
class Mask2FormerImageProcessorFast(MaskFormerImageProcessorFast) - added:
method Mask2FormerImageProcessorFast.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…, return_binary_maps: Optional[bool]=…) -> list[dict] - added:
method Mask2FormerImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method Mask2FormerImageProcessorFast.post_process_segmentation() - added:
method Mask2FormerImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor
- added:
export define_import_structure
- added:
class MaskFormerConfig(PretrainedConfig) - added:
constructor MaskFormerConfig(fpn_feature_size: int=…, mask_feature_size: int=…, no_object_weight: float=…, use_auxiliary_loss: bool=…, backbone_config: Optional[dict]=…, decoder_config: Optional[dict]=…, init_std: float=…, init_xavier_std: float=…, dice_weight: float=…, cross_entropy_weight: float=…, mask_weight: float=…, output_auxiliary_logits: Optional[bool]=…, backbone: Optional[str]=…, use_p… - added:
field MaskFormerConfig.attribute_map - added:
field MaskFormerConfig.backbones_supported - added:
field MaskFormerConfig.decoders_supported - added:
field MaskFormerConfig.model_type - added:
property MaskFormerConfig.sub_configs - added:
static method MaskFormerConfig.from_backbone_and_decoder_configs(backbone_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs)
- added:
class MaskFormerSwinConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor MaskFormerSwinConfig(image_size=…, patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, window_size=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, use_absolute_embeddings=…, initializer_range=…, layer_norm_eps=…, out_features=…, out_indices=…, **kwargs) - added:
field MaskFormerSwinConfig.attribute_map - added:
field MaskFormerSwinConfig.model_type
- added:
class MaskFormerFeatureExtractor(MaskFormerImageProcessor) - added:
constructor MaskFormerFeatureExtractor(*args, **kwargs) -> None
- added:
class MaskFormerImageProcessor(BaseImageProcessor) - added:
constructor MaskFormerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, size_divisor: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, num_labels: Optional[int]=…, pa… - added:
field MaskFormerImageProcessor.model_input_names - added:
method MaskFormerImageProcessor.convert_segmentation_map_to_binary_masks(segmentation_map: np.ndarray, instance_id_to_semantic_id: Optional[dict[int, int]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…) - added:
method MaskFormerImageProcessor.encode_inputs(pixel_values_list: list[ImageInput], segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, pad_size: Optional… - added:
method MaskFormerImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, pad_size: Optional[dict[str, int]]=…) -> BatchFeature - added:
method MaskFormerImageProcessor.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…, return_binary_maps: Optional[bool]=…) -> list[dict] - added:
method MaskFormerImageProcessor.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method MaskFormerImageProcessor.post_process_segmentation(outputs: MaskFormerForInstanceSegmentationOutput, target_size: Optional[tuple[int, int]]=…) -> torch.Tensor - added:
method MaskFormerImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method MaskFormerImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[dict[int, int]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, size_divisor: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, imag… - added:
method MaskFormerImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method MaskFormerImageProcessor.resize(image: np.ndarray, size: dict[str, int], size_divisor: int=…, resample: PILImageResampling=…, data_format=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method MaskFormerImageProcessor.to_dict() -> dict[str, Any]
- added:
class MaskFormerImageProcessorFast(BaseImageProcessorFast) - added:
constructor MaskFormerImageProcessorFast(**kwargs: Unpack[MaskFormerFastImageProcessorKwargs]) -> None - added:
field MaskFormerImageProcessorFast.default_to_square - added:
field MaskFormerImageProcessorFast.do_normalize - added:
field MaskFormerImageProcessorFast.do_pad - added:
field MaskFormerImageProcessorFast.do_reduce_labels - added:
field MaskFormerImageProcessorFast.do_rescale - added:
field MaskFormerImageProcessorFast.do_resize - added:
field MaskFormerImageProcessorFast.image_mean - added:
field MaskFormerImageProcessorFast.image_std - added:
field MaskFormerImageProcessorFast.model_input_names - added:
field MaskFormerImageProcessorFast.resample - added:
field MaskFormerImageProcessorFast.rescale_factor - added:
field MaskFormerImageProcessorFast.size - added:
field MaskFormerImageProcessorFast.size_divisor - added:
field MaskFormerImageProcessorFast.valid_kwargs - added:
method MaskFormerImageProcessorFast.pad(images: torch.Tensor, padded_size: tuple[int, int], segmentation_maps: Optional[torch.Tensor]=…, fill: int=…, ignore_index: int=…) -> BatchFeature - added:
method MaskFormerImageProcessorFast.post_process_instance_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…, return_binary_maps: Optional[bool]=…) -> list[dict] - added:
method MaskFormerImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method MaskFormerImageProcessorFast.post_process_segmentation(outputs: MaskFormerForInstanceSegmentationOutput, target_size: Optional[tuple[int, int]]=…) -> torch.Tensor - added:
method MaskFormerImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method MaskFormerImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, **kwargs: Unpack[MaskFormerFastImageProcessorKwargs]) -> BatchFeature - added:
method MaskFormerImageProcessorFast.reduce_label(labels: list[torch.Tensor]) - added:
method MaskFormerImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, size_divisor: int=…, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method MaskFormerImageProcessorFast.to_dict() -> dict[str, Any]
- added:
class MaskFormerForInstanceSegmentation(MaskFormerPreTrainedModel) - added:
class MaskFormerModel(MaskFormerPreTrainedModel) - added:
class MaskFormerPreTrainedModel(PreTrainedModel) - added:
constructor MaskFormerForInstanceSegmentation(config: MaskFormerConfig) - added:
constructor MaskFormerModel(config: MaskFormerConfig) - added:
field MaskFormerPreTrainedModel.base_model_prefix - added:
field MaskFormerPreTrainedModel.config: MaskFormerConfig - added:
field MaskFormerPreTrainedModel.main_input_name - added:
method MaskFormerForInstanceSegmentation.forward(pixel_values: Tensor, mask_labels: Optional[list[Tensor]]=…, class_labels: Optional[list[Tensor]]=…, pixel_mask: Optional[Tensor]=…, output_auxiliary_logits: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> MaskFormerForInstanceSegmentationOutput - added:
method MaskFormerForInstanceSegmentation.get_logits(outputs: MaskFormerModelOutput) -> tuple[Tensor, Tensor, dict[str, Tensor]] - added:
method MaskFormerForInstanceSegmentation.get_loss(loss_dict: dict[str, Tensor]) -> Tensor - added:
method MaskFormerForInstanceSegmentation.get_loss_dict(masks_queries_logits: Tensor, class_queries_logits: Tensor, mask_labels: Tensor, class_labels: Tensor, auxiliary_logits: dict[str, Tensor]) -> dict[str, Tensor] - added:
method MaskFormerModel.forward(pixel_values: Tensor, pixel_mask: Optional[Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> MaskFormerModelOutput
- added:
class MaskFormerSwinBackbone(MaskFormerSwinPreTrainedModel, BackboneMixin) - added:
class MaskFormerSwinModel(MaskFormerSwinPreTrainedModel) - added:
class MaskFormerSwinPreTrainedModel(PreTrainedModel) - added:
constructor MaskFormerSwinBackbone(config: MaskFormerSwinConfig) - added:
constructor MaskFormerSwinModel(config, add_pooling_layer=…) - added:
field MaskFormerSwinPreTrainedModel.base_model_prefix - added:
field MaskFormerSwinPreTrainedModel.config: MaskFormerSwinConfig - added:
field MaskFormerSwinPreTrainedModel.main_input_name - added:
field MaskFormerSwinPreTrainedModel.supports_gradient_checkpointing - added:
method MaskFormerSwinBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method MaskFormerSwinModel.forward(pixel_values=…, head_mask=…, output_attentions=…, output_hidden_states=…, interpolate_pos_encoding=…, return_dict=…) - added:
method MaskFormerSwinModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class MBartConfig(PretrainedConfig) - added:
class MBartOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor MBartConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, classifier_dropout=…, scale_e… - added:
field MBartConfig.attribute_map - added:
field MBartConfig.keys_to_ignore_at_inference - added:
field MBartConfig.model_type - added:
method MBartOnnxConfig.generate_dummy_inputs(tokenizer: PreTrainedTokenizer, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property MBartOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MBartOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxMBartForConditionalGeneration(FlaxMBartPreTrainedModel) - added:
class FlaxMBartForQuestionAnswering(FlaxMBartPreTrainedModel) - added:
class FlaxMBartForSequenceClassification(FlaxMBartPreTrainedModel) - added:
class FlaxMBartModel(FlaxMBartPreTrainedModel) - added:
class FlaxMBartPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxMBartPreTrainedModel(config: MBartConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxMBartForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxMBartForConditionalGeneration.module_class - added:
field FlaxMBartForQuestionAnswering.dtype - added:
field FlaxMBartForQuestionAnswering.module_class - added:
field FlaxMBartForSequenceClassification.dtype - added:
field FlaxMBartForSequenceClassification.module_class - added:
field FlaxMBartModel.config: MBartConfig - added:
field FlaxMBartModel.dtype: jnp.dtype - added:
field FlaxMBartModel.module_class - added:
field FlaxMBartPreTrainedModel.base_model_prefix: str - added:
field FlaxMBartPreTrainedModel.config_class - added:
field FlaxMBartPreTrainedModel.module_class: nn.Module - added:
method FlaxMBartForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Opti… - added:
method FlaxMBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxMBartForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxMBartPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict… - added:
method FlaxMBartPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxMBartPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxMBartPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class MBartForCausalLM(MBartPreTrainedModel, GenerationMixin) - added:
class MBartForConditionalGeneration(MBartPreTrainedModel, GenerationMixin) - added:
class MBartForQuestionAnswering(MBartPreTrainedModel) - added:
class MBartForSequenceClassification(MBartPreTrainedModel) - added:
class MBartModel(MBartPreTrainedModel) - added:
class MBartPreTrainedModel(PreTrainedModel) - added:
constructor MBartForCausalLM(config) - added:
constructor MBartForConditionalGeneration(config: MBartConfig) - added:
constructor MBartForQuestionAnswering(config) - added:
constructor MBartForSequenceClassification(config: MBartConfig, **kwargs) - added:
constructor MBartModel(config: MBartConfig) - added:
field MBartForConditionalGeneration.base_model_prefix - added:
field MBartPreTrainedModel.base_model_prefix - added:
field MBartPreTrainedModel.config: MBartConfig - added:
field MBartPreTrainedModel.supports_gradient_checkpointing - added:
method MBartForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method MBartForCausalLM.get_decoder() - added:
method MBartForCausalLM.get_input_embeddings() - added:
method MBartForCausalLM.set_decoder(decoder) - added:
method MBartForCausalLM.set_input_embeddings(value) - added:
method MBartForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[… - added:
method MBartForConditionalGeneration.get_decoder() - added:
method MBartForConditionalGeneration.get_encoder() - added:
method MBartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MBartForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method MBartForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTens… - added:
method MBartForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.… - added:
method MBartModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]… - added:
method MBartModel.get_encoder() - added:
method MBartModel.get_input_embeddings() - added:
method MBartModel.set_input_embeddings(value) - added:
property MBartPreTrainedModel.dummy_inputs
- added:
class TFMBartForConditionalGeneration(TFMBartPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFMBartModel(TFMBartPreTrainedModel) - added:
class TFMBartPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMBartForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFMBartModel(config: MBartConfig, *inputs, **kwargs) - added:
field TFMBartPreTrainedModel.base_model_prefix - added:
field TFMBartPreTrainedModel.config_class - added:
method TFMBartForConditionalGeneration.build(input_shape=…) - added:
method TFMBartForConditionalGeneration.call(input_ids: TFModelInputType=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: TFBaseModelOutput | None=…, past_k… - added:
method TFMBartForConditionalGeneration.get_bias() - added:
method TFMBartForConditionalGeneration.get_decoder() - added:
method TFMBartForConditionalGeneration.get_encoder() - added:
method TFMBartForConditionalGeneration.get_output_embeddings() - added:
method TFMBartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFMBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFMBartForConditionalGeneration.serving_output(output) - added:
method TFMBartForConditionalGeneration.set_bias(value) - added:
method TFMBartForConditionalGeneration.set_output_embeddings(value) - added:
method TFMBartModel.build(input_shape=…) - added:
method TFMBartModel.call(input_ids: TFModelInputType=…, attention_mask: tf.Tensor | None=…, decoder_input_ids: tf.Tensor | None=…, decoder_attention_mask: tf.Tensor | None=…, decoder_position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, decoder_head_mask: tf.Tensor | None=…, cross_attn_head_mask: tf.Tensor | None=…, encoder_outputs: tuple | TFBaseModelOutput | None=…, past_key_values:… - added:
method TFMBartModel.get_decoder() - added:
method TFMBartModel.get_encoder() - added:
method TFMBartModel.serving_output(output)
- added:
class MBartTokenizer(PreTrainedTokenizer) - added:
constructor MBartTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, tokenizer_file=…, src_lang=…, tgt_lang=…, sp_model_kwargs: Optional[dict[str, Any]]=…, additional_special_tokens=…, **kwargs) - added:
field MBartTokenizer.model_input_names - added:
field MBartTokenizer.prefix_tokens: list[int] - added:
field MBartTokenizer.suffix_tokens: list[int] - added:
field MBartTokenizer.vocab_files_names - added:
method MBartTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBartTokenizer.convert_tokens_to_string(tokens) - added:
method MBartTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBartTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MBartTokenizer.get_vocab() - added:
method MBartTokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method MBartTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MBartTokenizer.set_src_lang_special_tokens(src_lang) -> None - added:
method MBartTokenizer.set_tgt_lang_special_tokens(lang: str) -> None - added:
property MBartTokenizer.src_lang -> str - added:
property MBartTokenizer.vocab_size
- added:
class MBartTokenizerFast(PreTrainedTokenizerFast) - added:
constructor MBartTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, src_lang=…, tgt_lang=…, additional_special_tokens=…, **kwargs) - added:
field MBartTokenizerFast.model_input_names - added:
field MBartTokenizerFast.prefix_tokens: list[int] - added:
field MBartTokenizerFast.slow_tokenizer_class - added:
field MBartTokenizerFast.suffix_tokens: list[int] - added:
field MBartTokenizerFast.vocab_files_names - added:
method MBartTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBartTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBartTokenizerFast.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method MBartTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MBartTokenizerFast.set_src_lang_special_tokens(src_lang) -> None - added:
method MBartTokenizerFast.set_tgt_lang_special_tokens(lang: str) -> None - added:
property MBartTokenizerFast.src_lang -> str
- added:
export define_import_structure
- added:
class MBart50Tokenizer(PreTrainedTokenizer) - added:
constructor MBart50Tokenizer(vocab_file, src_lang=…, tgt_lang=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field MBart50Tokenizer.model_input_names - added:
field MBart50Tokenizer.prefix_tokens: list[int] - added:
field MBart50Tokenizer.suffix_tokens: list[int] - added:
field MBart50Tokenizer.vocab_files_names - added:
method MBart50Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBart50Tokenizer.convert_tokens_to_string(tokens) - added:
method MBart50Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MBart50Tokenizer.get_vocab() -> dict - added:
method MBart50Tokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method MBart50Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MBart50Tokenizer.set_src_lang_special_tokens(src_lang: str) -> None - added:
method MBart50Tokenizer.set_tgt_lang_special_tokens(tgt_lang: str) -> None - added:
property MBart50Tokenizer.src_lang -> str - added:
property MBart50Tokenizer.vocab_size -> int
- added:
class MBart50TokenizerFast(PreTrainedTokenizerFast) - added:
constructor MBart50TokenizerFast(vocab_file=…, src_lang=…, tgt_lang=…, tokenizer_file=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
field MBart50TokenizerFast.model_input_names - added:
field MBart50TokenizerFast.prefix_tokens: list[int] - added:
field MBart50TokenizerFast.slow_tokenizer_class - added:
field MBart50TokenizerFast.suffix_tokens: list[int] - added:
field MBart50TokenizerFast.vocab_files_names - added:
method MBart50TokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MBart50TokenizerFast.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method MBart50TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method MBart50TokenizerFast.set_src_lang_special_tokens(src_lang: str) -> None - added:
method MBart50TokenizerFast.set_tgt_lang_special_tokens(tgt_lang: str) -> None - added:
property MBart50TokenizerFast.src_lang -> str
- added:
export define_import_structure
- added:
class MegatronBertConfig(PretrainedConfig) - added:
constructor MegatronBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, position_embedding_type=…, use_cache=…, **kwargs) - added:
field MegatronBertConfig.model_type
- added:
class MegatronBertForCausalLM(MegatronBertPreTrainedModel, GenerationMixin) - added:
class MegatronBertForMaskedLM(MegatronBertPreTrainedModel) - added:
class MegatronBertForMultipleChoice(MegatronBertPreTrainedModel) - added:
class MegatronBertForNextSentencePrediction(MegatronBertPreTrainedModel) - added:
class MegatronBertForPreTraining(MegatronBertPreTrainedModel) - added:
class MegatronBertForQuestionAnswering(MegatronBertPreTrainedModel) - added:
class MegatronBertForSequenceClassification(MegatronBertPreTrainedModel) - added:
class MegatronBertForTokenClassification(MegatronBertPreTrainedModel) - added:
class MegatronBertModel(MegatronBertPreTrainedModel) - added:
class MegatronBertPreTrainedModel(PreTrainedModel) - added:
constructor MegatronBertForCausalLM(config) - added:
constructor MegatronBertForMaskedLM(config) - added:
constructor MegatronBertForMultipleChoice(config) - added:
constructor MegatronBertForNextSentencePrediction(config) - added:
constructor MegatronBertForPreTraining(config, add_binary_head=…) - added:
constructor MegatronBertForQuestionAnswering(config) - added:
constructor MegatronBertForSequenceClassification(config) - added:
constructor MegatronBertForTokenClassification(config) - added:
constructor MegatronBertModel(config, add_pooling_layer=…) - added:
field MegatronBertPreTrainedModel.base_model_prefix - added:
field MegatronBertPreTrainedModel.config: MegatronBertConfig - added:
field MegatronBertPreTrainedModel.load_tf_weights - added:
field MegatronBertPreTrainedModel.supports_gradient_checkpointing - added:
method MegatronBertForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.F… - added:
method MegatronBertForCausalLM.get_output_embeddings() - added:
method MegatronBertForCausalLM.set_output_embeddings(new_embeddings) - added:
method MegatronBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.F… - added:
method MegatronBertForMaskedLM.get_output_embeddings() - added:
method MegatronBertForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method MegatronBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method MegatronBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method MegatronBertForNextSentencePrediction.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outpu… - added:
method MegatronBertForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, next_sentence_label: Optional[torch.LongTensor]=…, ou… - added:
method MegatronBertForPreTraining.get_output_embeddings() - added:
method MegatronBertForPreTraining.set_output_embeddings(new_embeddings) - added:
method MegatronBertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTens… - added:
method MegatronBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outpu… - added:
method MegatronBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method MegatronBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTe… - added:
method MegatronBertModel.get_input_embeddings() - added:
method MegatronBertModel.set_input_embeddings(value)
- added:
function add_checkpointing_args(parser) - added:
function add_megatron_checkpoint_args(parser) - added:
function add_transformers_checkpoint_args(parser) - added:
function convert_checkpoint_from_megatron_to_transformers(args) - added:
function convert_checkpoint_from_transformers_to_megatron(args) - added:
function get_element_from_dict_by_path(d, path) - added:
function get_megatron_sharded_states(args, tp_size, pp_size, pp_rank) - added:
function main() - added:
function megatron_to_transformers_fix_query_key_value_ordering(param, checkpoint_version, num_splits, num_heads, hidden_size) - added:
function merge_transformers_sharded_states(path, num_checkpoints) - added:
function recursive_print(name, val, spaces=…) - added:
function transformers_to_megatron_fix_query_key_value_ordering(param, checkpoint_version, num_splits, num_heads, hidden_size) - added:
variable megatron_to_transformers - added:
variable tensor_parallel_params - added:
variable transformers_to_megatron
- added:
export define_import_structure
- added:
class MetaClip2Config(PretrainedConfig) - added:
class MetaClip2TextConfig(PretrainedConfig) - added:
class MetaClip2VisionConfig(PretrainedConfig) - added:
constructor MetaClip2Config(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
constructor MetaClip2TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor MetaClip2VisionConfig(hidden_size=…, intermediate_size=…, projection_dim=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field MetaClip2Config.model_type - added:
field MetaClip2Config.sub_configs - added:
field MetaClip2TextConfig.base_config_key - added:
field MetaClip2TextConfig.model_type - added:
field MetaClip2VisionConfig.base_config_key - added:
field MetaClip2VisionConfig.model_type
- added:
class MetaClip2ForImageClassification(MetaClip2PreTrainedModel) - added:
class MetaClip2Model(MetaClip2PreTrainedModel) - added:
class MetaClip2PreTrainedModel(PreTrainedModel) - added:
class MetaClip2TextModel(MetaClip2PreTrainedModel) - added:
class MetaClip2TextModelWithProjection(MetaClip2PreTrainedModel) - added:
class MetaClip2VisionModel(MetaClip2PreTrainedModel) - added:
class MetaClip2VisionModelWithProjection(MetaClip2PreTrainedModel) - added:
constructor MetaClip2ForImageClassification(config: MetaClip2Config) -> None - added:
constructor MetaClip2Model(config: MetaClip2Config) - added:
constructor MetaClip2TextModel(config: MetaClip2TextConfig) - added:
constructor MetaClip2TextModelWithProjection(config: MetaClip2TextConfig) - added:
constructor MetaClip2VisionModel(config: MetaClip2VisionConfig) - added:
constructor MetaClip2VisionModelWithProjection(config: MetaClip2VisionConfig) - added:
field MetaClip2ForImageClassification.main_input_name - added:
field MetaClip2Model.config: MetaClip2Config - added:
field MetaClip2PreTrainedModel.base_model_prefix - added:
field MetaClip2PreTrainedModel.config: MetaClip2Config - added:
field MetaClip2PreTrainedModel.supports_gradient_checkpointing - added:
field MetaClip2TextModel.config: MetaClip2TextConfig - added:
field MetaClip2TextModelWithProjection.config: MetaClip2TextConfig - added:
field MetaClip2VisionModel.config: MetaClip2VisionConfig - added:
field MetaClip2VisionModel.main_input_name - added:
field MetaClip2VisionModelWithProjection.config: MetaClip2VisionConfig - added:
field MetaClip2VisionModelWithProjection.main_input_name - added:
method MetaClip2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> ImageClassifierOutput - added:
method MetaClip2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> MetaClip2Output - added:
method MetaClip2Model.get_image_features(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method MetaClip2Model.get_text_features(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> torch.FloatTensor - added:
method MetaClip2TextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPooling - added:
method MetaClip2TextModel.get_input_embeddings() -> nn.Module - added:
method MetaClip2TextModel.set_input_embeddings(value) - added:
method MetaClip2TextModelWithProjection.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> MetaClip2TextModelOutput - added:
method MetaClip2TextModelWithProjection.get_input_embeddings() -> nn.Module - added:
method MetaClip2TextModelWithProjection.set_input_embeddings(value) - added:
method MetaClip2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> BaseModelOutputWithPooling - added:
method MetaClip2VisionModel.get_input_embeddings() -> nn.Module - added:
method MetaClip2VisionModelWithProjection.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> MetaClip2VisionModelOutput - added:
method MetaClip2VisionModelWithProjection.get_input_embeddings() -> nn.Module
- added:
class MetaClip2Config(CLIPConfig) - added:
class MetaClip2ForImageClassification(CLIPForImageClassification) - added:
class MetaClip2Model(CLIPModel) - added:
class MetaClip2PreTrainedModel(PreTrainedModel) - added:
class MetaClip2TextConfig(CLIPTextConfig) - added:
class MetaClip2TextModel(CLIPTextModel) - added:
class MetaClip2TextModelWithProjection(CLIPTextModelWithProjection) - added:
class MetaClip2VisionConfig(CLIPVisionConfig) - added:
class MetaClip2VisionModel(CLIPVisionModel) - added:
class MetaClip2VisionModelWithProjection(CLIPVisionModelWithProjection) - added:
constructor MetaClip2Model(config: MetaClip2Config) - added:
constructor MetaClip2TextModel(config: MetaClip2TextConfig) - added:
constructor MetaClip2TextModelWithProjection(config: MetaClip2TextConfig) - added:
field MetaClip2PreTrainedModel.base_model_prefix - added:
field MetaClip2PreTrainedModel.config: MetaClip2Config - added:
field MetaClip2PreTrainedModel.supports_gradient_checkpointing - added:
method MetaClip2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) - added:
method MetaClip2Model.get_image_features(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) - added:
method MetaClip2Model.get_text_features(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) - added:
method MetaClip2TextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) - added:
method MetaClip2TextModelWithProjection.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) - added:
method MetaClip2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) - added:
method MetaClip2VisionModelWithProjection.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…)
- added:
export define_import_structure
- added:
class MgpstrConfig(PretrainedConfig) - added:
constructor MgpstrConfig(image_size=…, patch_size=…, num_channels=…, max_token_length=…, num_character_labels=…, num_bpe_labels=…, num_wordpiece_labels=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, qkv_bias=…, distilled=…, layer_norm_eps=…, drop_rate=…, attn_drop_rate=…, drop_path_rate=…, output_a3_attentions=…, initializer_range=…, **kwargs) - added:
field MgpstrConfig.model_type
- added:
class MgpstrForSceneTextRecognition(MgpstrPreTrainedModel) - added:
class MgpstrModel(MgpstrPreTrainedModel) - added:
class MgpstrPreTrainedModel(PreTrainedModel) - added:
constructor MgpstrForSceneTextRecognition(config: MgpstrConfig) -> None - added:
constructor MgpstrModel(config: MgpstrConfig) - added:
field MgpstrForSceneTextRecognition.config: MgpstrConfig - added:
field MgpstrForSceneTextRecognition.main_input_name - added:
field MgpstrPreTrainedModel.base_model_prefix - added:
field MgpstrPreTrainedModel.config: MgpstrConfig - added:
method MgpstrForSceneTextRecognition.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_a3_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], MgpstrModelOutput] - added:
method MgpstrModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method MgpstrModel.get_input_embeddings() -> nn.Module
- added:
class MgpstrProcessor(ProcessorMixin) - added:
constructor MgpstrProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field MgpstrProcessor.attributes - added:
field MgpstrProcessor.char_tokenizer_class - added:
field MgpstrProcessor.image_processor_class - added:
method MgpstrProcessor.batch_decode(sequences) - added:
method MgpstrProcessor.bpe_decode(sequences) - added:
method MgpstrProcessor.char_decode(sequences) - added:
method MgpstrProcessor.wp_decode(sequences)
- added:
class MgpstrTokenizer(PreTrainedTokenizer) - added:
constructor MgpstrTokenizer(vocab_file, unk_token=…, bos_token=…, eos_token=…, pad_token=…, **kwargs) - added:
field MgpstrTokenizer.vocab_files_names - added:
method MgpstrTokenizer.get_vocab() - added:
method MgpstrTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MgpstrTokenizer.vocab_size
- added:
export define_import_structure
- added:
class MimiConfig(PretrainedConfig) - added:
constructor MimiConfig(sampling_rate=…, frame_rate=…, audio_channels=…, hidden_size=…, num_filters=…, num_residual_layers=…, upsampling_ratios=…, kernel_size=…, last_kernel_size=…, residual_kernel_size=…, dilation_growth_rate=…, use_causal_conv=…, pad_mode=…, compress=…, trim_right_ratio=…, codebook_size=…, codebook_dim=…, num_quantizers=…, use_conv_shortcut=…, vector_quantization_hidden_dimensio… - added:
field MimiConfig.model_type - added:
property MimiConfig.encodec_frame_rate -> int - added:
property MimiConfig.frame_rate -> float - added:
property MimiConfig.frame_size -> int - added:
property MimiConfig.num_codebooks -> int
- added:
class MimiModel(MimiPreTrainedModel) - added:
class MimiPreTrainedModel(PreTrainedModel) - added:
constructor MimiModel(config: MimiConfig) - added:
field MimiPreTrainedModel.base_model_prefix - added:
field MimiPreTrainedModel.config: MimiConfig - added:
field MimiPreTrainedModel.main_input_name - added:
field MimiPreTrainedModel.supports_gradient_checkpointing - added:
method MimiModel.decode(audio_codes: torch.Tensor, padding_mask: Optional[torch.Tensor]=…, decoder_past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor, torch.Tensor], MimiDecoderOutput] - added:
method MimiModel.encode(input_values: torch.Tensor, padding_mask: Optional[torch.Tensor]=…, num_quantizers: Optional[float]=…, encoder_past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, padding_cache: Optional[MimiConv1dPaddingCache]=…, use_streaming: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor, Optional[torch.Tensor]], MimiEncoderOutput] - added:
method MimiModel.forward(input_values: torch.Tensor, padding_mask: Optional[torch.Tensor]=…, num_quantizers: Optional[int]=…, audio_codes: Optional[torch.Tensor]=…, encoder_past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, decoder_past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor, torch.Tensor], MimiOut… - added:
method MimiModel.get_audio_codes_mask(padding_mask: torch.Tensor, padding_side: str=…) - added:
method MimiModel.get_encoded_length(input_length: torch.LongTensor) -> torch.LongTensor - added:
method MimiModel.get_encoder()
- added:
export define_import_structure
- added:
class MiniMaxConfig(PretrainedConfig) - added:
constructor MiniMaxConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, sliding_window=…, attention_dropout=…, num_experts_per_tok=…, num_lo… - added:
field MiniMaxConfig.base_model_pp_plan - added:
field MiniMaxConfig.base_model_tp_plan - added:
field MiniMaxConfig.keys_to_ignore_at_inference - added:
field MiniMaxConfig.model_type
- added:
class MiniMaxForCausalLM(MiniMaxPreTrainedModel, GenerationMixin) - added:
class MiniMaxForQuestionAnswering(GenericForQuestionAnswering, MiniMaxPreTrainedModel) - added:
class MiniMaxForSequenceClassification(GenericForSequenceClassification, MiniMaxPreTrainedModel) - added:
class MiniMaxForTokenClassification(GenericForTokenClassification, MiniMaxPreTrainedModel) - added:
class MiniMaxModel(MiniMaxPreTrainedModel) - added:
class MiniMaxPreTrainedModel(PreTrainedModel) - added:
constructor MiniMaxForCausalLM(config) - added:
constructor MiniMaxModel(config: MiniMaxConfig) - added:
field MiniMaxPreTrainedModel.base_model_prefix - added:
field MiniMaxPreTrainedModel.config: MiniMaxConfig - added:
field MiniMaxPreTrainedModel.supports_gradient_checkpointing - added:
method MiniMaxForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, log… - added:
method MiniMaxModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[MiniMaxCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Mo…
- added:
class MiniMaxConfig(MixtralConfig) - added:
class MiniMaxForCausalLM(MixtralForCausalLM) - added:
class MiniMaxForQuestionAnswering(MixtralForQuestionAnswering) - added:
class MiniMaxForSequenceClassification(MixtralForSequenceClassification) - added:
class MiniMaxForTokenClassification(MixtralForTokenClassification) - added:
class MiniMaxModel(MixtralModel) - added:
class MiniMaxPreTrainedModel(MixtralPreTrainedModel) - added:
constructor MiniMaxConfig(layer_types=…, block_size=…, full_attn_alpha_factor=…, full_attn_beta_factor=…, linear_attn_alpha_factor=…, linear_attn_beta_factor=…, mlp_alpha_factor=…, mlp_beta_factor=…, **super_kwargs) - added:
method MiniMaxForCausalLM.forward(**super_kwargs) - added:
method MiniMaxModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[MiniMaxCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Mo…
- added:
export define_import_structure
- added:
class MinistralConfig(PretrainedConfig) - added:
constructor MinistralConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, sliding_window=…, attention_dropout=…, layer_types=…, **kwargs) - added:
field MinistralConfig.base_model_pp_plan - added:
field MinistralConfig.base_model_tp_plan - added:
field MinistralConfig.keys_to_ignore_at_inference - added:
field MinistralConfig.model_type
- added:
class MinistralForCausalLM(MinistralPreTrainedModel, GenerationMixin) - added:
class MinistralForQuestionAnswering(GenericForQuestionAnswering, MinistralPreTrainedModel) - added:
class MinistralForSequenceClassification(GenericForSequenceClassification, MinistralPreTrainedModel) - added:
class MinistralForTokenClassification(GenericForTokenClassification, MinistralPreTrainedModel) - added:
class MinistralModel(MinistralPreTrainedModel) - added:
class MinistralPreTrainedModel(PreTrainedModel) - added:
constructor MinistralForCausalLM(config) - added:
constructor MinistralModel(config: MinistralConfig) - added:
field MinistralForQuestionAnswering.base_model_prefix - added:
field MinistralPreTrainedModel.base_model_prefix - added:
field MinistralPreTrainedModel.config: MinistralConfig - added:
field MinistralPreTrainedModel.supports_gradient_checkpointing - added:
method MinistralForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=… - added:
method MinistralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class MinistralConfig(MistralConfig, PretrainedConfig) - added:
class MinistralForCausalLM(Qwen2ForCausalLM) - added:
class MinistralForQuestionAnswering(Qwen2ForQuestionAnswering) - added:
class MinistralForSequenceClassification(Qwen2ForSequenceClassification) - added:
class MinistralForTokenClassification(Qwen2ForTokenClassification) - added:
class MinistralModel(Qwen2Model) - added:
class MinistralPreTrainedModel(Qwen2PreTrainedModel) - added:
constructor MinistralConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, sliding_window=…, attention_dropout=…, layer_types=…, **kwargs) - added:
constructor MinistralModel(config: MinistralConfig) - added:
field MinistralConfig.model_type - added:
method MinistralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
export define_import_structure
- added:
class MistralConfig(PretrainedConfig) - added:
constructor MistralConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, sliding_window=…, attention_dropout=…, **kwargs) - added:
field MistralConfig.base_model_pp_plan - added:
field MistralConfig.base_model_tp_plan - added:
field MistralConfig.keys_to_ignore_at_inference - added:
field MistralConfig.model_type
- added:
class FlaxMistralForCausalLM(FlaxMistralPreTrainedModel) - added:
class FlaxMistralModel(FlaxMistralPreTrainedModel) - added:
class FlaxMistralPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxMistralPreTrainedModel(config: MistralConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxMistralForCausalLM.module_class - added:
field FlaxMistralModel.module_class - added:
field FlaxMistralPreTrainedModel.base_model_prefix - added:
field FlaxMistralPreTrainedModel.config_class - added:
field FlaxMistralPreTrainedModel.module_class: nn.Module - added:
method FlaxMistralForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxMistralForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxMistralPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxMistralPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class MistralForCausalLM(MistralPreTrainedModel, GenerationMixin) - added:
class MistralForQuestionAnswering(GenericForQuestionAnswering, MistralPreTrainedModel) - added:
class MistralForSequenceClassification(GenericForSequenceClassification, MistralPreTrainedModel) - added:
class MistralForTokenClassification(GenericForTokenClassification, MistralPreTrainedModel) - added:
class MistralModel(MistralPreTrainedModel) - added:
class MistralPreTrainedModel(PreTrainedModel) - added:
constructor MistralForCausalLM(config) - added:
constructor MistralModel(config: MistralConfig) - added:
field MistralPreTrainedModel.base_model_prefix - added:
field MistralPreTrainedModel.config: MistralConfig - added:
field MistralPreTrainedModel.supports_gradient_checkpointing - added:
method MistralForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method MistralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class TFMistralForCausalLM(TFMistralPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFMistralForSequenceClassification(TFMistralPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFMistralModel(TFMistralPreTrainedModel) - added:
class TFMistralPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMistralForCausalLM(config, *inputs, **kwargs) - added:
constructor TFMistralForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFMistralModel(config: MistralConfig, *inputs, **kwargs) - added:
field TFMistralPreTrainedModel.base_model_prefix - added:
field TFMistralPreTrainedModel.config_class - added:
method TFMistralForCausalLM.build(input_shape=…) - added:
method TFMistralForCausalLM.call(input_ids: Optional[tf.Tensor]=…, attention_mask: Optional[tf.Tensor]=…, position_ids: Optional[tf.Tensor]=…, past_key_values: Optional[list[tf.Tensor]]=…, inputs_embeds: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> U… - added:
method TFMistralForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, **kwargs) - added:
method TFMistralForSequenceClassification.build(input_shape=…) - added:
method TFMistralForSequenceClassification.call(input_ids: Optional[tf.Tensor]=…, attention_mask: Optional[tf.Tensor]=…, position_ids: Optional[tf.Tensor]=…, past_key_values: Optional[list[tf.Tensor]]=…, inputs_embeds: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional… - added:
method TFMistralModel.build(input_shape=…) - added:
method TFMistralModel.call(input_ids: Optional[tf.Tensor]=…, attention_mask: Optional[tf.Tensor]=…, position_ids: Optional[tf.Tensor]=…, past_key_values: Optional[list[tf.Tensor]]=…, inputs_embeds: Optional[tf.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TFBaseModelOutputWithPast]
- added:
class MistralForCausalLM(LlamaForCausalLM) - added:
class MistralForQuestionAnswering(GenericForQuestionAnswering, MistralPreTrainedModel) - added:
class MistralForSequenceClassification(LlamaForSequenceClassification) - added:
class MistralForTokenClassification(LlamaForTokenClassification) - added:
class MistralModel(LlamaModel) - added:
class MistralPreTrainedModel(LlamaPreTrainedModel) - added:
method MistralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
export define_import_structure
- added:
class Mistral3Config(PretrainedConfig) - added:
constructor Mistral3Config(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, vision_feature_layer=…, multimodal_projector_bias=…, spatial_merge_size=…, **kwargs) - added:
field Mistral3Config.attribute_map - added:
field Mistral3Config.is_composition - added:
field Mistral3Config.model_type - added:
field Mistral3Config.sub_configs
- added:
class Mistral3ForConditionalGeneration(Mistral3PreTrainedModel, GenerationMixin) - added:
class Mistral3Model(Mistral3PreTrainedModel) - added:
class Mistral3PreTrainedModel(PreTrainedModel) - added:
constructor Mistral3ForConditionalGeneration(config: Mistral3Config) - added:
constructor Mistral3Model(config: Mistral3Config) - added:
field Mistral3PreTrainedModel.base_model_prefix - added:
field Mistral3PreTrainedModel.config: Mistral3Config - added:
field Mistral3PreTrainedModel.supports_gradient_checkpointing - added:
method Mistral3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bo… - added:
method Mistral3ForConditionalGeneration.get_decoder() - added:
method Mistral3ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, **kwargs) - added:
method Mistral3ForConditionalGeneration.get_input_embeddings() - added:
method Mistral3ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Mistral3ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Mistral3ForConditionalGeneration.set_decoder(decoder) - added:
method Mistral3ForConditionalGeneration.set_input_embeddings(value) - added:
method Mistral3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bo… - added:
method Mistral3Model.get_decoder() - added:
method Mistral3Model.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, **kwargs) - added:
method Mistral3Model.get_input_embeddings() - added:
method Mistral3Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Mistral3Model.set_decoder(decoder) - added:
method Mistral3Model.set_input_embeddings(value) - added:
property Mistral3ForConditionalGeneration.language_model - added:
property Mistral3ForConditionalGeneration.multi_modal_projector - added:
property Mistral3ForConditionalGeneration.vision_tower
- added:
class Mistral3ForConditionalGeneration(LlavaForConditionalGeneration) - added:
class Mistral3Model(LlavaModel) - added:
class Mistral3PreTrainedModel(LlavaPreTrainedModel) - added:
method Mistral3ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bo… - added:
method Mistral3ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, **kwargs) - added:
method Mistral3Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bo… - added:
method Mistral3Model.get_image_features(pixel_values: torch.FloatTensor, image_sizes: torch.Tensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, **kwargs)
- added:
export define_import_structure
- added:
class MixtralConfig(PretrainedConfig) - added:
constructor MixtralConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, sliding_window=…, attention_dropout=…, num_experts_per_tok=…, num_lo… - added:
field MixtralConfig.base_model_pp_plan - added:
field MixtralConfig.base_model_tp_plan - added:
field MixtralConfig.keys_to_ignore_at_inference - added:
field MixtralConfig.model_type
- added:
class MixtralForCausalLM(MixtralPreTrainedModel, GenerationMixin) - added:
class MixtralForQuestionAnswering(GenericForQuestionAnswering, MixtralPreTrainedModel) - added:
class MixtralForSequenceClassification(GenericForSequenceClassification, MixtralPreTrainedModel) - added:
class MixtralForTokenClassification(GenericForTokenClassification, MixtralPreTrainedModel) - added:
class MixtralModel(MixtralPreTrainedModel) - added:
class MixtralPreTrainedModel(PreTrainedModel) - added:
constructor MixtralForCausalLM(config) - added:
constructor MixtralModel(config: MixtralConfig) - added:
field MixtralPreTrainedModel.base_model_prefix - added:
field MixtralPreTrainedModel.config: MixtralConfig - added:
field MixtralPreTrainedModel.supports_gradient_checkpointing - added:
method MixtralForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, log… - added:
method MixtralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class MixtralForCausalLM(MistralForCausalLM) - added:
class MixtralForQuestionAnswering(MistralForQuestionAnswering) - added:
class MixtralForSequenceClassification(MistralForSequenceClassification) - added:
class MixtralForTokenClassification(MistralForTokenClassification) - added:
class MixtralModel(MistralModel) - added:
class MixtralPreTrainedModel(MistralPreTrainedModel) - added:
constructor MixtralForCausalLM(config) - added:
method MixtralForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, log… - added:
method MixtralModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
export define_import_structure
- added:
class MLCDVisionConfig(PretrainedConfig) - added:
constructor MLCDVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_groups=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field MLCDVisionConfig.base_config_key - added:
field MLCDVisionConfig.model_type
- added:
class MLCDPreTrainedModel(PreTrainedModel) - added:
class MLCDVisionModel(MLCDPreTrainedModel) - added:
constructor MLCDVisionModel(config: MLCDVisionConfig) - added:
field MLCDPreTrainedModel.base_model_prefix - added:
field MLCDPreTrainedModel.config: MLCDVisionConfig - added:
field MLCDPreTrainedModel.supports_gradient_checkpointing - added:
field MLCDVisionModel.config: MLCDVisionConfig - added:
field MLCDVisionModel.main_input_name - added:
method MLCDVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method MLCDVisionModel.get_input_embeddings() -> nn.Module
- added:
class MLCDPreTrainedModel(PreTrainedModel) - added:
class MLCDVisionConfig(PretrainedConfig) - added:
class MLCDVisionModel(CLIPVisionModel) - added:
constructor MLCDVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_groups=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field MLCDPreTrainedModel.base_model_prefix - added:
field MLCDPreTrainedModel.config: MLCDVisionConfig - added:
field MLCDPreTrainedModel.supports_gradient_checkpointing - added:
field MLCDVisionConfig.base_config_key - added:
field MLCDVisionConfig.model_type - added:
method MLCDVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling]
- added:
export define_import_structure
- added:
class MllamaConfig(PretrainedConfig) - added:
constructor MllamaConfig(vision_config=…, text_config=…, image_token_index=…, **kwargs) - added:
field MllamaConfig.attribute_map - added:
field MllamaConfig.model_type - added:
field MllamaConfig.sub_configs
- added:
class MllamaImageProcessor(BaseImageProcessor) - added:
constructor MllamaImageProcessor(do_convert_rgb: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, max_image_tiles: int=…, **kwargs) -> None - added:
field MllamaImageProcessor.model_input_names - added:
method MllamaImageProcessor.pad(image: np.ndarray, size: dict[str, int], aspect_ratio: tuple[int, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method MllamaImageProcessor.preprocess(images: ImageInput, do_convert_rgb: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_… - added:
method MllamaImageProcessor.resize(image: np.ndarray, size: dict[str, int], max_image_tiles: int, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> Union[np.ndarray, tuple[int, int]]
- added:
class MllamaForCausalLM(MllamaPreTrainedModel, GenerationMixin) - added:
class MllamaForConditionalGeneration(MllamaPreTrainedModel, GenerationMixin) - added:
class MllamaModel(MllamaPreTrainedModel) - added:
class MllamaPreTrainedModel(PreTrainedModel) - added:
class MllamaTextModel(MllamaPreTrainedModel) - added:
class MllamaVisionModel(MllamaPreTrainedModel) - added:
constructor MllamaForCausalLM(config) - added:
constructor MllamaForConditionalGeneration(config: MllamaConfig) - added:
constructor MllamaModel(config: MllamaConfig) - added:
constructor MllamaTextModel(config: MllamaTextConfig) - added:
constructor MllamaVisionModel(config: MllamaVisionConfig) - added:
field MllamaForCausalLM.base_model_prefix - added:
field MllamaForCausalLM.config: MllamaTextConfig - added:
field MllamaPreTrainedModel.base_model_prefix - added:
field MllamaPreTrainedModel.config: MllamaConfig - added:
field MllamaPreTrainedModel.supports_gradient_checkpointing - added:
field MllamaTextModel.base_model_prefix - added:
field MllamaTextModel.config: MllamaTextConfig - added:
field MllamaVisionModel.base_model_prefix - added:
field MllamaVisionModel.config: MllamaVisionConfig - added:
method MllamaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, cross_attention_states: Optional[torch.LongTensor]=…, cross_attention_mask: Optional[torch.LongTensor]=…, full_text_row_masked_out_mask: Optional[tuple[torch.Tensor, torch.Tensor]]=…, past_key_values: Optional[Union[Cache, list[torch.Float… - added:
method MllamaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, aspect_ratio_mask: Optional[torch.Tensor]=…, aspect_ratio_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, cross_attention_mask: Optional[torch.Tensor]=…, cross_attention_states: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTenso… - added:
method MllamaForConditionalGeneration.get_decoder() - added:
method MllamaForConditionalGeneration.get_input_embeddings() - added:
method MllamaForConditionalGeneration.prepare_inputs_for_generation(input_ids=…, inputs_embeds=…, attention_mask=…, position_ids=…, pixel_values=…, aspect_ratio_ids=…, aspect_ratio_mask=…, cross_attention_mask=…, past_key_values=…, use_cache=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method MllamaForConditionalGeneration.set_decoder(decoder) - added:
method MllamaForConditionalGeneration.set_input_embeddings(value) - added:
method MllamaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, aspect_ratio_mask: Optional[torch.Tensor]=…, aspect_ratio_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, cross_attention_mask: Optional[torch.Tensor]=…, cross_attention_states: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_valu… - added:
method MllamaModel.get_decoder() - added:
method MllamaModel.get_input_embeddings() - added:
method MllamaModel.set_decoder(decoder) - added:
method MllamaModel.set_input_embeddings(value) - added:
method MllamaTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, cross_attention_states: Optional[torch.FloatTensor]=…, cross_attention_mask: Optional[torch.Tensor]=…, full_text_row_masked_out_mask: Optional[tuple[torch.Tensor, torch.Tensor]]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTenso… - added:
method MllamaVisionModel.apply_class_embedding(hidden_state: torch.Tensor) -> torch.Tensor - added:
method MllamaVisionModel.forward(pixel_values: torch.Tensor, aspect_ratio_ids: torch.Tensor, aspect_ratio_mask: torch.Tensor, **kwargs) -> BaseModelOutput - added:
method MllamaVisionModel.get_input_embeddings() - added:
property MllamaForConditionalGeneration.language_model - added:
property MllamaForConditionalGeneration.vision_model
- added:
class MllamaProcessor(ProcessorMixin) - added:
constructor MllamaProcessor(image_processor, tokenizer, chat_template=…) - added:
field MllamaProcessor.attributes - added:
field MllamaProcessor.image_processor_class - added:
field MllamaProcessor.tokenizer_class - added:
method MllamaProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs) - added:
property MllamaProcessor.model_input_names
- added:
export define_import_structure
- added:
class MLukeTokenizer(PreTrainedTokenizer) - added:
constructor MLukeTokenizer(vocab_file, entity_vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, task=…, max_entity_length=…, max_mention_length=…, entity_token_1=…, entity_token_2=…, entity_unk_token=…, entity_pad_token=…, entity_mask_token=…, entity_mask2_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field MLukeTokenizer.model_input_names - added:
field MLukeTokenizer.vocab_files_names - added:
method MLukeTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MLukeTokenizer.convert_tokens_to_string(tokens) - added:
method MLukeTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MLukeTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MLukeTokenizer.get_vocab() - added:
method MLukeTokenizer.pad(encoded_inputs: Union[BatchEncoding, list[BatchEncoding], dict[str, EncodedInput], dict[str, list[EncodedInput]], list[dict[str, EncodedInput]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, max_entity_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_attention_mask: Optional[bool]=…, return_… - added:
method MLukeTokenizer.prepare_for_model(ids: list[int], pair_ids: Optional[list[int]]=…, entity_ids: Optional[list[int]]=…, pair_entity_ids: Optional[list[int]]=…, entity_token_spans: Optional[list[tuple[int, int]]]=…, pair_entity_token_spans: Optional[list[tuple[int, int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrateg… - added:
method MLukeTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str, str] - added:
property MLukeTokenizer.vocab_size
- added:
export define_import_structure
- added:
class MMGroundingDinoConfig(PretrainedConfig) - added:
constructor MMGroundingDinoConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, text_config=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…,… - added:
field MMGroundingDinoConfig.attribute_map - added:
field MMGroundingDinoConfig.model_type - added:
property MMGroundingDinoConfig.hidden_size -> int - added:
property MMGroundingDinoConfig.num_attention_heads -> int - added:
property MMGroundingDinoConfig.sub_configs
- added:
class MMGroundingDinoForObjectDetection(MMGroundingDinoPreTrainedModel) - added:
class MMGroundingDinoModel(MMGroundingDinoPreTrainedModel) - added:
class MMGroundingDinoPreTrainedModel(PreTrainedModel) - added:
constructor MMGroundingDinoForObjectDetection(config: MMGroundingDinoConfig) - added:
constructor MMGroundingDinoModel(config: MMGroundingDinoConfig) - added:
field MMGroundingDinoPreTrainedModel.base_model_prefix - added:
field MMGroundingDinoPreTrainedModel.config: MMGroundingDinoConfig - added:
field MMGroundingDinoPreTrainedModel.main_input_name - added:
method MMGroundingDinoForObjectDetection.forward(pixel_values: torch.FloatTensor, input_ids: torch.LongTensor, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, pixel_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[Union[MMGroundingDinoEncoderOutput, tuple]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retu… - added:
method MMGroundingDinoModel.forward(pixel_values: Tensor, input_ids: Tensor, token_type_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, pixel_mask: Optional[Tensor]=…, encoder_outputs=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method MMGroundingDinoModel.freeze_backbone() - added:
method MMGroundingDinoModel.generate_encoder_output_proposals(enc_output, padding_mask, spatial_shapes) - added:
method MMGroundingDinoModel.get_valid_ratio(mask) - added:
method MMGroundingDinoModel.unfreeze_backbone()
- added:
class MMGroundingDinoConfig(GroundingDinoConfig, PretrainedConfig) - added:
class MMGroundingDinoForObjectDetection(GroundingDinoForObjectDetection, MMGroundingDinoPreTrainedModel) - added:
class MMGroundingDinoModel(GroundingDinoModel, MMGroundingDinoPreTrainedModel) - added:
class MMGroundingDinoPreTrainedModel(GroundingDinoPreTrainedModel) - added:
constructor MMGroundingDinoConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, text_config=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…,… - added:
constructor MMGroundingDinoForObjectDetection(config: MMGroundingDinoConfig) - added:
constructor MMGroundingDinoModel(config: MMGroundingDinoConfig) - added:
field MMGroundingDinoConfig.model_type
- added:
export define_import_structure
- added:
class MobileBertConfig(PretrainedConfig) - added:
class MobileBertOnnxConfig(OnnxConfig) - added:
constructor MobileBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, embedding_size=…, trigram_input=…, use_bottleneck=…, intra_bottleneck_size=…, use_bottleneck_attention=…,… - added:
field MobileBertConfig.model_type - added:
property MobileBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class MobileBertForMaskedLM(MobileBertPreTrainedModel) - added:
class MobileBertForMultipleChoice(MobileBertPreTrainedModel) - added:
class MobileBertForNextSentencePrediction(MobileBertPreTrainedModel) - added:
class MobileBertForPreTraining(MobileBertPreTrainedModel) - added:
class MobileBertForQuestionAnswering(MobileBertPreTrainedModel) - added:
class MobileBertForSequenceClassification(MobileBertPreTrainedModel) - added:
class MobileBertForTokenClassification(MobileBertPreTrainedModel) - added:
class MobileBertLayer(nn.Module) - added:
class MobileBertModel(MobileBertPreTrainedModel) - added:
class MobileBertPreTrainedModel(PreTrainedModel) - added:
constructor MobileBertForMaskedLM(config) - added:
constructor MobileBertForMultipleChoice(config) - added:
constructor MobileBertForNextSentencePrediction(config) - added:
constructor MobileBertForPreTraining(config) - added:
constructor MobileBertForQuestionAnswering(config) - added:
constructor MobileBertForSequenceClassification(config) - added:
constructor MobileBertForTokenClassification(config) - added:
constructor MobileBertLayer(config) - added:
constructor MobileBertModel(config, add_pooling_layer=…) - added:
field MobileBertPreTrainedModel.base_model_prefix - added:
field MobileBertPreTrainedModel.config: MobileBertConfig - added:
field MobileBertPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_mobilebert(model, config, tf_checkpoint_path) - added:
method MobileBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states:… - added:
method MobileBertForMaskedLM.get_output_embeddings() - added:
method MobileBertForMaskedLM.resize_token_embeddings(new_num_tokens: Optional[int]=…) -> nn.Embedding - added:
method MobileBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method MobileBertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return… - added:
method MobileBertForNextSentencePrediction.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_… - added:
method MobileBertForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, next_sentence_label: Optional[torch.LongTensor]=…, outp… - added:
method MobileBertForPreTraining.get_output_embeddings() - added:
method MobileBertForPreTraining.resize_token_embeddings(new_num_tokens: Optional[int]=…) -> nn.Embedding - added:
method MobileBertForPreTraining.set_output_embeddings(new_embeddings) - added:
method MobileBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bo… - added:
method MobileBertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…… - added:
method MobileBertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, r… - added:
method MobileBertLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…) -> tuple[torch.Tensor] - added:
method MobileBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[boo… - added:
method MobileBertModel.get_input_embeddings() - added:
method MobileBertModel.set_input_embeddings(value)
- added:
class TFMobileBertForMaskedLM(TFMobileBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFMobileBertForMultipleChoice(TFMobileBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFMobileBertForNextSentencePrediction(TFMobileBertPreTrainedModel, TFNextSentencePredictionLoss) - added:
class TFMobileBertForPreTraining(TFMobileBertPreTrainedModel, TFMobileBertPreTrainingLoss) - added:
class TFMobileBertForQuestionAnswering(TFMobileBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFMobileBertForSequenceClassification(TFMobileBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFMobileBertForTokenClassification(TFMobileBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFMobileBertMainLayer(keras.layers.Layer) - added:
class TFMobileBertModel(TFMobileBertPreTrainedModel) - added:
class TFMobileBertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMobileBertForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFMobileBertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFMobileBertForNextSentencePrediction(config, *inputs, **kwargs) - added:
constructor TFMobileBertForPreTraining(config, *inputs, **kwargs) - added:
constructor TFMobileBertForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFMobileBertForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFMobileBertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFMobileBertMainLayer(config, add_pooling_layer=…, **kwargs) - added:
constructor TFMobileBertModel(config, *inputs, **kwargs) - added:
field TFMobileBertMainLayer.config_class - added:
field TFMobileBertPreTrainedModel.base_model_prefix - added:
field TFMobileBertPreTrainedModel.config_class - added:
method TFMobileBertForMaskedLM.build(input_shape=…) - added:
method TFMobileBertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFMobileBertForMaskedLM.get_lm_head() - added:
method TFMobileBertForMaskedLM.get_prefix_bias_name() - added:
method TFMobileBertForMaskedLM.tf_to_pt_weight_rename(tf_weight) - added:
method TFMobileBertForMultipleChoice.build(input_shape=…) - added:
method TFMobileBertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict… - added:
method TFMobileBertForNextSentencePrediction.build(input_shape=…) - added:
method TFMobileBertForNextSentencePrediction.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, ret… - added:
method TFMobileBertForPreTraining.build(input_shape=…) - added:
method TFMobileBertForPreTraining.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: b… - added:
method TFMobileBertForPreTraining.get_lm_head() - added:
method TFMobileBertForPreTraining.get_prefix_bias_name() - added:
method TFMobileBertForPreTraining.tf_to_pt_weight_rename(tf_weight) - added:
method TFMobileBertForQuestionAnswering.build(input_shape=…) - added:
method TFMobileBertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_d… - added:
method TFMobileBertForSequenceClassification.build(input_shape=…) - added:
method TFMobileBertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, ret… - added:
method TFMobileBertForTokenClassification.build(input_shape=…) - added:
method TFMobileBertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return… - added:
method TFMobileBertMainLayer.build(input_shape=…) - added:
method TFMobileBertMainLayer.call(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFMobileBertMainLayer.get_input_embeddings() - added:
method TFMobileBertMainLayer.set_input_embeddings(value) - added:
method TFMobileBertModel.build(input_shape=…) - added:
method TFMobileBertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | Non…
- added:
class MobileBertTokenizer(PreTrainedTokenizer) - added:
constructor MobileBertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field MobileBertTokenizer.vocab_files_names - added:
method MobileBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MobileBertTokenizer.convert_tokens_to_string(tokens) - added:
method MobileBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MobileBertTokenizer.get_vocab() - added:
method MobileBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MobileBertTokenizer.do_lower_case - added:
property MobileBertTokenizer.vocab_size
- added:
class MobileBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor MobileBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field MobileBertTokenizerFast.slow_tokenizer_class - added:
field MobileBertTokenizerFast.vocab_files_names - added:
method MobileBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method MobileBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class MobileNetV1Config(PretrainedConfig) - added:
class MobileNetV1OnnxConfig(OnnxConfig) - added:
constructor MobileNetV1Config(num_channels=…, image_size=…, depth_multiplier=…, min_depth=…, hidden_act=…, tf_padding=…, classifier_dropout_prob=…, initializer_range=…, layer_norm_eps=…, **kwargs) - added:
field MobileNetV1Config.model_type - added:
field MobileNetV1OnnxConfig.torch_onnx_minimum_version - added:
property MobileNetV1OnnxConfig.atol_for_validation -> float - added:
property MobileNetV1OnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MobileNetV1OnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class MobileNetV1FeatureExtractor(MobileNetV1ImageProcessor) - added:
constructor MobileNetV1FeatureExtractor(*args, **kwargs) -> None
- added:
class MobileNetV1ImageProcessor(BaseImageProcessor) - added:
constructor MobileNetV1ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field MobileNetV1ImageProcessor.model_input_names - added:
method MobileNetV1ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std… - added:
method MobileNetV1ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class MobileNetV1ImageProcessorFast(BaseImageProcessorFast) - added:
constructor MobileNetV1ImageProcessorFast(**kwargs: Unpack[DefaultFastImageProcessorKwargs]) -> None - added:
field MobileNetV1ImageProcessorFast.crop_size - added:
field MobileNetV1ImageProcessorFast.default_to_square - added:
field MobileNetV1ImageProcessorFast.do_center_crop - added:
field MobileNetV1ImageProcessorFast.do_normalize - added:
field MobileNetV1ImageProcessorFast.do_rescale - added:
field MobileNetV1ImageProcessorFast.do_resize - added:
field MobileNetV1ImageProcessorFast.image_mean - added:
field MobileNetV1ImageProcessorFast.image_std - added:
field MobileNetV1ImageProcessorFast.resample - added:
field MobileNetV1ImageProcessorFast.size
- added:
class MobileNetV1ForImageClassification(MobileNetV1PreTrainedModel) - added:
class MobileNetV1Model(MobileNetV1PreTrainedModel) - added:
class MobileNetV1PreTrainedModel(PreTrainedModel) - added:
constructor MobileNetV1ForImageClassification(config: MobileNetV1Config) -> None - added:
constructor MobileNetV1Model(config: MobileNetV1Config, add_pooling_layer: bool=…) - added:
field MobileNetV1PreTrainedModel.base_model_prefix - added:
field MobileNetV1PreTrainedModel.config: MobileNetV1Config - added:
field MobileNetV1PreTrainedModel.load_tf_weights - added:
field MobileNetV1PreTrainedModel.main_input_name - added:
field MobileNetV1PreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_mobilenet_v1(model, config, tf_checkpoint_path) - added:
method MobileNetV1ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method MobileNetV1Model.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class MobileNetV2Config(PretrainedConfig) - added:
class MobileNetV2OnnxConfig(OnnxConfig) - added:
constructor MobileNetV2Config(num_channels=…, image_size=…, depth_multiplier=…, depth_divisible_by=…, min_depth=…, expand_ratio=…, output_stride=…, first_layer_is_expansion=…, finegrained_output=…, hidden_act=…, tf_padding=…, classifier_dropout_prob=…, initializer_range=…, layer_norm_eps=…, semantic_loss_ignore_index=…, **kwargs) - added:
field MobileNetV2Config.model_type - added:
field MobileNetV2OnnxConfig.torch_onnx_minimum_version - added:
property MobileNetV2OnnxConfig.atol_for_validation -> float - added:
property MobileNetV2OnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MobileNetV2OnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class MobileNetV2FeatureExtractor(MobileNetV2ImageProcessor) - added:
constructor MobileNetV2FeatureExtractor(*args, **kwargs) -> None
- added:
class MobileNetV2ImageProcessor(BaseImageProcessor) - added:
constructor MobileNetV2ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_reduce_labels: bool=…, **kwarg… - added:
field MobileNetV2ImageProcessor.model_input_names - added:
method MobileNetV2ImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method MobileNetV2ImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optio… - added:
method MobileNetV2ImageProcessor.reduce_label(label: ImageInput) -> np.ndarray - added:
method MobileNetV2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class MobileNetV2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor MobileNetV2ImageProcessorFast(**kwargs: Unpack[MobileNetV2FastImageProcessorKwargs]) - added:
field MobileNetV2ImageProcessorFast.crop_size - added:
field MobileNetV2ImageProcessorFast.default_to_square - added:
field MobileNetV2ImageProcessorFast.do_center_crop - added:
field MobileNetV2ImageProcessorFast.do_normalize - added:
field MobileNetV2ImageProcessorFast.do_reduce_labels - added:
field MobileNetV2ImageProcessorFast.do_rescale - added:
field MobileNetV2ImageProcessorFast.do_resize - added:
field MobileNetV2ImageProcessorFast.image_mean - added:
field MobileNetV2ImageProcessorFast.image_std - added:
field MobileNetV2ImageProcessorFast.resample - added:
field MobileNetV2ImageProcessorFast.size - added:
field MobileNetV2ImageProcessorFast.valid_kwargs - added:
method MobileNetV2ImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method MobileNetV2ImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[MobileNetV2FastImageProcessorKwargs]) -> BatchFeature - added:
method MobileNetV2ImageProcessorFast.reduce_label(labels: list[torch.Tensor])
- added:
class MobileNetV2ForImageClassification(MobileNetV2PreTrainedModel) - added:
class MobileNetV2ForSemanticSegmentation(MobileNetV2PreTrainedModel) - added:
class MobileNetV2Model(MobileNetV2PreTrainedModel) - added:
class MobileNetV2PreTrainedModel(PreTrainedModel) - added:
constructor MobileNetV2ForImageClassification(config: MobileNetV2Config) -> None - added:
constructor MobileNetV2ForSemanticSegmentation(config: MobileNetV2Config) -> None - added:
constructor MobileNetV2Model(config: MobileNetV2Config, add_pooling_layer: bool=…) - added:
field MobileNetV2PreTrainedModel.base_model_prefix - added:
field MobileNetV2PreTrainedModel.config: MobileNetV2Config - added:
field MobileNetV2PreTrainedModel.load_tf_weights - added:
field MobileNetV2PreTrainedModel.main_input_name - added:
field MobileNetV2PreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_mobilenet_v2(model, config, tf_checkpoint_path) - added:
method MobileNetV2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method MobileNetV2ForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method MobileNetV2Model.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class MobileViTConfig(PretrainedConfig) - added:
class MobileViTOnnxConfig(OnnxConfig) - added:
constructor MobileViTConfig(num_channels=…, image_size=…, patch_size=…, hidden_sizes=…, neck_hidden_sizes=…, num_attention_heads=…, mlp_ratio=…, expand_ratio=…, hidden_act=…, conv_kernel_size=…, output_stride=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, classifier_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, aspp_out_channels=…, atrous_rates=…, aspp_dropout_prob… - added:
field MobileViTConfig.model_type - added:
field MobileViTOnnxConfig.torch_onnx_minimum_version - added:
property MobileViTOnnxConfig.atol_for_validation -> float - added:
property MobileViTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MobileViTOnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class MobileViTFeatureExtractor(MobileViTImageProcessor) - added:
constructor MobileViTFeatureExtractor(*args, **kwargs) -> None
- added:
class MobileViTImageProcessor(BaseImageProcessor) - added:
constructor MobileViTImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_flip_channel_order: bool=…, do_reduce_labels: bool=…, **kwargs) -> None - added:
field MobileViTImageProcessor.model_input_names - added:
method MobileViTImageProcessor.flip_channel_order(image: np.ndarray, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method MobileViTImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method MobileViTImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_flip_channel_order: Optional[bool]=…, do_reduce_… - added:
method MobileViTImageProcessor.reduce_label(label: ImageInput) -> np.ndarray - added:
method MobileViTImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class MobileViTImageProcessorFast(BaseImageProcessorFast) - added:
constructor MobileViTImageProcessorFast(**kwargs: Unpack[MobileVitFastImageProcessorKwargs]) - added:
field MobileViTImageProcessorFast.crop_size - added:
field MobileViTImageProcessorFast.default_to_square - added:
field MobileViTImageProcessorFast.do_center_crop - added:
field MobileViTImageProcessorFast.do_convert_rgb - added:
field MobileViTImageProcessorFast.do_flip_channel_order - added:
field MobileViTImageProcessorFast.do_normalize - added:
field MobileViTImageProcessorFast.do_reduce_labels - added:
field MobileViTImageProcessorFast.do_rescale - added:
field MobileViTImageProcessorFast.do_resize - added:
field MobileViTImageProcessorFast.resample - added:
field MobileViTImageProcessorFast.size - added:
field MobileViTImageProcessorFast.valid_kwargs - added:
method MobileViTImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method MobileViTImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[MobileVitFastImageProcessorKwargs]) -> BatchFeature - added:
method MobileViTImageProcessorFast.reduce_label(labels: list[torch.Tensor])
- added:
class MobileViTForImageClassification(MobileViTPreTrainedModel) - added:
class MobileViTForSemanticSegmentation(MobileViTPreTrainedModel) - added:
class MobileViTModel(MobileViTPreTrainedModel) - added:
class MobileViTPreTrainedModel(PreTrainedModel) - added:
constructor MobileViTForImageClassification(config: MobileViTConfig) -> None - added:
constructor MobileViTForSemanticSegmentation(config: MobileViTConfig) -> None - added:
constructor MobileViTModel(config: MobileViTConfig, expand_output: bool=…) - added:
field MobileViTPreTrainedModel.base_model_prefix - added:
field MobileViTPreTrainedModel.config: MobileViTConfig - added:
field MobileViTPreTrainedModel.main_input_name - added:
field MobileViTPreTrainedModel.supports_gradient_checkpointing - added:
method MobileViTForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method MobileViTForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method MobileViTModel.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
class TFMobileViTForImageClassification(TFMobileViTPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFMobileViTForSemanticSegmentation(TFMobileViTPreTrainedModel) - added:
class TFMobileViTModel(TFMobileViTPreTrainedModel) - added:
class TFMobileViTPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMobileViTForImageClassification(config: MobileViTConfig, *inputs, **kwargs) -> None - added:
constructor TFMobileViTForSemanticSegmentation(config: MobileViTConfig, **kwargs) -> None - added:
constructor TFMobileViTModel(config: MobileViTConfig, expand_output: bool=…, *inputs, **kwargs) - added:
field TFMobileViTPreTrainedModel.base_model_prefix - added:
field TFMobileViTPreTrainedModel.config_class - added:
field TFMobileViTPreTrainedModel.main_input_name - added:
method TFMobileViTForImageClassification.build(input_shape=…) - added:
method TFMobileViTForImageClassification.call(pixel_values: tf.Tensor | None=…, output_hidden_states: bool | None=…, labels: tf.Tensor | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFImageClassifierOutputWithNoAttention - added:
method TFMobileViTForSemanticSegmentation.build(input_shape=…) - added:
method TFMobileViTForSemanticSegmentation.call(pixel_values: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple | TFSemanticSegmenterOutputWithNoAttention - added:
method TFMobileViTForSemanticSegmentation.hf_compute_loss(logits, labels) - added:
method TFMobileViTModel.build(input_shape=…) - added:
method TFMobileViTModel.call(pixel_values: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple[tf.Tensor] | TFBaseModelOutputWithPooling
- added:
export define_import_structure
- added:
class MobileViTV2Config(PretrainedConfig) - added:
class MobileViTV2OnnxConfig(OnnxConfig) - added:
constructor MobileViTV2Config(num_channels=…, image_size=…, patch_size=…, expand_ratio=…, hidden_act=…, conv_kernel_size=…, output_stride=…, classifier_dropout_prob=…, initializer_range=…, layer_norm_eps=…, aspp_out_channels=…, atrous_rates=…, aspp_dropout_prob=…, semantic_loss_ignore_index=…, n_attn_blocks=…, base_attn_unit_dims=…, width_multiplier=…, ffn_multiplier=…, attn_dropout=…, ffn_dropou… - added:
field MobileViTV2Config.model_type - added:
field MobileViTV2OnnxConfig.torch_onnx_minimum_version - added:
property MobileViTV2OnnxConfig.atol_for_validation -> float - added:
property MobileViTV2OnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property MobileViTV2OnnxConfig.outputs -> Mapping[str, Mapping[int, str]]
- added:
class MobileViTV2ForImageClassification(MobileViTV2PreTrainedModel) - added:
class MobileViTV2ForSemanticSegmentation(MobileViTV2PreTrainedModel) - added:
class MobileViTV2Model(MobileViTV2PreTrainedModel) - added:
class MobileViTV2PreTrainedModel(PreTrainedModel) - added:
constructor MobileViTV2ForImageClassification(config: MobileViTV2Config) -> None - added:
constructor MobileViTV2ForSemanticSegmentation(config: MobileViTV2Config) -> None - added:
constructor MobileViTV2Model(config: MobileViTV2Config, expand_output: bool=…) - added:
field MobileViTV2PreTrainedModel.base_model_prefix - added:
field MobileViTV2PreTrainedModel.config: MobileViTV2Config - added:
field MobileViTV2PreTrainedModel.main_input_name - added:
field MobileViTV2PreTrainedModel.supports_gradient_checkpointing - added:
method MobileViTV2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method MobileViTV2ForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method MobileViTV2Model.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class ModernBertConfig(PretrainedConfig) - added:
constructor ModernBertConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, initializer_cutoff_factor=…, norm_eps=…, norm_bias=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, cls_token_id=…, sep_token_id=…, global_rope_theta=…, attention_bias=…, attention_dropout=…, global_attn_e… - added:
field ModernBertConfig.attribute_map - added:
field ModernBertConfig.keys_to_ignore_at_inference - added:
field ModernBertConfig.model_type - added:
method ModernBertConfig.to_dict()
- added:
class ModernBertForMaskedLM(ModernBertPreTrainedModel) - added:
class ModernBertForMultipleChoice(ModernBertPreTrainedModel) - added:
class ModernBertForQuestionAnswering(ModernBertPreTrainedModel) - added:
class ModernBertForSequenceClassification(ModernBertPreTrainedModel) - added:
class ModernBertForTokenClassification(ModernBertPreTrainedModel) - added:
class ModernBertModel(ModernBertPreTrainedModel) - added:
class ModernBertPreTrainedModel(PreTrainedModel) - added:
constructor ModernBertForMaskedLM(config: ModernBertConfig) - added:
constructor ModernBertForMultipleChoice(config: ModernBertConfig) - added:
constructor ModernBertForQuestionAnswering(config: ModernBertConfig) - added:
constructor ModernBertForSequenceClassification(config: ModernBertConfig) - added:
constructor ModernBertForTokenClassification(config: ModernBertConfig) - added:
constructor ModernBertModel(config: ModernBertConfig) - added:
field ModernBertPreTrainedModel.base_model_prefix - added:
field ModernBertPreTrainedModel.config: ModernBertConfig - added:
field ModernBertPreTrainedModel.supports_gradient_checkpointing - added:
method ModernBertForMaskedLM.compiled_head(output: torch.Tensor) -> torch.Tensor - added:
method ModernBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_size: Opt… - added:
method ModernBertForMaskedLM.get_output_embeddings() - added:
method ModernBertForMaskedLM.set_output_embeddings(new_embeddings: nn.Linear) - added:
method ModernBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_siz… - added:
method ModernBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, bat… - added:
method ModernBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, b… - added:
method ModernBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batc… - added:
method ModernBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_size: Optional[int]=…, seq_len: Optional[int]… - added:
method ModernBertModel.get_input_embeddings() - added:
method ModernBertModel.set_input_embeddings(value) - added:
method ModernBertPreTrainedModel.resize_token_embeddings(*args, **kwargs)
- added:
class ModernBertConfig(PretrainedConfig) - added:
class ModernBertForMaskedLM(ModernBertPreTrainedModel) - added:
class ModernBertForMultipleChoice(ModernBertPreTrainedModel) - added:
class ModernBertForQuestionAnswering(ModernBertPreTrainedModel) - added:
class ModernBertForSequenceClassification(ModernBertPreTrainedModel) - added:
class ModernBertForTokenClassification(ModernBertPreTrainedModel) - added:
class ModernBertModel(ModernBertPreTrainedModel) - added:
class ModernBertPreTrainedModel(PreTrainedModel) - added:
constructor ModernBertConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, initializer_cutoff_factor=…, norm_eps=…, norm_bias=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, cls_token_id=…, sep_token_id=…, global_rope_theta=…, attention_bias=…, attention_dropout=…, global_attn_e… - added:
constructor ModernBertForMaskedLM(config: ModernBertConfig) - added:
constructor ModernBertForMultipleChoice(config: ModernBertConfig) - added:
constructor ModernBertForQuestionAnswering(config: ModernBertConfig) - added:
constructor ModernBertForSequenceClassification(config: ModernBertConfig) - added:
constructor ModernBertForTokenClassification(config: ModernBertConfig) - added:
constructor ModernBertModel(config: ModernBertConfig) - added:
field ModernBertConfig.attribute_map - added:
field ModernBertConfig.keys_to_ignore_at_inference - added:
field ModernBertConfig.model_type - added:
field ModernBertPreTrainedModel.base_model_prefix - added:
field ModernBertPreTrainedModel.config: ModernBertConfig - added:
field ModernBertPreTrainedModel.supports_gradient_checkpointing - added:
method ModernBertConfig.to_dict() - added:
method ModernBertForMaskedLM.compiled_head(output: torch.Tensor) -> torch.Tensor - added:
method ModernBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_size: Opt… - added:
method ModernBertForMaskedLM.get_output_embeddings() - added:
method ModernBertForMaskedLM.set_output_embeddings(new_embeddings: nn.Linear) - added:
method ModernBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_siz… - added:
method ModernBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, bat… - added:
method ModernBertForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, b… - added:
method ModernBertForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batc… - added:
method ModernBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, sliding_window_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, indices: Optional[torch.Tensor]=…, cu_seqlens: Optional[torch.Tensor]=…, max_seqlen: Optional[int]=…, batch_size: Optional[int]=…, seq_len: Optional[int]… - added:
method ModernBertModel.get_input_embeddings() - added:
method ModernBertModel.set_input_embeddings(value) - added:
method ModernBertPreTrainedModel.resize_token_embeddings(*args, **kwargs)
- added:
export define_import_structure
- added:
class ModernBertDecoderConfig(PretrainedConfig) - added:
constructor ModernBertDecoderConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, initializer_cutoff_factor=…, norm_eps=…, norm_bias=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, cls_token_id=…, sep_token_id=…, global_rope_theta=…, attention_bias=…, attention_dropout=…, embedd… - added:
field ModernBertDecoderConfig.attribute_map - added:
field ModernBertDecoderConfig.keys_to_ignore_at_inference - added:
field ModernBertDecoderConfig.model_type
- added:
class ModernBertDecoderForCausalLM(ModernBertDecoderPreTrainedModel, GenerationMixin) - added:
class ModernBertDecoderForSequenceClassification(ModernBertDecoderPreTrainedModel) - added:
class ModernBertDecoderModel(ModernBertDecoderPreTrainedModel) - added:
class ModernBertDecoderPreTrainedModel(PreTrainedModel) - added:
constructor ModernBertDecoderForCausalLM(config: ModernBertDecoderConfig) - added:
constructor ModernBertDecoderForSequenceClassification(config: ModernBertDecoderConfig) - added:
constructor ModernBertDecoderModel(config: ModernBertDecoderConfig) - added:
field ModernBertDecoderPreTrainedModel.base_model_prefix - added:
field ModernBertDecoderPreTrainedModel.config: ModernBertDecoderConfig - added:
field ModernBertDecoderPreTrainedModel.supports_gradient_checkpointing - added:
method ModernBertDecoderForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs) -> Union[tuple, CausalLMOutputWithPast] - added:
method ModernBertDecoderForCausalLM.get_output_embeddings() - added:
method ModernBertDecoderForCausalLM.set_output_embeddings(new_embeddings) - added:
method ModernBertDecoderForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs) -> Union[tuple, SequenceClassifierOutputWithPast] - added:
method ModernBertDecoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) -> Union[tuple[torch.Tensor, ...], BaseModelOutputWithPast] - added:
method ModernBertDecoderModel.get_input_embeddings() - added:
method ModernBertDecoderModel.set_input_embeddings(value)
- added:
class ModernBertDecoderConfig(PretrainedConfig) - added:
class ModernBertDecoderForCausalLM(ModernBertDecoderPreTrainedModel, GenerationMixin) - added:
class ModernBertDecoderForSequenceClassification(ModernBertDecoderPreTrainedModel) - added:
class ModernBertDecoderModel(ModernBertDecoderPreTrainedModel) - added:
class ModernBertDecoderPreTrainedModel(ModernBertPreTrainedModel) - added:
constructor ModernBertDecoderConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, initializer_cutoff_factor=…, norm_eps=…, norm_bias=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, cls_token_id=…, sep_token_id=…, global_rope_theta=…, attention_bias=…, attention_dropout=…, embedd… - added:
constructor ModernBertDecoderForCausalLM(config: ModernBertDecoderConfig) - added:
constructor ModernBertDecoderForSequenceClassification(config: ModernBertDecoderConfig) - added:
constructor ModernBertDecoderModel(config: ModernBertDecoderConfig) - added:
field ModernBertDecoderConfig.attribute_map - added:
field ModernBertDecoderConfig.keys_to_ignore_at_inference - added:
field ModernBertDecoderConfig.model_type - added:
method ModernBertDecoderForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs) -> Union[tuple, CausalLMOutputWithPast] - added:
method ModernBertDecoderForCausalLM.get_output_embeddings() - added:
method ModernBertDecoderForCausalLM.set_output_embeddings(new_embeddings) - added:
method ModernBertDecoderForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs) -> Union[tuple, SequenceClassifierOutputWithPast] - added:
method ModernBertDecoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs) -> Union[tuple[torch.Tensor, ...], BaseModelOutputWithPast] - added:
method ModernBertDecoderModel.get_input_embeddings() - added:
method ModernBertDecoderModel.set_input_embeddings(value) - added:
method ModernBertDecoderPreTrainedModel.resize_token_embeddings(*args, **kwargs)
- added:
export define_import_structure
- added:
class MoonshineConfig(PretrainedConfig) - added:
constructor MoonshineConfig(vocab_size=…, hidden_size=…, intermediate_size=…, encoder_num_hidden_layers=…, decoder_num_hidden_layers=…, encoder_num_attention_heads=…, decoder_num_attention_heads=…, encoder_num_key_value_heads=…, decoder_num_key_value_heads=…, pad_head_dim_to_multiple_of=…, encoder_hidden_act=…, decoder_hidden_act=…, max_position_embeddings=…, initializer_range=…, decoder_start_to… - added:
field MoonshineConfig.attribute_map - added:
field MoonshineConfig.keys_to_ignore_at_inference - added:
field MoonshineConfig.model_type
- added:
class MoonshineForConditionalGeneration(MoonshinePreTrainedModel, GenerationMixin) - added:
class MoonshineModel(MoonshinePreTrainedModel) - added:
class MoonshinePreTrainedModel(PreTrainedModel) - added:
constructor MoonshineForConditionalGeneration(config: MoonshineConfig) - added:
constructor MoonshineModel(config: MoonshineConfig) - added:
field MoonshinePreTrainedModel.base_model_prefix - added:
field MoonshinePreTrainedModel.config: MoonshineConfig - added:
field MoonshinePreTrainedModel.main_input_name - added:
field MoonshinePreTrainedModel.supports_gradient_checkpointing - added:
method MoonshineForConditionalGeneration.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Union[EncoderDecoderCache, tuple[torch.FloatTensor]]]=…, decoder_inp… - added:
method MoonshineForConditionalGeneration.get_decoder() - added:
method MoonshineForConditionalGeneration.get_encoder() - added:
method MoonshineForConditionalGeneration.get_input_embeddings() -> nn.Module - added:
method MoonshineForConditionalGeneration.get_output_embeddings() - added:
method MoonshineForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method MoonshineModel.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Union[EncoderDecoderCache, tuple[torch.FloatTensor]]]=…, decoder_inputs_embeds: Optiona… - added:
method MoonshineModel.freeze_encoder() - added:
method MoonshineModel.get_encoder() - added:
method MoonshineModel.get_input_embeddings() - added:
method MoonshineModel.set_input_embeddings(value)
- added:
class MoonshineConfig(PretrainedConfig) - added:
class MoonshineForConditionalGeneration(MoonshinePreTrainedModel, GenerationMixin) - added:
class MoonshineModel(WhisperModel) - added:
class MoonshinePreTrainedModel(PreTrainedModel) - added:
constructor MoonshineConfig(vocab_size=…, hidden_size=…, intermediate_size=…, encoder_num_hidden_layers=…, decoder_num_hidden_layers=…, encoder_num_attention_heads=…, decoder_num_attention_heads=…, encoder_num_key_value_heads=…, decoder_num_key_value_heads=…, pad_head_dim_to_multiple_of=…, encoder_hidden_act=…, decoder_hidden_act=…, max_position_embeddings=…, initializer_range=…, decoder_start_to… - added:
constructor MoonshineForConditionalGeneration(config: MoonshineConfig) - added:
field MoonshineConfig.attribute_map - added:
field MoonshineConfig.keys_to_ignore_at_inference - added:
field MoonshineConfig.model_type - added:
field MoonshinePreTrainedModel.base_model_prefix - added:
field MoonshinePreTrainedModel.config: MoonshineConfig - added:
field MoonshinePreTrainedModel.main_input_name - added:
field MoonshinePreTrainedModel.supports_gradient_checkpointing - added:
method MoonshineForConditionalGeneration.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Union[EncoderDecoderCache, tuple[torch.FloatTensor]]]=…, decoder_inp… - added:
method MoonshineForConditionalGeneration.get_decoder() - added:
method MoonshineForConditionalGeneration.get_encoder() - added:
method MoonshineForConditionalGeneration.get_input_embeddings() -> nn.Module - added:
method MoonshineForConditionalGeneration.get_output_embeddings() - added:
method MoonshineForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method MoonshineModel.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Union[EncoderDecoderCache, tuple[torch.FloatTensor]]]=…, decoder_inputs_embeds: Optiona…
- added:
export define_import_structure
- added:
class MoshiConfig(PretrainedConfig) - added:
class MoshiDepthConfig(PretrainedConfig) - added:
constructor MoshiConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, audio_vocab_size=…, max_position_embeddings=…, rope_theta=…, hidden_act=…, head_dim=…, initializer_range=…, use_cache=…, sliding_window=…, attention_dropout=…, ffn_dim=…, rms_norm_eps=…, num_codebooks=…, tie_word_embeddings=…, **kwargs) - added:
constructor MoshiDepthConfig(vocab_size=…, hidden_size=…, input_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, audio_vocab_size=…, max_position_embeddings=…, hidden_act=…, head_dim=…, initializer_range=…, use_cache=…, sliding_window=…, attention_dropout=…, ffn_dim=…, rms_norm_eps=…, num_codebooks=…, tie_word_embeddings=…, **kwargs) - added:
field MoshiConfig.keys_to_ignore_at_inference - added:
field MoshiConfig.model_type - added:
field MoshiConfig.sub_configs - added:
field MoshiDepthConfig.keys_to_ignore_at_inference - added:
field MoshiDepthConfig.model_type - added:
property MoshiConfig.sampling_rate - added:
static method MoshiConfig.from_audio_encoder_config(audio_encoder_config: PretrainedConfig, **kwargs)
- added:
class MoshiForCausalLM(MoshiPreTrainedModel, GenerationMixin) - added:
class MoshiForConditionalGeneration(MoshiPreTrainedModel, GenerationMixin) - added:
class MoshiModel(MoshiPreTrainedModel) - added:
class MoshiPreTrainedModel(PreTrainedModel) - added:
constructor MoshiForCausalLM(config) - added:
constructor MoshiForConditionalGeneration(config: MoshiConfig) - added:
constructor MoshiModel(config: MoshiConfig) - added:
field MoshiForConditionalGeneration.config: MoshiConfig - added:
field MoshiForConditionalGeneration.main_input_name - added:
field MoshiForConditionalGeneration.supports_gradient_checkpointing - added:
field MoshiPreTrainedModel.base_model_prefix - added:
field MoshiPreTrainedModel.config: MoshiConfig - added:
field MoshiPreTrainedModel.main_input_name - added:
field MoshiPreTrainedModel.supports_gradient_checkpointing - added:
method MoshiForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optiona… - added:
method MoshiForConditionalGeneration.build_delay_pattern_mask(input_ids: torch.LongTensor, bos_token_id: int, pad_token_id: int, max_length: Optional[int]=…) - added:
method MoshiForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.BoolTensor]=…, user_input_values: Optional[torch.FloatTensor]=…, user_audio_codes: Optional[torch.Tensor]=…, moshi_input_values: Optional[torch.FloatTensor]=…, moshi_audio_codes: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor… - added:
method MoshiForConditionalGeneration.freeze_audio_encoder() - added:
method MoshiForConditionalGeneration.freeze_depth_decoder() - added:
method MoshiForConditionalGeneration.generate(input_ids: Optional[torch.LongTensor]=…, user_input_values: Optional[torch.FloatTensor]=…, user_audio_codes: Optional[torch.Tensor]=…, moshi_input_values: Optional[torch.FloatTensor]=…, moshi_audio_codes: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, return_audio_waveforms: Optional[bool]=…, return_audio_codes: Optional[bool]… - added:
method MoshiForConditionalGeneration.get_audio_encoder() - added:
method MoshiForConditionalGeneration.get_depth_decoder() - added:
method MoshiForConditionalGeneration.get_input_embeddings() - added:
method MoshiForConditionalGeneration.get_output_embeddings() - added:
method MoshiForConditionalGeneration.get_unconditional_inputs(num_samples=…) - added:
method MoshiForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, user_delay_pattern_mask=…, moshi_delay_pattern_mask=…, kwargs_depth_decoder=…, blank_user_audio_codes: Optional[torch.FloatTensor]=…, **kwargs) - added:
method MoshiForConditionalGeneration.set_input_embeddings(value) - added:
method MoshiForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method MoshiModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool… - added:
static method MoshiForConditionalGeneration.apply_delay_pattern_mask(input_ids, decoder_pad_token_mask)
- added:
export define_import_structure
- added:
class MPNetConfig(PretrainedConfig) - added:
constructor MPNetConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, relative_attention_num_buckets=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field MPNetConfig.model_type
- added:
class MPNetForMaskedLM(MPNetPreTrainedModel) - added:
class MPNetForMultipleChoice(MPNetPreTrainedModel) - added:
class MPNetForQuestionAnswering(MPNetPreTrainedModel) - added:
class MPNetForSequenceClassification(MPNetPreTrainedModel) - added:
class MPNetForTokenClassification(MPNetPreTrainedModel) - added:
class MPNetLayer(nn.Module) - added:
class MPNetModel(MPNetPreTrainedModel) - added:
class MPNetPreTrainedModel(PreTrainedModel) - added:
constructor MPNetForMaskedLM(config) - added:
constructor MPNetForMultipleChoice(config) - added:
constructor MPNetForQuestionAnswering(config) - added:
constructor MPNetForSequenceClassification(config) - added:
constructor MPNetForTokenClassification(config) - added:
constructor MPNetLayer(config) - added:
constructor MPNetModel(config, add_pooling_layer=…) - added:
field MPNetPreTrainedModel.base_model_prefix - added:
field MPNetPreTrainedModel.config: MPNetConfig - added:
method MPNetForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -… - added:
method MPNetForMaskedLM.get_output_embeddings() - added:
method MPNetForMaskedLM.set_output_embeddings(new_embeddings) - added:
method MPNetForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool… - added:
method MPNetForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hi… - added:
method MPNetForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optio… - added:
method MPNetForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional… - added:
method MPNetLayer.forward(hidden_states, attention_mask=…, head_mask=…, position_bias=…, output_attentions=…, **kwargs) - added:
method MPNetModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> Union[tuple[torch.Tensor], BaseM… - added:
method MPNetModel.get_input_embeddings() - added:
method MPNetModel.set_input_embeddings(value)
- added:
class TFMPNetEmbeddings(keras.layers.Layer) - added:
class TFMPNetForMaskedLM(TFMPNetPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFMPNetForMultipleChoice(TFMPNetPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFMPNetForQuestionAnswering(TFMPNetPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFMPNetForSequenceClassification(TFMPNetPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFMPNetForTokenClassification(TFMPNetPreTrainedModel, TFTokenClassificationLoss) - added:
class TFMPNetMainLayer(keras.layers.Layer) - added:
class TFMPNetModel(TFMPNetPreTrainedModel) - added:
class TFMPNetPreTrainedModel(TFPreTrainedModel) - added:
constructor TFMPNetEmbeddings(config, **kwargs) - added:
constructor TFMPNetForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFMPNetForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFMPNetForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFMPNetForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFMPNetForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFMPNetMainLayer(config, **kwargs) - added:
constructor TFMPNetModel(config, *inputs, **kwargs) - added:
field TFMPNetMainLayer.config_class - added:
field TFMPNetPreTrainedModel.base_model_prefix - added:
field TFMPNetPreTrainedModel.config_class - added:
method TFMPNetEmbeddings.build(input_shape=…) - added:
method TFMPNetEmbeddings.call(input_ids=…, position_ids=…, inputs_embeds=…, training=…) - added:
method TFMPNetEmbeddings.create_position_ids_from_input_ids(input_ids) - added:
method TFMPNetForMaskedLM.build(input_shape=…) - added:
method TFMPNetForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: tf.Tensor | None=…, training: bool=…) -> TFMaske… - added:
method TFMPNetForMaskedLM.get_lm_head() - added:
method TFMPNetForMaskedLM.get_prefix_bias_name() - added:
method TFMPNetForMultipleChoice.build(input_shape=…) - added:
method TFMPNetForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: tf.Tensor | None=…, training: bool=…) -> T… - added:
method TFMPNetForQuestionAnswering.build(input_shape=…) - added:
method TFMPNetForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.array | tf.Tensor | None=…, position_ids: np.array | tf.Tensor | None=…, head_mask: np.array | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: tf.Tensor | None=…, end_positions: t… - added:
method TFMPNetForSequenceClassification.build(input_shape=…) - added:
method TFMPNetForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.array | tf.Tensor | None=…, position_ids: np.array | tf.Tensor | None=…, head_mask: np.array | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: tf.Tensor | None=…, training: bool=…) ->… - added:
method TFMPNetForTokenClassification.build(input_shape=…) - added:
method TFMPNetForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: tf.Tensor | None=…, training: bool=…)… - added:
method TFMPNetMainLayer.build(input_shape=…) - added:
method TFMPNetMainLayer.call(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFMPNetMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFMPNetMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFMPNetModel.build(input_shape=…) - added:
method TFMPNetModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.array | tf.Tensor | None=…, position_ids: np.array | tf.Tensor | None=…, head_mask: np.array | tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutput | tuple[tf.Tensor]
- added:
class MPNetTokenizer(PreTrainedTokenizer) - added:
constructor MPNetTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field MPNetTokenizer.model_input_names - added:
field MPNetTokenizer.vocab_files_names - added:
method MPNetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MPNetTokenizer.convert_tokens_to_string(tokens) - added:
method MPNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MPNetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MPNetTokenizer.get_vocab() - added:
method MPNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MPNetTokenizer.do_lower_case - added:
property MPNetTokenizer.vocab_size
- added:
class MPNetTokenizerFast(PreTrainedTokenizerFast) - added:
constructor MPNetTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field MPNetTokenizerFast.model_input_names - added:
field MPNetTokenizerFast.slow_tokenizer_class - added:
field MPNetTokenizerFast.vocab_files_names - added:
method MPNetTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method MPNetTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MPNetTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MPNetTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class MptConfig(PretrainedConfig) - added:
constructor MptConfig(d_model: int=…, n_heads: int=…, n_layers: int=…, expansion_ratio: int=…, max_seq_len: int=…, vocab_size: int=…, resid_pdrop: float=…, layer_norm_epsilon: float=…, emb_pdrop: float=…, learned_pos_emb: bool=…, attn_config: MptAttentionConfig=…, init_device: str=…, logit_scale: Optional[Union[float, str]]=…, no_bias: bool=…, verbose: int=…, embedding_fraction: float=…, norm_typ… - added:
field MptConfig.attribute_map - added:
field MptConfig.model_type - added:
field MptConfig.sub_configs
- added:
class MptForCausalLM(MptPreTrainedModel, GenerationMixin) - added:
class MptForQuestionAnswering(MptPreTrainedModel) - added:
class MptForSequenceClassification(MptPreTrainedModel) - added:
class MptForTokenClassification(MptPreTrainedModel) - added:
class MptModel(MptPreTrainedModel) - added:
class MptPreTrainedModel(PreTrainedModel) - added:
constructor MptForCausalLM(config: MptConfig) - added:
constructor MptForQuestionAnswering(config) - added:
constructor MptForSequenceClassification(config: MptConfig) - added:
constructor MptForTokenClassification(config: MptConfig) - added:
constructor MptModel(config: MptConfig) - added:
constructor MptPreTrainedModel(*inputs, **kwargs) - added:
field MptPreTrainedModel.base_model_prefix - added:
field MptPreTrainedModel.config: MptConfig - added:
field MptPreTrainedModel.supports_gradient_checkpointing - added:
method MptForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.Tensor]… - added:
method MptForCausalLM.set_output_embeddings(new_embeddings: torch.Tensor) - added:
method MptForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, QuestionAn… - added:
method MptForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Ten… - added:
method MptForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **deprecated_arguments) ->… - added:
method MptModel.build_mpt_alibi_tensor(num_heads, sequence_length, alibi_bias_max=…, device=…) - added:
method MptModel.forward(input_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…, **kwargs) -> Union[tuple[torch.T… - added:
method MptModel.get_input_embeddings() - added:
method MptModel.set_input_embeddings(new_embeddings: torch.Tensor)
- added:
export define_import_structure
- added:
class MraConfig(PretrainedConfig) - added:
constructor MraConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, position_embedding_type=…, block_per_row=…, approx_mode=…, initial_prior_first_n_blocks=…, initial_prior_diagonal_n_blocks=…, pad… - added:
field MraConfig.model_type
- added:
class MraForMaskedLM(MraPreTrainedModel) - added:
class MraForMultipleChoice(MraPreTrainedModel) - added:
class MraForQuestionAnswering(MraPreTrainedModel) - added:
class MraForSequenceClassification(MraPreTrainedModel) - added:
class MraForTokenClassification(MraPreTrainedModel) - added:
class MraLayer(GradientCheckpointingLayer) - added:
class MraModel(MraPreTrainedModel) - added:
class MraPreTrainedModel(PreTrainedModel) - added:
constructor MraForMaskedLM(config) - added:
constructor MraForMultipleChoice(config) - added:
constructor MraForQuestionAnswering(config) - added:
constructor MraForSequenceClassification(config) - added:
constructor MraForTokenClassification(config) - added:
constructor MraLayer(config) - added:
constructor MraModel(config) - added:
field MraPreTrainedModel.base_model_prefix - added:
field MraPreTrainedModel.config: MraConfig - added:
field MraPreTrainedModel.supports_gradient_checkpointing - added:
method MraForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, MaskedLMO… - added:
method MraForMaskedLM.get_output_embeddings() - added:
method MraForMaskedLM.set_output_embeddings(new_embeddings) - added:
method MraForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Mul… - added:
method MraForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=… - added:
method MraForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tu… - added:
method MraForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple… - added:
method MraLayer.feed_forward_chunk(attention_output) - added:
method MraLayer.forward(hidden_states, attention_mask=…) - added:
method MraModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithCrossAttentions] - added:
method MraModel.get_input_embeddings() - added:
method MraModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class MT5Config(PretrainedConfig) - added:
class MT5OnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor MT5Config(vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, use_cache=…, tokenizer_class=…, tie_word_embeddings=…, pad_token_id=…, eos_token_id=…, decoder_start_token_id=…, cl… - added:
field MT5Config.attribute_map - added:
field MT5Config.keys_to_ignore_at_inference - added:
field MT5Config.model_type - added:
property MT5OnnxConfig.atol_for_validation -> float - added:
property MT5OnnxConfig.default_onnx_opset -> int - added:
property MT5OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxMT5EncoderModel(FlaxT5EncoderModel) - added:
class FlaxMT5ForConditionalGeneration(FlaxT5ForConditionalGeneration) - added:
class FlaxMT5Model(FlaxT5Model) - added:
field FlaxMT5EncoderModel.config_class - added:
field FlaxMT5EncoderModel.model_type - added:
field FlaxMT5ForConditionalGeneration.config_class - added:
field FlaxMT5ForConditionalGeneration.model_type - added:
field FlaxMT5Model.config_class - added:
field FlaxMT5Model.model_type
- added:
class MT5EncoderModel(MT5PreTrainedModel) - added:
class MT5ForConditionalGeneration(MT5PreTrainedModel, GenerationMixin) - added:
class MT5ForQuestionAnswering(MT5PreTrainedModel) - added:
class MT5ForSequenceClassification(MT5PreTrainedModel) - added:
class MT5ForTokenClassification(MT5PreTrainedModel) - added:
class MT5Model(MT5PreTrainedModel) - added:
class MT5PreTrainedModel(PreTrainedModel) - added:
constructor MT5EncoderModel(config: MT5Config) - added:
constructor MT5ForConditionalGeneration(config: MT5Config) - added:
constructor MT5ForQuestionAnswering(config: MT5Config) - added:
constructor MT5ForSequenceClassification(config: MT5Config) - added:
constructor MT5ForTokenClassification(config: MT5Config) - added:
constructor MT5Model(config: MT5Config) - added:
field MT5EncoderModel.config: MT5Config - added:
field MT5EncoderModel.model_type - added:
field MT5ForConditionalGeneration.config: MT5Config - added:
field MT5ForConditionalGeneration.model_type - added:
field MT5Model.config: MT5Config - added:
field MT5Model.model_type - added:
field MT5PreTrainedModel.base_model_prefix - added:
field MT5PreTrainedModel.config: MT5Config - added:
field MT5PreTrainedModel.is_parallelizable - added:
field MT5PreTrainedModel.load_tf_weights - added:
field MT5PreTrainedModel.supports_gradient_checkpointing - added:
method MT5EncoderModel.deparallelize() - added:
method MT5EncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method MT5EncoderModel.get_encoder() - added:
method MT5EncoderModel.get_input_embeddings() - added:
method MT5EncoderModel.parallelize(device_map=…) - added:
method MT5EncoderModel.set_input_embeddings(new_embeddings) - added:
method MT5ForConditionalGeneration.deparallelize() - added:
method MT5ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional… - added:
method MT5ForConditionalGeneration.get_encoder() - added:
method MT5ForConditionalGeneration.get_input_embeddings() - added:
method MT5ForConditionalGeneration.parallelize(device_map=…) - added:
method MT5ForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MT5ForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method MT5ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tup… - added:
method MT5ForQuestionAnswering.get_encoder() - added:
method MT5ForQuestionAnswering.get_input_embeddings() - added:
method MT5ForQuestionAnswering.set_input_embeddings(new_embeddings) - added:
method MT5ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.Fl… - added:
method MT5ForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], TokenClassifierOutput] - added:
method MT5Model.deparallelize() - added:
method MT5Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.… - added:
method MT5Model.get_encoder() - added:
method MT5Model.get_input_embeddings() - added:
method MT5Model.parallelize(device_map=…) - added:
method MT5Model.set_input_embeddings(new_embeddings) - added:
property MT5PreTrainedModel.dummy_inputs
- added:
class TFMT5EncoderModel(TFT5EncoderModel) - added:
class TFMT5ForConditionalGeneration(TFT5ForConditionalGeneration) - added:
class TFMT5Model(TFT5Model) - added:
field TFMT5EncoderModel.config_class - added:
field TFMT5EncoderModel.model_type - added:
field TFMT5ForConditionalGeneration.config_class - added:
field TFMT5ForConditionalGeneration.model_type - added:
field TFMT5Model.config_class - added:
field TFMT5Model.model_type
- added:
class MT5Tokenizer(T5Tokenizer)
- added:
class MT5TokenizerFast(T5TokenizerFast)
- added:
export define_import_structure
- added:
class MusicgenConfig(PretrainedConfig) - added:
class MusicgenDecoderConfig(PretrainedConfig) - added:
constructor MusicgenConfig(**kwargs) - added:
constructor MusicgenDecoderConfig(vocab_size=…, max_position_embeddings=…, num_hidden_layers=…, ffn_dim=…, num_attention_heads=…, layerdrop=…, use_cache=…, activation_function=…, hidden_size=…, dropout=…, attention_dropout=…, activation_dropout=…, initializer_factor=…, scale_embedding=…, num_codebooks=…, audio_channels=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, **kw… - added:
field MusicgenConfig.has_no_defaults_at_init - added:
field MusicgenConfig.model_type - added:
field MusicgenConfig.sub_configs - added:
field MusicgenDecoderConfig.base_config_key - added:
field MusicgenDecoderConfig.keys_to_ignore_at_inference - added:
field MusicgenDecoderConfig.model_type - added:
property MusicgenConfig.sampling_rate - added:
static method MusicgenConfig.from_sub_models_config(text_encoder_config: PretrainedConfig, audio_encoder_config: PretrainedConfig, decoder_config: MusicgenDecoderConfig, **kwargs)
- added:
class MusicgenForCausalLM(MusicgenPreTrainedModel, GenerationMixin) - added:
class MusicgenForConditionalGeneration(MusicgenPreTrainedModel, GenerationMixin) - added:
class MusicgenModel(MusicgenPreTrainedModel) - added:
class MusicgenPreTrainedModel(PreTrainedModel) - added:
constructor MusicgenForCausalLM(config: MusicgenDecoderConfig) - added:
constructor MusicgenForConditionalGeneration(config: Optional[MusicgenConfig]=…, text_encoder: Optional[PreTrainedModel]=…, audio_encoder: Optional[PreTrainedModel]=…, decoder: Optional[MusicgenForCausalLM]=…) - added:
constructor MusicgenModel(config: MusicgenDecoderConfig) - added:
field MusicgenForConditionalGeneration.base_model_prefix - added:
field MusicgenForConditionalGeneration.config: MusicgenConfig - added:
field MusicgenForConditionalGeneration.main_input_name - added:
field MusicgenForConditionalGeneration.supports_gradient_checkpointing - added:
field MusicgenPreTrainedModel.base_model_prefix - added:
field MusicgenPreTrainedModel.config: MusicgenDecoderConfig - added:
field MusicgenPreTrainedModel.supports_gradient_checkpointing - added:
method MusicgenForCausalLM.build_delay_pattern_mask(input_ids: torch.LongTensor, pad_token_id: int, max_length: Optional[int]=…) - added:
method MusicgenForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels… - added:
method MusicgenForCausalLM.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, synced_gpus: Optional[bool]=…, streamer: Optional[BaseStreamer]=…, **kwargs) - added:
method MusicgenForCausalLM.get_decoder() - added:
method MusicgenForCausalLM.get_input_embeddings() - added:
method MusicgenForCausalLM.get_output_embeddings() - added:
method MusicgenForCausalLM.prepare_inputs_for_generation(input_ids, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, head_mask=…, cross_attn_head_mask=…, past_key_values=…, use_cache=…, delay_pattern_mask=…, guidance_scale=…, **kwargs) - added:
method MusicgenForCausalLM.set_decoder(decoder) - added:
method MusicgenForCausalLM.set_input_embeddings(value) - added:
method MusicgenForCausalLM.set_output_embeddings(new_embeddings) - added:
method MusicgenForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.BoolTensor]=…, input_values: Optional[torch.FloatTensor]=…, padding_mask: Optional[torch.BoolTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values:… - added:
method MusicgenForConditionalGeneration.freeze_audio_encoder() - added:
method MusicgenForConditionalGeneration.freeze_text_encoder() - added:
method MusicgenForConditionalGeneration.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, synced_gpus: Optional[bool]=…, streamer: Optional[BaseStreamer]=…, **kwargs) - added:
method MusicgenForConditionalGeneration.get_audio_encoder() - added:
method MusicgenForConditionalGeneration.get_encoder() - added:
method MusicgenForConditionalGeneration.get_input_embeddings() - added:
method MusicgenForConditionalGeneration.get_output_embeddings() - added:
method MusicgenForConditionalGeneration.get_text_encoder() - added:
method MusicgenForConditionalGeneration.get_unconditional_inputs(num_samples=…) - added:
method MusicgenForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MusicgenForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, head_mask=…, decoder_attention_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, decoder_delay_pattern_mask=…, guidance_scale=…, cache_position=…, **kwargs) - added:
method MusicgenForConditionalGeneration.resize_token_embeddings(*args, **kwargs) - added:
method MusicgenForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method MusicgenForConditionalGeneration.tie_weights() - added:
method MusicgenModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: O… - added:
method MusicgenModel.get_input_embeddings() - added:
method MusicgenModel.set_input_embeddings(value) - added:
static method MusicgenForCausalLM.apply_delay_pattern_mask(input_ids, decoder_pad_token_mask) - added:
static method MusicgenForConditionalGeneration.from_sub_models_pretrained(text_encoder_pretrained_model_name_or_path: Optional[str]=…, audio_encoder_pretrained_model_name_or_path: Optional[str]=…, decoder_pretrained_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel
- added:
class MusicgenProcessor(ProcessorMixin) - added:
constructor MusicgenProcessor(feature_extractor, tokenizer) - added:
field MusicgenProcessor.feature_extractor_class - added:
field MusicgenProcessor.tokenizer_class - added:
method MusicgenProcessor.batch_decode(*args, **kwargs) - added:
method MusicgenProcessor.get_decoder_prompt_ids(task=…, language=…, no_timestamps=…)
- added:
export define_import_structure
- added:
class MusicgenMelodyConfig(PretrainedConfig) - added:
class MusicgenMelodyDecoderConfig(PretrainedConfig) - added:
constructor MusicgenMelodyConfig(num_chroma=…, chroma_length=…, **kwargs) - added:
constructor MusicgenMelodyDecoderConfig(vocab_size=…, max_position_embeddings=…, num_hidden_layers=…, ffn_dim=…, num_attention_heads=…, layerdrop=…, use_cache=…, activation_function=…, hidden_size=…, dropout=…, attention_dropout=…, activation_dropout=…, initializer_factor=…, scale_embedding=…, num_codebooks=…, audio_channels=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…… - added:
field MusicgenMelodyConfig.has_no_defaults_at_init - added:
field MusicgenMelodyConfig.model_type - added:
field MusicgenMelodyConfig.sub_configs - added:
field MusicgenMelodyDecoderConfig.base_config_key - added:
field MusicgenMelodyDecoderConfig.keys_to_ignore_at_inference - added:
field MusicgenMelodyDecoderConfig.model_type - added:
property MusicgenMelodyConfig.sampling_rate - added:
static method MusicgenMelodyConfig.from_sub_models_config(text_encoder_config: PretrainedConfig, audio_encoder_config: PretrainedConfig, decoder_config: MusicgenMelodyDecoderConfig, **kwargs)
- added:
class MusicgenMelodyFeatureExtractor(SequenceFeatureExtractor) - added:
constructor MusicgenMelodyFeatureExtractor(feature_size=…, sampling_rate=…, hop_length=…, chunk_length=…, n_fft=…, num_chroma=…, padding_value=…, return_attention_mask=…, stem_indices=…, **kwargs) - added:
field MusicgenMelodyFeatureExtractor.model_input_names - added:
method MusicgenMelodyFeatureExtractor.to_dict() -> dict[str, Any]
- added:
class MusicgenMelodyForCausalLM(MusicgenMelodyPreTrainedModel, GenerationMixin) - added:
class MusicgenMelodyForConditionalGeneration(PreTrainedModel, GenerationMixin) - added:
class MusicgenMelodyModel(MusicgenMelodyPreTrainedModel) - added:
class MusicgenMelodyPreTrainedModel(PreTrainedModel) - added:
constructor MusicgenMelodyForCausalLM(config: MusicgenMelodyDecoderConfig) - added:
constructor MusicgenMelodyForConditionalGeneration(config: MusicgenMelodyConfig=…, text_encoder: Optional[PreTrainedModel]=…, audio_encoder: Optional[PreTrainedModel]=…, decoder: Optional[MusicgenMelodyForCausalLM]=…) - added:
constructor MusicgenMelodyModel(config: MusicgenMelodyDecoderConfig) - added:
field MusicgenMelodyForConditionalGeneration.config: MusicgenMelodyConfig - added:
field MusicgenMelodyForConditionalGeneration.main_input_name - added:
field MusicgenMelodyForConditionalGeneration.supports_gradient_checkpointing - added:
field MusicgenMelodyPreTrainedModel.base_model_prefix - added:
field MusicgenMelodyPreTrainedModel.config: MusicgenMelodyDecoderConfig - added:
field MusicgenMelodyPreTrainedModel.supports_gradient_checkpointing - added:
method MusicgenMelodyForCausalLM.build_delay_pattern_mask(input_ids: torch.LongTensor, pad_token_id: int, max_length: Optional[int]=…) - added:
method MusicgenMelodyForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions:… - added:
method MusicgenMelodyForCausalLM.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, synced_gpus: Optional[bool]=…, streamer: Optional[BaseStreamer]=…, **kwargs) - added:
method MusicgenMelodyForCausalLM.get_decoder() - added:
method MusicgenMelodyForCausalLM.get_input_embeddings() - added:
method MusicgenMelodyForCausalLM.get_output_embeddings() - added:
method MusicgenMelodyForCausalLM.prepare_inputs_for_generation(input_ids, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, head_mask=…, past_key_values=…, use_cache=…, delay_pattern_mask=…, guidance_scale=…, **kwargs) - added:
method MusicgenMelodyForCausalLM.set_decoder(decoder) - added:
method MusicgenMelodyForCausalLM.set_input_embeddings(value) - added:
method MusicgenMelodyForCausalLM.set_output_embeddings(new_embeddings) - added:
method MusicgenMelodyForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.BoolTensor]=…, input_features: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, past_key_values: Optional[Cache]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, inputs_embeds: Opt… - added:
method MusicgenMelodyForConditionalGeneration.freeze_audio_encoder() - added:
method MusicgenMelodyForConditionalGeneration.freeze_text_encoder() - added:
method MusicgenMelodyForConditionalGeneration.generate(inputs: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, synced_gpus: Optional[bool]=…, streamer: Optional[BaseStreamer]=…, **kwargs) - added:
method MusicgenMelodyForConditionalGeneration.get_encoder() - added:
method MusicgenMelodyForConditionalGeneration.get_input_embeddings() - added:
method MusicgenMelodyForConditionalGeneration.get_output_embeddings() - added:
method MusicgenMelodyForConditionalGeneration.get_text_encoder() - added:
method MusicgenMelodyForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MusicgenMelodyForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, encoder_hidden_states=…, past_key_values=…, attention_mask=…, decoder_attention_mask=…, decoder_head_mask=…, use_cache=…, decoder_delay_pattern_mask=…, guidance_scale=…, **kwargs) - added:
method MusicgenMelodyForConditionalGeneration.resize_token_embeddings(*args, **kwargs) - added:
method MusicgenMelodyForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method MusicgenMelodyForConditionalGeneration.tie_weights() - added:
method MusicgenMelodyModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Option… - added:
method MusicgenMelodyModel.get_input_embeddings() - added:
method MusicgenMelodyModel.set_input_embeddings(value) - added:
static method MusicgenMelodyForCausalLM.apply_delay_pattern_mask(input_ids, decoder_pad_token_mask) - added:
static method MusicgenMelodyForConditionalGeneration.from_sub_models_pretrained(text_encoder_pretrained_model_name_or_path: Optional[str]=…, audio_encoder_pretrained_model_name_or_path: Optional[str]=…, decoder_pretrained_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel
- added:
class MusicgenMelodyProcessor(ProcessorMixin) - added:
constructor MusicgenMelodyProcessor(feature_extractor, tokenizer) - added:
field MusicgenMelodyProcessor.feature_extractor_class - added:
field MusicgenMelodyProcessor.tokenizer_class - added:
method MusicgenMelodyProcessor.batch_decode(*args, **kwargs) - added:
method MusicgenMelodyProcessor.get_decoder_prompt_ids(task=…, language=…, no_timestamps=…) - added:
method MusicgenMelodyProcessor.get_unconditional_inputs(num_samples=…, return_tensors=…)
- added:
export define_import_structure
- added:
class MvpConfig(PretrainedConfig) - added:
constructor MvpConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, classifier_dropout=…, scale_embedding=…, use_cache=…, pad_token_id… - added:
field MvpConfig.attribute_map - added:
field MvpConfig.keys_to_ignore_at_inference - added:
field MvpConfig.model_type
- added:
class MvpForCausalLM(MvpPreTrainedModel, GenerationMixin) - added:
class MvpForConditionalGeneration(MvpPreTrainedModel, GenerationMixin) - added:
class MvpForQuestionAnswering(MvpPreTrainedModel) - added:
class MvpForSequenceClassification(MvpPreTrainedModel) - added:
class MvpModel(MvpPreTrainedModel) - added:
class MvpPreTrainedModel(PreTrainedModel) - added:
constructor MvpForCausalLM(config) - added:
constructor MvpForConditionalGeneration(config: MvpConfig) - added:
constructor MvpForQuestionAnswering(config) - added:
constructor MvpForSequenceClassification(config: MvpConfig, **kwargs) - added:
constructor MvpModel(config: MvpConfig) - added:
field MvpPreTrainedModel.base_model_prefix - added:
field MvpPreTrainedModel.config: MvpConfig - added:
field MvpPreTrainedModel.supports_gradient_checkpointing - added:
method MvpForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Op… - added:
method MvpForCausalLM.get_decoder() - added:
method MvpForCausalLM.get_input_embeddings() - added:
method MvpForCausalLM.set_decoder(decoder) - added:
method MvpForCausalLM.set_input_embeddings(value) - added:
method MvpForCausalLM.set_lightweight_tuning() - added:
method MvpForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.Flo… - added:
method MvpForConditionalGeneration.get_decoder() - added:
method MvpForConditionalGeneration.get_encoder() - added:
method MvpForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method MvpForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method MvpForConditionalGeneration.set_lightweight_tuning() - added:
method MvpForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor… - added:
method MvpForQuestionAnswering.set_lightweight_tuning() - added:
method MvpForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.Fl… - added:
method MvpForSequenceClassification.set_lightweight_tuning() - added:
method MvpModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor]]=…, past_… - added:
method MvpModel.get_encoder() - added:
method MvpModel.get_input_embeddings() - added:
method MvpModel.set_input_embeddings(value) - added:
method MvpModel.set_lightweight_tuning() - added:
property MvpPreTrainedModel.dummy_inputs
- added:
class MvpTokenizer(PreTrainedTokenizer) - added:
constructor MvpTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field MvpTokenizer.model_input_names - added:
field MvpTokenizer.vocab_files_names - added:
method MvpTokenizer.bpe(token) - added:
method MvpTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MvpTokenizer.convert_tokens_to_string(tokens) - added:
method MvpTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MvpTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MvpTokenizer.get_vocab() - added:
method MvpTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method MvpTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MvpTokenizer.vocab_size
- added:
class MvpTokenizerFast(PreTrainedTokenizerFast) - added:
constructor MvpTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field MvpTokenizerFast.model_input_names - added:
field MvpTokenizerFast.slow_tokenizer_class - added:
field MvpTokenizerFast.vocab_files_names - added:
method MvpTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method MvpTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MvpTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MvpTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class MyT5Tokenizer(PreTrainedTokenizer) - added:
constructor MyT5Tokenizer(vocab_file, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, **kwargs) -> None - added:
field MyT5Tokenizer.model_input_names - added:
field MyT5Tokenizer.vocab_files_names - added:
method MyT5Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MyT5Tokenizer.convert_tokens_to_string(tokens) - added:
method MyT5Tokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method MyT5Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MyT5Tokenizer.get_vocab() - added:
method MyT5Tokenizer.morphological_decode(indices: list[str]) -> list[str] - added:
method MyT5Tokenizer.morphological_encode(indices: list[str]) -> list[str] - added:
method MyT5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property MyT5Tokenizer.vocab_size
- added:
export define_import_structure
- added:
class NemotronConfig(PretrainedConfig) - added:
constructor NemotronConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, head_dim=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, partial_rotary_factor=…, attention_bias=…, attention_dropout=…, mlp_bia… - added:
field NemotronConfig.keys_to_ignore_at_inference - added:
field NemotronConfig.model_type
- added:
class NemotronForCausalLM(NemotronPreTrainedModel, GenerationMixin) - added:
class NemotronForQuestionAnswering(GenericForQuestionAnswering, NemotronPreTrainedModel) - added:
class NemotronForSequenceClassification(GenericForSequenceClassification, NemotronPreTrainedModel) - added:
class NemotronForTokenClassification(GenericForTokenClassification, NemotronPreTrainedModel) - added:
class NemotronModel(NemotronPreTrainedModel) - added:
class NemotronPreTrainedModel(PreTrainedModel) - added:
constructor NemotronForCausalLM(config) - added:
constructor NemotronModel(config: NemotronConfig) - added:
field NemotronForQuestionAnswering.base_model_prefix - added:
field NemotronPreTrainedModel.base_model_prefix - added:
field NemotronPreTrainedModel.config: NemotronConfig - added:
field NemotronPreTrainedModel.supports_gradient_checkpointing - added:
method NemotronForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_state… - added:
method NemotronModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optiona…
- added:
export define_import_structure
- added:
class NllbTokenizer(PreTrainedTokenizer) - added:
constructor NllbTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, tokenizer_file=…, src_lang=…, tgt_lang=…, sp_model_kwargs: Optional[dict[str, Any]]=…, additional_special_tokens=…, legacy_behaviour=…, **kwargs) - added:
field NllbTokenizer.model_input_names - added:
field NllbTokenizer.prefix_tokens: list[int] - added:
field NllbTokenizer.suffix_tokens: list[int] - added:
field NllbTokenizer.vocab_files_names - added:
method NllbTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method NllbTokenizer.convert_tokens_to_string(tokens) - added:
method NllbTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method NllbTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method NllbTokenizer.get_vocab() - added:
method NllbTokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method NllbTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method NllbTokenizer.set_src_lang_special_tokens(src_lang) -> None - added:
method NllbTokenizer.set_tgt_lang_special_tokens(lang: str) -> None - added:
property NllbTokenizer.src_lang -> str - added:
property NllbTokenizer.vocab_size
- added:
class NllbTokenizerFast(PreTrainedTokenizerFast) - added:
constructor NllbTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, src_lang=…, tgt_lang=…, additional_special_tokens=…, legacy_behaviour=…, **kwargs) - added:
field NllbTokenizerFast.model_input_names - added:
field NllbTokenizerFast.prefix_tokens: list[int] - added:
field NllbTokenizerFast.slow_tokenizer_class - added:
field NllbTokenizerFast.suffix_tokens: list[int] - added:
field NllbTokenizerFast.vocab_files_names - added:
method NllbTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method NllbTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method NllbTokenizerFast.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method NllbTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method NllbTokenizerFast.set_src_lang_special_tokens(src_lang) -> None - added:
method NllbTokenizerFast.set_tgt_lang_special_tokens(lang: str) -> None - added:
property NllbTokenizerFast.src_lang -> str
- added:
export define_import_structure
- added:
class NllbMoeConfig(PretrainedConfig) - added:
constructor NllbMoeConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…, s… - added:
field NllbMoeConfig.attribute_map - added:
field NllbMoeConfig.keys_to_ignore_at_inference - added:
field NllbMoeConfig.model_type
- added:
class NllbMoeForConditionalGeneration(NllbMoePreTrainedModel, GenerationMixin) - added:
class NllbMoeModel(NllbMoePreTrainedModel) - added:
class NllbMoePreTrainedModel(PreTrainedModel) - added:
class NllbMoeSparseMLP(nn.Module) - added:
class NllbMoeTop2Router(nn.Module) - added:
constructor NllbMoeForConditionalGeneration(config: NllbMoeConfig) - added:
constructor NllbMoeModel(config: NllbMoeConfig) - added:
constructor NllbMoeSparseMLP(config: NllbMoeConfig, ffn_dim: int, expert_class: nn.Module=…) - added:
constructor NllbMoeTop2Router(config: NllbMoeConfig) - added:
field NllbMoeForConditionalGeneration.base_model_prefix - added:
field NllbMoePreTrainedModel.base_model_prefix - added:
field NllbMoePreTrainedModel.config: NllbMoeConfig - added:
field NllbMoePreTrainedModel.supports_gradient_checkpointing - added:
method NllbMoeForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tupl… - added:
method NllbMoeForConditionalGeneration.get_decoder() - added:
method NllbMoeForConditionalGeneration.get_encoder() - added:
method NllbMoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor… - added:
method NllbMoeModel.get_encoder() - added:
method NllbMoeModel.get_input_embeddings() - added:
method NllbMoeModel.set_input_embeddings(value) - added:
method NllbMoeSparseMLP.forward(hidden_states: torch.Tensor, padding_mask: Optional[torch.Tensor]=…) - added:
method NllbMoeTop2Router.forward(hidden_states: torch.Tensor, padding_mask: Optional[torch.LongTensor]=…) -> tuple - added:
method NllbMoeTop2Router.normalize_router_probabilities(router_probs, top_1_mask, top_2_mask) - added:
method NllbMoeTop2Router.route_tokens(router_logits: torch.Tensor, input_dtype: torch.dtype=…, padding_mask: Optional[torch.LongTensor]=…) -> tuple
- added:
export define_import_structure
- added:
class NougatImageProcessor(BaseImageProcessor) - added:
constructor NougatImageProcessor(do_crop_margin: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_thumbnail: bool=…, do_align_long_axis: bool=…, do_pad: bool=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -… - added:
field NougatImageProcessor.model_input_names - added:
method NougatImageProcessor.align_long_axis(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method NougatImageProcessor.crop_margin(image: np.ndarray, gray_threshold: int=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method NougatImageProcessor.pad_image(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method NougatImageProcessor.preprocess(images: ImageInput, do_crop_margin: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_thumbnail: Optional[bool]=…, do_align_long_axis: Optional[bool]=…, do_pad: Optional[bool]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[bool]… - added:
method NougatImageProcessor.python_bounding_rect(coordinates) - added:
method NougatImageProcessor.python_find_non_zero(image: np.ndarray) - added:
method NougatImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method NougatImageProcessor.thumbnail(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class NougatImageProcessorFast(BaseImageProcessorFast) - added:
constructor NougatImageProcessorFast(**kwargs: Unpack[NougatFastImageProcessorKwargs]) - added:
field NougatImageProcessorFast.do_align_long_axis: bool - added:
field NougatImageProcessorFast.do_crop_margin: bool - added:
field NougatImageProcessorFast.do_normalize: bool - added:
field NougatImageProcessorFast.do_pad: bool - added:
field NougatImageProcessorFast.do_rescale - added:
field NougatImageProcessorFast.do_resize: bool - added:
field NougatImageProcessorFast.do_thumbnail: bool - added:
field NougatImageProcessorFast.image_mean - added:
field NougatImageProcessorFast.image_std - added:
field NougatImageProcessorFast.resample - added:
field NougatImageProcessorFast.size - added:
field NougatImageProcessorFast.valid_kwargs - added:
method NougatImageProcessorFast.align_long_axis(image: torch.Tensor, size: SizeDict) -> torch.Tensor - added:
method NougatImageProcessorFast.crop_margin(image: torch.Tensor, gray_threshold: int=…) -> torch.Tensor - added:
method NougatImageProcessorFast.pad_images(image: torch.Tensor, size: SizeDict) -> torch.Tensor - added:
method NougatImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[NougatFastImageProcessorKwargs]) -> BatchFeature - added:
method NougatImageProcessorFast.python_bounding_rect(coordinates) - added:
method NougatImageProcessorFast.python_find_non_zero(image: torch.Tensor) - added:
method NougatImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method NougatImageProcessorFast.thumbnail(image: torch.Tensor, size: SizeDict) -> torch.Tensor
- added:
class NougatProcessor(ProcessorMixin) - added:
constructor NougatProcessor(image_processor, tokenizer) - added:
field NougatProcessor.attributes - added:
field NougatProcessor.image_processor_class - added:
field NougatProcessor.tokenizer_class - added:
method NougatProcessor.post_process_generation(*args, **kwargs)
- added:
class NougatTokenizerFast(PreTrainedTokenizerFast) - added:
constructor NougatTokenizerFast(vocab_file=…, tokenizer_file=…, clean_up_tokenization_spaces=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, **kwargs) - added:
field NougatTokenizerFast.model_input_names - added:
field NougatTokenizerFast.slow_tokenizer_class - added:
field NougatTokenizerFast.vocab_files_names - added:
method NougatTokenizerFast.correct_tables(generation: str) -> str - added:
method NougatTokenizerFast.post_process_generation(generation: Union[str, list[str]], fix_markdown: bool=…, num_workers: Optional[int]=…) -> Union[str, list[str]] - added:
method NougatTokenizerFast.post_process_single(generation: str, fix_markdown: bool=…) -> str - added:
method NougatTokenizerFast.remove_hallucinated_references(text: str) -> str
- added:
export define_import_structure
- added:
class NystromformerConfig(PretrainedConfig) - added:
constructor NystromformerConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, segment_means_seq_len=…, num_landmarks=…, conv_kernel_size=…, inv_coeff_init_option=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…,… - added:
field NystromformerConfig.model_type
- added:
class NystromformerForMaskedLM(NystromformerPreTrainedModel) - added:
class NystromformerForMultipleChoice(NystromformerPreTrainedModel) - added:
class NystromformerForQuestionAnswering(NystromformerPreTrainedModel) - added:
class NystromformerForSequenceClassification(NystromformerPreTrainedModel) - added:
class NystromformerForTokenClassification(NystromformerPreTrainedModel) - added:
class NystromformerLayer(GradientCheckpointingLayer) - added:
class NystromformerModel(NystromformerPreTrainedModel) - added:
class NystromformerPreTrainedModel(PreTrainedModel) - added:
constructor NystromformerForMaskedLM(config) - added:
constructor NystromformerForMultipleChoice(config) - added:
constructor NystromformerForQuestionAnswering(config) - added:
constructor NystromformerForSequenceClassification(config) - added:
constructor NystromformerForTokenClassification(config) - added:
constructor NystromformerLayer(config) - added:
constructor NystromformerModel(config) - added:
field NystromformerPreTrainedModel.base_model_prefix - added:
field NystromformerPreTrainedModel.config: NystromformerConfig - added:
field NystromformerPreTrainedModel.supports_gradient_checkpointing - added:
method NystromformerForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method NystromformerForMaskedLM.get_output_embeddings() - added:
method NystromformerForMaskedLM.set_output_embeddings(new_embeddings) - added:
method NystromformerForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidde… - added:
method NystromformerForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTen… - added:
method NystromformerForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outp… - added:
method NystromformerForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_… - added:
method NystromformerLayer.feed_forward_chunk(attention_output) - added:
method NystromformerLayer.forward(hidden_states, attention_mask=…, output_attentions=…) - added:
method NystromformerModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[… - added:
method NystromformerModel.get_input_embeddings() - added:
method NystromformerModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class OlmoConfig(PretrainedConfig) - added:
constructor OlmoConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, clip_qkv=…, **kwargs) - added:
field OlmoConfig.base_model_pp_plan - added:
field OlmoConfig.base_model_tp_plan - added:
field OlmoConfig.keys_to_ignore_at_inference - added:
field OlmoConfig.model_type
- added:
class OlmoForCausalLM(OlmoPreTrainedModel, GenerationMixin) - added:
class OlmoModel(OlmoPreTrainedModel) - added:
class OlmoPreTrainedModel(PreTrainedModel) - added:
constructor OlmoForCausalLM(config) - added:
constructor OlmoModel(config: OlmoConfig) - added:
field OlmoPreTrainedModel.base_model_prefix - added:
field OlmoPreTrainedModel.config: OlmoConfig - added:
field OlmoPreTrainedModel.supports_gradient_checkpointing - added:
method OlmoForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **… - added:
method OlmoModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class OlmoForCausalLM(LlamaForCausalLM) - added:
class OlmoModel(LlamaModel) - added:
constructor OlmoModel(config: OlmoConfig)
- added:
export define_import_structure
- added:
class Olmo2Config(PretrainedConfig) - added:
constructor Olmo2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, rms_norm_eps=…, **kwargs) - added:
field Olmo2Config.base_model_pp_plan - added:
field Olmo2Config.base_model_tp_plan - added:
field Olmo2Config.keys_to_ignore_at_inference - added:
field Olmo2Config.model_type
- added:
class Olmo2ForCausalLM(Olmo2PreTrainedModel, GenerationMixin) - added:
class Olmo2Model(Olmo2PreTrainedModel) - added:
class Olmo2PreTrainedModel(PreTrainedModel) - added:
constructor Olmo2ForCausalLM(config) - added:
constructor Olmo2Model(config: Olmo2Config) - added:
field Olmo2PreTrainedModel.base_model_prefix - added:
field Olmo2PreTrainedModel.config: Olmo2Config - added:
field Olmo2PreTrainedModel.supports_gradient_checkpointing - added:
method Olmo2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method Olmo2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Olmo2Config(OlmoConfig) - added:
class Olmo2ForCausalLM(OlmoForCausalLM) - added:
class Olmo2Model(OlmoModel) - added:
class Olmo2PreTrainedModel(LlamaPreTrainedModel) - added:
constructor Olmo2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, rms_norm_eps=…, **kwargs) - added:
constructor Olmo2Model(config: Olmo2Config) - added:
field Olmo2Config.base_model_pp_plan - added:
field Olmo2Config.base_model_tp_plan - added:
field Olmo2Config.model_type
- added:
export define_import_structure
- added:
class Olmo3Config(PretrainedConfig) - added:
constructor Olmo3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, rms_norm_eps=…, sliding_window=…, layer_typ… - added:
field Olmo3Config.base_model_pp_plan - added:
field Olmo3Config.base_model_tp_plan - added:
field Olmo3Config.keys_to_ignore_at_inference - added:
field Olmo3Config.model_type
- added:
class Olmo3ForCausalLM(Olmo3PreTrainedModel, GenerationMixin) - added:
class Olmo3Model(Olmo3PreTrainedModel) - added:
class Olmo3PreTrainedModel(PreTrainedModel) - added:
constructor Olmo3ForCausalLM(config) - added:
constructor Olmo3Model(config: Olmo3Config) - added:
field Olmo3PreTrainedModel.base_model_prefix - added:
field Olmo3PreTrainedModel.config: Olmo3Config - added:
field Olmo3PreTrainedModel.supports_gradient_checkpointing - added:
method Olmo3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method Olmo3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Olmo3Config(Olmo2Config) - added:
class Olmo3ForCausalLM(Olmo2ForCausalLM) - added:
class Olmo3Model(Olmo2Model) - added:
class Olmo3PreTrainedModel(Olmo2PreTrainedModel) - added:
constructor Olmo3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, rms_norm_eps=…, sliding_window=…, layer_typ… - added:
constructor Olmo3Model(config: Olmo3Config) - added:
field Olmo3Config.base_model_pp_plan - added:
field Olmo3Config.base_model_tp_plan - added:
field Olmo3Config.model_type - added:
method Olmo3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
export define_import_structure
- added:
class OlmoeConfig(PretrainedConfig) - added:
constructor OlmoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, clip_qkv=…, num_experts_per… - added:
field OlmoeConfig.keys_to_ignore_at_inference - added:
field OlmoeConfig.model_type
- added:
class OlmoeForCausalLM(OlmoePreTrainedModel, GenerationMixin) - added:
class OlmoeModel(OlmoePreTrainedModel) - added:
class OlmoePreTrainedModel(PreTrainedModel) - added:
constructor OlmoeForCausalLM(config) - added:
constructor OlmoeModel(config: OlmoeConfig) - added:
field OlmoePreTrainedModel.base_model_prefix - added:
field OlmoePreTrainedModel.config: OlmoeConfig - added:
field OlmoePreTrainedModel.supports_gradient_checkpointing - added:
method OlmoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_… - added:
method OlmoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Opti…
- added:
export define_import_structure
- added:
class OmDetTurboConfig(PretrainedConfig) - added:
constructor OmDetTurboConfig(text_config=…, backbone_config=…, use_timm_backbone=…, backbone=…, backbone_kwargs=…, use_pretrained_backbone=…, apply_layernorm_after_vision_backbone=…, image_size=…, disable_custom_kernels=…, layer_norm_eps=…, batch_norm_eps=…, init_std=…, text_projection_in_dim=…, text_projection_out_dim=…, task_encoder_hidden_dim=…, class_embed_dim=…, class_distance_type=…, num_qu… - added:
field OmDetTurboConfig.attribute_map - added:
field OmDetTurboConfig.model_type - added:
property OmDetTurboConfig.sub_configs
- added:
class OmDetTurboForObjectDetection(OmDetTurboPreTrainedModel) - added:
class OmDetTurboPreTrainedModel(PreTrainedModel) - added:
constructor OmDetTurboForObjectDetection(config: OmDetTurboConfig) - added:
field OmDetTurboPreTrainedModel.base_model_prefix - added:
field OmDetTurboPreTrainedModel.config: OmDetTurboConfig - added:
field OmDetTurboPreTrainedModel.main_input_name - added:
method OmDetTurboForObjectDetection.forward(pixel_values: torch.FloatTensor, classes_input_ids: torch.LongTensor, classes_attention_mask: torch.LongTensor, tasks_input_ids: torch.LongTensor, tasks_attention_mask: torch.LongTensor, classes_structure: torch.LongTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Opt… - added:
method OmDetTurboForObjectDetection.get_input_embeddings() - added:
method OmDetTurboForObjectDetection.resize_token_embeddings(new_num_tokens: Optional[int]=…, pad_to_multiple_of=…, mean_resizing: bool=…) -> nn.Embedding - added:
method OmDetTurboForObjectDetection.set_input_embeddings(value) - added:
method OmDetTurboPreTrainedModel.get_cached_class_embeddings(classes_input_ids, classes_attention_mask) - added:
method OmDetTurboPreTrainedModel.get_cached_task_embeddings(tasks_input_ids, tasks_attention_mask) - added:
method OmDetTurboPreTrainedModel.get_language_embedding(classes_input_ids, classes_attention_mask, tasks_input_ids, tasks_attention_mask, classes_structure)
- added:
class OmDetTurboProcessor(ProcessorMixin) - added:
constructor OmDetTurboProcessor(image_processor, tokenizer) - added:
field OmDetTurboProcessor.attributes - added:
field OmDetTurboProcessor.image_processor_class - added:
field OmDetTurboProcessor.tokenizer_class - added:
method OmDetTurboProcessor.post_process_grounded_object_detection(outputs: OmDetTurboObjectDetectionOutput, text_labels: Optional[Union[list[str], list[list[str]]]]=…, threshold: float=…, nms_threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…, max_num_det: Optional[int]=…) - added:
property OmDetTurboProcessor.model_input_names
- added:
export define_import_structure
- added:
class OneFormerConfig(PretrainedConfig) - added:
constructor OneFormerConfig(backbone_config: Optional[dict]=…, backbone: Optional[str]=…, use_pretrained_backbone: bool=…, use_timm_backbone: bool=…, backbone_kwargs: Optional[dict]=…, ignore_value: int=…, num_queries: int=…, no_object_weight: int=…, class_weight: float=…, mask_weight: float=…, dice_weight: float=…, contrastive_weight: float=…, contrastive_temperature: float=…, train_num_points:… - added:
field OneFormerConfig.attribute_map - added:
field OneFormerConfig.model_type - added:
property OneFormerConfig.sub_configs
- added:
class OneFormerImageProcessor(BaseImageProcessor) - added:
constructor OneFormerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, repo_path: Optional[str]=…, class_info_file: Optional… - added:
field OneFormerImageProcessor.model_input_names - added:
method OneFormerImageProcessor.convert_segmentation_map_to_binary_masks(segmentation_map: np.ndarray, instance_id_to_semantic_id: Optional[dict[int, int]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…) - added:
method OneFormerImageProcessor.encode_inputs(pixel_values_list: list[ImageInput], task_inputs: list[str], segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, input_data_format: Optional[Union[str, ChannelDimension]… - added:
method OneFormerImageProcessor.get_instance_annotations(label, num_class_obj) - added:
method OneFormerImageProcessor.get_panoptic_annotations(label, num_class_obj) - added:
method OneFormerImageProcessor.get_semantic_annotations(label, num_class_obj) - added:
method OneFormerImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method OneFormerImageProcessor.post_process_instance_segmentation(outputs, task_type: str=…, is_demo: bool=…, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) - added:
method OneFormerImageProcessor.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method OneFormerImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method OneFormerImageProcessor.preprocess(images: ImageInput, task_inputs: Optional[list[str]]=…, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[dict[int, int]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…,… - added:
method OneFormerImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method OneFormerImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method OneFormerImageProcessor.to_dict() -> dict[str, Any]
- added:
class OneFormerImageProcessorFast(BaseImageProcessorFast) - added:
constructor OneFormerImageProcessorFast(**kwargs: Unpack[OneFormerFastImageProcessorKwargs]) - added:
field OneFormerImageProcessorFast.class_info_file - added:
field OneFormerImageProcessorFast.crop_size - added:
field OneFormerImageProcessorFast.default_to_square - added:
field OneFormerImageProcessorFast.do_center_crop - added:
field OneFormerImageProcessorFast.do_convert_rgb - added:
field OneFormerImageProcessorFast.do_normalize - added:
field OneFormerImageProcessorFast.do_reduce_labels - added:
field OneFormerImageProcessorFast.do_rescale - added:
field OneFormerImageProcessorFast.do_resize - added:
field OneFormerImageProcessorFast.ignore_index - added:
field OneFormerImageProcessorFast.image_mean - added:
field OneFormerImageProcessorFast.image_std - added:
field OneFormerImageProcessorFast.model_input_names - added:
field OneFormerImageProcessorFast.num_labels - added:
field OneFormerImageProcessorFast.num_text - added:
field OneFormerImageProcessorFast.repo_path - added:
field OneFormerImageProcessorFast.resample - added:
field OneFormerImageProcessorFast.rescale_factor - added:
field OneFormerImageProcessorFast.size - added:
field OneFormerImageProcessorFast.valid_kwargs - added:
method OneFormerImageProcessorFast.convert_segmentation_map_to_binary_masks(segmentation_map: torch.Tensor, instance_id_to_semantic_id: Optional[dict[int, int]]=…, ignore_index: Optional[int]=…, do_reduce_labels: bool=…) - added:
method OneFormerImageProcessorFast.get_instance_annotations(label, num_class_obj) - added:
method OneFormerImageProcessorFast.get_panoptic_annotations(label, num_class_obj) - added:
method OneFormerImageProcessorFast.get_semantic_annotations(label, num_class_obj) - added:
method OneFormerImageProcessorFast.pad(images: list[torch.Tensor], return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…) -> BatchFeature - added:
method OneFormerImageProcessorFast.post_process_instance_segmentation(outputs, task_type: str=…, is_demo: bool=…, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, target_sizes: Optional[list[tuple[int, int]]]=…, return_coco_annotation: Optional[bool]=…) - added:
method OneFormerImageProcessorFast.post_process_panoptic_segmentation(outputs, threshold: float=…, mask_threshold: float=…, overlap_mask_area_threshold: float=…, label_ids_to_fuse: Optional[set[int]]=…, target_sizes: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method OneFormerImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…) -> torch.Tensor - added:
method OneFormerImageProcessorFast.preprocess(images: ImageInput, task_inputs: Optional[list[str]]=…, segmentation_maps: Optional[ImageInput]=…, instance_id_to_semantic_id: Optional[Union[list[dict[int, int]], dict[int, int]]]=…, **kwargs: Unpack[OneFormerFastImageProcessorKwargs]) -> BatchFeature
- added:
class OneFormerForUniversalSegmentation(OneFormerPreTrainedModel) - added:
class OneFormerModel(OneFormerPreTrainedModel) - added:
class OneFormerPreTrainedModel(PreTrainedModel) - added:
constructor OneFormerForUniversalSegmentation(config: OneFormerConfig) - added:
constructor OneFormerModel(config: OneFormerConfig) - added:
field OneFormerForUniversalSegmentation.main_input_name - added:
field OneFormerModel.main_input_name - added:
field OneFormerPreTrainedModel.base_model_prefix - added:
field OneFormerPreTrainedModel.config: OneFormerConfig - added:
field OneFormerPreTrainedModel.main_input_name - added:
method OneFormerForUniversalSegmentation.forward(pixel_values: Tensor, task_inputs: Tensor, text_inputs: Optional[Tensor]=…, mask_labels: Optional[list[Tensor]]=…, class_labels: Optional[list[Tensor]]=…, pixel_mask: Optional[Tensor]=…, output_auxiliary_logits: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> OneFormer… - added:
method OneFormerForUniversalSegmentation.get_loss(loss_dict: dict[str, Tensor]) -> Tensor - added:
method OneFormerForUniversalSegmentation.get_loss_dict(masks_queries_logits: Tensor, class_queries_logits: Tensor, contrastive_queries_logits: Tensor, mask_labels: Tensor, class_labels: Tensor, text_queries: Tensor, auxiliary_predictions: dict[str, Tensor], calculate_contrastive_loss: bool) -> dict[str, Tensor] - added:
method OneFormerModel.forward(pixel_values: Tensor, task_inputs: Tensor, text_inputs: Optional[Tensor]=…, pixel_mask: Optional[Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> OneFormerModelOutput
- added:
class OneFormerProcessor(ProcessorMixin) - added:
constructor OneFormerProcessor(image_processor=…, tokenizer=…, max_seq_length: int=…, task_seq_length: int=…, **kwargs) - added:
field OneFormerProcessor.attributes - added:
field OneFormerProcessor.image_processor_class - added:
field OneFormerProcessor.tokenizer_class - added:
method OneFormerProcessor.encode_inputs(images=…, task_inputs=…, segmentation_maps=…, **kwargs) - added:
method OneFormerProcessor.post_process_instance_segmentation(*args, **kwargs) - added:
method OneFormerProcessor.post_process_panoptic_segmentation(*args, **kwargs) - added:
method OneFormerProcessor.post_process_semantic_segmentation(*args, **kwargs)
- added:
export define_import_structure
- added:
class OpenAIGPTConfig(PretrainedConfig) - added:
constructor OpenAIGPTConfig(vocab_size=…, n_positions=…, n_embd=…, n_layer=…, n_head=…, afn=…, resid_pdrop=…, embd_pdrop=…, attn_pdrop=…, layer_norm_epsilon=…, initializer_range=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_proj_to_labels=…, summary_first_dropout=…, **kwargs) - added:
field OpenAIGPTConfig.attribute_map - added:
field OpenAIGPTConfig.model_type
- added:
class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel) - added:
class OpenAIGPTForSequenceClassification(OpenAIGPTPreTrainedModel) - added:
class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel, GenerationMixin) - added:
class OpenAIGPTModel(OpenAIGPTPreTrainedModel) - added:
class OpenAIGPTPreTrainedModel(PreTrainedModel) - added:
constructor OpenAIGPTDoubleHeadsModel(config) - added:
constructor OpenAIGPTForSequenceClassification(config) - added:
constructor OpenAIGPTLMHeadModel(config) - added:
constructor OpenAIGPTModel(config) - added:
field OpenAIGPTPreTrainedModel.base_model_prefix - added:
field OpenAIGPTPreTrainedModel.config: OpenAIGPTConfig - added:
field OpenAIGPTPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_openai_gpt(model, config, openai_checkpoint_folder_path) - added:
method OpenAIGPTDoubleHeadsModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, mc_token_ids: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, mc_labels:… - added:
method OpenAIGPTForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method OpenAIGPTLMHeadModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states:… - added:
method OpenAIGPTLMHeadModel.prepare_inputs_for_generation(input_ids: torch.LongTensor, **kwargs) -> dict[str, Any] - added:
method OpenAIGPTModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool… - added:
method OpenAIGPTModel.get_input_embeddings() - added:
method OpenAIGPTModel.set_input_embeddings(new_embeddings)
- added:
class TFOpenAIGPTDoubleHeadsModel(TFOpenAIGPTPreTrainedModel) - added:
class TFOpenAIGPTForSequenceClassification(TFOpenAIGPTPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFOpenAIGPTMainLayer(keras.layers.Layer) - added:
class TFOpenAIGPTModel(TFOpenAIGPTPreTrainedModel) - added:
class TFOpenAIGPTPreTrainedModel(TFPreTrainedModel) - added:
constructor TFOpenAIGPTDoubleHeadsModel(config, *inputs, **kwargs) - added:
constructor TFOpenAIGPTForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFOpenAIGPTLMHeadModel(config, *inputs, **kwargs) - added:
constructor TFOpenAIGPTMainLayer(config, *inputs, **kwargs) - added:
constructor TFOpenAIGPTModel(config, *inputs, **kwargs) - added:
field TFOpenAIGPTMainLayer.config_class - added:
field TFOpenAIGPTPreTrainedModel.base_model_prefix - added:
field TFOpenAIGPTPreTrainedModel.config_class - added:
method TFOpenAIGPTDoubleHeadsModel.build(input_shape=…) - added:
method TFOpenAIGPTDoubleHeadsModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, mc_token_ids: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, out… - added:
method TFOpenAIGPTForSequenceClassification.build(input_shape=…) - added:
method TFOpenAIGPTForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, retu… - added:
method TFOpenAIGPTLMHeadModel.build(input_shape=…) - added:
method TFOpenAIGPTLMHeadModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFOpenAIGPTLMHeadModel.get_output_embeddings() - added:
method TFOpenAIGPTLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - added:
method TFOpenAIGPTLMHeadModel.set_output_embeddings(value) - added:
method TFOpenAIGPTMainLayer.build(input_shape=…) - added:
method TFOpenAIGPTMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool |… - added:
method TFOpenAIGPTMainLayer.get_input_embeddings() - added:
method TFOpenAIGPTMainLayer.set_input_embeddings(value) - added:
method TFOpenAIGPTModel.build(input_shape=…) - added:
method TFOpenAIGPTModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None… - added:
property TFOpenAIGPTDoubleHeadsModel.input_signature
- added:
class OpenAIGPTTokenizer(PreTrainedTokenizer) - added:
constructor OpenAIGPTTokenizer(vocab_file, merges_file, unk_token=…, **kwargs) - added:
field OpenAIGPTTokenizer.model_input_names - added:
field OpenAIGPTTokenizer.vocab_files_names - added:
method OpenAIGPTTokenizer.bpe(token) - added:
method OpenAIGPTTokenizer.convert_tokens_to_string(tokens) - added:
method OpenAIGPTTokenizer.get_vocab() - added:
method OpenAIGPTTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property OpenAIGPTTokenizer.do_lower_case - added:
property OpenAIGPTTokenizer.vocab_size
- added:
class OpenAIGPTTokenizerFast(PreTrainedTokenizerFast) - added:
constructor OpenAIGPTTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, **kwargs) - added:
field OpenAIGPTTokenizerFast.model_input_names - added:
field OpenAIGPTTokenizerFast.slow_tokenizer_class - added:
field OpenAIGPTTokenizerFast.vocab_files_names - added:
method OpenAIGPTTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property OpenAIGPTTokenizerFast.do_lower_case
- added:
export define_import_structure
- added:
class OPTConfig(PretrainedConfig) - added:
constructor OPTConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, ffn_dim=…, max_position_embeddings=…, do_layer_norm_before=…, _remove_final_layer_norm=…, word_embed_proj_dim=…, dropout=…, attention_dropout=…, num_attention_heads=…, activation_function=…, layerdrop=…, init_std=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, enable_bias=…, layer_norm_elementwise_affine=…, **… - added:
field OPTConfig.keys_to_ignore_at_inference - added:
field OPTConfig.model_type
- added:
class FlaxOPTForCausalLM(FlaxOPTPreTrainedModel) - added:
class FlaxOPTModel(FlaxOPTPreTrainedModel) - added:
class FlaxOPTPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxOPTPreTrainedModel(config: OPTConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxOPTForCausalLM.module_class - added:
field FlaxOPTModel.config: OPTConfig - added:
field FlaxOPTModel.dtype: jnp.dtype - added:
field FlaxOPTModel.module_class - added:
field FlaxOPTPreTrainedModel.base_model_prefix: str - added:
field FlaxOPTPreTrainedModel.config_class - added:
field FlaxOPTPreTrainedModel.module_class: nn.Module - added:
method FlaxOPTForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxOPTForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxOPTPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxOPTPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class OPTForCausalLM(OPTPreTrainedModel, GenerationMixin) - added:
class OPTForQuestionAnswering(OPTPreTrainedModel) - added:
class OPTForSequenceClassification(OPTPreTrainedModel) - added:
class OPTModel(OPTPreTrainedModel) - added:
class OPTPreTrainedModel(PreTrainedModel) - added:
constructor OPTForCausalLM(config) - added:
constructor OPTForQuestionAnswering(config: OPTConfig) - added:
constructor OPTForSequenceClassification(config: OPTConfig) - added:
constructor OPTModel(config: OPTConfig) - added:
field OPTPreTrainedModel.base_model_prefix - added:
field OPTPreTrainedModel.config: OPTConfig - added:
field OPTPreTrainedModel.supports_gradient_checkpointing - added:
method OPTForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[b… - added:
method OPTForCausalLM.get_decoder() - added:
method OPTForCausalLM.get_input_embeddings() - added:
method OPTForCausalLM.set_decoder(decoder) - added:
method OPTForCausalLM.set_input_embeddings(value) - added:
method OPTForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional… - added:
method OPTForQuestionAnswering.get_input_embeddings() - added:
method OPTForQuestionAnswering.set_input_embeddings(value) - added:
method OPTForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…,… - added:
method OPTForSequenceClassification.get_input_embeddings() - added:
method OPTForSequenceClassification.set_input_embeddings(value) - added:
method OPTModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, position_ids: Optional[torch.LongTen… - added:
method OPTModel.get_input_embeddings() - added:
method OPTModel.set_input_embeddings(value)
- added:
class TFOPTForCausalLM(TFOPTPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFOPTModel(TFOPTPreTrainedModel) - added:
class TFOPTPreTrainedModel(TFPreTrainedModel) - added:
constructor TFOPTForCausalLM(config: OPTConfig, **kwargs) - added:
constructor TFOPTModel(config: OPTConfig, **kwargs) - added:
field TFOPTForCausalLM.config_class - added:
field TFOPTModel.config_class - added:
field TFOPTPreTrainedModel.base_model_prefix - added:
field TFOPTPreTrainedModel.config_class - added:
method TFOPTForCausalLM.build(input_shape=…) - added:
method TFOPTForCausalLM.call(input_ids: TFModelInputType | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor | None=…, use_cache: bool | None=…, output_attent… - added:
method TFOPTForCausalLM.get_output_embeddings() - added:
method TFOPTForCausalLM.prepare_inputs_for_generation(inputs, past_key_values=…, use_cache=…, **kwargs) - added:
method TFOPTForCausalLM.serving_output(output) - added:
method TFOPTModel.build(input_shape=…) - added:
method TFOPTModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, use_cache: bool | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training… - added:
method TFOPTModel.get_input_embeddings() - added:
method TFOPTModel.serving_output(output) - added:
method TFOPTModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class Ovis2Config(PretrainedConfig) - added:
class Ovis2VisionConfig(PretrainedConfig) - added:
constructor Ovis2Config(vision_config=…, text_config=…, image_token_id=…, visual_indicator_token_ids=…, vocab_size=…, hidden_size=…, **kwargs) - added:
constructor Ovis2VisionConfig(hidden_size: int=…, intermediate_size: int=…, num_hidden_layers: int=…, num_attention_heads: int=…, num_channels: int=…, image_size: int=…, patch_size: int=…, rms_norm_eps: float=…, attention_dropout: float=…, qkv_bias: bool=…, mlp_bias: bool=…, hidden_act=…, vocab_size=…, hidden_stride=…, num_visual_indicator_tokens=…, initializer_range=…, tokenize_function=…, **kwa… - added:
field Ovis2Config.model_type - added:
field Ovis2Config.sub_configs - added:
field Ovis2VisionConfig.base_config_key
- added:
class Ovis2ImageProcessor(BaseImageProcessor) - added:
constructor Ovis2ImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, crop_to_patches: bool=…, min_patches: int=…, max_patches: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, use_covering_… - added:
field Ovis2ImageProcessor.model_input_names - added:
method Ovis2ImageProcessor.crop_image_to_patches(images: np.ndarray, min_patches: int, max_patches: int, use_covering_area_grid: bool=…, patch_size: Optional[Union[tuple, int, dict]]=…, data_format: Optional[ChannelDimension]=…, covering_threshold: float=…) - added:
method Ovis2ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, crop_to_patches: Optional[bool]=…, min_patches: Optional[int]=…, max_patches: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float… - added:
method Ovis2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class Ovis2ImageProcessorFast(BaseImageProcessorFast) - added:
field Ovis2ImageProcessorFast.crop_to_patches - added:
field Ovis2ImageProcessorFast.default_to_square - added:
field Ovis2ImageProcessorFast.do_convert_rgb - added:
field Ovis2ImageProcessorFast.do_normalize - added:
field Ovis2ImageProcessorFast.do_rescale - added:
field Ovis2ImageProcessorFast.do_resize - added:
field Ovis2ImageProcessorFast.image_mean - added:
field Ovis2ImageProcessorFast.image_std - added:
field Ovis2ImageProcessorFast.max_patches - added:
field Ovis2ImageProcessorFast.min_patches - added:
field Ovis2ImageProcessorFast.resample - added:
field Ovis2ImageProcessorFast.size - added:
field Ovis2ImageProcessorFast.use_covering_area_grid - added:
field Ovis2ImageProcessorFast.valid_kwargs - added:
method Ovis2ImageProcessorFast.crop_image_to_patches(images: torch.Tensor, min_patches: int, max_patches: int, use_covering_area_grid: bool=…, covering_threshold: float=…, patch_size: Optional[Union[tuple, int, dict]]=…, interpolation: Optional[F.InterpolationMode]=…) - added:
method Ovis2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Ovis2ImageProcessorKwargs]) -> BatchFeature
- added:
class Ovis2ForConditionalGeneration(Ovis2PreTrainedModel, GenerationMixin) - added:
class Ovis2Model(Ovis2PreTrainedModel) - added:
class Ovis2PreTrainedModel(PreTrainedModel) - added:
constructor Ovis2ForConditionalGeneration(config: Ovis2Config) - added:
constructor Ovis2Model(config: Ovis2Config) - added:
field Ovis2PreTrainedModel.base_model_prefix - added:
field Ovis2PreTrainedModel.config: Ovis2Config - added:
field Ovis2PreTrainedModel.supports_gradient_checkpointing - added:
method Ovis2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]… - added:
method Ovis2ForConditionalGeneration.get_decoder() - added:
method Ovis2ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor) - added:
method Ovis2ForConditionalGeneration.get_input_embeddings() - added:
method Ovis2ForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Ovis2ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method Ovis2ForConditionalGeneration.set_decoder(decoder) - added:
method Ovis2ForConditionalGeneration.set_input_embeddings(value) - added:
method Ovis2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_s… - added:
method Ovis2Model.get_decoder() - added:
method Ovis2Model.get_image_features(pixel_values: torch.FloatTensor) -> torch.FloatTensor - added:
method Ovis2Model.get_input_embeddings() - added:
method Ovis2Model.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method Ovis2Model.set_decoder(decoder) - added:
method Ovis2Model.set_input_embeddings(value) - added:
property Ovis2ForConditionalGeneration.language_model - added:
property Ovis2ForConditionalGeneration.multi_modal_projector - added:
property Ovis2ForConditionalGeneration.vision_tower
- added:
class Ovis2ForConditionalGeneration(LlavaForConditionalGeneration, GenerationMixin) - added:
class Ovis2Model(LlavaModel) - added:
class Ovis2PreTrainedModel(PreTrainedModel) - added:
constructor Ovis2ForConditionalGeneration(config: Ovis2Config) - added:
constructor Ovis2Model(config: Ovis2Config) - added:
field Ovis2PreTrainedModel.base_model_prefix - added:
field Ovis2PreTrainedModel.config: Ovis2Config - added:
field Ovis2PreTrainedModel.supports_gradient_checkpointing - added:
method Ovis2ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]… - added:
method Ovis2ForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor) - added:
method Ovis2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_s… - added:
method Ovis2Model.get_image_features(pixel_values: torch.FloatTensor) -> torch.FloatTensor - added:
property Ovis2ForConditionalGeneration.multi_modal_projector
- added:
class Ovis2Processor(ProcessorMixin) - added:
constructor Ovis2Processor(image_processor=…, tokenizer=…, chat_template=…, image_token=…, image_seq_length=…, **kwargs) - added:
field Ovis2Processor.attributes - added:
field Ovis2Processor.image_processor_class - added:
field Ovis2Processor.tokenizer_class - added:
method Ovis2Processor.batch_decode(*args, **kwargs) - added:
method Ovis2Processor.decode(*args, **kwargs) - added:
property Ovis2Processor.model_input_names
- added:
export define_import_structure
- added:
class Owlv2Config(PretrainedConfig) - added:
class Owlv2TextConfig(PretrainedConfig) - added:
class Owlv2VisionConfig(PretrainedConfig) - added:
constructor Owlv2Config(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, return_dict=…, **kwargs) - added:
constructor Owlv2TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor Owlv2VisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field Owlv2Config.model_type - added:
field Owlv2Config.sub_configs - added:
field Owlv2TextConfig.base_config_key - added:
field Owlv2TextConfig.model_type - added:
field Owlv2VisionConfig.base_config_key - added:
field Owlv2VisionConfig.model_type - added:
static method Owlv2Config.from_text_vision_configs(text_config: dict, vision_config: dict, **kwargs)
- added:
class Owlv2ImageProcessor(BaseImageProcessor) - added:
constructor Owlv2ImageProcessor(do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_pad: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field Owlv2ImageProcessor.model_input_names - added:
method Owlv2ImageProcessor.pad(image: np.ndarray, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method Owlv2ImageProcessor.post_process_image_guided_detection(outputs, threshold=…, nms_threshold=…, target_sizes=…) - added:
method Owlv2ImageProcessor.post_process_object_detection(outputs: Owlv2ObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method Owlv2ImageProcessor.preprocess(images: ImageInput, do_pad: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, TensorType]]=…, da… - added:
method Owlv2ImageProcessor.resize(image: np.ndarray, size: dict[str, int], anti_aliasing: bool=…, anti_aliasing_sigma=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class Owlv2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Owlv2ImageProcessorFast(**kwargs: Unpack[Owlv2FastImageProcessorKwargs]) - added:
field Owlv2ImageProcessorFast.crop_size - added:
field Owlv2ImageProcessorFast.default_to_square - added:
field Owlv2ImageProcessorFast.do_center_crop - added:
field Owlv2ImageProcessorFast.do_convert_rgb - added:
field Owlv2ImageProcessorFast.do_normalize - added:
field Owlv2ImageProcessorFast.do_pad - added:
field Owlv2ImageProcessorFast.do_rescale - added:
field Owlv2ImageProcessorFast.do_resize - added:
field Owlv2ImageProcessorFast.image_mean - added:
field Owlv2ImageProcessorFast.image_std - added:
field Owlv2ImageProcessorFast.model_input_names - added:
field Owlv2ImageProcessorFast.resample - added:
field Owlv2ImageProcessorFast.rescale_factor - added:
field Owlv2ImageProcessorFast.size - added:
field Owlv2ImageProcessorFast.valid_kwargs - added:
method Owlv2ImageProcessorFast.pad(images: list[torch.Tensor], disable_grouping: Optional[bool], constant_value: float=…, **kwargs) -> list[torch.Tensor] - added:
method Owlv2ImageProcessorFast.post_process(outputs, target_sizes) - added:
method Owlv2ImageProcessorFast.post_process_image_guided_detection(outputs, threshold=…, nms_threshold=…, target_sizes=…) - added:
method Owlv2ImageProcessorFast.post_process_object_detection(outputs: Owlv2ObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method Owlv2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Owlv2FastImageProcessorKwargs]) - added:
method Owlv2ImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, anti_aliasing: bool=…, anti_aliasing_sigma=…, **kwargs) -> torch.Tensor
- added:
class Owlv2ForObjectDetection(Owlv2PreTrainedModel) - added:
class Owlv2Model(Owlv2PreTrainedModel) - added:
class Owlv2PreTrainedModel(PreTrainedModel) - added:
class Owlv2TextModel(Owlv2PreTrainedModel) - added:
class Owlv2VisionModel(Owlv2PreTrainedModel) - added:
constructor Owlv2ForObjectDetection(config: Owlv2Config) - added:
constructor Owlv2Model(config: Owlv2Config) - added:
constructor Owlv2TextModel(config: Owlv2TextConfig) - added:
constructor Owlv2VisionModel(config: Owlv2VisionConfig) - added:
field Owlv2ForObjectDetection.config: Owlv2Config - added:
field Owlv2Model.config: Owlv2Config - added:
field Owlv2PreTrainedModel.base_model_prefix - added:
field Owlv2PreTrainedModel.config: Owlv2Config - added:
field Owlv2PreTrainedModel.supports_gradient_checkpointing - added:
field Owlv2TextModel.config: Owlv2TextConfig - added:
field Owlv2VisionModel.config: Owlv2VisionConfig - added:
field Owlv2VisionModel.main_input_name - added:
method Owlv2ForObjectDetection.box_predictor(image_feats: torch.FloatTensor, feature_map: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method Owlv2ForObjectDetection.class_predictor(image_feats: torch.FloatTensor, query_embeds: Optional[torch.FloatTensor]=…, query_mask: Optional[torch.Tensor]=…) -> tuple[torch.FloatTensor] - added:
method Owlv2ForObjectDetection.compute_box_bias(num_patches_height: int, num_patches_width: int, feature_map: Optional[torch.FloatTensor]=…) -> torch.Tensor - added:
method Owlv2ForObjectDetection.embed_image_query(query_image_features: torch.FloatTensor, query_feature_map: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method Owlv2ForObjectDetection.forward(input_ids: torch.Tensor, pixel_values: torch.FloatTensor, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Owlv2ObjectDetectionOutput - added:
method Owlv2ForObjectDetection.image_embedder(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> tuple[torch.FloatTensor] - added:
method Owlv2ForObjectDetection.image_guided_detection(pixel_values: torch.FloatTensor, query_pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Owlv2ImageGuidedObjectDetectionOutput - added:
method Owlv2ForObjectDetection.image_text_embedder(input_ids: torch.Tensor, pixel_values: torch.FloatTensor, attention_mask: torch.Tensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> tuple[torch.FloatTensor] - added:
method Owlv2ForObjectDetection.objectness_predictor(image_features: torch.FloatTensor) -> torch.FloatTensor - added:
method Owlv2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_base_image_embeds: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Owlv2Output] - added:
method Owlv2Model.get_image_features(pixel_values: torch.Tensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method Owlv2Model.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method Owlv2TextModel.forward(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method Owlv2TextModel.get_input_embeddings() -> nn.Module - added:
method Owlv2TextModel.set_input_embeddings(value) - added:
method Owlv2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method Owlv2VisionModel.get_input_embeddings() -> nn.Module - added:
static method Owlv2ForObjectDetection.normalize_grid_corner_coordinates(num_patches_height: int, num_patches_width: int) -> torch.Tensor
- added:
class Owlv2ImageProcessorFast(OwlViTImageProcessorFast) - added:
constructor Owlv2ImageProcessorFast(**kwargs: Unpack[Owlv2FastImageProcessorKwargs]) - added:
field Owlv2ImageProcessorFast.crop_size - added:
field Owlv2ImageProcessorFast.do_center_crop - added:
field Owlv2ImageProcessorFast.do_normalize - added:
field Owlv2ImageProcessorFast.do_pad - added:
field Owlv2ImageProcessorFast.do_rescale - added:
field Owlv2ImageProcessorFast.do_resize - added:
field Owlv2ImageProcessorFast.image_mean - added:
field Owlv2ImageProcessorFast.image_std - added:
field Owlv2ImageProcessorFast.resample - added:
field Owlv2ImageProcessorFast.rescale_factor - added:
field Owlv2ImageProcessorFast.size - added:
field Owlv2ImageProcessorFast.valid_kwargs - added:
method Owlv2ImageProcessorFast.pad(images: list[torch.Tensor], disable_grouping: Optional[bool], constant_value: float=…, **kwargs) -> list[torch.Tensor] - added:
method Owlv2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Owlv2FastImageProcessorKwargs]) - added:
method Owlv2ImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, anti_aliasing: bool=…, anti_aliasing_sigma=…, **kwargs) -> torch.Tensor
- added:
class Owlv2Processor(ProcessorMixin) - added:
constructor Owlv2Processor(image_processor, tokenizer, **kwargs) - added:
field Owlv2Processor.attributes - added:
field Owlv2Processor.image_processor_class - added:
field Owlv2Processor.tokenizer_class - added:
method Owlv2Processor.post_process_grounded_object_detection(outputs: Owlv2ObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…, text_labels: Optional[list[list[str]]]=…) - added:
method Owlv2Processor.post_process_image_guided_detection(outputs: Owlv2ImageGuidedObjectDetectionOutput, threshold: float=…, nms_threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method Owlv2Processor.post_process_object_detection(*args, **kwargs)
- added:
export define_import_structure
- added:
class OwlViTConfig(PretrainedConfig) - added:
class OwlViTOnnxConfig(OnnxConfig) - added:
class OwlViTTextConfig(PretrainedConfig) - added:
class OwlViTVisionConfig(PretrainedConfig) - added:
constructor OwlViTConfig(text_config=…, vision_config=…, projection_dim=…, logit_scale_init_value=…, return_dict=…, **kwargs) - added:
constructor OwlViTTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor OwlViTVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, **kwargs) - added:
field OwlViTConfig.model_type - added:
field OwlViTConfig.sub_configs - added:
field OwlViTTextConfig.base_config_key - added:
field OwlViTTextConfig.model_type - added:
field OwlViTVisionConfig.base_config_key - added:
field OwlViTVisionConfig.model_type - added:
method OwlViTOnnxConfig.generate_dummy_inputs(processor: ProcessorMixin, batch_size: int=…, seq_length: int=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property OwlViTOnnxConfig.atol_for_validation -> float - added:
property OwlViTOnnxConfig.default_onnx_opset -> int - added:
property OwlViTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property OwlViTOnnxConfig.outputs -> Mapping[str, Mapping[int, str]] - added:
static method OwlViTConfig.from_text_vision_configs(text_config: dict, vision_config: dict, **kwargs)
- added:
class OwlViTFeatureExtractor(OwlViTImageProcessor) - added:
constructor OwlViTFeatureExtractor(*args, **kwargs) -> None
- added:
class OwlViTImageProcessor(BaseImageProcessor) - added:
constructor OwlViTImageProcessor(do_resize=…, size=…, resample=…, do_center_crop=…, crop_size=…, do_rescale=…, rescale_factor=…, do_normalize=…, image_mean=…, image_std=…, **kwargs) - added:
field OwlViTImageProcessor.model_input_names - added:
method OwlViTImageProcessor.center_crop(image: np.ndarray, crop_size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method OwlViTImageProcessor.post_process(outputs, target_sizes) - added:
method OwlViTImageProcessor.post_process_image_guided_detection(outputs, threshold=…, nms_threshold=…, target_sizes=…) - added:
method OwlViTImageProcessor.post_process_object_detection(outputs: OwlViTObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method OwlViTImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Opt… - added:
method OwlViTImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method OwlViTImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling.BICUBIC, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class OwlViTImageProcessorFast(BaseImageProcessorFast) - added:
field OwlViTImageProcessorFast.crop_size - added:
field OwlViTImageProcessorFast.default_to_square - added:
field OwlViTImageProcessorFast.do_center_crop - added:
field OwlViTImageProcessorFast.do_convert_rgb - added:
field OwlViTImageProcessorFast.do_normalize - added:
field OwlViTImageProcessorFast.do_rescale - added:
field OwlViTImageProcessorFast.do_resize - added:
field OwlViTImageProcessorFast.image_mean - added:
field OwlViTImageProcessorFast.image_std - added:
field OwlViTImageProcessorFast.model_input_names - added:
field OwlViTImageProcessorFast.resample - added:
field OwlViTImageProcessorFast.size - added:
method OwlViTImageProcessorFast.post_process(outputs, target_sizes) - added:
method OwlViTImageProcessorFast.post_process_image_guided_detection(outputs, threshold=…, nms_threshold=…, target_sizes=…) - added:
method OwlViTImageProcessorFast.post_process_object_detection(outputs: OwlViTObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…)
- added:
class OwlViTForObjectDetection(OwlViTPreTrainedModel) - added:
class OwlViTModel(OwlViTPreTrainedModel) - added:
class OwlViTPreTrainedModel(PreTrainedModel) - added:
class OwlViTTextModel(OwlViTPreTrainedModel) - added:
class OwlViTVisionModel(OwlViTPreTrainedModel) - added:
constructor OwlViTForObjectDetection(config: OwlViTConfig) - added:
constructor OwlViTModel(config: OwlViTConfig) - added:
constructor OwlViTTextModel(config: OwlViTTextConfig) - added:
constructor OwlViTVisionModel(config: OwlViTVisionConfig) - added:
field OwlViTForObjectDetection.config: OwlViTConfig - added:
field OwlViTModel.config: OwlViTConfig - added:
field OwlViTPreTrainedModel.base_model_prefix - added:
field OwlViTPreTrainedModel.config: OwlViTConfig - added:
field OwlViTPreTrainedModel.supports_gradient_checkpointing - added:
field OwlViTTextModel.config: OwlViTTextConfig - added:
field OwlViTVisionModel.config: OwlViTVisionConfig - added:
field OwlViTVisionModel.main_input_name - added:
method OwlViTForObjectDetection.box_predictor(image_feats: torch.FloatTensor, feature_map: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method OwlViTForObjectDetection.class_predictor(image_feats: torch.FloatTensor, query_embeds: Optional[torch.FloatTensor]=…, query_mask: Optional[torch.Tensor]=…) -> tuple[torch.FloatTensor] - added:
method OwlViTForObjectDetection.compute_box_bias(num_patches_height: int, num_patches_width: int, feature_map: Optional[torch.FloatTensor]=…) -> torch.Tensor - added:
method OwlViTForObjectDetection.embed_image_query(query_image_features: torch.FloatTensor, query_feature_map: torch.FloatTensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method OwlViTForObjectDetection.forward(input_ids: torch.Tensor, pixel_values: torch.FloatTensor, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> OwlViTObjectDetectionOutput - added:
method OwlViTForObjectDetection.image_embedder(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> tuple[torch.FloatTensor] - added:
method OwlViTForObjectDetection.image_guided_detection(pixel_values: torch.FloatTensor, query_pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> OwlViTImageGuidedObjectDetectionOutput - added:
method OwlViTForObjectDetection.image_text_embedder(input_ids: torch.Tensor, pixel_values: torch.FloatTensor, attention_mask: torch.Tensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…) -> tuple[torch.FloatTensor] - added:
method OwlViTModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_base_image_embeds: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, OwlViTOutpu… - added:
method OwlViTModel.get_image_features(pixel_values: torch.Tensor, interpolate_pos_encoding: bool=…) -> torch.FloatTensor - added:
method OwlViTModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method OwlViTTextModel.forward(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method OwlViTTextModel.get_input_embeddings() -> nn.Module - added:
method OwlViTTextModel.set_input_embeddings(value) - added:
method OwlViTVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method OwlViTVisionModel.get_input_embeddings() -> nn.Module - added:
static method OwlViTForObjectDetection.normalize_grid_corner_coordinates(num_patches_height: int, num_patches_width: int) -> torch.Tensor
- added:
class OwlViTProcessor(ProcessorMixin) - added:
constructor OwlViTProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field OwlViTProcessor.attributes - added:
field OwlViTProcessor.image_processor_class - added:
field OwlViTProcessor.tokenizer_class - added:
method OwlViTProcessor.post_process(*args, **kwargs) - added:
method OwlViTProcessor.post_process_grounded_object_detection(outputs: OwlViTObjectDetectionOutput, threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…, text_labels: Optional[list[list[str]]]=…) - added:
method OwlViTProcessor.post_process_image_guided_detection(outputs: OwlViTImageGuidedObjectDetectionOutput, threshold: float=…, nms_threshold: float=…, target_sizes: Optional[Union[TensorType, list[tuple]]]=…) - added:
method OwlViTProcessor.post_process_object_detection(*args, **kwargs) - added:
property OwlViTProcessor.feature_extractor - added:
property OwlViTProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class PaliGemmaConfig(PretrainedConfig) - added:
constructor PaliGemmaConfig(vision_config=…, text_config=…, image_token_index=…, vocab_size=…, projection_dim=…, hidden_size=…, **kwargs) - added:
field PaliGemmaConfig.attribute_map - added:
field PaliGemmaConfig.keys_to_ignore_at_inference - added:
field PaliGemmaConfig.model_type - added:
field PaliGemmaConfig.sub_configs
- added:
class PaliGemmaForConditionalGeneration(PaliGemmaPreTrainedModel, GenerationMixin) - added:
class PaliGemmaModel(PaliGemmaPreTrainedModel) - added:
class PaliGemmaPreTrainedModel(PreTrainedModel) - added:
constructor PaliGemmaForConditionalGeneration(config: PaliGemmaConfig) - added:
constructor PaliGemmaModel(config: PaliGemmaConfig) - added:
field PaliGemmaModel.accepts_loss_kwargs - added:
field PaliGemmaPreTrainedModel.base_model_prefix - added:
field PaliGemmaPreTrainedModel.config: PaliGemmaConfig - added:
field PaliGemmaPreTrainedModel.supports_gradient_checkpointing - added:
method PaliGemmaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labe… - added:
method PaliGemmaForConditionalGeneration.get_decoder() - added:
method PaliGemmaForConditionalGeneration.get_image_features(pixel_values) - added:
method PaliGemmaForConditionalGeneration.get_input_embeddings() - added:
method PaliGemmaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, cache_position=…, position_ids=…, pixel_values=…, attention_mask=…, token_type_ids=…, use_cache=…, logits_to_keep=…, labels=…, **kwargs) - added:
method PaliGemmaForConditionalGeneration.set_decoder(decoder) - added:
method PaliGemmaForConditionalGeneration.set_input_embeddings(value) - added:
method PaliGemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.… - added:
method PaliGemmaModel.get_decoder() - added:
method PaliGemmaModel.get_image_features(pixel_values: torch.FloatTensor) - added:
method PaliGemmaModel.get_input_embeddings() - added:
method PaliGemmaModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method PaliGemmaModel.set_decoder(decoder) - added:
method PaliGemmaModel.set_input_embeddings(value) - added:
property PaliGemmaForConditionalGeneration.language_model - added:
property PaliGemmaForConditionalGeneration.multi_modal_projector - added:
property PaliGemmaForConditionalGeneration.vision_tower
- added:
class PaliGemmaProcessor(ProcessorMixin) - added:
constructor PaliGemmaProcessor(image_processor=…, tokenizer=…, chat_template=…, **kwargs) - added:
field PaliGemmaProcessor.attributes - added:
field PaliGemmaProcessor.image_processor_class - added:
field PaliGemmaProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class ParakeetCTCConfig(PretrainedConfig) - added:
class ParakeetEncoderConfig(PretrainedConfig) - added:
constructor ParakeetCTCConfig(vocab_size=…, ctc_loss_reduction=…, ctc_zero_infinity=…, encoder_config: Union[dict, ParakeetEncoderConfig]=…, pad_token_id=…, **kwargs) - added:
constructor ParakeetEncoderConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, attention_bias=…, conv_kernel_size=…, subsampling_factor=…, subsampling_conv_channels=…, num_mel_bins=…, subsampling_conv_kernel_size=…, subsampling_conv_stride=…, dropout=…, dropout_positions=…, layerdrop=…, activation_dropout=…, attention_dropout=…, max_position_embedd… - added:
field ParakeetCTCConfig.model_type - added:
field ParakeetCTCConfig.sub_configs - added:
field ParakeetEncoderConfig.keys_to_ignore_at_inference - added:
field ParakeetEncoderConfig.model_type - added:
static method ParakeetCTCConfig.from_encoder_config(encoder_config: ParakeetEncoderConfig, **kwargs)
- added:
class ParakeetFeatureExtractor(SequenceFeatureExtractor) - added:
constructor ParakeetFeatureExtractor(feature_size=…, sampling_rate=…, hop_length=…, n_fft=…, win_length=…, preemphasis=…, padding_value=…, **kwargs) - added:
field ParakeetFeatureExtractor.model_input_names
- added:
class ParakeetEncoder(ParakeetPreTrainedModel) - added:
class ParakeetForCTC(ParakeetPreTrainedModel) - added:
class ParakeetPreTrainedModel(PreTrainedModel) - added:
constructor ParakeetEncoder(config: ParakeetEncoderConfig) - added:
constructor ParakeetForCTC(config: ParakeetCTCConfig) - added:
field ParakeetEncoder.base_model_prefix - added:
field ParakeetEncoder.config: ParakeetEncoderConfig - added:
field ParakeetForCTC.config: ParakeetCTCConfig - added:
field ParakeetPreTrainedModel.base_model_prefix - added:
field ParakeetPreTrainedModel.config: ParakeetCTCConfig - added:
field ParakeetPreTrainedModel.main_input_name - added:
field ParakeetPreTrainedModel.supports_gradient_checkpointing - added:
method ParakeetEncoder.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method ParakeetForCTC.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> CausalLMOutput - added:
method ParakeetForCTC.generate(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, return_dict_in_generate: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[ParakeetGenerateOutput, torch.LongTensor]
- added:
class ParakeetEncoder(ParakeetPreTrainedModel) - added:
class ParakeetForCTC(ParakeetPreTrainedModel) - added:
class ParakeetPreTrainedModel(PreTrainedModel) - added:
constructor ParakeetEncoder(config: ParakeetEncoderConfig) - added:
constructor ParakeetForCTC(config: ParakeetCTCConfig) - added:
field ParakeetEncoder.base_model_prefix - added:
field ParakeetEncoder.config: ParakeetEncoderConfig - added:
field ParakeetForCTC.config: ParakeetCTCConfig - added:
field ParakeetPreTrainedModel.base_model_prefix - added:
field ParakeetPreTrainedModel.config: ParakeetCTCConfig - added:
field ParakeetPreTrainedModel.main_input_name - added:
field ParakeetPreTrainedModel.supports_gradient_checkpointing - added:
method ParakeetEncoder.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method ParakeetForCTC.forward(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> CausalLMOutput - added:
method ParakeetForCTC.generate(input_features: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, return_dict_in_generate: bool=…, **kwargs: Unpack[TransformersKwargs]) -> Union[ParakeetGenerateOutput, torch.LongTensor]
- added:
class ParakeetProcessor(ProcessorMixin) - added:
field ParakeetProcessor.attributes - added:
field ParakeetProcessor.feature_extractor_class - added:
field ParakeetProcessor.tokenizer_class - added:
property ParakeetProcessor.model_input_names
- added:
class ParakeetTokenizerFast(PreTrainedTokenizerFast)
- added:
export define_import_structure
- added:
class PatchTSMixerConfig(PretrainedConfig) - added:
constructor PatchTSMixerConfig(context_length: int=…, patch_length: int=…, num_input_channels: int=…, patch_stride: int=…, num_parallel_samples: int=…, d_model: int=…, expansion_factor: int=…, num_layers: int=…, dropout: float=…, mode: str=…, gated_attn: bool=…, norm_mlp: str=…, self_attn: bool=…, self_attn_heads: int=…, use_positional_encoding: bool=…, positional_encoding_type: str=…, scaling: O… - added:
field PatchTSMixerConfig.attribute_map - added:
field PatchTSMixerConfig.model_type
- added:
class PatchTSMixerForPrediction(PatchTSMixerPreTrainedModel) - added:
class PatchTSMixerForPretraining(PatchTSMixerPreTrainedModel) - added:
class PatchTSMixerForRegression(PatchTSMixerPreTrainedModel) - added:
class PatchTSMixerForTimeSeriesClassification(PatchTSMixerPreTrainedModel) - added:
class PatchTSMixerModel(PatchTSMixerPreTrainedModel) - added:
class PatchTSMixerPreTrainedModel(PreTrainedModel) - added:
constructor PatchTSMixerForPrediction(config: PatchTSMixerConfig) - added:
constructor PatchTSMixerForPretraining(config: PatchTSMixerConfig) - added:
constructor PatchTSMixerForRegression(config: PatchTSMixerConfig) - added:
constructor PatchTSMixerForTimeSeriesClassification(config: PatchTSMixerConfig) - added:
constructor PatchTSMixerModel(config: PatchTSMixerConfig, mask_input: bool=…) - added:
field PatchTSMixerPreTrainedModel.base_model_prefix - added:
field PatchTSMixerPreTrainedModel.config: PatchTSMixerConfig - added:
field PatchTSMixerPreTrainedModel.main_input_name - added:
field PatchTSMixerPreTrainedModel.supports_gradient_checkpointing - added:
method PatchTSMixerForPrediction.forward(past_values: torch.Tensor, observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_loss: bool=…, return_dict: Optional[bool]=…) -> PatchTSMixerForPredictionOutput - added:
method PatchTSMixerForPrediction.generate(past_values: torch.Tensor, observed_mask: Optional[torch.Tensor]=…) -> SamplePatchTSMixerPredictionOutput - added:
method PatchTSMixerForPretraining.forward(past_values: torch.Tensor, observed_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_loss: bool=…, return_dict: Optional[bool]=…) -> PatchTSMixerForPreTrainingOutput - added:
method PatchTSMixerForRegression.forward(past_values: torch.Tensor, target_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_loss: bool=…, return_dict: Optional[bool]=…) -> PatchTSMixerForRegressionOutput - added:
method PatchTSMixerForRegression.generate(past_values: torch.Tensor) -> SamplePatchTSMixerRegressionOutput - added:
method PatchTSMixerForTimeSeriesClassification.forward(past_values: torch.Tensor, target_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_loss: bool=…, return_dict: Optional[bool]=…) -> PatchTSMixerForTimeSeriesClassificationOutput - added:
method PatchTSMixerModel.forward(past_values: torch.Tensor, observed_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> PatchTSMixerModelOutput
- added:
export define_import_structure
- added:
class PatchTSTConfig(PretrainedConfig) - added:
constructor PatchTSTConfig(num_input_channels: int=…, context_length: int=…, distribution_output: str=…, loss: str=…, patch_length: int=…, patch_stride: int=…, num_hidden_layers: int=…, d_model: int=…, num_attention_heads: int=…, share_embedding: bool=…, channel_attention: bool=…, ffn_dim: int=…, norm_type: str=…, norm_eps: float=…, attention_dropout: float=…, positional_dropout: float=…, path_dr… - added:
field PatchTSTConfig.attribute_map - added:
field PatchTSTConfig.model_type
- added:
class PatchTSTForClassification(PatchTSTPreTrainedModel) - added:
class PatchTSTForPrediction(PatchTSTPreTrainedModel) - added:
class PatchTSTForPretraining(PatchTSTPreTrainedModel) - added:
class PatchTSTForRegression(PatchTSTPreTrainedModel) - added:
class PatchTSTModel(PatchTSTPreTrainedModel) - added:
class PatchTSTPreTrainedModel(PreTrainedModel) - added:
constructor PatchTSTForClassification(config: PatchTSTConfig) - added:
constructor PatchTSTForPrediction(config: PatchTSTConfig) - added:
constructor PatchTSTForPretraining(config: PatchTSTConfig) - added:
constructor PatchTSTForRegression(config: PatchTSTConfig) - added:
constructor PatchTSTModel(config: PatchTSTConfig) - added:
field PatchTSTPreTrainedModel.base_model_prefix - added:
field PatchTSTPreTrainedModel.config: PatchTSTConfig - added:
field PatchTSTPreTrainedModel.main_input_name - added:
field PatchTSTPreTrainedModel.supports_gradient_checkpointing - added:
method PatchTSTForClassification.forward(past_values: torch.Tensor, target_values: Optional[torch.Tensor]=…, past_observed_mask: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, PatchTSTForClassificationOutput] - added:
method PatchTSTForPrediction.forward(past_values: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, PatchTSTForPredictionOutput] - added:
method PatchTSTForPrediction.generate(past_values: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…) -> SamplePatchTSTOutput - added:
method PatchTSTForPretraining.forward(past_values: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, PatchTSTForPretrainingOutput] - added:
method PatchTSTForRegression.forward(past_values: torch.Tensor, target_values: Optional[torch.Tensor]=…, past_observed_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, PatchTSTForRegressionOutput] - added:
method PatchTSTForRegression.generate(past_values: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…) -> SamplePatchTSTOutput - added:
method PatchTSTModel.forward(past_values: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, PatchTSTModelOutput]
- added:
export define_import_structure
- added:
class PegasusConfig(PretrainedConfig) - added:
constructor PegasusConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…, s… - added:
field PegasusConfig.attribute_map - added:
field PegasusConfig.keys_to_ignore_at_inference - added:
field PegasusConfig.model_type - added:
property PegasusConfig.hidden_size -> int - added:
property PegasusConfig.num_attention_heads -> int
- added:
class FlaxPegasusForConditionalGeneration(FlaxPegasusPreTrainedModel) - added:
class FlaxPegasusModel(FlaxPegasusPreTrainedModel) - added:
class FlaxPegasusPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxPegasusPreTrainedModel(config: PegasusConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxPegasusForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxPegasusForConditionalGeneration.module_class - added:
field FlaxPegasusModel.config: PegasusConfig - added:
field FlaxPegasusModel.dtype: jnp.dtype - added:
field FlaxPegasusModel.module_class - added:
field FlaxPegasusPreTrainedModel.base_model_prefix: str - added:
field FlaxPegasusPreTrainedModel.config_class - added:
field FlaxPegasusPreTrainedModel.module_class: nn.Module - added:
method FlaxPegasusForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, deterministic: bool=…, pa… - added:
method FlaxPegasusForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxPegasusForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxPegasusPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[di… - added:
method FlaxPegasusPreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, position_ids: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxPegasusPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxPegasusPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class PegasusForCausalLM(PegasusPreTrainedModel, GenerationMixin) - added:
class PegasusForConditionalGeneration(PegasusPreTrainedModel, GenerationMixin) - added:
class PegasusModel(PegasusPreTrainedModel) - added:
class PegasusPreTrainedModel(PreTrainedModel) - added:
constructor PegasusForCausalLM(config) - added:
constructor PegasusForConditionalGeneration(config: PegasusConfig) - added:
constructor PegasusModel(config: PegasusConfig) - added:
field PegasusForConditionalGeneration.base_model_prefix - added:
field PegasusPreTrainedModel.base_model_prefix - added:
field PegasusPreTrainedModel.config: PegasusConfig - added:
field PegasusPreTrainedModel.supports_gradient_checkpointing - added:
method PegasusForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels… - added:
method PegasusForCausalLM.get_decoder() - added:
method PegasusForCausalLM.get_input_embeddings() - added:
method PegasusForCausalLM.get_position_embeddings() -> nn.Embedding - added:
method PegasusForCausalLM.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PegasusForCausalLM.set_decoder(decoder) - added:
method PegasusForCausalLM.set_input_embeddings(value) - added:
method PegasusForConditionalGeneration.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTenso… - added:
method PegasusForConditionalGeneration.get_decoder() - added:
method PegasusForConditionalGeneration.get_encoder() - added:
method PegasusForConditionalGeneration.get_position_embeddings() -> tuple[nn.Embedding] - added:
method PegasusForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method PegasusForConditionalGeneration.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PegasusForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method PegasusModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_val… - added:
method PegasusModel.get_encoder() - added:
method PegasusModel.get_input_embeddings() - added:
method PegasusModel.get_position_embeddings() -> tuple[nn.Embedding] - added:
method PegasusModel.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PegasusModel.set_input_embeddings(value)
- added:
class TFPegasusForConditionalGeneration(TFPegasusPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFPegasusModel(TFPegasusPreTrainedModel) - added:
class TFPegasusPreTrainedModel(TFPreTrainedModel) - added:
constructor TFPegasusForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFPegasusModel(config: PegasusConfig, *inputs, **kwargs) - added:
field TFPegasusPreTrainedModel.base_model_prefix - added:
field TFPegasusPreTrainedModel.config_class - added:
method TFPegasusForConditionalGeneration.build(input_shape=…) - added:
method TFPegasusForConditionalGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_… - added:
method TFPegasusForConditionalGeneration.get_bias() - added:
method TFPegasusForConditionalGeneration.get_decoder() - added:
method TFPegasusForConditionalGeneration.get_encoder() - added:
method TFPegasusForConditionalGeneration.get_output_embeddings() - added:
method TFPegasusForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFPegasusForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFPegasusForConditionalGeneration.serving_output(output) - added:
method TFPegasusForConditionalGeneration.set_bias(value) - added:
method TFPegasusForConditionalGeneration.set_output_embeddings(value) - added:
method TFPegasusModel.build(input_shape=…) - added:
method TFPegasusModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.… - added:
method TFPegasusModel.get_decoder() - added:
method TFPegasusModel.get_encoder() - added:
method TFPegasusModel.serving_output(output)
- added:
class PegasusTokenizer(PreTrainedTokenizer) - added:
constructor PegasusTokenizer(vocab_file, pad_token=…, eos_token=…, unk_token=…, mask_token=…, mask_token_sent=…, additional_special_tokens=…, offset=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field PegasusTokenizer.model_input_names - added:
field PegasusTokenizer.vocab_files_names - added:
method PegasusTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method PegasusTokenizer.convert_tokens_to_string(tokens) - added:
method PegasusTokenizer.get_special_tokens_mask(token_ids_0: list, token_ids_1: Optional[list]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method PegasusTokenizer.get_vocab() -> dict[str, int] - added:
method PegasusTokenizer.num_special_tokens_to_add(pair=…) - added:
method PegasusTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property PegasusTokenizer.vocab_size -> int
- added:
class PegasusTokenizerFast(PreTrainedTokenizerFast) - added:
constructor PegasusTokenizerFast(vocab_file=…, tokenizer_file=…, pad_token=…, eos_token=…, unk_token=…, mask_token=…, mask_token_sent=…, additional_special_tokens=…, offset=…, **kwargs) - added:
field PegasusTokenizerFast.model_input_names - added:
field PegasusTokenizerFast.slow_tokenizer_class - added:
field PegasusTokenizerFast.vocab_files_names - added:
method PegasusTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method PegasusTokenizerFast.get_special_tokens_mask(token_ids_0: list, token_ids_1: Optional[list]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method PegasusTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class PegasusXConfig(PretrainedConfig) - added:
constructor PegasusXConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…,… - added:
field PegasusXConfig.attribute_map - added:
field PegasusXConfig.keys_to_ignore_at_inference - added:
field PegasusXConfig.model_type - added:
property PegasusXConfig.hidden_size -> int - added:
property PegasusXConfig.num_attention_heads -> int
- added:
class PegasusXForConditionalGeneration(PegasusXPreTrainedModel, GenerationMixin) - added:
class PegasusXModel(PegasusXPreTrainedModel) - added:
class PegasusXPreTrainedModel(PreTrainedModel) - added:
constructor PegasusXForConditionalGeneration(config: PegasusXConfig) - added:
constructor PegasusXModel(config: PegasusXConfig) - added:
field PegasusXForConditionalGeneration.base_model_prefix - added:
field PegasusXPreTrainedModel.base_model_prefix - added:
field PegasusXPreTrainedModel.config: PegasusXConfig - added:
field PegasusXPreTrainedModel.supports_gradient_checkpointing - added:
method PegasusXForConditionalGeneration.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, decoder_inputs_embeds: Optional[torch.Tensor]… - added:
method PegasusXForConditionalGeneration.get_decoder() - added:
method PegasusXForConditionalGeneration.get_encoder() - added:
method PegasusXForConditionalGeneration.get_position_embeddings() -> tuple[nn.Embedding] - added:
method PegasusXForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method PegasusXForConditionalGeneration.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PegasusXModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, decoder_inputs_embeds: Optional[torch.Tensor]=…, use_cache: Opti… - added:
method PegasusXModel.get_encoder() - added:
method PegasusXModel.get_input_embeddings() - added:
method PegasusXModel.get_position_embeddings() -> tuple[nn.Embedding] - added:
method PegasusXModel.resize_position_embeddings(new_num_position_embeddings: int) - added:
method PegasusXModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class PerceiverConfig(PretrainedConfig) - added:
class PerceiverOnnxConfig(OnnxConfig) - added:
constructor PerceiverConfig(num_latents=…, d_latents=…, d_model=…, num_blocks=…, num_self_attends_per_block=…, num_self_attention_heads=…, num_cross_attention_heads=…, qk_channels=…, v_channels=…, cross_attention_shape_for_attention=…, self_attention_widening_factor=…, cross_attention_widening_factor=…, hidden_act=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, use_query… - added:
field PerceiverConfig.model_type - added:
method PerceiverOnnxConfig.generate_dummy_inputs(preprocessor: Union[PreTrainedTokenizerBase, FeatureExtractionMixin], batch_size: int=…, seq_length: int=…, num_choices: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…) -> Mapping[str, Any] - added:
property PerceiverOnnxConfig.atol_for_validation -> float - added:
property PerceiverOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class PerceiverFeatureExtractor(PerceiverImageProcessor) - added:
constructor PerceiverFeatureExtractor(*args, **kwargs) -> None
- added:
class PerceiverImageProcessor(BaseImageProcessor) - added:
constructor PerceiverImageProcessor(do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field PerceiverImageProcessor.model_input_names - added:
method PerceiverImageProcessor.center_crop(image: np.ndarray, crop_size: dict[str, int], size: Optional[int]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method PerceiverImageProcessor.preprocess(images: ImageInput, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std:… - added:
method PerceiverImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class PerceiverImageProcessorFast(BaseImageProcessorFast) - added:
field PerceiverImageProcessorFast.crop_size - added:
field PerceiverImageProcessorFast.do_center_crop - added:
field PerceiverImageProcessorFast.do_normalize - added:
field PerceiverImageProcessorFast.do_rescale - added:
field PerceiverImageProcessorFast.do_resize - added:
field PerceiverImageProcessorFast.image_mean - added:
field PerceiverImageProcessorFast.image_std - added:
field PerceiverImageProcessorFast.resample - added:
field PerceiverImageProcessorFast.size - added:
method PerceiverImageProcessorFast.center_crop(image: torch.Tensor, crop_size: dict[str, int], size: dict[str, int], **kwargs) -> torch.Tensor
- added:
class PerceiverForImageClassificationConvProcessing(PerceiverPreTrainedModel) - added:
class PerceiverForImageClassificationFourier(PerceiverPreTrainedModel) - added:
class PerceiverForImageClassificationLearned(PerceiverPreTrainedModel) - added:
class PerceiverForMaskedLM(PerceiverPreTrainedModel) - added:
class PerceiverForMultimodalAutoencoding(PerceiverPreTrainedModel) - added:
class PerceiverForOpticalFlow(PerceiverPreTrainedModel) - added:
class PerceiverForSequenceClassification(PerceiverPreTrainedModel) - added:
class PerceiverLayer(nn.Module) - added:
class PerceiverModel(PerceiverPreTrainedModel) - added:
class PerceiverPreTrainedModel(PreTrainedModel) - added:
constructor PerceiverForImageClassificationConvProcessing(config) - added:
constructor PerceiverForImageClassificationFourier(config) - added:
constructor PerceiverForImageClassificationLearned(config) - added:
constructor PerceiverForMaskedLM(config: PerceiverConfig) - added:
constructor PerceiverForMultimodalAutoencoding(config: PerceiverConfig) - added:
constructor PerceiverForOpticalFlow(config) - added:
constructor PerceiverForSequenceClassification(config) - added:
constructor PerceiverLayer(config, is_cross_attention=…, qk_channels=…, v_channels=…, num_heads=…, q_dim=…, kv_dim=…, widening_factor=…, use_query_residual=…) - added:
constructor PerceiverModel(config, decoder: Optional[PerceiverAbstractDecoder]=…, input_preprocessor: PreprocessorType=…, output_postprocessor: PostprocessorType=…) - added:
field PerceiverPreTrainedModel.base_model_prefix - added:
field PerceiverPreTrainedModel.config: PerceiverConfig - added:
field PerceiverPreTrainedModel.main_input_name - added:
method PerceiverForImageClassificationConvProcessing.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, pixel_values: Optional[torch.Tensor]=…) -> Union[tuple, PerceiverClassifierOutput] - added:
method PerceiverForImageClassificationFourier.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, pixel_values: Optional[torch.Tensor]=…) -> Union[tuple, PerceiverClassifierOutput] - added:
method PerceiverForImageClassificationLearned.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…, pixel_values: Optional[torch.Tensor]=…) -> Union[tuple,… - added:
method PerceiverForMaskedLM.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, input_ids: Optional[torch.Tensor]=…) -> Union[tuple, PerceiverMaskedLMOutput] - added:
method PerceiverForMultimodalAutoencoding.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, subsampled_output_points: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, PerceiverClassi… - added:
method PerceiverForOpticalFlow.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, PerceiverClassifierOutput] - added:
method PerceiverForSequenceClassification.forward(inputs: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…, input_ids: Optional[torch.Tensor]=…) -> Union[tuple, PerceiverClassifierOutput] - added:
method PerceiverLayer.feed_forward_chunk(attention_output) - added:
method PerceiverLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs: Optional[torch.FloatTensor]=…, inputs_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…) -> tuple[torch.Tensor] - added:
method PerceiverModel.forward(inputs: torch.FloatTensor, attention_mask: Optional[torch.FloatTensor]=…, subsampled_output_points: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, PerceiverModelOutput] - added:
method PerceiverModel.get_input_embeddings() - added:
method PerceiverModel.set_input_embeddings(value)
- added:
class PerceiverTokenizer(PreTrainedTokenizer) - added:
constructor PerceiverTokenizer(pad_token=…, bos_token=…, eos_token=…, mask_token=…, cls_token=…, sep_token=…, model_max_length=…, **kwargs) -> None - added:
field PerceiverTokenizer.model_input_names - added:
method PerceiverTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method PerceiverTokenizer.convert_tokens_to_string(tokens) - added:
method PerceiverTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method PerceiverTokenizer.get_vocab() -> dict[str, int] - added:
method PerceiverTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property PerceiverTokenizer.vocab_size
- added:
export define_import_structure
- added:
class PerceptionLMConfig(PretrainedConfig) - added:
constructor PerceptionLMConfig(vision_config=…, text_config=…, vision_use_cls_token=…, projector_pooling_ratio=…, image_token_id=…, video_token_id=…, **kwargs) - added:
field PerceptionLMConfig.model_type - added:
field PerceptionLMConfig.sub_configs
- added:
class PerceptionLMImageProcessorFast(BaseImageProcessorFast) - added:
constructor PerceptionLMImageProcessorFast(**kwargs: Unpack[PerceptionLMFastImageProcessorKwargs]) -> None - added:
field PerceptionLMImageProcessorFast.do_center_crop - added:
field PerceptionLMImageProcessorFast.do_convert_rgb - added:
field PerceptionLMImageProcessorFast.do_normalize - added:
field PerceptionLMImageProcessorFast.do_rescale - added:
field PerceptionLMImageProcessorFast.do_resize - added:
field PerceptionLMImageProcessorFast.image_mean - added:
field PerceptionLMImageProcessorFast.image_std - added:
field PerceptionLMImageProcessorFast.resample - added:
field PerceptionLMImageProcessorFast.size - added:
field PerceptionLMImageProcessorFast.valid_kwargs - added:
method PerceptionLMImageProcessorFast.preprocess(images, **kwargs: Unpack[PerceptionLMFastImageProcessorKwargs]) -> BatchFeature - added:
method PerceptionLMImageProcessorFast.resize(image: np.ndarray, tile_size: int, max_num_tiles: int, resample: PILImageResampling=…, input_data_format: Optional[Union[str, ChannelDimension]]=…)
- added:
class PerceptionLMForConditionalGeneration(PerceptionLMPreTrainedModel, GenerationMixin) - added:
class PerceptionLMModel(PerceptionLMPreTrainedModel) - added:
class PerceptionLMPreTrainedModel(PreTrainedModel) - added:
constructor PerceptionLMForConditionalGeneration(config: PerceptionLMConfig) - added:
constructor PerceptionLMModel(config: PerceptionLMConfig) - added:
field PerceptionLMPreTrainedModel.base_model_prefix - added:
field PerceptionLMPreTrainedModel.config: PerceptionLMConfig - added:
field PerceptionLMPreTrainedModel.supports_gradient_checkpointing - added:
method PerceptionLMForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use… - added:
method PerceptionLMForConditionalGeneration.get_decoder() - added:
method PerceptionLMForConditionalGeneration.get_input_embeddings() - added:
method PerceptionLMForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method PerceptionLMForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_values_videos=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method PerceptionLMForConditionalGeneration.set_decoder(decoder) - added:
method PerceptionLMForConditionalGeneration.set_input_embeddings(value) - added:
method PerceptionLMModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[boo… - added:
method PerceptionLMModel.get_decoder() - added:
method PerceptionLMModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs) - added:
method PerceptionLMModel.get_input_embeddings() - added:
method PerceptionLMModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method PerceptionLMModel.set_decoder(decoder) - added:
method PerceptionLMModel.set_input_embeddings(value)
- added:
class PerceptionLMForConditionalGeneration(LlavaForConditionalGeneration) - added:
class PerceptionLMModel(LlavaModel) - added:
class PerceptionLMPreTrainedModel(LlavaPreTrainedModel) - added:
constructor PerceptionLMModel(config: PerceptionLMConfig) - added:
field PerceptionLMPreTrainedModel.base_model_prefix - added:
method PerceptionLMForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use… - added:
method PerceptionLMForConditionalGeneration.get_image_features(**kwargs) - added:
method PerceptionLMForConditionalGeneration.language_model() - added:
method PerceptionLMForConditionalGeneration.multi_modal_projector() - added:
method PerceptionLMForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, pixel_values_videos=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method PerceptionLMForConditionalGeneration.vision_tower() - added:
method PerceptionLMModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[boo… - added:
method PerceptionLMModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs) - added:
method PerceptionLMModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…)
- added:
class PerceptionLMProcessor(ProcessorMixin) - added:
constructor PerceptionLMProcessor(video_processor=…, image_processor=…, tokenizer=…, patch_size=…, chat_template=…, pooling_ratio=…, **kwargs) - added:
field PerceptionLMProcessor.attributes - added:
field PerceptionLMProcessor.image_processor_class - added:
field PerceptionLMProcessor.tokenizer_class - added:
field PerceptionLMProcessor.video_processor_class
- added:
class PerceptionLMVideoProcessor(BaseVideoProcessor) - added:
constructor PerceptionLMVideoProcessor(**kwargs: Unpack[PerceptionLMFastVideoProcessorInitKwargs]) - added:
field PerceptionLMVideoProcessor.do_center_crop - added:
field PerceptionLMVideoProcessor.do_convert_rgb - added:
field PerceptionLMVideoProcessor.do_normalize - added:
field PerceptionLMVideoProcessor.do_rescale - added:
field PerceptionLMVideoProcessor.do_resize - added:
field PerceptionLMVideoProcessor.image_mean - added:
field PerceptionLMVideoProcessor.image_std - added:
field PerceptionLMVideoProcessor.model_input_names - added:
field PerceptionLMVideoProcessor.resample - added:
field PerceptionLMVideoProcessor.size - added:
field PerceptionLMVideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class PersimmonConfig(PretrainedConfig) - added:
constructor PersimmonConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, qk_layernorm=…, hidden_dropout=…, attention_dropout=…, partial_rotary_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **k… - added:
field PersimmonConfig.keys_to_ignore_at_inference - added:
field PersimmonConfig.model_type
- added:
class PersimmonForCausalLM(PersimmonPreTrainedModel, GenerationMixin) - added:
class PersimmonForSequenceClassification(GenericForSequenceClassification, PersimmonPreTrainedModel) - added:
class PersimmonForTokenClassification(GenericForTokenClassification, PersimmonPreTrainedModel) - added:
class PersimmonModel(PersimmonPreTrainedModel) - added:
class PersimmonPreTrainedModel(PreTrainedModel) - added:
constructor PersimmonForCausalLM(config) - added:
constructor PersimmonModel(config: PersimmonConfig) - added:
field PersimmonPreTrainedModel.base_model_prefix - added:
field PersimmonPreTrainedModel.config: PersimmonConfig - added:
field PersimmonPreTrainedModel.supports_gradient_checkpointing - added:
method PersimmonForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_posi… - added:
method PersimmonModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs…
- added:
export define_import_structure
- added:
class PhiConfig(PretrainedConfig) - added:
constructor PhiConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, resid_pdrop=…, embd_pdrop=…, attention_dropout=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, partial_rotary_factor=…, qk_layernorm=…, bos_token_id=…, eos_tok… - added:
field PhiConfig.base_model_pp_plan - added:
field PhiConfig.base_model_tp_plan - added:
field PhiConfig.keys_to_ignore_at_inference - added:
field PhiConfig.model_type
- added:
class PhiForCausalLM(PhiPreTrainedModel, GenerationMixin) - added:
class PhiForSequenceClassification(GenericForSequenceClassification, PhiPreTrainedModel) - added:
class PhiForTokenClassification(GenericForTokenClassification, PhiPreTrainedModel) - added:
class PhiModel(PhiPreTrainedModel) - added:
class PhiPreTrainedModel(PreTrainedModel) - added:
constructor PhiForCausalLM(config) - added:
constructor PhiModel(config: PhiConfig) - added:
field PhiPreTrainedModel.base_model_prefix - added:
field PhiPreTrainedModel.config: PhiConfig - added:
field PhiPreTrainedModel.supports_gradient_checkpointing - added:
method PhiForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **k… - added:
method PhiModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpa…
- added:
class PhiForCausalLM(LlamaForCausalLM) - added:
class PhiForSequenceClassification(LlamaForSequenceClassification) - added:
class PhiForTokenClassification(LlamaForTokenClassification) - added:
class PhiModel(LlamaModel) - added:
constructor PhiForCausalLM(config) - added:
constructor PhiModel(config: PhiConfig) - added:
method PhiModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpa…
- added:
export define_import_structure
- added:
class Phi3Config(PretrainedConfig) - added:
constructor Phi3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, resid_pdrop=…, embd_pdrop=…, attention_dropout=…, hidden_act=…, max_position_embeddings=…, original_max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, partial_rotary_factor=…, bos_… - added:
field Phi3Config.base_model_pp_plan - added:
field Phi3Config.base_model_tp_plan - added:
field Phi3Config.keys_to_ignore_at_inference - added:
field Phi3Config.model_type
- added:
class Phi3ForCausalLM(Phi3PreTrainedModel, GenerationMixin) - added:
class Phi3ForSequenceClassification(GenericForSequenceClassification, Phi3PreTrainedModel) - added:
class Phi3ForTokenClassification(GenericForTokenClassification, Phi3PreTrainedModel) - added:
class Phi3Model(Phi3PreTrainedModel) - added:
class Phi3PreTrainedModel(PreTrainedModel) - added:
constructor Phi3ForCausalLM(config) - added:
constructor Phi3Model(config: Phi3Config) - added:
field Phi3PreTrainedModel.base_model_prefix - added:
field Phi3PreTrainedModel.config: Phi3Config - added:
field Phi3PreTrainedModel.supports_gradient_checkpointing - added:
method Phi3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, **… - added:
method Phi3ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, **kwargs) - added:
method Phi3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Phi3ForCausalLM(MistralForCausalLM) - added:
class Phi3ForSequenceClassification(MistralForSequenceClassification) - added:
class Phi3ForTokenClassification(MistralForTokenClassification) - added:
class Phi3PreTrainedModel(MistralPreTrainedModel) - added:
method Phi3ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, **kwargs)
- added:
export define_import_structure
- added:
class Phi4MultimodalAudioConfig(PretrainedConfig) - added:
class Phi4MultimodalConfig(PretrainedConfig) - added:
class Phi4MultimodalVisionConfig(PretrainedConfig) - added:
constructor Phi4MultimodalAudioConfig(hidden_size: int=…, intermediate_size: int=…, num_blocks: int=…, num_attention_heads: int=…, activation: str=…, chunk_size: int=…, left_chunk: int=…, dropout_rate: float=…, ext_pw_out_channel: int=…, depthwise_separable_out_channel: int=…, depthwise_multiplier: int=…, kernel_size: int=…, conv_activation: str=…, input_size: int=…, conv_glu_type: str=…, time_re… - added:
constructor Phi4MultimodalConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, resid_pdrop=…, embd_pdrop=…, attention_dropout=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, partial_rotary_factor=…, bos_token_id=…, eos_token_id=…… - added:
constructor Phi4MultimodalVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, crop_size: int=…, image_token_id: int=…, feature_layer: int=…, **kwargs) - added:
field Phi4MultimodalAudioConfig.model_type - added:
field Phi4MultimodalConfig.base_model_pp_plan - added:
field Phi4MultimodalConfig.base_model_tp_plan - added:
field Phi4MultimodalConfig.keys_to_ignore_at_inference - added:
field Phi4MultimodalConfig.model_type - added:
field Phi4MultimodalConfig.sub_configs - added:
field Phi4MultimodalVisionConfig.base_config_key - added:
field Phi4MultimodalVisionConfig.model_type
- added:
class Phi4MultimodalFeatureExtractor(SequenceFeatureExtractor) - added:
constructor Phi4MultimodalFeatureExtractor(feature_size: int=…, sampling_rate: int=…, hop_length: int=…, n_fft: int=…, win_length: int=…, preemphasis: float=…, padding_value: float=…, audio_compression_rate: int=…, audio_downsample_rate: int=…, audio_feat_stride: int=…, mel_min_frequency: float=…, mel_max_frequency: float=…, **kwargs) - added:
field Phi4MultimodalFeatureExtractor.model_input_names
- added:
class Phi4MultimodalImageProcessorFast(BaseImageProcessorFast) - added:
constructor Phi4MultimodalImageProcessorFast(**kwargs: Unpack[Phi4MultimodalFastImageProcessorKwargs]) - added:
field Phi4MultimodalImageProcessorFast.do_convert_rgb - added:
field Phi4MultimodalImageProcessorFast.do_normalize - added:
field Phi4MultimodalImageProcessorFast.do_rescale - added:
field Phi4MultimodalImageProcessorFast.do_resize - added:
field Phi4MultimodalImageProcessorFast.dynamic_hd - added:
field Phi4MultimodalImageProcessorFast.image_mean - added:
field Phi4MultimodalImageProcessorFast.image_std - added:
field Phi4MultimodalImageProcessorFast.model_input_names - added:
field Phi4MultimodalImageProcessorFast.patch_size - added:
field Phi4MultimodalImageProcessorFast.resample - added:
field Phi4MultimodalImageProcessorFast.size - added:
field Phi4MultimodalImageProcessorFast.valid_kwargs - added:
method Phi4MultimodalImageProcessorFast.dynamic_preprocess(image, image_size, patch_size, mask_size, max_num=…, min_num=…) - added:
method Phi4MultimodalImageProcessorFast.find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size) - added:
method Phi4MultimodalImageProcessorFast.pad_mask_to_max_num_crops(masks, max_crops=…) - added:
method Phi4MultimodalImageProcessorFast.pad_to_max_num_crops(images, max_crops=…) - added:
method Phi4MultimodalImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Phi4MultimodalFastImageProcessorKwargs]) -> BatchFeature
- added:
class Phi4MultimodalAudioModel(Phi4MultimodalAudioPreTrainedModel) - added:
class Phi4MultimodalAudioPreTrainedModel(PreTrainedModel) - added:
class Phi4MultimodalForCausalLM(Phi4MultimodalPreTrainedModel, GenerationMixin) - added:
class Phi4MultimodalModel(Phi4MultimodalPreTrainedModel) - added:
class Phi4MultimodalPreTrainedModel(PreTrainedModel) - added:
class Phi4MultimodalVisionModel(Phi4MultimodalVisionPreTrainedModel) - added:
class Phi4MultimodalVisionPreTrainedModel(PreTrainedModel) - added:
constructor Phi4MultimodalAudioModel(config: Phi4MultimodalAudioConfig) - added:
constructor Phi4MultimodalForCausalLM(config) - added:
constructor Phi4MultimodalModel(config: Phi4MultimodalConfig) - added:
constructor Phi4MultimodalVisionModel(config: Phi4MultimodalVisionConfig) - added:
field Phi4MultimodalAudioPreTrainedModel.config: Phi4MultimodalAudioConfig - added:
field Phi4MultimodalAudioPreTrainedModel.supports_gradient_checkpointing - added:
field Phi4MultimodalPreTrainedModel.base_model_prefix - added:
field Phi4MultimodalPreTrainedModel.config: Phi4MultimodalConfig - added:
field Phi4MultimodalPreTrainedModel.supports_gradient_checkpointing - added:
field Phi4MultimodalVisionModel.config: Phi4MultimodalVisionConfig - added:
field Phi4MultimodalVisionModel.main_input_name - added:
field Phi4MultimodalVisionPreTrainedModel.base_model_prefix - added:
field Phi4MultimodalVisionPreTrainedModel.config: Phi4MultimodalVisionConfig - added:
field Phi4MultimodalVisionPreTrainedModel.supports_gradient_checkpointing - added:
method Phi4MultimodalAudioModel.calculate_hs_mask(hidden_states, device, mask) - added:
method Phi4MultimodalAudioModel.forward(hidden_states: torch.Tensor, mask: Optional[torch.Tensor]) - added:
method Phi4MultimodalAudioModel.forward_embeddings(hidden_states, masks) - added:
method Phi4MultimodalForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, image_attention_mask=…, audio_input_features: Optional[… - added:
method Phi4MultimodalForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, image_pixel_values=…, image_sizes=…, image_attention_mask=…, audio_input_features=…, audio_embed_sizes=…, audio_attention_mask=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, **kwargs) - added:
method Phi4MultimodalModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, image_attention_mask=…, audio_input_features: Optional[torch.… - added:
method Phi4MultimodalVisionModel.forward(pixel_values, patch_attention_mask: Optional[torch.BoolTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Phi4MultimodalVisionModel.get_input_embeddings() -> nn.Module
- added:
class Phi4MultimodalAudioConfig(PretrainedConfig) - added:
class Phi4MultimodalAudioModel(Phi4MultimodalAudioPreTrainedModel) - added:
class Phi4MultimodalAudioPreTrainedModel(PreTrainedModel) - added:
class Phi4MultimodalConfig(Phi3Config) - added:
class Phi4MultimodalForCausalLM(Phi3ForCausalLM) - added:
class Phi4MultimodalModel(Phi3Model) - added:
class Phi4MultimodalPreTrainedModel(Phi3PreTrainedModel) - added:
class Phi4MultimodalVisionConfig(SiglipVisionConfig) - added:
class Phi4MultimodalVisionModel(Phi4MultimodalVisionPreTrainedModel) - added:
class Phi4MultimodalVisionPreTrainedModel(SiglipPreTrainedModel) - added:
constructor Phi4MultimodalAudioConfig(hidden_size: int=…, intermediate_size: int=…, num_blocks: int=…, num_attention_heads: int=…, activation: str=…, chunk_size: int=…, left_chunk: int=…, dropout_rate: float=…, ext_pw_out_channel: int=…, depthwise_separable_out_channel: int=…, depthwise_multiplier: int=…, kernel_size: int=…, conv_activation: str=…, input_size: int=…, conv_glu_type: str=…, time_re… - added:
constructor Phi4MultimodalAudioModel(config: Phi4MultimodalAudioConfig) - added:
constructor Phi4MultimodalConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, resid_pdrop=…, embd_pdrop=…, attention_dropout=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, partial_rotary_factor=…, bos_token_id=…, eos_token_id=…… - added:
constructor Phi4MultimodalForCausalLM(config) - added:
constructor Phi4MultimodalModel(config: Phi4MultimodalConfig) - added:
constructor Phi4MultimodalVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, crop_size: int=…, image_token_id: int=…, feature_layer: int=…, **kwargs) - added:
constructor Phi4MultimodalVisionModel(config: Phi4MultimodalVisionConfig) - added:
field Phi4MultimodalAudioConfig.model_type - added:
field Phi4MultimodalAudioPreTrainedModel.config: Phi4MultimodalAudioConfig - added:
field Phi4MultimodalAudioPreTrainedModel.supports_gradient_checkpointing - added:
field Phi4MultimodalConfig.sub_configs - added:
field Phi4MultimodalVisionConfig.model_type - added:
field Phi4MultimodalVisionModel.config: Phi4MultimodalVisionConfig - added:
field Phi4MultimodalVisionModel.main_input_name - added:
field Phi4MultimodalVisionPreTrainedModel.base_model_prefix - added:
field Phi4MultimodalVisionPreTrainedModel.config: Phi4MultimodalVisionConfig - added:
field Phi4MultimodalVisionPreTrainedModel.supports_gradient_checkpointing - added:
method Phi4MultimodalAudioModel.calculate_hs_mask(hidden_states, device, mask) - added:
method Phi4MultimodalAudioModel.forward(hidden_states: torch.Tensor, mask: Optional[torch.Tensor]) - added:
method Phi4MultimodalAudioModel.forward_embeddings(hidden_states, masks) - added:
method Phi4MultimodalForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, image_attention_mask=…, audio_input_features: Optional[… - added:
method Phi4MultimodalForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, image_pixel_values=…, image_sizes=…, image_attention_mask=…, audio_input_features=…, audio_embed_sizes=…, audio_attention_mask=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, **kwargs) - added:
method Phi4MultimodalModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_pixel_values: Optional[torch.FloatTensor]=…, image_sizes: Optional[torch.LongTensor]=…, image_attention_mask=…, audio_input_features: Optional[torch.… - added:
method Phi4MultimodalVisionModel.forward(pixel_values, patch_attention_mask: Optional[torch.BoolTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Phi4MultimodalVisionModel.get_input_embeddings() -> nn.Module
- added:
class Phi4MultimodalProcessor(ProcessorMixin) - added:
constructor Phi4MultimodalProcessor(image_processor, audio_processor, tokenizer, **kwargs) - added:
field Phi4MultimodalProcessor.attributes - added:
field Phi4MultimodalProcessor.audio_processor_class - added:
field Phi4MultimodalProcessor.image_processor_class - added:
field Phi4MultimodalProcessor.tokenizer_class
- added:
export define_import_structure
- added:
class PhimoeConfig(PretrainedConfig) - added:
constructor PhimoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, sliding_window=…, attention_dropout=…, num_experts_per_tok=…, num… - added:
field PhimoeConfig.keys_to_ignore_at_inference - added:
field PhimoeConfig.model_type
- added:
class PhimoeForCausalLM(PhimoePreTrainedModel, GenerationMixin) - added:
class PhimoeForSequenceClassification(GenericForSequenceClassification, PhimoePreTrainedModel) - added:
class PhimoeModel(PhimoePreTrainedModel) - added:
class PhimoePreTrainedModel(PreTrainedModel) - added:
constructor PhimoeForCausalLM(config) - added:
constructor PhimoeModel(config: PhimoeConfig) - added:
field PhimoePreTrainedModel.base_model_prefix - added:
field PhimoePreTrainedModel.config: PhimoeConfig - added:
field PhimoePreTrainedModel.supports_gradient_checkpointing - added:
method PhimoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router… - added:
method PhimoeForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, logits_to_keep=…, **kwargs) - added:
method PhimoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: O…
- added:
export define_import_structure
- added:
class PhobertTokenizer(PreTrainedTokenizer) - added:
constructor PhobertTokenizer(vocab_file, merges_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
field PhobertTokenizer.vocab_files_names - added:
method PhobertTokenizer.add_from_file(f) - added:
method PhobertTokenizer.bpe(token) - added:
method PhobertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method PhobertTokenizer.convert_tokens_to_string(tokens) - added:
method PhobertTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method PhobertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method PhobertTokenizer.get_vocab() - added:
method PhobertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property PhobertTokenizer.vocab_size
- added:
export define_import_structure
- added:
class Pix2StructConfig(PretrainedConfig) - added:
class Pix2StructTextConfig(PretrainedConfig) - added:
class Pix2StructVisionConfig(PretrainedConfig) - added:
constructor Pix2StructConfig(text_config=…, vision_config=…, initializer_factor=…, initializer_range=…, is_vqa=…, tie_word_embeddings=…, is_encoder_decoder=…, **kwargs) - added:
constructor Pix2StructTextConfig(vocab_size=…, hidden_size=…, d_kv=…, d_ff=…, num_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, dense_act_fn=…, decoder_start_token_id=…, use_cache=…, pad_token_id=…, eos_token_id=…, tie_word_embeddings=…, is_decoder=…, **kwargs) - added:
constructor Pix2StructVisionConfig(hidden_size=…, patch_embed_hidden_size=…, d_ff=…, d_kv=…, num_hidden_layers=…, num_attention_heads=…, dense_act_fn=…, layer_norm_eps=…, dropout_rate=…, attention_dropout=…, initializer_range=…, initializer_factor=…, seq_len=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, **kwargs) - added:
field Pix2StructConfig.model_type - added:
field Pix2StructConfig.sub_configs - added:
field Pix2StructTextConfig.attribute_map - added:
field Pix2StructTextConfig.keys_to_ignore_at_inference - added:
field Pix2StructTextConfig.model_type - added:
field Pix2StructVisionConfig.model_type
- added:
class Pix2StructImageProcessor(BaseImageProcessor) - added:
constructor Pix2StructImageProcessor(do_convert_rgb: bool=…, do_normalize: bool=…, patch_size: Optional[dict[str, int]]=…, max_patches: int=…, is_vqa: bool=…, **kwargs) -> None - added:
field Pix2StructImageProcessor.model_input_names - added:
method Pix2StructImageProcessor.extract_flattened_patches(image: np.ndarray, max_patches: int, patch_size: dict, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Pix2StructImageProcessor.normalize(image: np.ndarray, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method Pix2StructImageProcessor.preprocess(images: ImageInput, header_text: Optional[str]=…, do_convert_rgb: Optional[bool]=…, do_normalize: Optional[bool]=…, max_patches: Optional[int]=…, patch_size: Optional[dict[str, int]]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> ImageInput
- added:
class Pix2StructForConditionalGeneration(Pix2StructPreTrainedModel, GenerationMixin) - added:
class Pix2StructPreTrainedModel(PreTrainedModel) - added:
class Pix2StructTextModel(Pix2StructPreTrainedModel) - added:
class Pix2StructVisionModel(Pix2StructPreTrainedModel) - added:
constructor Pix2StructForConditionalGeneration(config: Pix2StructConfig) - added:
constructor Pix2StructTextModel(config) - added:
constructor Pix2StructVisionModel(config: Pix2StructVisionConfig) - added:
field Pix2StructForConditionalGeneration.config: Pix2StructConfig - added:
field Pix2StructForConditionalGeneration.main_input_name - added:
field Pix2StructPreTrainedModel.config: Pix2StructConfig - added:
field Pix2StructTextModel.config: Pix2StructTextConfig - added:
field Pix2StructTextModel.supports_gradient_checkpointing - added:
field Pix2StructVisionModel.config: Pix2StructVisionConfig - added:
field Pix2StructVisionModel.main_input_name - added:
field Pix2StructVisionModel.supports_gradient_checkpointing - added:
method Pix2StructForConditionalGeneration.forward(flattened_patches: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_o… - added:
method Pix2StructForConditionalGeneration.get_encoder() - added:
method Pix2StructForConditionalGeneration.get_input_embeddings() - added:
method Pix2StructForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method Pix2StructForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method Pix2StructForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method Pix2StructTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]… - added:
method Pix2StructTextModel.set_input_embeddings(new_embeddings) - added:
method Pix2StructVisionModel.forward(flattened_patches: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method Pix2StructVisionModel.get_input_embeddings() - added:
property Pix2StructPreTrainedModel.dummy_inputs
- added:
class Pix2StructProcessor(ProcessorMixin) - added:
constructor Pix2StructProcessor(image_processor, tokenizer) - added:
field Pix2StructProcessor.attributes - added:
field Pix2StructProcessor.image_processor_class - added:
field Pix2StructProcessor.tokenizer_class - added:
property Pix2StructProcessor.model_input_names
- added:
export define_import_structure
- added:
class PixtralVisionConfig(PretrainedConfig) - added:
constructor PixtralVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, attention_dropout=…, rope_theta=…, initializer_range=…, **kwargs) - added:
field PixtralVisionConfig.model_type
- added:
class PixtralImageProcessor(BaseImageProcessor) - added:
constructor PixtralImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, patch_size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field PixtralImageProcessor.model_input_names - added:
method PixtralImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, patch_size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]… - added:
method PixtralImageProcessor.resize(image: np.ndarray, size: dict[str, int], patch_size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class PixtralImageProcessorFast(BaseImageProcessorFast) - added:
constructor PixtralImageProcessorFast(**kwargs: Unpack[PixtralFastImageProcessorKwargs]) - added:
field PixtralImageProcessorFast.default_to_square - added:
field PixtralImageProcessorFast.do_convert_rgb - added:
field PixtralImageProcessorFast.do_normalize - added:
field PixtralImageProcessorFast.do_rescale - added:
field PixtralImageProcessorFast.do_resize - added:
field PixtralImageProcessorFast.image_mean - added:
field PixtralImageProcessorFast.image_std - added:
field PixtralImageProcessorFast.model_input_names - added:
field PixtralImageProcessorFast.patch_size - added:
field PixtralImageProcessorFast.resample - added:
field PixtralImageProcessorFast.size - added:
field PixtralImageProcessorFast.valid_kwargs - added:
method PixtralImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[PixtralFastImageProcessorKwargs]) -> BatchFeature - added:
method PixtralImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, patch_size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor
- added:
class PixtralPreTrainedModel(PreTrainedModel) - added:
class PixtralVisionModel(PixtralPreTrainedModel) - added:
constructor PixtralVisionModel(config) - added:
field PixtralPreTrainedModel.base_model_prefix - added:
field PixtralPreTrainedModel.config: PixtralVisionConfig - added:
field PixtralPreTrainedModel.main_input_name - added:
field PixtralPreTrainedModel.supports_gradient_checkpointing - added:
field PixtralVisionModel.base_model_prefix - added:
method PixtralVisionModel.forward(pixel_values: torch.Tensor, image_sizes: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…, *args, **kwargs: Unpack[FlashAttentionKwargs]) -> Union[tuple, BaseModelOutput] - added:
method PixtralVisionModel.get_input_embeddings()
- added:
class PixtralProcessor(ProcessorMixin) - added:
constructor PixtralProcessor(image_processor=…, tokenizer=…, patch_size: int=…, spatial_merge_size: int=…, chat_template=…, image_token=…, image_break_token=…, image_end_token=…, **kwargs) - added:
field PixtralProcessor.attributes - added:
field PixtralProcessor.image_processor_class - added:
field PixtralProcessor.tokenizer_class - added:
property PixtralProcessor.model_input_names
- added:
export define_import_structure
- added:
class PLBartConfig(PretrainedConfig) - added:
constructor PLBartConfig(vocab_size=…, max_position_embeddings=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, classifier_dropout=…, scale_… - added:
field PLBartConfig.attribute_map - added:
field PLBartConfig.keys_to_ignore_at_inference - added:
field PLBartConfig.model_type
- added:
class PLBartForCausalLM(PLBartPreTrainedModel, GenerationMixin) - added:
class PLBartForConditionalGeneration(PLBartPreTrainedModel, GenerationMixin) - added:
class PLBartForSequenceClassification(PLBartPreTrainedModel) - added:
class PLBartModel(PLBartPreTrainedModel) - added:
class PLBartPreTrainedModel(PreTrainedModel) - added:
constructor PLBartForCausalLM(config) - added:
constructor PLBartForConditionalGeneration(config: PLBartConfig) - added:
constructor PLBartForSequenceClassification(config: PLBartConfig, **kwargs) - added:
constructor PLBartModel(config: PLBartConfig) - added:
field PLBartForConditionalGeneration.base_model_prefix - added:
field PLBartPreTrainedModel.base_model_prefix - added:
field PLBartPreTrainedModel.config: PLBartConfig - added:
field PLBartPreTrainedModel.supports_gradient_checkpointing - added:
method PLBartForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels:… - added:
method PLBartForCausalLM.get_decoder() - added:
method PLBartForCausalLM.get_input_embeddings() - added:
method PLBartForCausalLM.set_decoder(decoder) - added:
method PLBartForCausalLM.set_input_embeddings(value) - added:
method PLBartForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.LongTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[to… - added:
method PLBartForConditionalGeneration.get_decoder() - added:
method PLBartForConditionalGeneration.get_encoder() - added:
method PLBartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method PLBartForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method PLBartForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch… - added:
method PLBartModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.LongTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor]]=…… - added:
method PLBartModel.get_encoder() - added:
method PLBartModel.get_input_embeddings() - added:
method PLBartModel.set_input_embeddings(value)
- added:
class PLBartForCausalLM(BartForCausalLM) - added:
class PLBartForConditionalGeneration(PLBartPreTrainedModel, GenerationMixin) - added:
class PLBartForSequenceClassification(BigBirdPegasusForSequenceClassification) - added:
class PLBartModel(PLBartPreTrainedModel) - added:
class PLBartPreTrainedModel(PreTrainedModel) - added:
constructor PLBartForConditionalGeneration(config: PLBartConfig) - added:
constructor PLBartModel(config: PLBartConfig) - added:
field PLBartForConditionalGeneration.base_model_prefix - added:
field PLBartPreTrainedModel.base_model_prefix - added:
field PLBartPreTrainedModel.config: PLBartConfig - added:
field PLBartPreTrainedModel.supports_gradient_checkpointing - added:
method PLBartForCausalLM.forward(**super_kwargs) - added:
method PLBartForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.LongTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[to… - added:
method PLBartForConditionalGeneration.get_decoder() - added:
method PLBartForConditionalGeneration.get_encoder() - added:
method PLBartForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method PLBartForConditionalGeneration.resize_token_embeddings(new_num_tokens: int, pad_to_multiple_of: Optional[int]=…, mean_resizing: bool=…) -> nn.Embedding - added:
method PLBartForSequenceClassification.forward(**super_kwargs) - added:
method PLBartModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.LongTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.FloatTensor]]=…… - added:
method PLBartModel.get_encoder() - added:
method PLBartModel.get_input_embeddings() - added:
method PLBartModel.set_input_embeddings(value)
- added:
class PLBartTokenizer(PreTrainedTokenizer) - added:
constructor PLBartTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, language_codes=…, tokenizer_file=…, src_lang=…, tgt_lang=…, sp_model_kwargs: Optional[dict[str, Any]]=…, additional_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field PLBartTokenizer.model_input_names - added:
field PLBartTokenizer.prefix_tokens: list[int] - added:
field PLBartTokenizer.suffix_tokens: list[int] - added:
field PLBartTokenizer.vocab_files_names - added:
method PLBartTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method PLBartTokenizer.convert_tokens_to_string(tokens) - added:
method PLBartTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method PLBartTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method PLBartTokenizer.get_vocab() - added:
method PLBartTokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method PLBartTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method PLBartTokenizer.set_src_lang_special_tokens(src_lang) -> None - added:
method PLBartTokenizer.set_tgt_lang_special_tokens(lang: str) -> None - added:
property PLBartTokenizer.src_lang -> str - added:
property PLBartTokenizer.vocab_size
- added:
export define_import_structure
- added:
class PoolFormerConfig(PretrainedConfig) - added:
class PoolFormerOnnxConfig(OnnxConfig) - added:
constructor PoolFormerConfig(num_channels=…, patch_size=…, stride=…, pool_size=…, mlp_ratio=…, depths=…, hidden_sizes=…, patch_sizes=…, strides=…, padding=…, num_encoder_blocks=…, drop_path_rate=…, hidden_act=…, use_layer_scale=…, layer_scale_init_value=…, initializer_range=…, **kwargs) - added:
field PoolFormerConfig.model_type - added:
field PoolFormerOnnxConfig.torch_onnx_minimum_version - added:
property PoolFormerOnnxConfig.atol_for_validation -> float - added:
property PoolFormerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class PoolFormerFeatureExtractor(PoolFormerImageProcessor) - added:
constructor PoolFormerFeatureExtractor(*args, **kwargs) -> None
- added:
class PoolFormerImageProcessor(BaseImageProcessor) - added:
constructor PoolFormerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, crop_pct: int=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, rescale_factor: Union[int, float]=…, do_rescale: bool=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field PoolFormerImageProcessor.model_input_names - added:
method PoolFormerImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, crop_pct: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float,… - added:
method PoolFormerImageProcessor.resize(image: np.ndarray, size: dict[str, int], crop_pct: Optional[float]=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class PoolFormerImageProcessorFast(BaseImageProcessorFast) - added:
constructor PoolFormerImageProcessorFast(**kwargs: Unpack[PoolFormerFastImageProcessorKwargs]) - added:
field PoolFormerImageProcessorFast.crop_pct - added:
field PoolFormerImageProcessorFast.crop_size - added:
field PoolFormerImageProcessorFast.default_to_square - added:
field PoolFormerImageProcessorFast.do_center_crop - added:
field PoolFormerImageProcessorFast.do_normalize - added:
field PoolFormerImageProcessorFast.do_rescale - added:
field PoolFormerImageProcessorFast.do_resize - added:
field PoolFormerImageProcessorFast.image_mean - added:
field PoolFormerImageProcessorFast.image_std - added:
field PoolFormerImageProcessorFast.resample - added:
field PoolFormerImageProcessorFast.size - added:
field PoolFormerImageProcessorFast.valid_kwargs - added:
method PoolFormerImageProcessorFast.center_crop(image: torch.Tensor, size: SizeDict, **kwargs) -> torch.Tensor - added:
method PoolFormerImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[PoolFormerFastImageProcessorKwargs]) -> BatchFeature - added:
method PoolFormerImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, crop_pct: Optional[float]=…, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor
- added:
class PoolFormerForImageClassification(PoolFormerPreTrainedModel) - added:
class PoolFormerModel(PoolFormerPreTrainedModel) - added:
class PoolFormerPreTrainedModel(PreTrainedModel) - added:
constructor PoolFormerForImageClassification(config) - added:
constructor PoolFormerModel(config) - added:
field PoolFormerPreTrainedModel.base_model_prefix - added:
field PoolFormerPreTrainedModel.config: PoolFormerConfig - added:
field PoolFormerPreTrainedModel.main_input_name - added:
method PoolFormerForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method PoolFormerModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithNoAttention] - added:
method PoolFormerModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class Pop2PianoConfig(PretrainedConfig) - added:
constructor Pop2PianoConfig(vocab_size=…, composer_vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, use_cache=…, pad_token_id=…, eos_token_id=…, dense_act_fn=…, **kwargs) - added:
field Pop2PianoConfig.keys_to_ignore_at_inference - added:
field Pop2PianoConfig.model_type
- added:
class Pop2PianoFeatureExtractor(SequenceFeatureExtractor) - added:
constructor Pop2PianoFeatureExtractor(sampling_rate: int=…, padding_value: int=…, window_size: int=…, hop_length: int=…, min_frequency: float=…, feature_size: int=…, num_bars: int=…, **kwargs) - added:
field Pop2PianoFeatureExtractor.model_input_names - added:
method Pop2PianoFeatureExtractor.extract_rhythm(audio: np.ndarray) - added:
method Pop2PianoFeatureExtractor.interpolate_beat_times(beat_times: numpy.ndarray, steps_per_beat: numpy.ndarray, n_extend: numpy.ndarray) - added:
method Pop2PianoFeatureExtractor.mel_spectrogram(sequence: np.ndarray) - added:
method Pop2PianoFeatureExtractor.pad(inputs: BatchFeature, is_batched: bool, return_attention_mask: bool, return_tensors: Optional[Union[str, TensorType]]=…) - added:
method Pop2PianoFeatureExtractor.preprocess_mel(audio: np.ndarray, beatstep: np.ndarray)
- added:
class Pop2PianoForConditionalGeneration(Pop2PianoPreTrainedModel, GenerationMixin) - added:
class Pop2PianoPreTrainedModel(PreTrainedModel) - added:
constructor Pop2PianoForConditionalGeneration(config: Pop2PianoConfig) - added:
field Pop2PianoPreTrainedModel.base_model_prefix - added:
field Pop2PianoPreTrainedModel.config: Pop2PianoConfig - added:
field Pop2PianoPreTrainedModel.is_parallelizable - added:
field Pop2PianoPreTrainedModel.supports_gradient_checkpointing - added:
method Pop2PianoForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Op… - added:
method Pop2PianoForConditionalGeneration.generate(input_features, attention_mask=…, composer=…, generation_config=…, **kwargs) - added:
method Pop2PianoForConditionalGeneration.get_encoder() - added:
method Pop2PianoForConditionalGeneration.get_input_embeddings() - added:
method Pop2PianoForConditionalGeneration.get_mel_conditioner_outputs(input_features: torch.FloatTensor, composer: str, generation_config: GenerationConfig, attention_mask: Optional[torch.FloatTensor]=…) - added:
method Pop2PianoForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method Pop2PianoForConditionalGeneration.set_input_embeddings(new_embeddings)
- added:
class Pop2PianoProcessor(ProcessorMixin) - added:
constructor Pop2PianoProcessor(feature_extractor, tokenizer) - added:
field Pop2PianoProcessor.attributes - added:
field Pop2PianoProcessor.feature_extractor_class - added:
field Pop2PianoProcessor.tokenizer_class - added:
method Pop2PianoProcessor.batch_decode(token_ids, feature_extractor_output: BatchFeature, return_midi: bool=…) -> BatchEncoding - added:
method Pop2PianoProcessor.save_pretrained(save_directory, **kwargs) - added:
static method Pop2PianoProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
class Pop2PianoTokenizer(PreTrainedTokenizer) - added:
constructor Pop2PianoTokenizer(vocab, default_velocity=…, num_bars=…, unk_token=…, eos_token=…, pad_token=…, bos_token=…, **kwargs) - added:
field Pop2PianoTokenizer.model_input_names - added:
field Pop2PianoTokenizer.vocab_files_names - added:
method Pop2PianoTokenizer.batch_decode(token_ids, feature_extractor_output: BatchFeature, return_midi: bool=…) - added:
method Pop2PianoTokenizer.batch_encode_plus(notes: Union[np.ndarray, list[pretty_midi.Note]], truncation_strategy: Optional[TruncationStrategy]=…, max_length: Optional[int]=…, **kwargs) -> BatchEncoding - added:
method Pop2PianoTokenizer.encode_plus(notes: Union[np.ndarray, list[pretty_midi.Note]], truncation_strategy: Optional[TruncationStrategy]=…, max_length: Optional[int]=…, **kwargs) -> BatchEncoding - added:
method Pop2PianoTokenizer.get_vocab() - added:
method Pop2PianoTokenizer.notes_to_midi(notes: np.ndarray, beatstep: np.ndarray, offset_sec: int=…) - added:
method Pop2PianoTokenizer.relative_batch_tokens_ids_to_midi(tokens: np.ndarray, beatstep: np.ndarray, beat_offset_idx: int=…, bars_per_batch: int=…, cutoff_time_idx: int=…) - added:
method Pop2PianoTokenizer.relative_batch_tokens_ids_to_notes(tokens: np.ndarray, beat_offset_idx: int, bars_per_batch: int, cutoff_time_idx: int) - added:
method Pop2PianoTokenizer.relative_tokens_ids_to_notes(tokens: np.ndarray, start_idx: float, cutoff_time_idx: Optional[float]=…) - added:
method Pop2PianoTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property Pop2PianoTokenizer.vocab_size
- added:
export PromptDepthAnythingConfig - added:
export PromptDepthAnythingForDepthEstimation - added:
export PromptDepthAnythingImageProcessor - added:
export PromptDepthAnythingImageProcessorFast - added:
export PromptDepthAnythingPreTrainedModel - added:
export define_import_structure
- added:
class PromptDepthAnythingConfig(PretrainedConfig) - added:
constructor PromptDepthAnythingConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, patch_size=…, initializer_range=…, reassemble_hidden_size=…, reassemble_factors=…, neck_hidden_sizes=…, fusion_hidden_size=…, head_in_index=…, head_hidden_size=…, depth_estimation_type=…, max_depth=…, **kwargs) - added:
field PromptDepthAnythingConfig.model_type - added:
method PromptDepthAnythingConfig.to_dict() - added:
property PromptDepthAnythingConfig.sub_configs
- added:
class PromptDepthAnythingImageProcessor(BaseImageProcessor) - added:
constructor PromptDepthAnythingImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, size_divisor: Opti… - added:
field PromptDepthAnythingImageProcessor.model_input_names - added:
method PromptDepthAnythingImageProcessor.pad_image(image: np.ndarray, size_divisor: int, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method PromptDepthAnythingImageProcessor.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method PromptDepthAnythingImageProcessor.preprocess(images: ImageInput, prompt_depth: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[int]=…, keep_aspect_ratio: Optional[bool]=…, ensure_multiple_of: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Uni… - added:
method PromptDepthAnythingImageProcessor.resize(image: np.ndarray, size: dict[str, int], keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class PromptDepthAnythingImageProcessorFast(BaseImageProcessorFast) - added:
constructor PromptDepthAnythingImageProcessorFast(**kwargs: Unpack[PromptDepthAnythingFastImageProcessorKwargs]) - added:
field PromptDepthAnythingImageProcessorFast.do_normalize - added:
field PromptDepthAnythingImageProcessorFast.do_pad - added:
field PromptDepthAnythingImageProcessorFast.do_rescale - added:
field PromptDepthAnythingImageProcessorFast.do_resize - added:
field PromptDepthAnythingImageProcessorFast.ensure_multiple_of - added:
field PromptDepthAnythingImageProcessorFast.image_mean - added:
field PromptDepthAnythingImageProcessorFast.image_std - added:
field PromptDepthAnythingImageProcessorFast.keep_aspect_ratio - added:
field PromptDepthAnythingImageProcessorFast.model_input_names - added:
field PromptDepthAnythingImageProcessorFast.prompt_scale_to_meter - added:
field PromptDepthAnythingImageProcessorFast.resample - added:
field PromptDepthAnythingImageProcessorFast.size - added:
field PromptDepthAnythingImageProcessorFast.size_divisor - added:
field PromptDepthAnythingImageProcessorFast.valid_kwargs - added:
method PromptDepthAnythingImageProcessorFast.pad_image(image: torch.Tensor, size_divisor: int) -> torch.Tensor - added:
method PromptDepthAnythingImageProcessorFast.post_process_depth_estimation(outputs: DepthEstimatorOutput, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…) -> list[dict[str, TensorType]] - added:
method PromptDepthAnythingImageProcessorFast.preprocess(images: ImageInput, prompt_depth: Optional[ImageInput]=…, **kwargs: Unpack[PromptDepthAnythingFastImageProcessorKwargs]) -> BatchFeature - added:
method PromptDepthAnythingImageProcessorFast.resize_with_aspect_ratio(image: torch.Tensor, size: SizeDict, keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, interpolation: Optional[F.InterpolationMode]=…) -> torch.Tensor
- added:
class PromptDepthAnythingForDepthEstimation(PromptDepthAnythingPreTrainedModel) - added:
class PromptDepthAnythingPreTrainedModel(PreTrainedModel) - added:
constructor PromptDepthAnythingForDepthEstimation(config) - added:
field PromptDepthAnythingPreTrainedModel.base_model_prefix - added:
field PromptDepthAnythingPreTrainedModel.config: PromptDepthAnythingConfig - added:
field PromptDepthAnythingPreTrainedModel.main_input_name - added:
field PromptDepthAnythingPreTrainedModel.supports_gradient_checkpointing - added:
method PromptDepthAnythingForDepthEstimation.forward(pixel_values: torch.FloatTensor, prompt_depth: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthEstimatorOutput]
- added:
class PromptDepthAnythingConfig(DepthAnythingConfig) - added:
class PromptDepthAnythingForDepthEstimation(DepthAnythingForDepthEstimation) - added:
class PromptDepthAnythingPreTrainedModel(PreTrainedModel) - added:
field PromptDepthAnythingConfig.model_type - added:
field PromptDepthAnythingPreTrainedModel.base_model_prefix - added:
field PromptDepthAnythingPreTrainedModel.config: PromptDepthAnythingConfig - added:
field PromptDepthAnythingPreTrainedModel.main_input_name - added:
field PromptDepthAnythingPreTrainedModel.supports_gradient_checkpointing - added:
method PromptDepthAnythingForDepthEstimation.forward(pixel_values: torch.FloatTensor, prompt_depth: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthEstimatorOutput]
- added:
export define_import_structure
- added:
class ProphetNetConfig(PretrainedConfig) - added:
constructor ProphetNetConfig(activation_dropout: Optional[float]=…, activation_function: Optional[Union[str, Callable]]=…, vocab_size: Optional[int]=…, hidden_size: Optional[int]=…, encoder_ffn_dim: Optional[int]=…, num_encoder_layers: Optional[int]=…, num_encoder_attention_heads: Optional[int]=…, decoder_ffn_dim: Optional[int]=…, num_decoder_layers: Optional[int]=…, num_decoder_attention_heads:… - added:
field ProphetNetConfig.attribute_map - added:
field ProphetNetConfig.keys_to_ignore_at_inference - added:
field ProphetNetConfig.model_type - added:
property ProphetNetConfig.num_hidden_layers -> int
- added:
class ProphetNetDecoder(ProphetNetPreTrainedModel) - added:
class ProphetNetEncoder(ProphetNetPreTrainedModel) - added:
class ProphetNetForCausalLM(ProphetNetPreTrainedModel, GenerationMixin) - added:
class ProphetNetForConditionalGeneration(ProphetNetPreTrainedModel, GenerationMixin) - added:
class ProphetNetModel(ProphetNetPreTrainedModel) - added:
class ProphetNetPreTrainedModel(PreTrainedModel) - added:
constructor ProphetNetDecoder(config: ProphetNetConfig, word_embeddings: Optional[nn.Embedding]=…) - added:
constructor ProphetNetEncoder(config: ProphetNetConfig, word_embeddings: Optional[nn.Embedding]=…) - added:
constructor ProphetNetForCausalLM(config: ProphetNetConfig) - added:
constructor ProphetNetForConditionalGeneration(config: ProphetNetConfig) - added:
constructor ProphetNetModel(config: ProphetNetConfig) - added:
field ProphetNetPreTrainedModel.base_model_prefix - added:
field ProphetNetPreTrainedModel.config: ProphetNetConfig - added:
field ProphetNetPreTrainedModel.supports_gradient_checkpointing - added:
method ProphetNetDecoder.compute_buffered_relative_buckets(position_ids) - added:
method ProphetNetDecoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, use_cache: Optional[bool]=… - added:
method ProphetNetDecoder.get_input_embeddings() - added:
method ProphetNetDecoder.prepare_attention_mask(hidden_states, attention_mask) - added:
method ProphetNetDecoder.prepare_predict_attention_mask(hidden_states, attention_mask) - added:
method ProphetNetDecoder.set_input_embeddings(value) - added:
method ProphetNetEncoder.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method ProphetNetEncoder.get_input_embeddings() - added:
method ProphetNetEncoder.set_input_embeddings(value) - added:
method ProphetNetForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch… - added:
method ProphetNetForCausalLM.get_decoder() - added:
method ProphetNetForCausalLM.get_input_embeddings() - added:
method ProphetNetForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, head_mask=…, use_cache=…, **kwargs) - added:
method ProphetNetForCausalLM.set_decoder(decoder) - added:
method ProphetNetForCausalLM.set_input_embeddings(value) - added:
method ProphetNetForConditionalGeneration.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[torch.Tensor]=…… - added:
method ProphetNetForConditionalGeneration.get_decoder() - added:
method ProphetNetForConditionalGeneration.get_encoder() - added:
method ProphetNetForConditionalGeneration.get_input_embeddings() - added:
method ProphetNetForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method ProphetNetModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple]=…, past_key_values: Optiona… - added:
method ProphetNetModel.get_encoder() - added:
method ProphetNetModel.get_input_embeddings() - added:
method ProphetNetModel.set_input_embeddings(value)
- added:
class ProphetNetTokenizer(PreTrainedTokenizer) - added:
constructor ProphetNetTokenizer(vocab_file: str, do_lower_case: Optional[bool]=…, do_basic_tokenize: Optional[bool]=…, never_split: Optional[Iterable]=…, unk_token: Optional[str]=…, sep_token: Optional[str]=…, x_sep_token: Optional[str]=…, pad_token: Optional[str]=…, mask_token: Optional[str]=…, tokenize_chinese_chars: Optional[bool]=…, strip_accents: Optional[bool]=…, clean_up_tokenization_space… - added:
field ProphetNetTokenizer.model_input_names: list[str] - added:
field ProphetNetTokenizer.vocab_files_names - added:
method ProphetNetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method ProphetNetTokenizer.convert_tokens_to_string(tokens: str) - added:
method ProphetNetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: Optional[bool]=…) -> list[int] - added:
method ProphetNetTokenizer.get_vocab() - added:
method ProphetNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ProphetNetTokenizer.vocab_size
- added:
export define_import_structure
- added:
class PvtConfig(PretrainedConfig) - added:
class PvtOnnxConfig(OnnxConfig) - added:
constructor PvtConfig(image_size: int=…, num_channels: int=…, num_encoder_blocks: int=…, depths: list[int]=…, sequence_reduction_ratios: list[int]=…, hidden_sizes: list[int]=…, patch_sizes: list[int]=…, strides: list[int]=…, num_attention_heads: list[int]=…, mlp_ratios: list[int]=…, hidden_act: Mapping[str, Callable]=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…, initiali… - added:
field PvtConfig.model_type - added:
field PvtOnnxConfig.torch_onnx_minimum_version - added:
property PvtOnnxConfig.atol_for_validation -> float - added:
property PvtOnnxConfig.default_onnx_opset -> int - added:
property PvtOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class PvtImageProcessor(BaseImageProcessor) - added:
constructor PvtImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field PvtImageProcessor.model_input_names - added:
method PvtImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, Tens… - added:
method PvtImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class PvtImageProcessorFast(BaseImageProcessorFast) - added:
field PvtImageProcessorFast.crop_size - added:
field PvtImageProcessorFast.default_to_square - added:
field PvtImageProcessorFast.do_center_crop - added:
field PvtImageProcessorFast.do_convert_rgb - added:
field PvtImageProcessorFast.do_normalize - added:
field PvtImageProcessorFast.do_rescale - added:
field PvtImageProcessorFast.do_resize - added:
field PvtImageProcessorFast.image_mean - added:
field PvtImageProcessorFast.image_std - added:
field PvtImageProcessorFast.model_input_names - added:
field PvtImageProcessorFast.resample - added:
field PvtImageProcessorFast.size
- added:
class PvtForImageClassification(PvtPreTrainedModel) - added:
class PvtModel(PvtPreTrainedModel) - added:
class PvtPreTrainedModel(PreTrainedModel) - added:
constructor PvtForImageClassification(config: PvtConfig) -> None - added:
constructor PvtModel(config: PvtConfig) - added:
field PvtPreTrainedModel.base_model_prefix - added:
field PvtPreTrainedModel.config: PvtConfig - added:
field PvtPreTrainedModel.main_input_name - added:
method PvtForImageClassification.forward(pixel_values: Optional[torch.Tensor], labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method PvtModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
export define_import_structure
- added:
class PvtV2Config(BackboneConfigMixin, PretrainedConfig) - added:
constructor PvtV2Config(image_size: Union[int, tuple[int, int]]=…, num_channels: int=…, num_encoder_blocks: int=…, depths: list[int]=…, sr_ratios: list[int]=…, hidden_sizes: list[int]=…, patch_sizes: list[int]=…, strides: list[int]=…, num_attention_heads: list[int]=…, mlp_ratios: list[int]=…, hidden_act: Union[str, Callable]=…, hidden_dropout_prob: float=…, attention_probs_dropout_prob: float=…,… - added:
field PvtV2Config.model_type
- added:
class PvtV2Backbone(PvtV2Model, BackboneMixin) - added:
class PvtV2ForImageClassification(PvtV2PreTrainedModel) - added:
class PvtV2Model(PvtV2PreTrainedModel) - added:
class PvtV2PreTrainedModel(PreTrainedModel) - added:
constructor PvtV2Backbone(config: PvtV2Config) - added:
constructor PvtV2ForImageClassification(config: PvtV2Config) -> None - added:
constructor PvtV2Model(config: PvtV2Config) - added:
field PvtV2PreTrainedModel.base_model_prefix - added:
field PvtV2PreTrainedModel.config: PvtV2Config - added:
field PvtV2PreTrainedModel.main_input_name - added:
field PvtV2PreTrainedModel.supports_gradient_checkpointing - added:
method PvtV2Backbone.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method PvtV2ForImageClassification.forward(pixel_values: Optional[torch.Tensor], labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method PvtV2Model.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
export define_import_structure
- added:
class Qwen2Config(PretrainedConfig) - added:
constructor Qwen2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, use_sliding_window=…, sliding_window=…, max_window_layers=…, layer_types=…, attention_dropout=…, **kwargs) - added:
field Qwen2Config.base_model_pp_plan - added:
field Qwen2Config.base_model_tp_plan - added:
field Qwen2Config.keys_to_ignore_at_inference - added:
field Qwen2Config.model_type
- added:
class Qwen2ForCausalLM(Qwen2PreTrainedModel, GenerationMixin) - added:
class Qwen2ForQuestionAnswering(GenericForQuestionAnswering, Qwen2PreTrainedModel) - added:
class Qwen2ForSequenceClassification(GenericForSequenceClassification, Qwen2PreTrainedModel) - added:
class Qwen2ForTokenClassification(GenericForTokenClassification, Qwen2PreTrainedModel) - added:
class Qwen2Model(Qwen2PreTrainedModel) - added:
class Qwen2PreTrainedModel(PreTrainedModel) - added:
class Qwen2RMSNorm(nn.Module) - added:
constructor Qwen2ForCausalLM(config) - added:
constructor Qwen2Model(config: Qwen2Config) - added:
constructor Qwen2RMSNorm(hidden_size, eps: float=…) -> None - added:
field Qwen2ForQuestionAnswering.base_model_prefix - added:
field Qwen2PreTrainedModel.base_model_prefix - added:
field Qwen2PreTrainedModel.config: Qwen2Config - added:
field Qwen2PreTrainedModel.supports_gradient_checkpointing - added:
method Qwen2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method Qwen2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method Qwen2RMSNorm.extra_repr() - added:
method Qwen2RMSNorm.forward(hidden_states: torch.Tensor) -> torch.Tensor
- added:
class Qwen2ForCausalLM(LlamaForCausalLM) - added:
class Qwen2ForQuestionAnswering(LlamaForQuestionAnswering) - added:
class Qwen2ForSequenceClassification(LlamaForSequenceClassification) - added:
class Qwen2ForTokenClassification(LlamaForTokenClassification) - added:
class Qwen2Model(MistralModel) - added:
class Qwen2PreTrainedModel(LlamaPreTrainedModel) - added:
class Qwen2RMSNorm(nn.Module) - added:
class Qwen2RMSNorm(nn.RMSNorm) - added:
constructor Qwen2Model(config: Qwen2Config) - added:
constructor Qwen2RMSNorm(hidden_size, eps: float=…) -> None - added:
method Qwen2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method Qwen2RMSNorm.extra_repr() - added:
method Qwen2RMSNorm.forward(hidden_states: torch.Tensor) -> torch.Tensor
- added:
class Qwen2Tokenizer(PreTrainedTokenizer) - added:
constructor Qwen2Tokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, clean_up_tokenization_spaces=…, split_special_tokens=…, **kwargs) - added:
field Qwen2Tokenizer.model_input_names - added:
field Qwen2Tokenizer.vocab_files_names - added:
method Qwen2Tokenizer.bpe(token) - added:
method Qwen2Tokenizer.convert_tokens_to_string(tokens) - added:
method Qwen2Tokenizer.decode(token_ids, skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, spaces_between_special_tokens: bool=…, **kwargs) -> str - added:
method Qwen2Tokenizer.get_vocab() - added:
method Qwen2Tokenizer.prepare_for_tokenization(text, **kwargs) - added:
method Qwen2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property Qwen2Tokenizer.vocab_size -> int
- added:
class Qwen2TokenizerFast(PreTrainedTokenizerFast) - added:
constructor Qwen2TokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, **kwargs) - added:
field Qwen2TokenizerFast.model_input_names - added:
field Qwen2TokenizerFast.slow_tokenizer_class - added:
field Qwen2TokenizerFast.vocab_files_names - added:
method Qwen2TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class Qwen2_5OmniConfig(PretrainedConfig) - added:
class Qwen2_5OmniTalkerConfig(PretrainedConfig) - added:
class Qwen2_5OmniThinkerConfig(PretrainedConfig) - added:
class Qwen2_5OmniToken2WavConfig(PretrainedConfig) - added:
constructor Qwen2_5OmniConfig(thinker_config=…, talker_config=…, token2wav_config=…, enable_audio_output: bool=…, **kwargs) - added:
constructor Qwen2_5OmniTalkerConfig(audio_token_index=…, image_token_index=…, video_token_index=…, vocab_size=…, tts_text_start_token_id=…, tts_text_end_token_id=…, tts_text_pad_token_id=…, tts_codec_start_token_id=…, tts_codec_end_token_id=…, tts_codec_pad_token_id=…, tts_codec_mask_token_id=…, vision_start_token_id=…, vision_end_token_id=…, embedding_size=…, hidden_size=…, intermediate_size=…,… - added:
constructor Qwen2_5OmniThinkerConfig(audio_config=…, vision_config=…, text_config=…, audio_token_index=…, image_token_index=…, video_token_index=…, position_id_per_seconds=…, seconds_per_chunk=…, audio_start_token_id=…, audio_end_token_id=…, user_token_id=…, initializer_range=…, **kwargs) - added:
constructor Qwen2_5OmniToken2WavConfig(dit_config=…, bigvgan_config=…, **kwargs) - added:
field Qwen2_5OmniConfig.model_type - added:
field Qwen2_5OmniConfig.sub_configs - added:
field Qwen2_5OmniTalkerConfig.attribute_map - added:
field Qwen2_5OmniTalkerConfig.model_type - added:
field Qwen2_5OmniThinkerConfig.attribute_map - added:
field Qwen2_5OmniThinkerConfig.model_type - added:
field Qwen2_5OmniThinkerConfig.sub_configs - added:
field Qwen2_5OmniToken2WavConfig.model_type - added:
field Qwen2_5OmniToken2WavConfig.sub_configs - added:
method Qwen2_5OmniConfig.get_text_config(*args, **kwargs)
- added:
class Qwen2_5OmniForConditionalGeneration(Qwen2_5OmniPreTrainedModel, GenerationMixin) - added:
class Qwen2_5OmniPreTrainedModel(PreTrainedModel) - added:
class Qwen2_5OmniPreTrainedModelForConditionalGeneration(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniTalkerForConditionalGeneration(Qwen2_5OmniPreTrainedModelForConditionalGeneration, GenerationMixin) - added:
class Qwen2_5OmniTalkerModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniThinkerForConditionalGeneration(Qwen2_5OmniPreTrainedModelForConditionalGeneration, GenerationMixin) - added:
class Qwen2_5OmniThinkerTextModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniToken2WavBigVGANModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniToken2WavDiTModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniToken2WavModel(Qwen2_5OmniPreTrainedModel) - added:
constructor Qwen2_5OmniForConditionalGeneration(config) - added:
constructor Qwen2_5OmniTalkerForConditionalGeneration(config: Qwen2_5OmniTalkerConfig) - added:
constructor Qwen2_5OmniTalkerModel(config: Qwen2_5OmniTalkerConfig) - added:
constructor Qwen2_5OmniThinkerForConditionalGeneration(config: Qwen2_5OmniThinkerConfig) - added:
constructor Qwen2_5OmniThinkerTextModel(config: Qwen2_5OmniTextConfig) - added:
constructor Qwen2_5OmniToken2WavBigVGANModel(config: Qwen2_5OmniBigVGANConfig) - added:
constructor Qwen2_5OmniToken2WavDiTModel(config: Qwen2_5OmniDiTConfig) - added:
constructor Qwen2_5OmniToken2WavModel(config: Qwen2_5OmniToken2WavConfig) - added:
field Qwen2_5OmniForConditionalGeneration.config: Qwen2_5OmniConfig - added:
field Qwen2_5OmniPreTrainedModel.base_model_prefix - added:
field Qwen2_5OmniPreTrainedModel.config: Qwen2_5OmniConfig - added:
field Qwen2_5OmniPreTrainedModel.supports_gradient_checkpointing - added:
field Qwen2_5OmniTalkerForConditionalGeneration.base_model_prefix - added:
field Qwen2_5OmniTalkerForConditionalGeneration.config: Qwen2_5OmniTalkerConfig - added:
field Qwen2_5OmniTalkerModel.config: Qwen2_5OmniTalkerConfig - added:
field Qwen2_5OmniThinkerForConditionalGeneration.base_model_prefix - added:
field Qwen2_5OmniThinkerForConditionalGeneration.config: Qwen2_5OmniThinkerConfig - added:
field Qwen2_5OmniThinkerTextModel.config: Qwen2_5OmniTextConfig - added:
field Qwen2_5OmniToken2WavBigVGANModel.config: Qwen2_5OmniBigVGANConfig - added:
field Qwen2_5OmniToken2WavDiTModel.config: Qwen2_5OmniDiTConfig - added:
field Qwen2_5OmniToken2WavModel.base_model_prefix - added:
field Qwen2_5OmniToken2WavModel.config: Qwen2_5OmniToken2WavConfig - added:
method Qwen2_5OmniForConditionalGeneration.disable_talker() - added:
method Qwen2_5OmniForConditionalGeneration.enable_talker() - added:
method Qwen2_5OmniForConditionalGeneration.generate(input_ids: Optional[torch.Tensor]=…, speaker: str=…, use_audio_in_video: bool=…, return_audio: Optional[bool]=…, thinker_max_new_tokens: int=…, talker_max_new_tokens: int=…, talker_do_sample: bool=…, talker_top_k: int=…, talker_top_p: float=…, talker_temperature: float=…, talker_eos_token_id: list[int]=…, talker_repetition_penalty: float=…, **kw… - added:
method Qwen2_5OmniForConditionalGeneration.load_speakers(path) - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_chunked_index(token_indices: torch.Tensor, tokens_per_chunk: int, remove_index: int) -> list[tuple[int, int]] - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[int], grid_hs: list[int], grid_ws: list[int]) - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tenso… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, thinker_reply_part: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, rope_deltas: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_posi… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen2_5OmniTalkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, input_text_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, thinker_reply_part=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, input_audio_features=…, audio_feature_attention_mask=…, audio_feature_lengths=…, use_audio_in_v… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2_5OmniTalkerModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position:… - added:
method Qwen2_5OmniThinkerForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, feature_attention_mask… - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor, feature_attention_mask: Optional[torch.LongTensor]=…, audio_feature_lengths: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, input_features=…, feature_attention_mask=…, use_audio_in_video=…, video_second_per_grid=…, **kwargs) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2_5OmniThinkerTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_posi… - added:
method Qwen2_5OmniToken2WavBigVGANModel.amplitude_to_db(amplitude, min_db_level) - added:
method Qwen2_5OmniToken2WavBigVGANModel.forward(mel_spectrogram) - added:
method Qwen2_5OmniToken2WavBigVGANModel.normalize_spectrogram(spectrogram, max_value, min_db) - added:
method Qwen2_5OmniToken2WavBigVGANModel.process_mel_spectrogram(mel_spectrogram) - added:
method Qwen2_5OmniToken2WavDiTModel.forward(hidden_states, condition_vector, speaker_embedding, quantized_code, time_step, drop_audio_conditioning=…, drop_code=…, apply_cfg=…) - added:
method Qwen2_5OmniToken2WavDiTModel.sample(conditioning_vector, reference_mel_spectrogram, quantized_code, num_steps=…, guidance_scale=…, sway_coefficient=…) - added:
method Qwen2_5OmniToken2WavModel.forward(code, conditioning, reference_mel, num_steps=…, guidance_scale=…, sway_coefficient=…, **kwargs) - added:
static method Qwen2_5OmniForConditionalGeneration.from_pretrained(pretrained_model_name_or_path, *model_args, config=…, cache_dir=…, ignore_mismatched_sizes=…, force_download=…, local_files_only=…, token=…, revision=…, use_safetensors=…, weights_only=…, **kwargs)
- added:
class Qwen2_5OmniConfig(PretrainedConfig) - added:
class Qwen2_5OmniForConditionalGeneration(Qwen2_5OmniPreTrainedModel, GenerationMixin) - added:
class Qwen2_5OmniPreTrainedModel(Qwen2_5_VLPreTrainedModel) - added:
class Qwen2_5OmniPreTrainedModelForConditionalGeneration(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniTalkerConfig(PretrainedConfig) - added:
class Qwen2_5OmniTalkerForConditionalGeneration(Qwen2_5OmniPreTrainedModelForConditionalGeneration, GenerationMixin) - added:
class Qwen2_5OmniTalkerModel(Qwen2_5_VLTextModel) - added:
class Qwen2_5OmniThinkerConfig(PretrainedConfig) - added:
class Qwen2_5OmniThinkerForConditionalGeneration(Qwen2_5OmniPreTrainedModelForConditionalGeneration, GenerationMixin) - added:
class Qwen2_5OmniThinkerTextModel(Qwen2_5_VLTextModel) - added:
class Qwen2_5OmniToken2WavBigVGANModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniToken2WavConfig(PretrainedConfig) - added:
class Qwen2_5OmniToken2WavDiTModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen2_5OmniToken2WavModel(Qwen2_5OmniPreTrainedModel) - added:
constructor Qwen2_5OmniConfig(thinker_config=…, talker_config=…, token2wav_config=…, enable_audio_output: bool=…, **kwargs) - added:
constructor Qwen2_5OmniForConditionalGeneration(config) - added:
constructor Qwen2_5OmniTalkerConfig(audio_token_index=…, image_token_index=…, video_token_index=…, vocab_size=…, tts_text_start_token_id=…, tts_text_end_token_id=…, tts_text_pad_token_id=…, tts_codec_start_token_id=…, tts_codec_end_token_id=…, tts_codec_pad_token_id=…, tts_codec_mask_token_id=…, vision_start_token_id=…, vision_end_token_id=…, embedding_size=…, hidden_size=…, intermediate_size=…,… - added:
constructor Qwen2_5OmniTalkerForConditionalGeneration(config: Qwen2_5OmniTalkerConfig) - added:
constructor Qwen2_5OmniTalkerModel(config: Qwen2_5OmniTalkerConfig) - added:
constructor Qwen2_5OmniThinkerConfig(audio_config=…, vision_config=…, text_config=…, audio_token_index=…, image_token_index=…, video_token_index=…, position_id_per_seconds=…, seconds_per_chunk=…, audio_start_token_id=…, audio_end_token_id=…, user_token_id=…, initializer_range=…, **kwargs) - added:
constructor Qwen2_5OmniThinkerForConditionalGeneration(config: Qwen2_5OmniThinkerConfig) - added:
constructor Qwen2_5OmniThinkerTextModel(config: Qwen2_5OmniTextConfig) - added:
constructor Qwen2_5OmniToken2WavBigVGANModel(config: Qwen2_5OmniBigVGANConfig) - added:
constructor Qwen2_5OmniToken2WavConfig(dit_config=…, bigvgan_config=…, **kwargs) - added:
constructor Qwen2_5OmniToken2WavDiTModel(config: Qwen2_5OmniDiTConfig) - added:
constructor Qwen2_5OmniToken2WavModel(config: Qwen2_5OmniToken2WavConfig) - added:
field Qwen2_5OmniConfig.model_type - added:
field Qwen2_5OmniConfig.sub_configs - added:
field Qwen2_5OmniForConditionalGeneration.config: Qwen2_5OmniConfig - added:
field Qwen2_5OmniPreTrainedModel.config: Qwen2_5OmniConfig - added:
field Qwen2_5OmniTalkerConfig.attribute_map - added:
field Qwen2_5OmniTalkerConfig.model_type - added:
field Qwen2_5OmniTalkerForConditionalGeneration.base_model_prefix - added:
field Qwen2_5OmniTalkerForConditionalGeneration.config: Qwen2_5OmniTalkerConfig - added:
field Qwen2_5OmniTalkerModel.config: Qwen2_5OmniTalkerConfig - added:
field Qwen2_5OmniThinkerConfig.attribute_map - added:
field Qwen2_5OmniThinkerConfig.model_type - added:
field Qwen2_5OmniThinkerConfig.sub_configs - added:
field Qwen2_5OmniThinkerForConditionalGeneration.base_model_prefix - added:
field Qwen2_5OmniThinkerForConditionalGeneration.config: Qwen2_5OmniThinkerConfig - added:
field Qwen2_5OmniThinkerTextModel.config: Qwen2_5OmniTextConfig - added:
field Qwen2_5OmniToken2WavBigVGANModel.config: Qwen2_5OmniBigVGANConfig - added:
field Qwen2_5OmniToken2WavConfig.model_type - added:
field Qwen2_5OmniToken2WavConfig.sub_configs - added:
field Qwen2_5OmniToken2WavDiTModel.config: Qwen2_5OmniDiTConfig - added:
field Qwen2_5OmniToken2WavModel.base_model_prefix - added:
field Qwen2_5OmniToken2WavModel.config: Qwen2_5OmniToken2WavConfig - added:
method Qwen2_5OmniConfig.get_text_config(*args, **kwargs) - added:
method Qwen2_5OmniForConditionalGeneration.disable_talker() - added:
method Qwen2_5OmniForConditionalGeneration.enable_talker() - added:
method Qwen2_5OmniForConditionalGeneration.generate(input_ids: Optional[torch.Tensor]=…, speaker: str=…, use_audio_in_video: bool=…, return_audio: Optional[bool]=…, thinker_max_new_tokens: int=…, talker_max_new_tokens: int=…, talker_do_sample: bool=…, talker_top_k: int=…, talker_top_p: float=…, talker_temperature: float=…, talker_eos_token_id: list[int]=…, talker_repetition_penalty: float=…, **kw… - added:
method Qwen2_5OmniForConditionalGeneration.load_speakers(path) - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_chunked_index(token_indices: torch.Tensor, tokens_per_chunk: int, remove_index: int) -> list[tuple[int, int]] - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[int], grid_hs: list[int], grid_ws: list[int]) - added:
method Qwen2_5OmniPreTrainedModelForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tenso… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, thinker_reply_part: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, rope_deltas: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_posi… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen2_5OmniTalkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, input_text_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, thinker_reply_part=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, input_audio_features=…, audio_feature_attention_mask=…, audio_feature_lengths=…, use_audio_in_v… - added:
method Qwen2_5OmniTalkerForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, feature_attention_mask… - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor, feature_attention_mask: Optional[torch.LongTensor]=…, audio_feature_lengths: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, input_features=…, feature_attention_mask=…, use_audio_in_video=…, video_second_per_grid=…, **kwargs) - added:
method Qwen2_5OmniThinkerForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2_5OmniToken2WavBigVGANModel.amplitude_to_db(amplitude, min_db_level) - added:
method Qwen2_5OmniToken2WavBigVGANModel.forward(mel_spectrogram) - added:
method Qwen2_5OmniToken2WavBigVGANModel.normalize_spectrogram(spectrogram, max_value, min_db) - added:
method Qwen2_5OmniToken2WavBigVGANModel.process_mel_spectrogram(mel_spectrogram) - added:
method Qwen2_5OmniToken2WavDiTModel.forward(hidden_states, condition_vector, speaker_embedding, quantized_code, time_step, drop_audio_conditioning=…, drop_code=…, apply_cfg=…) - added:
method Qwen2_5OmniToken2WavDiTModel.sample(conditioning_vector, reference_mel_spectrogram, quantized_code, num_steps=…, guidance_scale=…, sway_coefficient=…) - added:
method Qwen2_5OmniToken2WavModel.forward(code, conditioning, reference_mel, num_steps=…, guidance_scale=…, sway_coefficient=…, **kwargs) - added:
static method Qwen2_5OmniForConditionalGeneration.from_pretrained(pretrained_model_name_or_path, *model_args, config=…, cache_dir=…, ignore_mismatched_sizes=…, force_download=…, local_files_only=…, token=…, revision=…, use_safetensors=…, weights_only=…, **kwargs)
- added:
class Qwen2_5OmniProcessor(ProcessorMixin) - added:
constructor Qwen2_5OmniProcessor(image_processor=…, video_processor=…, feature_extractor=…, tokenizer=…, chat_template=…) - added:
field Qwen2_5OmniProcessor.attributes - added:
field Qwen2_5OmniProcessor.feature_extractor_class - added:
field Qwen2_5OmniProcessor.image_processor_class - added:
field Qwen2_5OmniProcessor.tokenizer_class - added:
field Qwen2_5OmniProcessor.video_processor_class - added:
method Qwen2_5OmniProcessor.apply_chat_template(conversations, chat_template=…, **kwargs) - added:
method Qwen2_5OmniProcessor.get_chunked_index(token_indices: np.ndarray, tokens_per_chunk: int) -> list[tuple[int, int]] - added:
method Qwen2_5OmniProcessor.replace_multimodal_special_tokens(text, audio_lengths, image_grid_thw, video_grid_thw, video_second_per_grid, use_audio_in_video, position_id_per_seconds, seconds_per_chunk) - added:
property Qwen2_5OmniProcessor.model_input_names
- added:
export define_import_structure
- added:
class Qwen2_5_VLConfig(PretrainedConfig) - added:
class Qwen2_5_VLTextConfig(PretrainedConfig) - added:
constructor Qwen2_5_VLConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, vision_end_token_id=…, **kwargs) - added:
constructor Qwen2_5_VLTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, use_sliding_window=…, sliding_window=…, max_window_layers=…, layer_types=…, attention_dropout=…, rope_scaling=…, **kwargs) - added:
field Qwen2_5_VLConfig.keys_to_ignore_at_inference - added:
field Qwen2_5_VLConfig.model_type - added:
field Qwen2_5_VLConfig.sub_configs - added:
field Qwen2_5_VLTextConfig.base_config_key - added:
field Qwen2_5_VLTextConfig.base_model_pp_plan - added:
field Qwen2_5_VLTextConfig.base_model_tp_plan - added:
field Qwen2_5_VLTextConfig.keys_to_ignore_at_inference - added:
field Qwen2_5_VLTextConfig.model_type
- added:
class Qwen2_5_VLForConditionalGeneration(Qwen2_5_VLPreTrainedModel, GenerationMixin) - added:
class Qwen2_5_VLModel(Qwen2_5_VLPreTrainedModel) - added:
class Qwen2_5_VLPreTrainedModel(PreTrainedModel) - added:
class Qwen2_5_VLTextModel(Qwen2_5_VLPreTrainedModel) - added:
constructor Qwen2_5_VLForConditionalGeneration(config) - added:
constructor Qwen2_5_VLModel(config) - added:
constructor Qwen2_5_VLTextModel(config: Qwen2_5_VLTextConfig) - added:
field Qwen2_5_VLForConditionalGeneration.accepts_loss_kwargs - added:
field Qwen2_5_VLModel.accepts_loss_kwargs - added:
field Qwen2_5_VLModel.base_model_prefix - added:
field Qwen2_5_VLModel.config: Qwen2_5_VLConfig - added:
field Qwen2_5_VLPreTrainedModel.base_model_prefix - added:
field Qwen2_5_VLPreTrainedModel.config: Qwen2_5_VLConfig - added:
field Qwen2_5_VLPreTrainedModel.supports_gradient_checkpointing - added:
field Qwen2_5_VLTextModel.config: Qwen2_5_VLTextConfig - added:
method Qwen2_5_VLForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]… - added:
method Qwen2_5_VLForConditionalGeneration.get_decoder() - added:
method Qwen2_5_VLForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5_VLForConditionalGeneration.get_input_embeddings() - added:
method Qwen2_5_VLForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5_VLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, second_per_grid_ts=…, **kwargs) - added:
method Qwen2_5_VLForConditionalGeneration.set_decoder(decoder) - added:
method Qwen2_5_VLForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2_5_VLModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, pixel_values: Optional… - added:
method Qwen2_5_VLModel.get_decoder() - added:
method Qwen2_5_VLModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5_VLModel.get_input_embeddings() - added:
method Qwen2_5_VLModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen2_5_VLModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, second_per_grid_ts: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen2_5_VLModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2_5_VLModel.set_decoder(decoder) - added:
method Qwen2_5_VLModel.set_input_embeddings(value) - added:
method Qwen2_5_VLTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Op… - added:
property Qwen2_5_VLForConditionalGeneration.language_model - added:
property Qwen2_5_VLForConditionalGeneration.visual
- added:
class Qwen2_5_VLConfig(Qwen2VLConfig) - added:
class Qwen2_5_VLForConditionalGeneration(Qwen2VLForConditionalGeneration) - added:
class Qwen2_5_VLModel(Qwen2VLModel) - added:
class Qwen2_5_VLPreTrainedModel(Qwen2VLPreTrainedModel) - added:
class Qwen2_5_VLProcessor(Qwen2VLProcessor) - added:
class Qwen2_5_VLTextConfig(Qwen2VLTextConfig) - added:
constructor Qwen2_5_VLModel(config) - added:
field Qwen2_5_VLConfig.model_type - added:
field Qwen2_5_VLConfig.sub_configs - added:
field Qwen2_5_VLForConditionalGeneration.accepts_loss_kwargs - added:
field Qwen2_5_VLModel.accepts_loss_kwargs - added:
field Qwen2_5_VLModel.base_model_prefix - added:
field Qwen2_5_VLModel.config: Qwen2_5_VLConfig - added:
field Qwen2_5_VLProcessor.image_processor_class - added:
field Qwen2_5_VLTextConfig.model_type - added:
method Qwen2_5_VLForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]… - added:
method Qwen2_5_VLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, second_per_grid_ts=…, **kwargs) - added:
method Qwen2_5_VLModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, pixel_values: Optional… - added:
method Qwen2_5_VLModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, second_per_grid_ts: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
property Qwen2_5_VLProcessor.model_input_names
- added:
class Qwen2_5_VLProcessor(ProcessorMixin) - added:
constructor Qwen2_5_VLProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
field Qwen2_5_VLProcessor.attributes - added:
field Qwen2_5_VLProcessor.image_processor_class - added:
field Qwen2_5_VLProcessor.tokenizer_class - added:
field Qwen2_5_VLProcessor.video_processor_class - added:
method Qwen2_5_VLProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs) - added:
property Qwen2_5_VLProcessor.model_input_names
- added:
export define_import_structure
- added:
class Qwen2AudioConfig(PretrainedConfig) - added:
class Qwen2AudioEncoderConfig(PretrainedConfig) - added:
constructor Qwen2AudioConfig(audio_config=…, text_config=…, audio_token_index=…, **kwargs) - added:
constructor Qwen2AudioEncoderConfig(num_mel_bins=…, encoder_layers=…, encoder_attention_heads=…, encoder_ffn_dim=…, encoder_layerdrop=…, d_model=…, dropout=…, attention_dropout=…, activation_function=…, activation_dropout=…, scale_embedding=…, initializer_range=…, max_source_positions=…, **kwargs) - added:
field Qwen2AudioConfig.attribute_map - added:
field Qwen2AudioConfig.model_type - added:
field Qwen2AudioConfig.sub_configs - added:
field Qwen2AudioEncoderConfig.model_type
- added:
class Qwen2AudioEncoder(Qwen2AudioPreTrainedModel) - added:
class Qwen2AudioForConditionalGeneration(Qwen2AudioPreTrainedModel, GenerationMixin) - added:
class Qwen2AudioPreTrainedModel(PreTrainedModel) - added:
constructor Qwen2AudioEncoder(config: Qwen2AudioEncoderConfig) - added:
constructor Qwen2AudioForConditionalGeneration(config: Qwen2AudioConfig) - added:
field Qwen2AudioEncoder.config: Qwen2AudioEncoderConfig - added:
field Qwen2AudioEncoder.main_input_name - added:
field Qwen2AudioPreTrainedModel.base_model_prefix - added:
field Qwen2AudioPreTrainedModel.config: Qwen2AudioConfig - added:
field Qwen2AudioPreTrainedModel.supports_gradient_checkpointing - added:
method Qwen2AudioEncoder.forward(input_features, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method Qwen2AudioEncoder.get_input_embeddings() -> nn.Module - added:
method Qwen2AudioEncoder.set_input_embeddings(value: nn.Module) - added:
method Qwen2AudioForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, feature_attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_c… - added:
method Qwen2AudioForConditionalGeneration.get_decoder() - added:
method Qwen2AudioForConditionalGeneration.get_input_embeddings() - added:
method Qwen2AudioForConditionalGeneration.get_output_embeddings() - added:
method Qwen2AudioForConditionalGeneration.prepare_inputs_for_generation(*args, **kwargs) - added:
method Qwen2AudioForConditionalGeneration.set_decoder(decoder) - added:
method Qwen2AudioForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2AudioForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
property Qwen2AudioForConditionalGeneration.padding_side
- added:
class Qwen2AudioProcessor(ProcessorMixin) - added:
constructor Qwen2AudioProcessor(feature_extractor=…, tokenizer=…, chat_template=…, audio_token=…, audio_bos_token=…, audio_eos_token=…) - added:
field Qwen2AudioProcessor.attributes - added:
field Qwen2AudioProcessor.feature_extractor_class - added:
field Qwen2AudioProcessor.tokenizer_class - added:
property Qwen2AudioProcessor.default_chat_template - added:
property Qwen2AudioProcessor.model_input_names
- added:
export define_import_structure
- added:
class Qwen2MoeConfig(PretrainedConfig) - added:
constructor Qwen2MoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, use_sliding_window=…, sliding_window=…, max_window_layers=…, attention_dropout=…, decoder_sparse_step=…, moe_in… - added:
field Qwen2MoeConfig.base_model_pp_plan - added:
field Qwen2MoeConfig.base_model_tp_plan - added:
field Qwen2MoeConfig.keys_to_ignore_at_inference - added:
field Qwen2MoeConfig.model_type
- added:
class Qwen2MoeForCausalLM(Qwen2MoePreTrainedModel, GenerationMixin) - added:
class Qwen2MoeForQuestionAnswering(GenericForQuestionAnswering, Qwen2MoePreTrainedModel) - added:
class Qwen2MoeForSequenceClassification(GenericForSequenceClassification, Qwen2MoePreTrainedModel) - added:
class Qwen2MoeForTokenClassification(GenericForTokenClassification, Qwen2MoePreTrainedModel) - added:
class Qwen2MoeModel(Qwen2MoePreTrainedModel) - added:
class Qwen2MoePreTrainedModel(PreTrainedModel) - added:
constructor Qwen2MoeForCausalLM(config) - added:
constructor Qwen2MoeModel(config: Qwen2MoeConfig) - added:
field Qwen2MoePreTrainedModel.base_model_prefix - added:
field Qwen2MoePreTrainedModel.config: Qwen2MoeConfig - added:
field Qwen2MoePreTrainedModel.supports_gradient_checkpointing - added:
method Qwen2MoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_rout… - added:
method Qwen2MoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position:…
- added:
export define_import_structure
- added:
class Qwen2VLConfig(PretrainedConfig) - added:
class Qwen2VLTextConfig(PretrainedConfig) - added:
constructor Qwen2VLConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, vision_end_token_id=…, **kwargs) - added:
constructor Qwen2VLTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, use_sliding_window=…, sliding_window=…, max_window_layers=…, layer_types=…, attention_dropout=…, rope_scaling=…, **kwargs) - added:
field Qwen2VLConfig.keys_to_ignore_at_inference - added:
field Qwen2VLConfig.model_type - added:
field Qwen2VLConfig.sub_configs - added:
field Qwen2VLTextConfig.base_config_key - added:
field Qwen2VLTextConfig.base_model_pp_plan - added:
field Qwen2VLTextConfig.base_model_tp_plan - added:
field Qwen2VLTextConfig.keys_to_ignore_at_inference - added:
field Qwen2VLTextConfig.model_type
- added:
class Qwen2VLImageProcessor(BaseImageProcessor) - added:
constructor Qwen2VLImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, min_pixels: Optional[int]=…, max_pixels: Optional[int]=…, patch_size: int=…,… - added:
field Qwen2VLImageProcessor.model_input_names - added:
method Qwen2VLImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Qwen2VLImageProcessor.preprocess(images: ImageInput, videos: Optional[VideoInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, min_pixels: Optional[int]=…, max_pixels: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]…
- added:
class Qwen2VLImageProcessorFast(BaseImageProcessorFast) - added:
constructor Qwen2VLImageProcessorFast(**kwargs: Unpack[Qwen2VLFastImageProcessorKwargs]) - added:
field Qwen2VLImageProcessorFast.do_convert_rgb - added:
field Qwen2VLImageProcessorFast.do_normalize - added:
field Qwen2VLImageProcessorFast.do_rescale - added:
field Qwen2VLImageProcessorFast.do_resize - added:
field Qwen2VLImageProcessorFast.image_mean - added:
field Qwen2VLImageProcessorFast.image_std - added:
field Qwen2VLImageProcessorFast.max_pixels - added:
field Qwen2VLImageProcessorFast.merge_size - added:
field Qwen2VLImageProcessorFast.min_pixels - added:
field Qwen2VLImageProcessorFast.model_input_names - added:
field Qwen2VLImageProcessorFast.patch_size - added:
field Qwen2VLImageProcessorFast.resample - added:
field Qwen2VLImageProcessorFast.size - added:
field Qwen2VLImageProcessorFast.temporal_patch_size - added:
field Qwen2VLImageProcessorFast.valid_kwargs - added:
method Qwen2VLImageProcessorFast.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method Qwen2VLImageProcessorFast.preprocess(images: ImageInput, videos: Optional[VideoInput]=…, **kwargs: Unpack[Qwen2VLFastImageProcessorKwargs]) -> BatchFeature
- added:
class Qwen2VLForConditionalGeneration(Qwen2VLPreTrainedModel, GenerationMixin) - added:
class Qwen2VLModel(Qwen2VLPreTrainedModel) - added:
class Qwen2VLPreTrainedModel(PreTrainedModel) - added:
class Qwen2VLTextModel(Qwen2VLPreTrainedModel) - added:
constructor Qwen2VLForConditionalGeneration(config) - added:
constructor Qwen2VLModel(config: Qwen2VLConfig) - added:
constructor Qwen2VLTextModel(config: Qwen2VLTextConfig) - added:
field Qwen2VLModel.accepts_loss_kwargs - added:
field Qwen2VLModel.base_model_prefix - added:
field Qwen2VLPreTrainedModel.base_model_prefix - added:
field Qwen2VLPreTrainedModel.config: Qwen2VLConfig - added:
field Qwen2VLPreTrainedModel.supports_gradient_checkpointing - added:
field Qwen2VLTextModel.config: Qwen2VLTextConfig - added:
method Qwen2VLForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…,… - added:
method Qwen2VLForConditionalGeneration.get_decoder() - added:
method Qwen2VLForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2VLForConditionalGeneration.get_input_embeddings() - added:
method Qwen2VLForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2VLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Qwen2VLForConditionalGeneration.set_decoder(decoder) - added:
method Qwen2VLForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen2VLModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, pixel_values: Optional[to… - added:
method Qwen2VLModel.get_decoder() - added:
method Qwen2VLModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2VLModel.get_input_embeddings() - added:
method Qwen2VLModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen2VLModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen2VLModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen2VLModel.set_decoder(decoder) - added:
method Qwen2VLModel.set_input_embeddings(value) - added:
method Qwen2VLTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_position: Optio… - added:
property Qwen2VLForConditionalGeneration.language_model - added:
property Qwen2VLForConditionalGeneration.visual
- added:
class Qwen2VLProcessor(ProcessorMixin) - added:
constructor Qwen2VLProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
field Qwen2VLProcessor.attributes - added:
field Qwen2VLProcessor.image_processor_class - added:
field Qwen2VLProcessor.tokenizer_class - added:
field Qwen2VLProcessor.video_processor_class - added:
method Qwen2VLProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs)
- added:
class Qwen2VLVideoProcessor(BaseVideoProcessor) - added:
constructor Qwen2VLVideoProcessor(**kwargs: Unpack[Qwen2VLVideoProcessorInitKwargs]) - added:
field Qwen2VLVideoProcessor.do_convert_rgb - added:
field Qwen2VLVideoProcessor.do_normalize - added:
field Qwen2VLVideoProcessor.do_rescale - added:
field Qwen2VLVideoProcessor.do_resize - added:
field Qwen2VLVideoProcessor.do_sample_frames - added:
field Qwen2VLVideoProcessor.image_mean - added:
field Qwen2VLVideoProcessor.image_std - added:
field Qwen2VLVideoProcessor.max_frames - added:
field Qwen2VLVideoProcessor.max_pixels - added:
field Qwen2VLVideoProcessor.merge_size - added:
field Qwen2VLVideoProcessor.min_frames - added:
field Qwen2VLVideoProcessor.min_pixels - added:
field Qwen2VLVideoProcessor.model_input_names - added:
field Qwen2VLVideoProcessor.patch_size - added:
field Qwen2VLVideoProcessor.resample - added:
field Qwen2VLVideoProcessor.size - added:
field Qwen2VLVideoProcessor.temporal_patch_size - added:
field Qwen2VLVideoProcessor.valid_kwargs - added:
method Qwen2VLVideoProcessor.get_num_of_video_patches(num_frames: int, height: int, width: int, videos_kwargs=…) - added:
method Qwen2VLVideoProcessor.sample_frames(metadata: VideoMetadata, temporal_patch_size: Optional[int]=…, min_frames: Optional[int]=…, max_frames: Optional[int]=…, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, **kwargs)
- added:
export define_import_structure
- added:
class Qwen3Config(PretrainedConfig) - added:
constructor Qwen3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, use_sliding_window=…, sliding_window=…, max_window_layers=…, layer_types=…, attentio… - added:
field Qwen3Config.base_model_pp_plan - added:
field Qwen3Config.base_model_tp_plan - added:
field Qwen3Config.keys_to_ignore_at_inference - added:
field Qwen3Config.model_type
- added:
class Qwen3ForCausalLM(Qwen3PreTrainedModel, GenerationMixin) - added:
class Qwen3ForQuestionAnswering(GenericForQuestionAnswering, Qwen3PreTrainedModel) - added:
class Qwen3ForSequenceClassification(GenericForSequenceClassification, Qwen3PreTrainedModel) - added:
class Qwen3ForTokenClassification(GenericForTokenClassification, Qwen3PreTrainedModel) - added:
class Qwen3Model(Qwen3PreTrainedModel) - added:
class Qwen3PreTrainedModel(PreTrainedModel) - added:
constructor Qwen3ForCausalLM(config) - added:
constructor Qwen3Model(config: Qwen3Config) - added:
field Qwen3ForQuestionAnswering.base_model_prefix - added:
field Qwen3PreTrainedModel.base_model_prefix - added:
field Qwen3PreTrainedModel.config: Qwen3Config - added:
field Qwen3PreTrainedModel.supports_gradient_checkpointing - added:
method Qwen3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…, *… - added:
method Qwen3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class Qwen3ForCausalLM(Qwen2ForCausalLM) - added:
class Qwen3ForQuestionAnswering(Qwen2ForQuestionAnswering) - added:
class Qwen3ForSequenceClassification(Qwen2ForSequenceClassification) - added:
class Qwen3ForTokenClassification(Qwen2ForTokenClassification) - added:
class Qwen3Model(Qwen2Model) - added:
class Qwen3PreTrainedModel(Qwen2PreTrainedModel) - added:
method Qwen3ForCausalLM.forward(**super_kwargs: Unpack[TransformersKwargs]) -> CausalLMOutputWithPast
- added:
export define_import_structure
- added:
class Qwen3MoeConfig(PretrainedConfig) - added:
constructor Qwen3MoeConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, use_sliding_window=…, sliding_window=…, attention_dropout=…, decoder_sparse_step=…, moe_inter… - added:
field Qwen3MoeConfig.base_model_pp_plan - added:
field Qwen3MoeConfig.base_model_tp_plan - added:
field Qwen3MoeConfig.keys_to_ignore_at_inference - added:
field Qwen3MoeConfig.model_type
- added:
class Qwen3MoeForCausalLM(Qwen3MoePreTrainedModel, GenerationMixin) - added:
class Qwen3MoeForQuestionAnswering(GenericForQuestionAnswering, Qwen3MoePreTrainedModel) - added:
class Qwen3MoeForSequenceClassification(GenericForSequenceClassification, Qwen3MoePreTrainedModel) - added:
class Qwen3MoeForTokenClassification(GenericForTokenClassification, Qwen3MoePreTrainedModel) - added:
class Qwen3MoeModel(Qwen3MoePreTrainedModel) - added:
class Qwen3MoePreTrainedModel(PreTrainedModel) - added:
constructor Qwen3MoeForCausalLM(config) - added:
constructor Qwen3MoeModel(config: Qwen3MoeConfig) - added:
field Qwen3MoeForQuestionAnswering.base_model_prefix - added:
field Qwen3MoePreTrainedModel.base_model_prefix - added:
field Qwen3MoePreTrainedModel.config: Qwen3MoeConfig - added:
field Qwen3MoePreTrainedModel.supports_gradient_checkpointing - added:
method Qwen3MoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, lo… - added:
method Qwen3MoeModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class Qwen3MoeForCausalLM(MixtralForCausalLM) - added:
class Qwen3MoeForQuestionAnswering(LlamaForQuestionAnswering) - added:
class Qwen3MoeForSequenceClassification(LlamaForSequenceClassification) - added:
class Qwen3MoeForTokenClassification(LlamaForTokenClassification) - added:
class Qwen3MoeModel(MixtralModel) - added:
constructor Qwen3MoeForCausalLM(config) - added:
method Qwen3MoeForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, lo…
- added:
export define_import_structure
- added:
class Qwen3NextConfig(PretrainedConfig) - added:
constructor Qwen3NextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, partial_rotary_factor=…, attention_bias=…, attention_dropout=…, head_dim=…, linear_conv_kernel_dim=…, linear_k… - added:
field Qwen3NextConfig.base_model_pp_plan - added:
field Qwen3NextConfig.base_model_tp_plan - added:
field Qwen3NextConfig.keys_to_ignore_at_inference - added:
field Qwen3NextConfig.model_type
- added:
class Qwen3NextForCausalLM(Qwen3NextPreTrainedModel, GenerationMixin) - added:
class Qwen3NextForQuestionAnswering(GenericForQuestionAnswering, Qwen3NextPreTrainedModel) - added:
class Qwen3NextForSequenceClassification(GenericForSequenceClassification, Qwen3NextPreTrainedModel) - added:
class Qwen3NextForTokenClassification(GenericForTokenClassification, Qwen3NextPreTrainedModel) - added:
class Qwen3NextModel(Qwen3NextPreTrainedModel) - added:
class Qwen3NextPreTrainedModel(PreTrainedModel) - added:
constructor Qwen3NextForCausalLM(config) - added:
constructor Qwen3NextModel(config: Qwen3NextConfig) - added:
field Qwen3NextForQuestionAnswering.base_model_prefix - added:
field Qwen3NextPreTrainedModel.base_model_prefix - added:
field Qwen3NextPreTrainedModel.config: Qwen3NextConfig - added:
field Qwen3NextPreTrainedModel.supports_gradient_checkpointing - added:
method Qwen3NextForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Qwen3NextDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.… - added:
method Qwen3NextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
class Qwen3NextForCausalLM(MixtralForCausalLM) - added:
class Qwen3NextForQuestionAnswering(LlamaForQuestionAnswering) - added:
class Qwen3NextForSequenceClassification(LlamaForSequenceClassification) - added:
class Qwen3NextForTokenClassification(LlamaForTokenClassification) - added:
class Qwen3NextModel(Qwen3NextPreTrainedModel) - added:
class Qwen3NextPreTrainedModel(PreTrainedModel) - added:
constructor Qwen3NextForCausalLM(config) - added:
constructor Qwen3NextModel(config: Qwen3NextConfig) - added:
field Qwen3NextPreTrainedModel.base_model_prefix - added:
field Qwen3NextPreTrainedModel.config: Qwen3NextConfig - added:
field Qwen3NextPreTrainedModel.supports_gradient_checkpointing - added:
method Qwen3NextForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Qwen3NextDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_router_logits: Optional[bool]=…, cache_position: Optional[torch.… - added:
method Qwen3NextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> MoeModelOutputWithPast
- added:
export define_import_structure
- added:
class Qwen3OmniMoeConfig(PretrainedConfig) - added:
class Qwen3OmniMoeTalkerConfig(PretrainedConfig) - added:
class Qwen3OmniMoeThinkerConfig(PretrainedConfig) - added:
constructor Qwen3OmniMoeConfig(thinker_config=…, talker_config=…, code2wav_config=…, enable_audio_output=…, im_start_token_id=…, im_end_token_id=…, tts_pad_token_id=…, tts_bos_token_id=…, tts_eos_token_id=…, system_token_id=…, user_token_id=…, assistant_token_id=…, **kwargs) - added:
constructor Qwen3OmniMoeTalkerConfig(code_predictor_config=…, text_config=…, num_code_groups=…, thinker_hidden_size=…, codec_eos_token_id=…, accept_hidden_layer=…, codec_nothink_id=…, codec_think_bos_id=…, codec_think_eos_id=…, codec_pad_id=…, codec_bos_id=…, audio_token_id=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, position_id_per_seconds=…, audio_start_token_id=…, speaker_i… - added:
constructor Qwen3OmniMoeThinkerConfig(audio_config=…, vision_config=…, text_config=…, audio_token_id=…, image_token_id=…, video_token_id=…, position_id_per_seconds=…, audio_start_token_id=…, user_token_id=…, initializer_range=…, **kwargs) - added:
field Qwen3OmniMoeConfig.model_type - added:
field Qwen3OmniMoeConfig.sub_configs - added:
field Qwen3OmniMoeTalkerConfig.sub_configs - added:
field Qwen3OmniMoeThinkerConfig.attribute_map - added:
field Qwen3OmniMoeThinkerConfig.model_type - added:
field Qwen3OmniMoeThinkerConfig.sub_configs - added:
method Qwen3OmniMoeConfig.get_text_config(decoder=…) -> PretrainedConfig
- added:
class Qwen3OmniMoeCode2Wav(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeCode2WavDecoderBlock(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeCode2WavTransformerModel(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeForConditionalGeneration(Qwen3OmniMoePreTrainedModel, GenerationMixin) - added:
class Qwen3OmniMoePreTrainedModel(PreTrainedModel) - added:
class Qwen3OmniMoePreTrainedModelForConditionalGeneration(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeTalkerCodePredictorModel(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration(Qwen3OmniMoePreTrainedModel, GenerationMixin) - added:
class Qwen3OmniMoeTalkerForConditionalGeneration(Qwen3OmniMoeThinkerTextPreTrainedModel, GenerationMixin) - added:
class Qwen3OmniMoeTalkerModel(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeThinkerForConditionalGeneration(Qwen3OmniMoePreTrainedModelForConditionalGeneration, GenerationMixin) - added:
class Qwen3OmniMoeThinkerTextModel(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeThinkerTextPreTrainedModel(PreTrainedModel) - added:
constructor Qwen3OmniMoeCode2Wav(config: Qwen3OmniMoeCode2WavConfig) - added:
constructor Qwen3OmniMoeCode2WavDecoderBlock(config: Qwen3OmniMoeCode2WavConfig, layer_idx) - added:
constructor Qwen3OmniMoeCode2WavTransformerModel(config: Qwen3OmniMoeCode2WavConfig) - added:
constructor Qwen3OmniMoeForConditionalGeneration(config: Qwen3OmniMoeConfig) - added:
constructor Qwen3OmniMoeTalkerCodePredictorModel(config: Qwen3OmniMoeTalkerCodePredictorConfig) - added:
constructor Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration(config: Qwen3OmniMoeTalkerCodePredictorConfig) - added:
constructor Qwen3OmniMoeTalkerForConditionalGeneration(config: Qwen3OmniMoeTalkerConfig) - added:
constructor Qwen3OmniMoeTalkerModel(config: Qwen3OmniMoeTalkerTextConfig) - added:
constructor Qwen3OmniMoeThinkerForConditionalGeneration(config) - added:
constructor Qwen3OmniMoeThinkerTextModel(config: Qwen3OmniMoeTextConfig) - added:
field Qwen3OmniMoeForConditionalGeneration.config_class - added:
field Qwen3OmniMoePreTrainedModel.base_model_prefix - added:
field Qwen3OmniMoePreTrainedModel.config: Qwen3OmniMoeConfig - added:
field Qwen3OmniMoePreTrainedModel.supports_gradient_checkpointing - added:
field Qwen3OmniMoeTalkerCodePredictorModel.base_model_prefix - added:
field Qwen3OmniMoeTalkerCodePredictorModel.config_class - added:
field Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.base_model_prefix - added:
field Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.config_class - added:
field Qwen3OmniMoeTalkerForConditionalGeneration.base_model_prefix - added:
field Qwen3OmniMoeTalkerForConditionalGeneration.config_class - added:
field Qwen3OmniMoeTalkerModel.base_model_prefix - added:
field Qwen3OmniMoeTalkerModel.config: Qwen3OmniMoeTextConfig - added:
field Qwen3OmniMoeTalkerModel.config_class - added:
field Qwen3OmniMoeThinkerForConditionalGeneration.base_model_prefix - added:
field Qwen3OmniMoeThinkerForConditionalGeneration.config: Qwen3OmniMoeThinkerConfig - added:
field Qwen3OmniMoeThinkerTextModel.config: Qwen3OmniMoeTextConfig - added:
field Qwen3OmniMoeThinkerTextModel.config_class - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.base_model_prefix - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.config - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.config_class - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.supports_gradient_checkpointing - added:
method Qwen3OmniMoeCode2Wav.chunked_decode(codes, chunk_size=…, left_context_size=…) - added:
method Qwen3OmniMoeCode2Wav.forward(codes) - added:
method Qwen3OmniMoeCode2WavDecoderBlock.forward(hidden) - added:
method Qwen3OmniMoeCode2WavTransformerModel.forward(input_ids=…, attention_mask=…, position_ids=…, past_key_values=…, inputs_embeds=…, use_cache=…, cache_position=…, **kwargs) -> BaseModelOutputWithPast - added:
method Qwen3OmniMoeForConditionalGeneration.disable_talker() - added:
method Qwen3OmniMoeForConditionalGeneration.enable_talker() - added:
method Qwen3OmniMoeForConditionalGeneration.generate(input_ids: Optional[torch.Tensor]=…, speaker: str=…, use_audio_in_video: bool=…, return_audio: Optional[bool]=…, thinker_max_new_tokens: int=…, thinker_eos_token_id: int=…, talker_max_new_tokens: int=…, talker_do_sample: bool=…, talker_top_k: int=…, talker_top_p: float=…, talker_temperature: float=…, talker_repetition_penalty: float=…, **kwargs) - added:
method Qwen3OmniMoePreTrainedModelForConditionalGeneration.get_chunked_index(token_indices: torch.Tensor, tokens_per_chunk: int, remove_index: int) -> list[tuple[int, int]] - added:
method Qwen3OmniMoePreTrainedModelForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[torch.Tensor], grid_hs: list[torch.Tensor], grid_ws: list[torch.Tensor]) - added:
method Qwen3OmniMoePreTrainedModelForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tens… - added:
method Qwen3OmniMoeTalkerCodePredictorModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method Qwen3OmniMoeTalkerCodePredictorModel.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.forward(input_ids=…, attention_mask=…, position_ids=…, past_key_values=…, inputs_embeds=…, labels=…, use_cache=…, cache_position=…, generation_steps=…, **kwargs) -> CausalLMOutputWithPast - added:
method Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.forward(input_ids=…, attention_mask=…, use_audio_in_video=…, audio_feature_lengths=…, video_second_per_grid=…, image_grid_thw=…, video_grid_thw=…, position_ids=…, past_key_values=…, inputs_embeds=…, labels=…, use_cache=…, output_router_logits=…, cache_position=…, residual_codes=…, trailing_text_hidden=…, tts_pad_embed=…, generation_step=…, talker_… - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[torch.Tensor], grid_hs: list[torch.Tensor], grid_ws: list[torch.Tensor]) - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, **kwargs) - added:
method Qwen3OmniMoeTalkerModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, visual_pos_masks: Optional[torch.Tensor]=…, deepstack_visual_embeds: Optiona… - added:
method Qwen3OmniMoeTalkerModel.get_input_embeddings() - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.forward(input_ids=…, input_features=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, attention_mask=…, feature_attention_mask=…, audio_feature_lengths=…, position_ids=…, past_key_values=…, inputs_embeds=…, rope_deltas=…, labels=…, use_cache=…, output_router_logits: Optional[bool]=…, use_audio_in_video=…, cache_positio… - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor, feature_attention_mask: Optional[torch.LongTensor]=…, audio_feature_lengths: Optional[torch.LongTensor]=…) - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, input_features=…, feature_attention_mask=…, use_audio_in_video=…, video_second_per_grid=…, **kwargs) - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen3OmniMoeThinkerTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, visual_pos_masks: Optional[torch.Tensor]=…, deepstack_visual_embeds: Op…
- added:
class Qwen3OmniMoeCode2Wav(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeCode2WavDecoderBlock(Qwen3OmniMoePreTrainedModel) - added:
class Qwen3OmniMoeCode2WavTransformerModel(Qwen3Model) - added:
class Qwen3OmniMoeConfig(PretrainedConfig) - added:
class Qwen3OmniMoeForConditionalGeneration(Qwen3OmniMoePreTrainedModel, GenerationMixin) - added:
class Qwen3OmniMoePreTrainedModel(Qwen2_5OmniPreTrainedModel) - added:
class Qwen3OmniMoePreTrainedModelForConditionalGeneration(Qwen2_5OmniPreTrainedModelForConditionalGeneration) - added:
class Qwen3OmniMoeProcessor(Qwen2_5OmniProcessor, ProcessorMixin) - added:
class Qwen3OmniMoeTalkerCodePredictorModel(Qwen3Model) - added:
class Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration(Qwen3ForCausalLM) - added:
class Qwen3OmniMoeTalkerConfig(PretrainedConfig) - added:
class Qwen3OmniMoeTalkerForConditionalGeneration(Qwen3MoeForCausalLM) - added:
class Qwen3OmniMoeTalkerModel(Qwen3VLMoeTextModel) - added:
class Qwen3OmniMoeThinkerConfig(Qwen2_5OmniThinkerConfig) - added:
class Qwen3OmniMoeThinkerForConditionalGeneration(Qwen2_5OmniThinkerForConditionalGeneration) - added:
class Qwen3OmniMoeThinkerTextModel(Qwen3VLMoeTextModel) - added:
class Qwen3OmniMoeThinkerTextPreTrainedModel(Qwen3MoePreTrainedModel) - added:
constructor Qwen3OmniMoeCode2Wav(config: Qwen3OmniMoeCode2WavConfig) - added:
constructor Qwen3OmniMoeCode2WavDecoderBlock(config: Qwen3OmniMoeCode2WavConfig, layer_idx) - added:
constructor Qwen3OmniMoeCode2WavTransformerModel(config: Qwen3OmniMoeCode2WavConfig) - added:
constructor Qwen3OmniMoeConfig(thinker_config=…, talker_config=…, code2wav_config=…, enable_audio_output=…, im_start_token_id=…, im_end_token_id=…, tts_pad_token_id=…, tts_bos_token_id=…, tts_eos_token_id=…, system_token_id=…, user_token_id=…, assistant_token_id=…, **kwargs) - added:
constructor Qwen3OmniMoeForConditionalGeneration(config: Qwen3OmniMoeConfig) - added:
constructor Qwen3OmniMoeTalkerCodePredictorModel(config: Qwen3OmniMoeTalkerCodePredictorConfig) - added:
constructor Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration(config: Qwen3OmniMoeTalkerCodePredictorConfig) - added:
constructor Qwen3OmniMoeTalkerConfig(code_predictor_config=…, text_config=…, num_code_groups=…, thinker_hidden_size=…, codec_eos_token_id=…, accept_hidden_layer=…, codec_nothink_id=…, codec_think_bos_id=…, codec_think_eos_id=…, codec_pad_id=…, codec_bos_id=…, audio_token_id=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, position_id_per_seconds=…, audio_start_token_id=…, speaker_i… - added:
constructor Qwen3OmniMoeTalkerForConditionalGeneration(config: Qwen3OmniMoeTalkerConfig) - added:
constructor Qwen3OmniMoeTalkerModel(config: Qwen3OmniMoeTalkerTextConfig) - added:
constructor Qwen3OmniMoeThinkerConfig(audio_config=…, vision_config=…, text_config=…, audio_token_id=…, image_token_id=…, video_token_id=…, position_id_per_seconds=…, audio_start_token_id=…, user_token_id=…, initializer_range=…, **kwargs) - added:
constructor Qwen3OmniMoeThinkerForConditionalGeneration(config) - added:
constructor Qwen3OmniMoeThinkerTextModel(config: Qwen3OmniMoeTextConfig) - added:
field Qwen3OmniMoeConfig.model_type - added:
field Qwen3OmniMoeConfig.sub_configs - added:
field Qwen3OmniMoeForConditionalGeneration.config_class - added:
field Qwen3OmniMoeTalkerCodePredictorModel.base_model_prefix - added:
field Qwen3OmniMoeTalkerCodePredictorModel.config_class - added:
field Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.base_model_prefix - added:
field Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.config_class - added:
field Qwen3OmniMoeTalkerConfig.sub_configs - added:
field Qwen3OmniMoeTalkerForConditionalGeneration.base_model_prefix - added:
field Qwen3OmniMoeTalkerForConditionalGeneration.config_class - added:
field Qwen3OmniMoeTalkerModel.base_model_prefix - added:
field Qwen3OmniMoeTalkerModel.config_class - added:
field Qwen3OmniMoeThinkerConfig.attribute_map - added:
field Qwen3OmniMoeThinkerConfig.model_type - added:
field Qwen3OmniMoeThinkerTextModel.config_class - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.config - added:
field Qwen3OmniMoeThinkerTextPreTrainedModel.config_class - added:
method Qwen3OmniMoeCode2Wav.chunked_decode(codes, chunk_size=…, left_context_size=…) - added:
method Qwen3OmniMoeCode2Wav.forward(codes) - added:
method Qwen3OmniMoeCode2WavDecoderBlock.forward(hidden) - added:
method Qwen3OmniMoeCode2WavTransformerModel.forward(input_ids=…, attention_mask=…, position_ids=…, past_key_values=…, inputs_embeds=…, use_cache=…, cache_position=…, **kwargs) - added:
method Qwen3OmniMoeConfig.get_text_config(decoder=…) -> PretrainedConfig - added:
method Qwen3OmniMoeForConditionalGeneration.disable_talker() - added:
method Qwen3OmniMoeForConditionalGeneration.enable_talker() - added:
method Qwen3OmniMoeForConditionalGeneration.generate(input_ids: Optional[torch.Tensor]=…, speaker: str=…, use_audio_in_video: bool=…, return_audio: Optional[bool]=…, thinker_max_new_tokens: int=…, thinker_eos_token_id: int=…, talker_max_new_tokens: int=…, talker_do_sample: bool=…, talker_top_k: int=…, talker_top_p: float=…, talker_temperature: float=…, talker_repetition_penalty: float=…, **kwargs) - added:
method Qwen3OmniMoePreTrainedModelForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[torch.Tensor], grid_hs: list[torch.Tensor], grid_ws: list[torch.Tensor]) - added:
method Qwen3OmniMoePreTrainedModelForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tens… - added:
method Qwen3OmniMoeProcessor.apply_chat_template(conversations, chat_template=…, **kwargs) - added:
method Qwen3OmniMoeProcessor.replace_multimodal_special_tokens(text, audio_lengths, image_grid_thw, video_grid_thw, video_second_per_grid, use_audio_in_video, position_id_per_seconds, seconds_per_chunk) - added:
method Qwen3OmniMoeTalkerCodePredictorModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast - added:
method Qwen3OmniMoeTalkerCodePredictorModel.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.forward(input_ids=…, attention_mask=…, position_ids=…, past_key_values=…, inputs_embeds=…, labels=…, use_cache=…, cache_position=…, generation_steps=…, **kwargs) - added:
method Qwen3OmniMoeTalkerCodePredictorModelForConditionalGeneration.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.forward(input_ids=…, attention_mask=…, use_audio_in_video=…, audio_feature_lengths=…, video_second_per_grid=…, image_grid_thw=…, video_grid_thw=…, position_ids=…, past_key_values=…, inputs_embeds=…, labels=…, use_cache=…, output_router_logits=…, cache_position=…, residual_codes=…, trailing_text_hidden=…, tts_pad_embed=…, generation_step=…, talker_… - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_input_embeddings() - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_llm_pos_ids_for_vision(start_idx: int, vision_idx: int, spatial_merge_size: int, t_index: list[torch.Tensor], grid_hs: list[torch.Tensor], grid_ws: list[torch.Tensor]) - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, use_audio_in_video: bool=…, audio_seqlens: Optional[torch.LongTensor]=…, second_per_grids: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen3OmniMoeTalkerForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, **kwargs) - added:
method Qwen3OmniMoeTalkerModel.get_input_embeddings() - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.forward(input_ids=…, input_features=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, attention_mask=…, feature_attention_mask=…, audio_feature_lengths=…, position_ids=…, past_key_values=…, inputs_embeds=…, rope_deltas=…, labels=…, use_cache=…, output_router_logits: Optional[bool]=…, use_audio_in_video=…, cache_positio… - added:
method Qwen3OmniMoeThinkerForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor, feature_attention_mask: Optional[torch.LongTensor]=…, audio_feature_lengths: Optional[torch.LongTensor]=…)
- added:
class Qwen3OmniMoeProcessor(ProcessorMixin) - added:
constructor Qwen3OmniMoeProcessor(image_processor=…, video_processor=…, feature_extractor=…, tokenizer=…, chat_template=…) - added:
field Qwen3OmniMoeProcessor.attributes - added:
field Qwen3OmniMoeProcessor.feature_extractor_class - added:
field Qwen3OmniMoeProcessor.image_processor_class - added:
field Qwen3OmniMoeProcessor.tokenizer_class - added:
field Qwen3OmniMoeProcessor.video_processor_class - added:
method Qwen3OmniMoeProcessor.apply_chat_template(conversations, chat_template=…, **kwargs) - added:
method Qwen3OmniMoeProcessor.get_chunked_index(token_indices: np.ndarray, tokens_per_chunk: int) -> list[tuple[int, int]] - added:
method Qwen3OmniMoeProcessor.replace_multimodal_special_tokens(text, audio_lengths, image_grid_thw, video_grid_thw, video_second_per_grid, use_audio_in_video, position_id_per_seconds, seconds_per_chunk) - added:
property Qwen3OmniMoeProcessor.model_input_names
- added:
export define_import_structure
- added:
class Qwen3VLConfig(PretrainedConfig) - added:
class Qwen3VLTextConfig(PretrainedConfig) - added:
constructor Qwen3VLConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, vision_end_token_id=…, tie_word_embeddings=…, **kwargs) - added:
constructor Qwen3VLTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
field Qwen3VLConfig.keys_to_ignore_at_inference - added:
field Qwen3VLConfig.model_type - added:
field Qwen3VLConfig.sub_configs - added:
field Qwen3VLTextConfig.base_config_key - added:
field Qwen3VLTextConfig.model_type
- added:
class Qwen3VLForConditionalGeneration(Qwen3VLPreTrainedModel, GenerationMixin) - added:
class Qwen3VLModel(Qwen3VLPreTrainedModel) - added:
class Qwen3VLPreTrainedModel(PreTrainedModel) - added:
class Qwen3VLTextModel(Qwen3VLPreTrainedModel) - added:
class Qwen3VLVisionModel(Qwen3VLPreTrainedModel) - added:
constructor Qwen3VLForConditionalGeneration(config) - added:
constructor Qwen3VLModel(config) - added:
constructor Qwen3VLTextModel(config: Qwen3VLTextConfig) - added:
constructor Qwen3VLVisionModel(config, *inputs, **kwargs) -> None - added:
field Qwen3VLForConditionalGeneration.accepts_loss_kwargs - added:
field Qwen3VLForConditionalGeneration.config: Qwen3VLConfig - added:
field Qwen3VLModel.accepts_loss_kwargs - added:
field Qwen3VLModel.base_model_prefix - added:
field Qwen3VLModel.config: Qwen3VLConfig - added:
field Qwen3VLPreTrainedModel.base_model_prefix - added:
field Qwen3VLPreTrainedModel.config: Qwen3VLConfig - added:
field Qwen3VLPreTrainedModel.supports_gradient_checkpointing - added:
field Qwen3VLTextModel.config: Qwen3VLTextConfig - added:
field Qwen3VLVisionModel.config: Qwen3VLVisionConfig - added:
method Qwen3VLForConditionalGeneration.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optiona… - added:
method Qwen3VLForConditionalGeneration.get_decoder() - added:
method Qwen3VLForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLForConditionalGeneration.get_input_embeddings() - added:
method Qwen3VLForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Qwen3VLForConditionalGeneration.set_decoder(decoder) - added:
method Qwen3VLForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen3VLModel.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.Lon… - added:
method Qwen3VLModel.get_decoder() - added:
method Qwen3VLModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLModel.get_input_embeddings() - added:
method Qwen3VLModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen3VLModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen3VLModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLModel.set_decoder(decoder) - added:
method Qwen3VLModel.set_input_embeddings(value) - added:
method Qwen3VLTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, visual_pos_masks: Optional[torch.Tensor]=…, deepstack_visual_embeds: Optional[list[… - added:
method Qwen3VLVisionModel.fast_pos_embed_interpolate(grid_thw) - added:
method Qwen3VLVisionModel.forward(hidden_states: torch.Tensor, grid_thw: torch.Tensor, **kwargs) -> torch.Tensor - added:
method Qwen3VLVisionModel.rot_pos_emb(grid_thw: torch.Tensor) -> torch.Tensor - added:
property Qwen3VLForConditionalGeneration.language_model - added:
property Qwen3VLForConditionalGeneration.visual
- added:
class Qwen3VLConfig(PretrainedConfig) - added:
class Qwen3VLForConditionalGeneration(Qwen2_5_VLForConditionalGeneration) - added:
class Qwen3VLModel(Qwen2_5_VLModel) - added:
class Qwen3VLPreTrainedModel(Qwen2VLPreTrainedModel) - added:
class Qwen3VLProcessor(Qwen2VLProcessor) - added:
class Qwen3VLTextConfig(PretrainedConfig) - added:
class Qwen3VLTextModel(Qwen3VLPreTrainedModel, Qwen3Model) - added:
class Qwen3VLVisionModel(Qwen3VLPreTrainedModel) - added:
constructor Qwen3VLConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, vision_end_token_id=…, tie_word_embeddings=…, **kwargs) - added:
constructor Qwen3VLModel(config) - added:
constructor Qwen3VLProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
constructor Qwen3VLTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_dropout=…, **kwargs) - added:
constructor Qwen3VLTextModel(config: Qwen3VLTextConfig) - added:
constructor Qwen3VLVisionModel(config, *inputs, **kwargs) -> None - added:
field Qwen3VLConfig.keys_to_ignore_at_inference - added:
field Qwen3VLConfig.model_type - added:
field Qwen3VLConfig.sub_configs - added:
field Qwen3VLForConditionalGeneration.config: Qwen3VLConfig - added:
field Qwen3VLModel.config: Qwen3VLConfig - added:
field Qwen3VLPreTrainedModel.config: Qwen3VLConfig - added:
field Qwen3VLTextConfig.base_config_key - added:
field Qwen3VLTextConfig.model_type - added:
field Qwen3VLTextModel.config: Qwen3VLTextConfig - added:
field Qwen3VLVisionModel.config: Qwen3VLVisionConfig - added:
method Qwen3VLForConditionalGeneration.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optiona… - added:
method Qwen3VLForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Qwen3VLModel.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.Lon… - added:
method Qwen3VLModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen3VLModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, visual_pos_masks: Optional[torch.Tensor]=…, deepstack_visual_embeds: Optional[list[… - added:
method Qwen3VLVisionModel.fast_pos_embed_interpolate(grid_thw) - added:
method Qwen3VLVisionModel.forward(hidden_states: torch.Tensor, grid_thw: torch.Tensor, **kwargs) -> torch.Tensor - added:
method Qwen3VLVisionModel.rot_pos_emb(grid_thw: torch.Tensor) -> torch.Tensor
- added:
class Qwen3VLProcessor(ProcessorMixin) - added:
constructor Qwen3VLProcessor(image_processor=…, tokenizer=…, video_processor=…, chat_template=…, **kwargs) - added:
field Qwen3VLProcessor.attributes - added:
field Qwen3VLProcessor.image_processor_class - added:
field Qwen3VLProcessor.tokenizer_class - added:
field Qwen3VLProcessor.video_processor_class - added:
method Qwen3VLProcessor.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, clean_up_tokenization_spaces=…, **kwargs)
- added:
class Qwen3VLVideoProcessor(BaseVideoProcessor) - added:
constructor Qwen3VLVideoProcessor(**kwargs: Unpack[Qwen3VLVideoProcessorInitKwargs]) - added:
field Qwen3VLVideoProcessor.do_convert_rgb - added:
field Qwen3VLVideoProcessor.do_normalize - added:
field Qwen3VLVideoProcessor.do_rescale - added:
field Qwen3VLVideoProcessor.do_resize - added:
field Qwen3VLVideoProcessor.do_sample_frames - added:
field Qwen3VLVideoProcessor.fps - added:
field Qwen3VLVideoProcessor.image_mean - added:
field Qwen3VLVideoProcessor.image_std - added:
field Qwen3VLVideoProcessor.max_frames - added:
field Qwen3VLVideoProcessor.merge_size - added:
field Qwen3VLVideoProcessor.min_frames - added:
field Qwen3VLVideoProcessor.model_input_names - added:
field Qwen3VLVideoProcessor.patch_size - added:
field Qwen3VLVideoProcessor.resample - added:
field Qwen3VLVideoProcessor.size - added:
field Qwen3VLVideoProcessor.temporal_patch_size - added:
field Qwen3VLVideoProcessor.valid_kwargs - added:
method Qwen3VLVideoProcessor.sample_frames(metadata: VideoMetadata, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, **kwargs)
- added:
export define_import_structure
- added:
class Qwen3VLMoeConfig(PretrainedConfig) - added:
class Qwen3VLMoeTextConfig(PretrainedConfig) - added:
constructor Qwen3VLMoeConfig(text_config=…, vision_config=…, image_token_id=…, video_token_id=…, vision_start_token_id=…, vision_end_token_id=…, tie_word_embeddings=…, **kwargs) - added:
constructor Qwen3VLMoeTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, decoder_sparse_step=…, moe_intermediate_size=…, num_experts_per_tok=…, num_experts… - added:
field Qwen3VLMoeConfig.keys_to_ignore_at_inference - added:
field Qwen3VLMoeConfig.model_type - added:
field Qwen3VLMoeConfig.sub_configs - added:
field Qwen3VLMoeTextConfig.base_config_key - added:
field Qwen3VLMoeTextConfig.base_model_pp_plan - added:
field Qwen3VLMoeTextConfig.base_model_tp_plan - added:
field Qwen3VLMoeTextConfig.keys_to_ignore_at_inference - added:
field Qwen3VLMoeTextConfig.model_type
- added:
class Qwen3VLMoeForConditionalGeneration(Qwen3VLMoePreTrainedModel, GenerationMixin) - added:
class Qwen3VLMoeModel(Qwen3VLMoePreTrainedModel) - added:
class Qwen3VLMoePreTrainedModel(PreTrainedModel) - added:
class Qwen3VLMoeTextModel(Qwen3VLMoePreTrainedModel) - added:
class Qwen3VLMoeVisionModel(Qwen3VLMoePreTrainedModel) - added:
constructor Qwen3VLMoeForConditionalGeneration(config) - added:
constructor Qwen3VLMoeModel(config) - added:
constructor Qwen3VLMoeTextModel(config: Qwen3VLMoeTextConfig) - added:
constructor Qwen3VLMoeVisionModel(config, *inputs, **kwargs) -> None - added:
field Qwen3VLMoeForConditionalGeneration.accepts_loss_kwargs - added:
field Qwen3VLMoeForConditionalGeneration.config: Qwen3VLMoeConfig - added:
field Qwen3VLMoeModel.accepts_loss_kwargs - added:
field Qwen3VLMoeModel.base_model_prefix - added:
field Qwen3VLMoeModel.config: Qwen3VLMoeConfig - added:
field Qwen3VLMoePreTrainedModel.base_model_prefix - added:
field Qwen3VLMoePreTrainedModel.config: Qwen3VLMoeConfig - added:
field Qwen3VLMoePreTrainedModel.supports_gradient_checkpointing - added:
field Qwen3VLMoeTextModel.config: Qwen3VLMoeTextConfig - added:
field Qwen3VLMoeVisionModel.config: Qwen3VLMoeVisionConfig - added:
method Qwen3VLMoeForConditionalGeneration.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Opti… - added:
method Qwen3VLMoeForConditionalGeneration.get_decoder() - added:
method Qwen3VLMoeForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLMoeForConditionalGeneration.get_input_embeddings() - added:
method Qwen3VLMoeForConditionalGeneration.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLMoeForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, pixel_values=…, pixel_values_videos=…, image_grid_thw=…, video_grid_thw=…, **kwargs) - added:
method Qwen3VLMoeForConditionalGeneration.set_decoder(decoder) - added:
method Qwen3VLMoeForConditionalGeneration.set_input_embeddings(value) - added:
method Qwen3VLMoeModel.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.… - added:
method Qwen3VLMoeModel.get_decoder() - added:
method Qwen3VLMoeModel.get_image_features(pixel_values: torch.FloatTensor, image_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLMoeModel.get_input_embeddings() - added:
method Qwen3VLMoeModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method Qwen3VLMoeModel.get_rope_index(input_ids: Optional[torch.LongTensor]=…, image_grid_thw: Optional[torch.LongTensor]=…, video_grid_thw: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Qwen3VLMoeModel.get_video_features(pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor]=…) - added:
method Qwen3VLMoeModel.set_decoder(decoder) - added:
method Qwen3VLMoeModel.set_input_embeddings(value) - added:
method Qwen3VLMoeTextModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, visual_pos_masks: Optional[torch.Tensor]=…, deepstack_visual_embeds: Optional[li… - added:
method Qwen3VLMoeVisionModel.fast_pos_embed_interpolate(grid_thw) - added:
method Qwen3VLMoeVisionModel.forward(hidden_states: torch.Tensor, grid_thw: torch.Tensor, **kwargs) -> torch.Tensor - added:
method Qwen3VLMoeVisionModel.rot_pos_emb(grid_thw: torch.Tensor) -> torch.Tensor - added:
property Qwen3VLMoeForConditionalGeneration.language_model - added:
property Qwen3VLMoeForConditionalGeneration.visual
- added:
class Qwen3VLMoeConfig(Qwen3VLConfig) - added:
class Qwen3VLMoeForConditionalGeneration(Qwen3VLForConditionalGeneration) - added:
class Qwen3VLMoeModel(Qwen3VLModel) - added:
class Qwen3VLMoePreTrainedModel(Qwen3MoePreTrainedModel) - added:
class Qwen3VLMoeTextConfig(PretrainedConfig) - added:
class Qwen3VLMoeTextModel(Qwen3VLTextModel) - added:
class Qwen3VLMoeVisionModel(Qwen3VLVisionModel) - added:
constructor Qwen3VLMoeTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, decoder_sparse_step=…, moe_intermediate_size=…, num_experts_per_tok=…, num_experts… - added:
field Qwen3VLMoeConfig.model_type - added:
field Qwen3VLMoeConfig.sub_configs - added:
field Qwen3VLMoePreTrainedModel.config: Qwen3VLMoeConfig - added:
field Qwen3VLMoeTextConfig.base_config_key - added:
field Qwen3VLMoeTextConfig.base_model_pp_plan - added:
field Qwen3VLMoeTextConfig.base_model_tp_plan - added:
field Qwen3VLMoeTextConfig.keys_to_ignore_at_inference - added:
field Qwen3VLMoeTextConfig.model_type - added:
method Qwen3VLMoeForConditionalGeneration.forward(input_ids: torch.LongTensor=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.Tensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, image_grid_thw: Opti…
- added:
export define_import_structure
- added:
class RagConfig(PretrainedConfig) - added:
constructor RagConfig(vocab_size=…, is_encoder_decoder=…, prefix=…, bos_token_id=…, pad_token_id=…, eos_token_id=…, decoder_start_token_id=…, title_sep=…, doc_sep=…, n_docs=…, max_combined_length=…, retrieval_vector_size=…, retrieval_batch_size=…, dataset=…, dataset_split=…, index_name=…, index_path=…, passages_path=…, use_dummy_dataset=…, reduce_loss=…, label_smoothing=…, do_deduplication=…, exc… - added:
field RagConfig.has_no_defaults_at_init - added:
field RagConfig.model_type - added:
static method RagConfig.from_question_encoder_generator_configs(question_encoder_config: PretrainedConfig, generator_config: PretrainedConfig, **kwargs) -> PretrainedConfig
- added:
class RagModel(RagPreTrainedModel) - added:
class RagPreTrainedModel(PreTrainedModel) - added:
class RagSequenceForGeneration(RagPreTrainedModel) - added:
class RagTokenForGeneration(RagPreTrainedModel, GenerationMixin) - added:
constructor RagModel(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional[RagRetriever]=…, **kwargs) - added:
constructor RagSequenceForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional[RagRetriever]=…, **kwargs) - added:
constructor RagTokenForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional[RagRetriever]=…, **kwargs) - added:
field RagPreTrainedModel.base_model_prefix - added:
field RagPreTrainedModel.config: RagConfig - added:
method RagModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, past_key_values: Optional[Cache]=…, doc_scores: Optional[torch.FloatTensor]=…, context_input_ids: Optional[torch.LongTensor]=…,… - added:
method RagSequenceForGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.Tensor]]]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, past_key_values: Optional[Cache]=…, context_input_ids: Optional[torch.LongTensor]=…, context_attention_mask: Optiona… - added:
method RagSequenceForGeneration.generate(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, context_input_ids: Optional[torch.LongTensor]=…, context_attention_mask: Optional[torch.LongTensor]=…, doc_scores: Optional[torch.FloatTensor]=…, do_deduplication: Optional[bool]=…, num_return_sequences: Optional[int]=…, num_beams: Optional[int]=…, n_docs: Optional[int]=… - added:
method RagSequenceForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, exclude_bos_score=…, n_docs=…) - added:
method RagSequenceForGeneration.set_context_encoder_for_training(ctx_encoder: PreTrainedModel) - added:
method RagSequenceForGeneration.set_retriever(retriever: RagRetriever) - added:
method RagTokenForGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.Tensor]]]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, past_key_values: Optional[Cache]=…, context_input_ids: Optional[torch.LongTensor]=…, context_attention_mask: Optio… - added:
method RagTokenForGeneration.generate(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, context_input_ids: Optional[torch.LongTensor]=…, context_attention_mask: Optional[torch.LongTensor]=…, doc_scores: Optional[torch.FloatTensor]=…, n_docs: Optional[int]=…, generation_config: Optional[GenerationConfig]=…, prefix_allowed_tokens_fn: Optional[Callable[[int, torc… - added:
method RagTokenForGeneration.get_input_embeddings() - added:
method RagTokenForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, n_docs=…) - added:
method RagTokenForGeneration.get_output_embeddings() - added:
method RagTokenForGeneration.marginalize(seq_logits, doc_scores, n_docs=…) - added:
method RagTokenForGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, use_cache=…, encoder_outputs=…, doc_scores=…, n_docs=…, **kwargs) - added:
method RagTokenForGeneration.set_context_encoder_for_training(ctx_encoder: PreTrainedModel) - added:
method RagTokenForGeneration.set_output_embeddings(new_embeddings) - added:
method RagTokenForGeneration.set_retriever(retriever: RagRetriever) - added:
method RagTokenForGeneration.shift_tokens_right(input_ids, start_token_id=…) - added:
property RagSequenceForGeneration.generator - added:
property RagSequenceForGeneration.question_encoder - added:
property RagSequenceForGeneration.retriever - added:
property RagTokenForGeneration.generator - added:
property RagTokenForGeneration.question_encoder - added:
property RagTokenForGeneration.retriever - added:
static method RagPreTrainedModel.from_pretrained_question_encoder_generator(question_encoder_pretrained_model_name_or_path: Optional[str]=…, generator_pretrained_model_name_or_path: Optional[str]=…, retriever: RagRetriever=…, **kwargs) -> PreTrainedModel
- added:
class TFRagModel(TFRagPreTrainedModel) - added:
class TFRagPreTrainedModel(TFPreTrainedModel) - added:
class TFRagSequenceForGeneration(TFRagPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFRagTokenForGeneration(TFRagPreTrainedModel, TFCausalLanguageModelingLoss) - added:
constructor TFRagModel(config: PretrainedConfig | None=…, question_encoder: TFPreTrainedModel | None=…, generator: TFPreTrainedModel | None=…, retriever: RagRetriever | None=…, load_weight_prefix: str | None=…, **kwargs) - added:
constructor TFRagSequenceForGeneration(config: PretrainedConfig | None=…, question_encoder: TFPreTrainedModel | None=…, generator: TFPreTrainedModel | None=…, retriever: RagRetriever | None=…, **kwargs) - added:
constructor TFRagTokenForGeneration(config: PretrainedConfig | None=…, question_encoder: TFPreTrainedModel | None=…, generator: TFPreTrainedModel | None=…, retriever: RagRetriever | None=…, **kwargs) - added:
field TFRagModel.load_weight_prefix - added:
field TFRagPreTrainedModel.base_model_prefix - added:
field TFRagPreTrainedModel.config_class - added:
field TFRagSequenceForGeneration.load_weight_prefix - added:
field TFRagTokenForGeneration.load_weight_prefix - added:
method TFRagModel.build(input_shape=…) - added:
method TFRagModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, doc_scores: np.ndarray | tf.Tensor | None=…, context_input_ids: np… - added:
method TFRagModel.set_retriever(retriever: RagRetriever) - added:
method TFRagSequenceForGeneration.build(input_shape=…) - added:
method TFRagSequenceForGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, doc_scores: np.ndarray | tf.Tensor | None=…, conte… - added:
method TFRagSequenceForGeneration.generate(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, context_input_ids=…, context_attention_mask=…, doc_scores=…, do_deduplication=…, num_return_sequences=…, num_beams=…, n_docs=…, **model_kwargs) - added:
method TFRagSequenceForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, exclude_bos_score=…, n_docs=…) - added:
method TFRagSequenceForGeneration.set_retriever(retriever: RagRetriever) - added:
method TFRagTokenForGeneration.build(input_shape=…) - added:
method TFRagTokenForGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, doc_scores: np.ndarray | tf.Tensor | None=…, context_… - added:
method TFRagTokenForGeneration.generate(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, context_input_ids=…, context_attention_mask=…, doc_scores=…, n_docs=…, generation_config=…, logits_processor=…, **kwargs) - added:
method TFRagTokenForGeneration.get_input_embeddings() - added:
method TFRagTokenForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, n_docs=…) - added:
method TFRagTokenForGeneration.get_output_embeddings() - added:
method TFRagTokenForGeneration.hf_compute_loss(labels, y_pred, smooth_epsilon=…, from_logits=…, reduce_loss=…) - added:
method TFRagTokenForGeneration.marginalize(seq_logits, doc_scores, n_docs=…) - added:
method TFRagTokenForGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, use_cache=…, encoder_outputs=…, doc_scores=…, n_docs=…, **kwargs) - added:
method TFRagTokenForGeneration.set_retriever(retriever: RagRetriever) - added:
method TFRagTokenForGeneration.shift_tokens_right(input_ids, start_token_id=…) - added:
property TFRagSequenceForGeneration.generator - added:
property TFRagSequenceForGeneration.question_encoder - added:
property TFRagSequenceForGeneration.retriever - added:
property TFRagTokenForGeneration.generator - added:
property TFRagTokenForGeneration.question_encoder - added:
property TFRagTokenForGeneration.retriever - added:
static method TFRagPreTrainedModel.from_pretrained_question_encoder_generator(question_encoder_pretrained_model_name_or_path: str | None=…, generator_pretrained_model_name_or_path: str | None=…, retriever: RagRetriever=…, *model_args, **kwargs) -> TFPreTrainedModel
- added:
class RagRetriever - added:
constructor RagRetriever(config, question_encoder_tokenizer, generator_tokenizer, index=…, init_retrieval=…) - added:
method RagRetriever.init_retrieval() - added:
method RagRetriever.postprocess_docs(docs, input_strings, prefix, n_docs, return_tensors=…) - added:
method RagRetriever.retrieve(question_hidden_states: np.ndarray, n_docs: int) -> tuple[np.ndarray, np.ndarray, list[dict]] - added:
method RagRetriever.save_pretrained(save_directory) - added:
method RagRetriever.set_ctx_encoder_tokenizer(ctx_encoder_tokenizer: PreTrainedTokenizer) - added:
static method RagRetriever.from_pretrained(retriever_name_or_path, indexed_dataset=…, **kwargs)
- added:
class RagTokenizer - added:
constructor RagTokenizer(question_encoder, generator) - added:
method RagTokenizer.batch_decode(*args, **kwargs) - added:
method RagTokenizer.decode(*args, **kwargs) - added:
method RagTokenizer.prepare_seq2seq_batch(src_texts: list[str], tgt_texts: Optional[list[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: Optional[str]=…, truncation: bool=…, **kwargs) -> BatchEncoding - added:
method RagTokenizer.save_pretrained(save_directory) - added:
static method RagTokenizer.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
export define_import_structure
- added:
class RecurrentGemmaConfig(PretrainedConfig) - added:
constructor RecurrentGemmaConfig(num_hidden_layers=…, vocab_size=…, hidden_size=…, intermediate_size=…, num_attention_heads=…, lru_width=…, attention_window_size=…, conv1d_width=…, logits_soft_cap=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, hidden_activation=…, partial_rotary_factor=…, rope_theta=…, block_types=…, attention_dropout=…, num_key_value_heads=…, att… - added:
field RecurrentGemmaConfig.model_type - added:
property RecurrentGemmaConfig.layers_block_type
- added:
class RecurrentGemmaForCausalLM(RecurrentGemmaPreTrainedModel, GenerationMixin) - added:
class RecurrentGemmaModel(RecurrentGemmaPreTrainedModel) - added:
class RecurrentGemmaPreTrainedModel(PreTrainedModel) - added:
constructor RecurrentGemmaForCausalLM(config) - added:
constructor RecurrentGemmaModel(config: RecurrentGemmaConfig) - added:
field RecurrentGemmaPreTrainedModel.base_model_prefix - added:
field RecurrentGemmaPreTrainedModel.config: RecurrentGemmaConfig - added:
field RecurrentGemmaPreTrainedModel.supports_gradient_checkpointing - added:
method RecurrentGemmaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, use_cache: Optional[bool]=…, *… - added:
method RecurrentGemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithNoAttentio… - added:
method RecurrentGemmaPreTrainedModel.reset_cache(batch, device, dtype)
- added:
export define_import_structure
- added:
class ReformerConfig(PretrainedConfig) - added:
constructor ReformerConfig(attention_head_size=…, attn_layers=…, axial_norm_std=…, axial_pos_embds=…, axial_pos_shape=…, axial_pos_embds_dim=…, chunk_size_lm_head=…, eos_token_id=…, feed_forward_size=…, hash_seed=…, hidden_act=…, hidden_dropout_prob=…, hidden_size=…, initializer_range=…, is_decoder=…, layer_norm_eps=…, local_num_chunks_before=…, local_num_chunks_after=…, local_attention_probs_dro… - added:
field ReformerConfig.attribute_map - added:
field ReformerConfig.keys_to_ignore_at_inference - added:
field ReformerConfig.model_type
- added:
class ReformerAttention(nn.Module) - added:
class ReformerForMaskedLM(ReformerPreTrainedModel) - added:
class ReformerForQuestionAnswering(ReformerPreTrainedModel) - added:
class ReformerForSequenceClassification(ReformerPreTrainedModel) - added:
class ReformerLayer(nn.Module) - added:
class ReformerModel(ReformerPreTrainedModel) - added:
class ReformerModelWithLMHead(ReformerPreTrainedModel, GenerationMixin) - added:
class ReformerPreTrainedModel(PreTrainedModel) - added:
constructor ReformerAttention(config, layer_id=…) - added:
constructor ReformerForMaskedLM(config) - added:
constructor ReformerForQuestionAnswering(config) - added:
constructor ReformerForSequenceClassification(config) - added:
constructor ReformerLayer(config, layer_id=…) - added:
constructor ReformerModel(config) - added:
constructor ReformerModelWithLMHead(config) - added:
field ReformerPreTrainedModel.base_model_prefix - added:
field ReformerPreTrainedModel.config: ReformerConfig - added:
method ReformerAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…, buckets=…, cache_position=…) - added:
method ReformerForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, num_hashes: Optional[int]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]… - added:
method ReformerForMaskedLM.get_output_embeddings() - added:
method ReformerForMaskedLM.set_output_embeddings(new_embeddings) - added:
method ReformerForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, num_hashes: Optional[int]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, outpu… - added:
method ReformerForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, num_hashes: Optional[int]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict:… - added:
method ReformerLayer.backward_pass(next_attn_output, hidden_states, grad_attn_output, grad_hidden_states, attention_mask=…, head_mask=…, buckets=…) - added:
method ReformerLayer.forward(prev_attn_output, hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…) - added:
method ReformerModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, num_hashes: Optional[int]=…, past_buckets_states: Optional[list[tuple[torch.Tensor]]]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_attentio… - added:
method ReformerModel.get_input_embeddings() - added:
method ReformerModel.set_input_embeddings(value) - added:
method ReformerModelWithLMHead.forward(input_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, num_hashes: Optional[int]=…, past_buckets_states: Optional[list[tuple[torch.Tensor]]]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, outpu… - added:
method ReformerModelWithLMHead.get_output_embeddings() - added:
method ReformerModelWithLMHead.prepare_inputs_for_generation(input_ids, past_key_values=…, use_cache=…, num_hashes=…, **kwargs) - added:
method ReformerModelWithLMHead.set_output_embeddings(new_embeddings) - added:
property ReformerPreTrainedModel.dummy_inputs
- added:
class ReformerTokenizer(PreTrainedTokenizer) - added:
constructor ReformerTokenizer(vocab_file, eos_token=…, unk_token=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field ReformerTokenizer.model_input_names - added:
field ReformerTokenizer.vocab_files_names - added:
method ReformerTokenizer.convert_tokens_to_string(tokens) - added:
method ReformerTokenizer.get_vocab() -> dict[str, int] - added:
method ReformerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property ReformerTokenizer.vocab_size
- added:
class ReformerTokenizerFast(PreTrainedTokenizerFast) - added:
constructor ReformerTokenizerFast(vocab_file=…, tokenizer_file=…, eos_token=…, unk_token=…, additional_special_tokens=…, **kwargs) - added:
field ReformerTokenizerFast.model_input_names - added:
field ReformerTokenizerFast.slow_tokenizer_class - added:
field ReformerTokenizerFast.vocab_files_names - added:
method ReformerTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class RegNetConfig(PretrainedConfig) - added:
constructor RegNetConfig(num_channels=…, embedding_size=…, hidden_sizes=…, depths=…, groups_width=…, layer_type=…, hidden_act=…, **kwargs) - added:
field RegNetConfig.layer_types - added:
field RegNetConfig.model_type
- added:
class FlaxRegNetForImageClassification(FlaxRegNetPreTrainedModel) - added:
class FlaxRegNetModel(FlaxRegNetPreTrainedModel) - added:
class FlaxRegNetPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxRegNetPreTrainedModel(config: RegNetConfig, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxRegNetForImageClassification.module_class - added:
field FlaxRegNetModel.module_class - added:
field FlaxRegNetPreTrainedModel.base_model_prefix - added:
field FlaxRegNetPreTrainedModel.config_class - added:
field FlaxRegNetPreTrainedModel.main_input_name - added:
field FlaxRegNetPreTrainedModel.module_class: nn.Module - added:
method FlaxRegNetPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class RegNetForImageClassification(RegNetPreTrainedModel) - added:
class RegNetModel(RegNetPreTrainedModel) - added:
class RegNetPreTrainedModel(PreTrainedModel) - added:
constructor RegNetForImageClassification(config) - added:
constructor RegNetModel(config) - added:
field RegNetPreTrainedModel.base_model_prefix - added:
field RegNetPreTrainedModel.config: RegNetConfig - added:
field RegNetPreTrainedModel.main_input_name - added:
method RegNetForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention - added:
method RegNetModel.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
class TFRegNetForImageClassification(TFRegNetPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFRegNetModel(TFRegNetPreTrainedModel) - added:
class TFRegNetPreTrainedModel(TFPreTrainedModel) - added:
constructor TFRegNetForImageClassification(config: RegNetConfig, *inputs, **kwargs) - added:
constructor TFRegNetModel(config: RegNetConfig, *inputs, **kwargs) - added:
field TFRegNetPreTrainedModel.base_model_prefix - added:
field TFRegNetPreTrainedModel.config_class - added:
field TFRegNetPreTrainedModel.main_input_name - added:
method TFRegNetForImageClassification.build(input_shape=…) - added:
method TFRegNetForImageClassification.call(pixel_values: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[TFSequenceClassifierOutput, tuple[tf.Tensor]] - added:
method TFRegNetModel.build(input_shape=…) - added:
method TFRegNetModel.call(pixel_values: tf.Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[TFBaseModelOutputWithPoolingAndNoAttention, tuple[tf.Tensor]] - added:
property TFRegNetPreTrainedModel.input_signature
- added:
export define_import_structure
- added:
class RemBertConfig(PretrainedConfig) - added:
class RemBertOnnxConfig(OnnxConfig) - added:
constructor RemBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, input_embedding_size=…, output_embedding_size=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, classifier_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_… - added:
field RemBertConfig.model_type - added:
property RemBertOnnxConfig.atol_for_validation -> float - added:
property RemBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class RemBertForCausalLM(RemBertPreTrainedModel, GenerationMixin) - added:
class RemBertForMaskedLM(RemBertPreTrainedModel) - added:
class RemBertForMultipleChoice(RemBertPreTrainedModel) - added:
class RemBertForQuestionAnswering(RemBertPreTrainedModel) - added:
class RemBertForSequenceClassification(RemBertPreTrainedModel) - added:
class RemBertForTokenClassification(RemBertPreTrainedModel) - added:
class RemBertLayer(GradientCheckpointingLayer) - added:
class RemBertModel(RemBertPreTrainedModel) - added:
class RemBertPreTrainedModel(PreTrainedModel) - added:
constructor RemBertForCausalLM(config) - added:
constructor RemBertForMaskedLM(config) - added:
constructor RemBertForMultipleChoice(config) - added:
constructor RemBertForQuestionAnswering(config) - added:
constructor RemBertForSequenceClassification(config) - added:
constructor RemBertForTokenClassification(config) - added:
constructor RemBertLayer(config, layer_idx=…) - added:
constructor RemBertModel(config, add_pooling_layer=…) - added:
field RemBertPreTrainedModel.base_model_prefix - added:
field RemBertPreTrainedModel.config: RemBertConfig - added:
field RemBertPreTrainedModel.load_tf_weights - added:
field RemBertPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_rembert(model, config, tf_checkpoint_path) - added:
method RemBertForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTe… - added:
method RemBertForCausalLM.get_output_embeddings() - added:
method RemBertForCausalLM.set_output_embeddings(new_embeddings) - added:
method RemBertForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTe… - added:
method RemBertForMaskedLM.get_output_embeddings() - added:
method RemBertForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method RemBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method RemBertForMultipleChoice.forward(input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_st… - added:
method RemBertForQuestionAnswering.forward(input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]… - added:
method RemBertForSequenceClassification.forward(input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method RemBertForTokenClassification.forward(input_ids: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidd… - added:
method RemBertLayer.feed_forward_chunk(attention_output) - added:
method RemBertLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…) -> tuple[torch.Tensor] - added:
method RemBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=… - added:
method RemBertModel.get_input_embeddings() - added:
method RemBertModel.set_input_embeddings(value) - added:
static method RemBertForMaskedLM.can_generate() -> bool
- added:
class TFRemBertForCausalLM(TFRemBertPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFRemBertForMaskedLM(TFRemBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFRemBertForMultipleChoice(TFRemBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFRemBertForQuestionAnswering(TFRemBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFRemBertForSequenceClassification(TFRemBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFRemBertForTokenClassification(TFRemBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFRemBertLayer(keras.layers.Layer) - added:
class TFRemBertModel(TFRemBertPreTrainedModel) - added:
class TFRemBertPreTrainedModel(TFPreTrainedModel) - added:
constructor TFRemBertForCausalLM(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertForMaskedLM(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertForMultipleChoice(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertForQuestionAnswering(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertForSequenceClassification(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertForTokenClassification(config: RemBertConfig, *inputs, **kwargs) - added:
constructor TFRemBertLayer(config: RemBertConfig, **kwargs) - added:
constructor TFRemBertModel(config: RemBertConfig, *inputs, **kwargs) - added:
field TFRemBertPreTrainedModel.base_model_prefix - added:
field TFRemBertPreTrainedModel.config_class - added:
method TFRemBertForCausalLM.build(input_shape=…) - added:
method TFRemBertForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray… - added:
method TFRemBertForCausalLM.get_lm_head() -> keras.layers.Layer - added:
method TFRemBertForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFRemBertForMaskedLM.build(input_shape=…) - added:
method TFRemBertForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool |… - added:
method TFRemBertForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFRemBertForMultipleChoice.build(input_shape=…) - added:
method TFRemBertForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: b… - added:
method TFRemBertForQuestionAnswering.build(input_shape=…) - added:
method TFRemBertForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict… - added:
method TFRemBertForSequenceClassification.build(input_shape=…) - added:
method TFRemBertForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return… - added:
method TFRemBertForTokenClassification.build(input_shape=…) - added:
method TFRemBertForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFRemBertLayer.build(input_shape=…) - added:
method TFRemBertLayer.call(hidden_states: tf.Tensor, attention_mask: tf.Tensor, head_mask: tf.Tensor, encoder_hidden_states: tf.Tensor | None, encoder_attention_mask: tf.Tensor | None, past_key_value: tuple[tf.Tensor] | None, output_attentions: bool, training: bool=…) -> tuple[tf.Tensor] - added:
method TFRemBertModel.build(input_shape=…) - added:
method TFRemBertModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.T…
- added:
class RemBertTokenizer(PreTrainedTokenizer) - added:
constructor RemBertTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field RemBertTokenizer.vocab_files_names - added:
method RemBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RemBertTokenizer.convert_tokens_to_string(tokens) - added:
method RemBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RemBertTokenizer.get_vocab() - added:
method RemBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RemBertTokenizer.vocab_size
- added:
class RemBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor RemBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field RemBertTokenizerFast.slow_tokenizer_class - added:
field RemBertTokenizerFast.vocab_files_names - added:
method RemBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RemBertTokenizerFast.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RemBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class ResNetConfig(BackboneConfigMixin, PretrainedConfig) - added:
class ResNetOnnxConfig(OnnxConfig) - added:
constructor ResNetConfig(num_channels=…, embedding_size=…, hidden_sizes=…, depths=…, layer_type=…, hidden_act=…, downsample_in_first_stage=…, downsample_in_bottleneck=…, out_features=…, out_indices=…, **kwargs) - added:
field ResNetConfig.layer_types - added:
field ResNetConfig.model_type - added:
field ResNetOnnxConfig.torch_onnx_minimum_version - added:
property ResNetOnnxConfig.atol_for_validation -> float - added:
property ResNetOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxResNetForImageClassification(FlaxResNetPreTrainedModel) - added:
class FlaxResNetModel(FlaxResNetPreTrainedModel) - added:
class FlaxResNetPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxResNetPreTrainedModel(config: ResNetConfig, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxResNetForImageClassification.module_class - added:
field FlaxResNetModel.module_class - added:
field FlaxResNetPreTrainedModel.base_model_prefix - added:
field FlaxResNetPreTrainedModel.config_class - added:
field FlaxResNetPreTrainedModel.main_input_name - added:
field FlaxResNetPreTrainedModel.module_class: nn.Module - added:
method FlaxResNetPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class ResNetBackbone(ResNetPreTrainedModel, BackboneMixin) - added:
class ResNetForImageClassification(ResNetPreTrainedModel) - added:
class ResNetModel(ResNetPreTrainedModel) - added:
class ResNetPreTrainedModel(PreTrainedModel) - added:
constructor ResNetBackbone(config) - added:
constructor ResNetForImageClassification(config) - added:
constructor ResNetModel(config) - added:
field ResNetBackbone.has_attentions - added:
field ResNetPreTrainedModel.base_model_prefix - added:
field ResNetPreTrainedModel.config: ResNetConfig - added:
field ResNetPreTrainedModel.main_input_name - added:
method ResNetBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method ResNetForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention - added:
method ResNetModel.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BaseModelOutputWithPoolingAndNoAttention
- added:
class TFResNetForImageClassification(TFResNetPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFResNetModel(TFResNetPreTrainedModel) - added:
class TFResNetPreTrainedModel(TFPreTrainedModel) - added:
constructor TFResNetForImageClassification(config: ResNetConfig, **kwargs) -> None - added:
constructor TFResNetModel(config: ResNetConfig, **kwargs) -> None - added:
field TFResNetPreTrainedModel.base_model_prefix - added:
field TFResNetPreTrainedModel.config_class - added:
field TFResNetPreTrainedModel.main_input_name - added:
method TFResNetForImageClassification.build(input_shape=…) - added:
method TFResNetForImageClassification.call(pixel_values: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tuple[tf.Tensor], TFImageClassifierOutputWithNoAttention] - added:
method TFResNetForImageClassification.classifier(x: tf.Tensor) -> tf.Tensor - added:
method TFResNetModel.build(input_shape=…) - added:
method TFResNetModel.call(pixel_values: tf.Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tuple[tf.Tensor], TFBaseModelOutputWithPoolingAndNoAttention] - added:
property TFResNetPreTrainedModel.input_signature
- added:
export define_import_structure
- added:
class RobertaConfig(PretrainedConfig) - added:
class RobertaOnnxConfig(OnnxConfig) - added:
constructor RobertaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field RobertaConfig.model_type - added:
property RobertaOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxRobertaForCausalLM(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaForMaskedLM(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaForMultipleChoice(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaForQuestionAnswering(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaForSequenceClassification(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaForTokenClassification(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaModel(FlaxRobertaPreTrainedModel) - added:
class FlaxRobertaPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxRobertaPreTrainedModel(config: RobertaConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxRobertaForCausalLM.module_class - added:
field FlaxRobertaForMaskedLM.module_class - added:
field FlaxRobertaForMultipleChoice.module_class - added:
field FlaxRobertaForQuestionAnswering.module_class - added:
field FlaxRobertaForSequenceClassification.module_class - added:
field FlaxRobertaForTokenClassification.module_class - added:
field FlaxRobertaModel.module_class - added:
field FlaxRobertaPreTrainedModel.base_model_prefix - added:
field FlaxRobertaPreTrainedModel.config_class - added:
field FlaxRobertaPreTrainedModel.module_class: nn.Module - added:
method FlaxRobertaForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxRobertaForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxRobertaPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxRobertaPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxRobertaPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class RobertaForCausalLM(RobertaPreTrainedModel, GenerationMixin) - added:
class RobertaForMaskedLM(RobertaPreTrainedModel) - added:
class RobertaForMultipleChoice(RobertaPreTrainedModel) - added:
class RobertaForQuestionAnswering(RobertaPreTrainedModel) - added:
class RobertaForSequenceClassification(RobertaPreTrainedModel) - added:
class RobertaForTokenClassification(RobertaPreTrainedModel) - added:
class RobertaModel(RobertaPreTrainedModel) - added:
class RobertaPreTrainedModel(PreTrainedModel) - added:
constructor RobertaForCausalLM(config) - added:
constructor RobertaForMaskedLM(config) - added:
constructor RobertaForMultipleChoice(config) - added:
constructor RobertaForQuestionAnswering(config) - added:
constructor RobertaForSequenceClassification(config) - added:
constructor RobertaForTokenClassification(config) - added:
constructor RobertaModel(config, add_pooling_layer=…) - added:
field RobertaPreTrainedModel.base_model_prefix - added:
field RobertaPreTrainedModel.config: RobertaConfig - added:
field RobertaPreTrainedModel.supports_gradient_checkpointing - added:
method RobertaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method RobertaForCausalLM.get_output_embeddings() - added:
method RobertaForCausalLM.set_output_embeddings(new_embeddings) - added:
method RobertaForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method RobertaForMaskedLM.get_output_embeddings() - added:
method RobertaForMaskedLM.set_output_embeddings(new_embeddings) - added:
method RobertaForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_stat… - added:
method RobertaForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…… - added:
method RobertaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method RobertaForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method RobertaModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=… - added:
method RobertaModel.get_input_embeddings() - added:
method RobertaModel.set_input_embeddings(value)
- added:
class TFRobertaForCausalLM(TFRobertaPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassificationLoss) - added:
class TFRobertaMainLayer(keras.layers.Layer) - added:
class TFRobertaModel(TFRobertaPreTrainedModel) - added:
class TFRobertaPreTrainedModel(TFPreTrainedModel) - added:
constructor TFRobertaForCausalLM(config: RobertaConfig, *inputs, **kwargs) - added:
constructor TFRobertaForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFRobertaForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFRobertaForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFRobertaForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFRobertaForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFRobertaMainLayer(config, add_pooling_layer=…, **kwargs) - added:
constructor TFRobertaModel(config, *inputs, **kwargs) - added:
field TFRobertaMainLayer.config_class - added:
field TFRobertaPreTrainedModel.base_model_prefix - added:
field TFRobertaPreTrainedModel.config_class - added:
method TFRobertaForCausalLM.build(input_shape=…) - added:
method TFRobertaForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray… - added:
method TFRobertaForCausalLM.get_lm_head() - added:
method TFRobertaForCausalLM.get_prefix_bias_name() - added:
method TFRobertaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFRobertaForMaskedLM.build(input_shape=…) - added:
method TFRobertaForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool |… - added:
method TFRobertaForMaskedLM.get_lm_head() - added:
method TFRobertaForMaskedLM.get_prefix_bias_name() - added:
method TFRobertaForMultipleChoice.build(input_shape=…) - added:
method TFRobertaForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: b… - added:
method TFRobertaForQuestionAnswering.build(input_shape=…) - added:
method TFRobertaForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict… - added:
method TFRobertaForSequenceClassification.build(input_shape=…) - added:
method TFRobertaForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return… - added:
method TFRobertaForTokenClassification.build(input_shape=…) - added:
method TFRobertaForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_di… - added:
method TFRobertaMainLayer.build(input_shape=…) - added:
method TFRobertaMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray |… - added:
method TFRobertaMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFRobertaMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFRobertaModel.build(input_shape=…) - added:
method TFRobertaModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.T…
- added:
class RobertaTokenizer(PreTrainedTokenizer) - added:
constructor RobertaTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - added:
field RobertaTokenizer.model_input_names - added:
field RobertaTokenizer.vocab_files_names - added:
method RobertaTokenizer.bpe(token) - added:
method RobertaTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RobertaTokenizer.convert_tokens_to_string(tokens) - added:
method RobertaTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RobertaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RobertaTokenizer.get_vocab() - added:
method RobertaTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method RobertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RobertaTokenizer.vocab_size
- added:
class RobertaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor RobertaTokenizerFast(vocab_file=…, merges_file=…, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - added:
field RobertaTokenizerFast.model_input_names - added:
field RobertaTokenizerFast.slow_tokenizer_class - added:
field RobertaTokenizerFast.vocab_files_names - added:
method RobertaTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method RobertaTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RobertaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RobertaTokenizerFast.mask_token -> str
- added:
export define_import_structure
- added:
class RobertaPreLayerNormConfig(PretrainedConfig) - added:
class RobertaPreLayerNormOnnxConfig(OnnxConfig) - added:
constructor RobertaPreLayerNormConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **k… - added:
field RobertaPreLayerNormConfig.model_type - added:
property RobertaPreLayerNormOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxRobertaPreLayerNormForCausalLM(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormForMaskedLM(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormForMultipleChoice(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormForQuestionAnswering(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormForSequenceClassification(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormForTokenClassification(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormModel(FlaxRobertaPreLayerNormPreTrainedModel) - added:
class FlaxRobertaPreLayerNormPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxRobertaPreLayerNormPreTrainedModel(config: RobertaPreLayerNormConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxRobertaPreLayerNormForCausalLM.module_class - added:
field FlaxRobertaPreLayerNormForMaskedLM.module_class - added:
field FlaxRobertaPreLayerNormForMultipleChoice.module_class - added:
field FlaxRobertaPreLayerNormForQuestionAnswering.module_class - added:
field FlaxRobertaPreLayerNormForSequenceClassification.module_class - added:
field FlaxRobertaPreLayerNormForTokenClassification.module_class - added:
field FlaxRobertaPreLayerNormModel.module_class - added:
field FlaxRobertaPreLayerNormPreTrainedModel.base_model_prefix - added:
field FlaxRobertaPreLayerNormPreTrainedModel.config_class - added:
field FlaxRobertaPreLayerNormPreTrainedModel.module_class: nn.Module - added:
method FlaxRobertaPreLayerNormForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxRobertaPreLayerNormForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxRobertaPreLayerNormPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxRobertaPreLayerNormPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxRobertaPreLayerNormPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class RobertaPreLayerNormForCausalLM(RobertaPreLayerNormPreTrainedModel, GenerationMixin) - added:
class RobertaPreLayerNormForMaskedLM(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormForMultipleChoice(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormForQuestionAnswering(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormForSequenceClassification(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormForTokenClassification(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormModel(RobertaPreLayerNormPreTrainedModel) - added:
class RobertaPreLayerNormPreTrainedModel(PreTrainedModel) - added:
constructor RobertaPreLayerNormForCausalLM(config) - added:
constructor RobertaPreLayerNormForMaskedLM(config) - added:
constructor RobertaPreLayerNormForMultipleChoice(config) - added:
constructor RobertaPreLayerNormForQuestionAnswering(config) - added:
constructor RobertaPreLayerNormForSequenceClassification(config) - added:
constructor RobertaPreLayerNormForTokenClassification(config) - added:
constructor RobertaPreLayerNormModel(config, add_pooling_layer=…) - added:
field RobertaPreLayerNormPreTrainedModel.base_model_prefix - added:
field RobertaPreLayerNormPreTrainedModel.config: RobertaPreLayerNormConfig - added:
field RobertaPreLayerNormPreTrainedModel.supports_gradient_checkpointing - added:
method RobertaPreLayerNormForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[… - added:
method RobertaPreLayerNormForCausalLM.get_output_embeddings() - added:
method RobertaPreLayerNormForCausalLM.set_output_embeddings(new_embeddings) - added:
method RobertaPreLayerNormForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[… - added:
method RobertaPreLayerNormForMaskedLM.get_output_embeddings() - added:
method RobertaPreLayerNormForMaskedLM.set_output_embeddings(new_embeddings) - added:
method RobertaPreLayerNormForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output… - added:
method RobertaPreLayerNormForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.L… - added:
method RobertaPreLayerNormForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…… - added:
method RobertaPreLayerNormForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, o… - added:
method RobertaPreLayerNormModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Opti… - added:
method RobertaPreLayerNormModel.get_input_embeddings() - added:
method RobertaPreLayerNormModel.set_input_embeddings(value)
- added:
class TFRobertaPreLayerNormForCausalLM(TFRobertaPreLayerNormPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFRobertaPreLayerNormForMaskedLM(TFRobertaPreLayerNormPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFRobertaPreLayerNormForMultipleChoice(TFRobertaPreLayerNormPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFRobertaPreLayerNormForQuestionAnswering(TFRobertaPreLayerNormPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFRobertaPreLayerNormForSequenceClassification(TFRobertaPreLayerNormPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFRobertaPreLayerNormForTokenClassification(TFRobertaPreLayerNormPreTrainedModel, TFTokenClassificationLoss) - added:
class TFRobertaPreLayerNormMainLayer(keras.layers.Layer) - added:
class TFRobertaPreLayerNormModel(TFRobertaPreLayerNormPreTrainedModel) - added:
class TFRobertaPreLayerNormPreTrainedModel(TFPreTrainedModel) - added:
constructor TFRobertaPreLayerNormForCausalLM(config: RobertaPreLayerNormConfig, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFRobertaPreLayerNormMainLayer(config, add_pooling_layer=…, **kwargs) - added:
constructor TFRobertaPreLayerNormModel(config, *inputs, **kwargs) - added:
field TFRobertaPreLayerNormMainLayer.config_class - added:
field TFRobertaPreLayerNormPreTrainedModel.base_model_prefix - added:
field TFRobertaPreLayerNormPreTrainedModel.config_class - added:
method TFRobertaPreLayerNormForCausalLM.build(input_shape=…) - added:
method TFRobertaPreLayerNormForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask:… - added:
method TFRobertaPreLayerNormForCausalLM.get_lm_head() - added:
method TFRobertaPreLayerNormForCausalLM.get_prefix_bias_name() - added:
method TFRobertaPreLayerNormForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFRobertaPreLayerNormForMaskedLM.build(input_shape=…) - added:
method TFRobertaPreLayerNormForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_d… - added:
method TFRobertaPreLayerNormForMaskedLM.get_lm_head() - added:
method TFRobertaPreLayerNormForMaskedLM.get_prefix_bias_name() - added:
method TFRobertaPreLayerNormForMultipleChoice.build(input_shape=…) - added:
method TFRobertaPreLayerNormForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, re… - added:
method TFRobertaPreLayerNormForQuestionAnswering.build(input_shape=…) - added:
method TFRobertaPreLayerNormForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…,… - added:
method TFRobertaPreLayerNormForSequenceClassification.build(input_shape=…) - added:
method TFRobertaPreLayerNormForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | No… - added:
method TFRobertaPreLayerNormForTokenClassification.build(input_shape=…) - added:
method TFRobertaPreLayerNormForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=… - added:
method TFRobertaPreLayerNormMainLayer.build(input_shape=…) - added:
method TFRobertaPreLayerNormMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: n… - added:
method TFRobertaPreLayerNormMainLayer.get_input_embeddings() -> keras.layers.Layer - added:
method TFRobertaPreLayerNormMainLayer.set_input_embeddings(value: tf.Variable) - added:
method TFRobertaPreLayerNormModel.build(input_shape=…) - added:
method TFRobertaPreLayerNormModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.nd…
- added:
export define_import_structure
- added:
class RoCBertConfig(PretrainedConfig) - added:
constructor RoCBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, position_embedding_type=…, classifier_dropout=…, enable_pronunciation=…, enable_shape=…, pronunc… - added:
field RoCBertConfig.model_type
- added:
class RoCBertForCausalLM(RoCBertPreTrainedModel, GenerationMixin) - added:
class RoCBertForMaskedLM(RoCBertPreTrainedModel) - added:
class RoCBertForMultipleChoice(RoCBertPreTrainedModel) - added:
class RoCBertForPreTraining(RoCBertPreTrainedModel) - added:
class RoCBertForQuestionAnswering(RoCBertPreTrainedModel) - added:
class RoCBertForSequenceClassification(RoCBertPreTrainedModel) - added:
class RoCBertForTokenClassification(RoCBertPreTrainedModel) - added:
class RoCBertLayer(GradientCheckpointingLayer) - added:
class RoCBertModel(RoCBertPreTrainedModel) - added:
class RoCBertPreTrainedModel(PreTrainedModel) - added:
constructor RoCBertForCausalLM(config) - added:
constructor RoCBertForMaskedLM(config) - added:
constructor RoCBertForMultipleChoice(config) - added:
constructor RoCBertForPreTraining(config) - added:
constructor RoCBertForQuestionAnswering(config) - added:
constructor RoCBertForSequenceClassification(config) - added:
constructor RoCBertForTokenClassification(config) - added:
constructor RoCBertLayer(config, layer_idx=…) - added:
constructor RoCBertModel(config, add_pooling_layer=…) - added:
field RoCBertPreTrainedModel.base_model_prefix - added:
field RoCBertPreTrainedModel.config: RoCBertConfig - added:
field RoCBertPreTrainedModel.load_tf_weights - added:
field RoCBertPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_roc_bert(model, config, tf_checkpoint_path) - added:
method RoCBertForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_ma… - added:
method RoCBertForCausalLM.get_output_embeddings() - added:
method RoCBertForCausalLM.prepare_inputs_for_generation(input_ids, input_shape_ids=…, input_pronunciation_ids=…, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method RoCBertForCausalLM.set_output_embeddings(new_embeddings) - added:
method RoCBertForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[… - added:
method RoCBertForMaskedLM.get_output_embeddings() - added:
method RoCBertForMaskedLM.prepare_inputs_for_generation(input_ids, input_shape_ids=…, input_pronunciation_ids=…, attention_mask=…, **model_kwargs) - added:
method RoCBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method RoCBertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Ten… - added:
method RoCBertForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, attack_input_ids: Optional[torch.Tensor]=…, attack_input_shape_ids: Optional[torch.Tensor]=…, attack_input_pronunciation_ids: Optional[torch.Tensor… - added:
method RoCBertForPreTraining.get_output_embeddings() - added:
method RoCBertForPreTraining.set_output_embeddings(new_embeddings) - added:
method RoCBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Option… - added:
method RoCBertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[t… - added:
method RoCBertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torc… - added:
method RoCBertLayer.feed_forward_chunk(attention_output) - added:
method RoCBertLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, output_attentions: Optional[bool]=…, cache_position: Optional[torch.Tensor]=…) -> tuple[torch.Tensor] - added:
method RoCBertModel.forward(input_ids: Optional[torch.Tensor]=…, input_shape_ids: Optional[torch.Tensor]=…, input_pronunciation_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.… - added:
method RoCBertModel.get_input_embeddings() - added:
method RoCBertModel.get_pronunciation_embeddings() - added:
method RoCBertModel.get_shape_embeddings() - added:
method RoCBertModel.set_input_embeddings(value) - added:
method RoCBertModel.set_pronunciation_embeddings(value) - added:
method RoCBertModel.set_shape_embeddings(value)
- added:
class RoCBertTokenizer(PreTrainedTokenizer) - added:
constructor RoCBertTokenizer(vocab_file, word_shape_file, word_pronunciation_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RoCBertTokenizer.vocab_files_names - added:
method RoCBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, cls_token_id: Optional[int]=…, sep_token_id: Optional[int]=…) -> list[int] - added:
method RoCBertTokenizer.convert_tokens_to_pronunciation_ids(tokens: Union[str, list[str]]) -> Union[int, list[int]] - added:
method RoCBertTokenizer.convert_tokens_to_shape_ids(tokens: Union[str, list[str]]) -> Union[int, list[int]] - added:
method RoCBertTokenizer.convert_tokens_to_string(tokens) - added:
method RoCBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RoCBertTokenizer.get_vocab() - added:
method RoCBertTokenizer.prepare_for_model(ids: list[int], shape_ids: list[int], pronunciation_ids: list[int], pair_ids: Optional[list[int]]=…, pair_shape_ids: Optional[list[int]]=…, pair_pronunciation_ids: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…,… - added:
method RoCBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str, str, str] - added:
property RoCBertTokenizer.do_lower_case - added:
property RoCBertTokenizer.vocab_size
- added:
export define_import_structure
- added:
class RoFormerConfig(PretrainedConfig) - added:
class RoFormerOnnxConfig(OnnxConfig) - added:
constructor RoFormerConfig(vocab_size=…, embedding_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, rotary_value=…, use_cache=…, **kwargs) - added:
field RoFormerConfig.model_type - added:
property RoFormerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxRoFormerForMaskedLM(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerForMultipleChoice(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerForQuestionAnswering(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerForSequenceClassification(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerForTokenClassification(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerModel(FlaxRoFormerPreTrainedModel) - added:
class FlaxRoFormerPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxRoFormerPreTrainedModel(config: RoFormerConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxRoFormerForMaskedLM.module_class - added:
field FlaxRoFormerForMultipleChoice.module_class - added:
field FlaxRoFormerForQuestionAnswering.module_class - added:
field FlaxRoFormerForSequenceClassification.module_class - added:
field FlaxRoFormerForTokenClassification.module_class - added:
field FlaxRoFormerModel.module_class - added:
field FlaxRoFormerPreTrainedModel.base_model_prefix - added:
field FlaxRoFormerPreTrainedModel.config_class - added:
field FlaxRoFormerPreTrainedModel.module_class: nn.Module - added:
method FlaxRoFormerPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class RoFormerForCausalLM(RoFormerPreTrainedModel, GenerationMixin) - added:
class RoFormerForMaskedLM(RoFormerPreTrainedModel) - added:
class RoFormerForMultipleChoice(RoFormerPreTrainedModel) - added:
class RoFormerForQuestionAnswering(RoFormerPreTrainedModel) - added:
class RoFormerForSequenceClassification(RoFormerPreTrainedModel) - added:
class RoFormerForTokenClassification(RoFormerPreTrainedModel) - added:
class RoFormerLayer(GradientCheckpointingLayer) - added:
class RoFormerModel(RoFormerPreTrainedModel) - added:
class RoFormerPreTrainedModel(PreTrainedModel) - added:
constructor RoFormerForCausalLM(config) - added:
constructor RoFormerForMaskedLM(config) - added:
constructor RoFormerForMultipleChoice(config) - added:
constructor RoFormerForQuestionAnswering(config) - added:
constructor RoFormerForSequenceClassification(config) - added:
constructor RoFormerForTokenClassification(config) - added:
constructor RoFormerLayer(config, layer_idx=…) - added:
constructor RoFormerModel(config) - added:
field RoFormerPreTrainedModel.base_model_prefix - added:
field RoFormerPreTrainedModel.config: RoFormerConfig - added:
field RoFormerPreTrainedModel.load_tf_weights - added:
field RoFormerPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_roformer(model, config, tf_checkpoint_path) - added:
method RoFormerForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[to… - added:
method RoFormerForCausalLM.get_output_embeddings() - added:
method RoFormerForCausalLM.set_output_embeddings(new_embeddings) - added:
method RoFormerForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor… - added:
method RoFormerForMaskedLM.get_output_embeddings() - added:
method RoFormerForMaskedLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method RoFormerForMaskedLM.set_output_embeddings(new_embeddings) - added:
method RoFormerForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional… - added:
method RoFormerForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outp… - added:
method RoFormerForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
method RoFormerForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Opt… - added:
method RoFormerLayer.feed_forward_chunk(attention_output) - added:
method RoFormerLayer.forward(hidden_states, attention_mask=…, sinusoidal_pos=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, past_key_values=…, output_attentions=…, cache_position=…) - added:
method RoFormerModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…, past_key_values: Optional[Cache]=…, use… - added:
method RoFormerModel.get_input_embeddings() - added:
method RoFormerModel.set_input_embeddings(value)
- added:
class TFRoFormerForCausalLM(TFRoFormerPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFRoFormerForMaskedLM(TFRoFormerPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFRoFormerForMultipleChoice(TFRoFormerPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFRoFormerForQuestionAnswering(TFRoFormerPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFRoFormerForSequenceClassification(TFRoFormerPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFRoFormerForTokenClassification(TFRoFormerPreTrainedModel, TFTokenClassificationLoss) - added:
class TFRoFormerLayer(keras.layers.Layer) - added:
class TFRoFormerModel(TFRoFormerPreTrainedModel) - added:
class TFRoFormerPreTrainedModel(TFPreTrainedModel) - added:
constructor TFRoFormerForCausalLM(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerForMaskedLM(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerForMultipleChoice(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerForQuestionAnswering(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerForSequenceClassification(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerForTokenClassification(config: RoFormerConfig, *inputs, **kwargs) - added:
constructor TFRoFormerLayer(config: RoFormerConfig, **kwargs) - added:
constructor TFRoFormerModel(config: RoFormerConfig, *inputs, **kwargs) - added:
field TFRoFormerPreTrainedModel.base_model_prefix - added:
field TFRoFormerPreTrainedModel.config_class - added:
method TFRoFormerForCausalLM.build(input_shape=…) - added:
method TFRoFormerForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=… - added:
method TFRoFormerForCausalLM.get_lm_head() -> keras.layers.Layer - added:
method TFRoFormerForMaskedLM.build(input_shape=…) - added:
method TFRoFormerForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=… - added:
method TFRoFormerForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFRoFormerForMultipleChoice.build(input_shape=…) - added:
method TFRoFormerForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor |… - added:
method TFRoFormerForQuestionAnswering.build(input_shape=…) - added:
method TFRoFormerForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, start_positions: np.ndarray |… - added:
method TFRoFormerForSequenceClassification.build(input_shape=…) - added:
method TFRoFormerForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.… - added:
method TFRoFormerForTokenClassification.build(input_shape=…) - added:
method TFRoFormerForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Ten… - added:
method TFRoFormerLayer.build(input_shape=…) - added:
method TFRoFormerLayer.call(hidden_states: tf.Tensor, attention_mask: tf.Tensor, sinusoidal_pos: tf.Tensor, head_mask: tf.Tensor, output_attentions: bool, training: bool=…) -> tuple[tf.Tensor] - added:
method TFRoFormerModel.build(input_shape=…) - added:
method TFRoFormerModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFBaseModelOutpu…
- added:
class RoFormerTokenizer(PreTrainedTokenizer) - added:
constructor RoFormerTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RoFormerTokenizer.vocab_files_names - added:
method RoFormerTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method RoFormerTokenizer.convert_tokens_to_string(tokens) - added:
method RoFormerTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method RoFormerTokenizer.get_vocab() - added:
method RoFormerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property RoFormerTokenizer.do_lower_case - added:
property RoFormerTokenizer.vocab_size
- added:
class RoFormerTokenizerFast(PreTrainedTokenizerFast) - added:
constructor RoFormerTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field RoFormerTokenizerFast.slow_tokenizer_class - added:
field RoFormerTokenizerFast.vocab_files_names - added:
method RoFormerTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method RoFormerTokenizerFast.save_pretrained(save_directory, legacy_format=…, filename_prefix=…, push_to_hub=…, **kwargs) - added:
method RoFormerTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
class JiebaPreTokenizer - added:
constructor JiebaPreTokenizer(vocab) -> None - added:
method JiebaPreTokenizer.jieba_split(i: int, normalized_string: NormalizedString) -> list[NormalizedString] - added:
method JiebaPreTokenizer.pre_tokenize(pretok: PreTokenizedString)
- added:
export define_import_structure
- added:
class RTDetrConfig(PretrainedConfig) - added:
constructor RTDetrConfig(initializer_range=…, initializer_bias_prior_prob=…, layer_norm_eps=…, batch_norm_eps=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, freeze_backbone_batch_norms=…, backbone_kwargs=…, encoder_hidden_dim=…, encoder_in_channels=…, feat_strides=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, dropout=…, activation_dropout=…,… - added:
field RTDetrConfig.attribute_map - added:
field RTDetrConfig.layer_types - added:
field RTDetrConfig.model_type - added:
property RTDetrConfig.hidden_size -> int - added:
property RTDetrConfig.num_attention_heads -> int - added:
property RTDetrConfig.sub_configs - added:
static method RTDetrConfig.from_backbone_configs(backbone_config: PretrainedConfig, **kwargs)
- added:
class RTDetrResNetConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor RTDetrResNetConfig(num_channels=…, embedding_size=…, hidden_sizes=…, depths=…, layer_type=…, hidden_act=…, downsample_in_first_stage=…, downsample_in_bottleneck=…, out_features=…, out_indices=…, **kwargs) - added:
field RTDetrResNetConfig.layer_types - added:
field RTDetrResNetConfig.model_type
- added:
class RTDetrImageProcessor(BaseImageProcessor) - added:
constructor RTDetrImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: bool=…, do_pad: bool=…, pad_size: Opt… - added:
field RTDetrImageProcessor.model_input_names - added:
method RTDetrImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method RTDetrImageProcessor.pad(images: list[np.ndarray], annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_size: Optio… - added:
method RTDetrImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, use_focal_loss: bool=…) - added:
method RTDetrImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method RTDetrImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[… - added:
method RTDetrImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method RTDetrImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method RTDetrImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict
- added:
class RTDetrImageProcessorFast(BaseImageProcessorFast) - added:
constructor RTDetrImageProcessorFast(**kwargs: Unpack[RTDetrFastImageProcessorKwargs]) -> None - added:
field RTDetrImageProcessorFast.default_to_square - added:
field RTDetrImageProcessorFast.do_convert_annotations - added:
field RTDetrImageProcessorFast.do_normalize - added:
field RTDetrImageProcessorFast.do_pad - added:
field RTDetrImageProcessorFast.do_rescale - added:
field RTDetrImageProcessorFast.do_resize - added:
field RTDetrImageProcessorFast.format - added:
field RTDetrImageProcessorFast.image_mean - added:
field RTDetrImageProcessorFast.image_std - added:
field RTDetrImageProcessorFast.model_input_names - added:
field RTDetrImageProcessorFast.resample - added:
field RTDetrImageProcessorFast.size - added:
field RTDetrImageProcessorFast.valid_kwargs - added:
method RTDetrImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method RTDetrImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method RTDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, use_focal_loss: bool=…) - added:
method RTDetrImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method RTDetrImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[RTDetrFastImageProcessorKwargs]) -> BatchFeature - added:
method RTDetrImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method RTDetrImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…)
- added:
class RTDetrForObjectDetection(RTDetrPreTrainedModel) - added:
class RTDetrModel(RTDetrPreTrainedModel) - added:
class RTDetrPreTrainedModel(PreTrainedModel) - added:
constructor RTDetrForObjectDetection(config: RTDetrConfig) - added:
constructor RTDetrModel(config: RTDetrConfig) - added:
field RTDetrPreTrainedModel.base_model_prefix - added:
field RTDetrPreTrainedModel.config: RTDetrConfig - added:
field RTDetrPreTrainedModel.main_input_name - added:
method RTDetrForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]… - added:
method RTDetrModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[… - added:
method RTDetrModel.freeze_backbone() - added:
method RTDetrModel.generate_anchors(spatial_shapes=…, grid_size=…, device=…, dtype=…) - added:
method RTDetrModel.get_encoder() - added:
method RTDetrModel.unfreeze_backbone()
- added:
class RTDetrResNetBackbone(RTDetrResNetPreTrainedModel, BackboneMixin) - added:
class RTDetrResNetPreTrainedModel(PreTrainedModel) - added:
constructor RTDetrResNetBackbone(config) - added:
field RTDetrResNetBackbone.has_attentions - added:
field RTDetrResNetPreTrainedModel.base_model_prefix - added:
field RTDetrResNetPreTrainedModel.config: RTDetrResNetConfig - added:
field RTDetrResNetPreTrainedModel.main_input_name - added:
method RTDetrResNetBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput
- added:
class RTDetrImageProcessorFast(DetrImageProcessorFast) - added:
constructor RTDetrImageProcessorFast(**kwargs: Unpack[RTDetrFastImageProcessorKwargs]) -> None - added:
field RTDetrImageProcessorFast.default_to_square - added:
field RTDetrImageProcessorFast.do_convert_annotations - added:
field RTDetrImageProcessorFast.do_normalize - added:
field RTDetrImageProcessorFast.do_pad - added:
field RTDetrImageProcessorFast.do_rescale - added:
field RTDetrImageProcessorFast.do_resize - added:
field RTDetrImageProcessorFast.format - added:
field RTDetrImageProcessorFast.image_mean - added:
field RTDetrImageProcessorFast.image_std - added:
field RTDetrImageProcessorFast.model_input_names - added:
field RTDetrImageProcessorFast.resample - added:
field RTDetrImageProcessorFast.size - added:
field RTDetrImageProcessorFast.valid_kwargs - added:
method RTDetrImageProcessorFast.from_dict() - added:
method RTDetrImageProcessorFast.post_process() - added:
method RTDetrImageProcessorFast.post_process_instance() - added:
method RTDetrImageProcessorFast.post_process_instance_segmentation() - added:
method RTDetrImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, use_focal_loss: bool=…) - added:
method RTDetrImageProcessorFast.post_process_panoptic() - added:
method RTDetrImageProcessorFast.post_process_panoptic_segmentation() - added:
method RTDetrImageProcessorFast.post_process_segmentation() - added:
method RTDetrImageProcessorFast.post_process_semantic_segmentation() - added:
method RTDetrImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method RTDetrImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[RTDetrFastImageProcessorKwargs]) -> BatchFeature
- added:
export define_import_structure
- added:
class RTDetrV2Config(PretrainedConfig) - added:
constructor RTDetrV2Config(initializer_range=…, initializer_bias_prior_prob=…, layer_norm_eps=…, batch_norm_eps=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, freeze_backbone_batch_norms=…, backbone_kwargs=…, encoder_hidden_dim=…, encoder_in_channels=…, feat_strides=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, dropout=…, activation_dropout=… - added:
field RTDetrV2Config.attribute_map - added:
field RTDetrV2Config.layer_types - added:
field RTDetrV2Config.model_type - added:
property RTDetrV2Config.sub_configs - added:
static method RTDetrV2Config.from_backbone_configs(backbone_config: PretrainedConfig, **kwargs)
- added:
class RTDetrV2ForObjectDetection(RTDetrV2PreTrainedModel) - added:
class RTDetrV2Model(RTDetrV2PreTrainedModel) - added:
class RTDetrV2PreTrainedModel(PreTrainedModel) - added:
constructor RTDetrV2ForObjectDetection(config: RTDetrV2Config) - added:
constructor RTDetrV2Model(config: RTDetrV2Config) - added:
field RTDetrV2PreTrainedModel.base_model_prefix - added:
field RTDetrV2PreTrainedModel.config: RTDetrV2Config - added:
field RTDetrV2PreTrainedModel.main_input_name - added:
method RTDetrV2ForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[boo… - added:
method RTDetrV2Model.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Unio… - added:
method RTDetrV2Model.freeze_backbone() - added:
method RTDetrV2Model.generate_anchors(spatial_shapes=…, grid_size=…, device=…, dtype=…) - added:
method RTDetrV2Model.get_encoder() - added:
method RTDetrV2Model.unfreeze_backbone()
- added:
class RTDetrV2Config(PretrainedConfig) - added:
class RTDetrV2ForObjectDetection(RTDetrForObjectDetection, RTDetrV2PreTrainedModel) - added:
class RTDetrV2Model(RTDetrModel) - added:
class RTDetrV2PreTrainedModel(RTDetrPreTrainedModel) - added:
constructor RTDetrV2Config(initializer_range=…, initializer_bias_prior_prob=…, layer_norm_eps=…, batch_norm_eps=…, backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, freeze_backbone_batch_norms=…, backbone_kwargs=…, encoder_hidden_dim=…, encoder_in_channels=…, feat_strides=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, dropout=…, activation_dropout=… - added:
constructor RTDetrV2ForObjectDetection(config: RTDetrV2Config) - added:
constructor RTDetrV2Model(config: RTDetrV2Config) - added:
field RTDetrV2Config.attribute_map - added:
field RTDetrV2Config.layer_types - added:
field RTDetrV2Config.model_type - added:
property RTDetrV2Config.sub_configs - added:
static method RTDetrV2Config.from_backbone_configs(backbone_config: PretrainedConfig, **kwargs)
- added:
export define_import_structure
- added:
class RwkvConfig(PretrainedConfig) - added:
constructor RwkvConfig(vocab_size=…, context_length=…, hidden_size=…, num_hidden_layers=…, attention_hidden_size=…, intermediate_size=…, layer_norm_epsilon=…, bos_token_id=…, eos_token_id=…, rescale_every=…, tie_word_embeddings=…, use_cache=…, **kwargs) - added:
field RwkvConfig.attribute_map - added:
field RwkvConfig.model_type
- added:
class RwkvForCausalLM(RwkvPreTrainedModel, GenerationMixin) - added:
class RwkvModel(RwkvPreTrainedModel) - added:
class RwkvPreTrainedModel(PreTrainedModel) - added:
constructor RwkvForCausalLM(config) - added:
constructor RwkvModel(config) - added:
field RwkvPreTrainedModel.base_model_prefix - added:
field RwkvPreTrainedModel.config: RwkvConfig - added:
field RwkvPreTrainedModel.supports_gradient_checkpointing - added:
method RwkvForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, state: Optional[list[torch.FloatTensor]]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> Uni… - added:
method RwkvForCausalLM.get_output_embeddings() - added:
method RwkvForCausalLM.prepare_inputs_for_generation(input_ids, state=…, inputs_embeds=…, use_cache=…, **kwargs) - added:
method RwkvForCausalLM.set_output_embeddings(new_embeddings) - added:
method RwkvModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, state: Optional[list[torch.FloatTensor]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, RwkvOutput] - added:
method RwkvModel.get_input_embeddings() - added:
method RwkvModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class SamConfig(PretrainedConfig) - added:
class SamMaskDecoderConfig(PretrainedConfig) - added:
class SamPromptEncoderConfig(PretrainedConfig) - added:
class SamVisionConfig(PretrainedConfig) - added:
constructor SamConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, **kwargs) - added:
constructor SamMaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, layer_norm_eps=…, **kwargs) - added:
constructor SamPromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, **kwargs) - added:
constructor SamVisionConfig(hidden_size=…, output_channels=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, mlp_ratio=…, use_abs_pos=…, use_rel_pos=…, window_size=…, global_attn_indexes=…, num_pos_feats=…, mlp_dim=…, **kwargs) - added:
field SamConfig.model_type - added:
field SamConfig.sub_configs - added:
field SamMaskDecoderConfig.base_config_key - added:
field SamPromptEncoderConfig.base_config_key - added:
field SamVisionConfig.base_config_key - added:
field SamVisionConfig.model_type
- added:
class SamImageProcessor(BaseImageProcessor) - added:
constructor SamImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, mask_size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, pad_size: Optional[int]=…, mask_pad_size: Optiona… - added:
field SamImageProcessor.model_input_names - added:
method SamImageProcessor.filter_masks(masks, iou_scores, original_size, cropped_box_image, pred_iou_thresh=…, stability_score_thresh=…, mask_threshold=…, stability_score_offset=…, return_tensors=…) - added:
method SamImageProcessor.generate_crop_boxes(image, target_size, crop_n_layers: int=…, overlap_ratio: float=…, points_per_crop: Optional[int]=…, crop_n_points_downscale_factor: Optional[list[int]]=…, device: Optional[torch.device]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, return_tensors: str=…) - added:
method SamImageProcessor.pad_image(image: np.ndarray, pad_size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method SamImageProcessor.post_process_for_mask_generation(all_masks, all_scores, all_boxes, crops_nms_thresh, return_tensors=…) - added:
method SamImageProcessor.post_process_masks(masks, original_sizes, reshaped_input_sizes, mask_threshold=…, binarize=…, pad_size=…, return_tensors=…) - added:
method SamImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, mask_size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]… - added:
method SamImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class SamImageProcessorFast(BaseImageProcessorFast) - added:
constructor SamImageProcessorFast(**kwargs: Unpack[SamFastImageProcessorKwargs]) - added:
field SamImageProcessorFast.do_convert_rgb - added:
field SamImageProcessorFast.do_normalize - added:
field SamImageProcessorFast.do_pad - added:
field SamImageProcessorFast.do_rescale - added:
field SamImageProcessorFast.do_resize - added:
field SamImageProcessorFast.image_mean - added:
field SamImageProcessorFast.image_std - added:
field SamImageProcessorFast.mask_pad_size - added:
field SamImageProcessorFast.mask_size - added:
field SamImageProcessorFast.pad_size - added:
field SamImageProcessorFast.resample - added:
field SamImageProcessorFast.size - added:
field SamImageProcessorFast.valid_kwargs - added:
method SamImageProcessorFast.filter_masks(masks, iou_scores, original_size, cropped_box_image, pred_iou_thresh=…, stability_score_thresh=…, mask_threshold=…, stability_score_offset=…) - added:
method SamImageProcessorFast.generate_crop_boxes(image: torch.Tensor, target_size, crop_n_layers: int=…, overlap_ratio: float=…, points_per_crop: Optional[int]=…, crop_n_points_downscale_factor: Optional[list[int]]=…, device: Optional[torch.device]=…) - added:
method SamImageProcessorFast.post_process_for_mask_generation(all_masks, all_scores, all_boxes, crops_nms_thresh) - added:
method SamImageProcessorFast.post_process_masks(masks, original_sizes, reshaped_input_sizes, mask_threshold=…, binarize=…, pad_size=…) - added:
method SamImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[SamFastImageProcessorKwargs]) -> BatchFeature - added:
method SamImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F_t.InterpolationMode], **kwargs) -> torch.Tensor
- added:
class SamModel(SamPreTrainedModel) - added:
class SamPreTrainedModel(PreTrainedModel) - added:
class SamVisionModel(SamPreTrainedModel) - added:
constructor SamModel(config: SamConfig) - added:
constructor SamVisionModel(config: SamVisionConfig) - added:
field SamPreTrainedModel.base_model_prefix - added:
field SamPreTrainedModel.config: SamConfig - added:
field SamPreTrainedModel.main_input_name - added:
field SamPreTrainedModel.supports_gradient_checkpointing - added:
field SamVisionModel.config: SamVisionConfig - added:
field SamVisionModel.main_input_name - added:
method SamModel.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, target_embedding: Optional… - added:
method SamModel.get_image_embeddings(pixel_values, **kwargs: Unpack[TransformersKwargs]) - added:
method SamModel.get_image_wide_positional_embeddings() - added:
method SamModel.get_input_embeddings() - added:
method SamModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) - added:
method SamVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, SamVisionEncoderOutput] - added:
method SamVisionModel.get_input_embeddings() -> nn.Module
- added:
class TFSamModel(TFSamPreTrainedModel) - added:
class TFSamPreTrainedModel(TFPreTrainedModel) - added:
class TFSamVisionModel(TFSamPreTrainedModel) - added:
constructor TFSamModel(config, **kwargs) - added:
constructor TFSamVisionModel(config: SamVisionConfig, **kwargs) - added:
field TFSamPreTrainedModel.base_model_prefix - added:
field TFSamPreTrainedModel.config_class - added:
field TFSamPreTrainedModel.main_input_name - added:
field TFSamVisionModel.config_class - added:
field TFSamVisionModel.main_input_name - added:
method TFSamModel.build(input_shape=…) - added:
method TFSamModel.call(pixel_values: TFModelInputType | None=…, input_points: tf.Tensor | None=…, input_labels: tf.Tensor | None=…, input_boxes: tf.Tensor | None=…, input_masks: tf.Tensor | None=…, image_embeddings: tf.Tensor | None=…, multimask_output: bool=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…, **kwargs) -> TFSamIm… - added:
method TFSamModel.get_image_embeddings(pixel_values, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…) - added:
method TFSamModel.get_image_wide_positional_embeddings() - added:
method TFSamModel.get_input_embeddings() - added:
method TFSamModel.get_prompt_embeddings(input_points: tf.Tensor | None=…, input_labels: tf.Tensor | None=…, input_boxes: tf.Tensor | None=…, input_masks: tf.Tensor | None=…) - added:
method TFSamModel.serving_output(output: TFSamImageSegmentationOutput) -> TFSamImageSegmentationOutput - added:
method TFSamVisionModel.build(input_shape=…) - added:
method TFSamVisionModel.call(pixel_values: TFModelInputType | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…, **kwargs) -> TFSamVisionEncoderOutput | tuple[tf.Tensor] - added:
method TFSamVisionModel.get_input_embeddings()
- added:
class SamProcessor(ProcessorMixin) - added:
constructor SamProcessor(image_processor) - added:
field SamProcessor.attributes - added:
field SamProcessor.image_processor_class - added:
method SamProcessor.post_process_masks(*args, **kwargs) - added:
property SamProcessor.model_input_names
- added:
export define_import_structure
- added:
class Sam2Config(PretrainedConfig) - added:
class Sam2HieraDetConfig(PretrainedConfig) - added:
class Sam2MaskDecoderConfig(PretrainedConfig) - added:
class Sam2PromptEncoderConfig(PretrainedConfig) - added:
class Sam2VisionConfig(PretrainedConfig) - added:
constructor Sam2Config(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, **kwargs) - added:
constructor Sam2HieraDetConfig(hidden_size=…, num_attention_heads=…, num_channels=…, image_size=…, patch_kernel_size=…, patch_stride=…, patch_padding=…, query_stride=…, window_positional_embedding_background_size=…, num_query_pool_stages=…, blocks_per_stage=…, embed_dim_per_stage=…, num_attention_heads_per_stage=…, window_size_per_stage=…, global_attention_blocks=…, mlp_ratio=…, hidden_act=…, lay… - added:
constructor Sam2MaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, dynamic_multimask_via_stability=…, dynamic_multimask_stability_delta=…, dynamic_multimask_stability_thresh=…, **kwargs) - added:
constructor Sam2PromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, scale=…, **kwargs) - added:
constructor Sam2VisionConfig(backbone_config=…, backbone_channel_list=…, backbone_feature_sizes=…, fpn_hidden_size=…, fpn_kernel_size=…, fpn_stride=…, fpn_padding=…, fpn_top_down_levels=…, num_feature_levels=…, hidden_act=…, layer_norm_eps=…, initializer_range=…, **kwargs) - added:
field Sam2Config.model_type - added:
field Sam2Config.sub_configs - added:
field Sam2HieraDetConfig.base_config_key - added:
field Sam2HieraDetConfig.model_type - added:
field Sam2MaskDecoderConfig.base_config_key - added:
field Sam2PromptEncoderConfig.base_config_key - added:
field Sam2VisionConfig.base_config_key - added:
field Sam2VisionConfig.model_type - added:
field Sam2VisionConfig.sub_configs
- added:
class Sam2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Sam2ImageProcessorFast(**kwargs: Unpack[Sam2FastImageProcessorKwargs]) - added:
field Sam2ImageProcessorFast.do_convert_rgb - added:
field Sam2ImageProcessorFast.do_normalize - added:
field Sam2ImageProcessorFast.do_pad - added:
field Sam2ImageProcessorFast.do_rescale - added:
field Sam2ImageProcessorFast.do_resize - added:
field Sam2ImageProcessorFast.image_mean - added:
field Sam2ImageProcessorFast.image_std - added:
field Sam2ImageProcessorFast.mask_pad_size - added:
field Sam2ImageProcessorFast.mask_size - added:
field Sam2ImageProcessorFast.pad_size - added:
field Sam2ImageProcessorFast.resample - added:
field Sam2ImageProcessorFast.size - added:
field Sam2ImageProcessorFast.valid_kwargs - added:
method Sam2ImageProcessorFast.filter_masks(masks, iou_scores, original_size, cropped_box_image, pred_iou_thresh=…, stability_score_thresh=…, mask_threshold=…, stability_score_offset=…) - added:
method Sam2ImageProcessorFast.generate_crop_boxes(image: torch.Tensor, target_size, crop_n_layers: int=…, overlap_ratio: float=…, points_per_crop: Optional[int]=…, crop_n_points_downscale_factor: Optional[list[int]]=…, device: Optional[torch.device]=…) - added:
method Sam2ImageProcessorFast.pad_image() - added:
method Sam2ImageProcessorFast.post_process_for_mask_generation(all_masks, all_scores, all_boxes, crops_nms_thresh) - added:
method Sam2ImageProcessorFast.post_process_masks(masks, original_sizes, mask_threshold=…, binarize=…, max_hole_area=…, max_sprinkle_area=…, apply_non_overlapping_constraints=…, **kwargs) - added:
method Sam2ImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[Sam2FastImageProcessorKwargs]) -> BatchFeature
- added:
class Sam2HieraDetModel(Sam2PreTrainedModel) - added:
class Sam2Model(Sam2PreTrainedModel) - added:
class Sam2PreTrainedModel(PreTrainedModel) - added:
class Sam2VisionModel(Sam2PreTrainedModel) - added:
constructor Sam2HieraDetModel(config: Sam2HieraDetConfig) - added:
constructor Sam2Model(config: Sam2Config) - added:
constructor Sam2VisionModel(config: Sam2VisionConfig) - added:
field Sam2HieraDetModel.config_class - added:
field Sam2HieraDetModel.main_input_name - added:
field Sam2PreTrainedModel.base_model_prefix - added:
field Sam2PreTrainedModel.config_class - added:
field Sam2PreTrainedModel.main_input_name - added:
field Sam2VisionModel.config_class - added:
field Sam2VisionModel.main_input_name - added:
method Sam2HieraDetModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Sam2HieraDetModelOutput] - added:
method Sam2HieraDetModel.get_input_embeddings() - added:
method Sam2Model.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, target_embedding: Optiona… - added:
method Sam2Model.get_image_embeddings(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> list[torch.Tensor] - added:
method Sam2Model.get_image_features(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> tuple[list[torch.Tensor], list[torch.Tensor], Optional[tuple[torch.FloatTensor, ...]], Optional[tuple[torch.FloatTensor, ...]]] - added:
method Sam2Model.get_image_wide_positional_embeddings() -> torch.Tensor - added:
method Sam2Model.get_input_embeddings() - added:
method Sam2Model.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) - added:
method Sam2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Sam2VisionEncoderOutput] - added:
method Sam2VisionModel.get_input_embeddings()
- added:
class Sam2HieraDetModel(Sam2PreTrainedModel) - added:
class Sam2ImageProcessorFast(SamImageProcessorFast) - added:
class Sam2Model(SamModel) - added:
class Sam2PreTrainedModel(PreTrainedModel) - added:
class Sam2VisionModel(Sam2PreTrainedModel) - added:
constructor Sam2HieraDetModel(config: Sam2HieraDetConfig) - added:
constructor Sam2ImageProcessorFast(**kwargs: Unpack[Sam2FastImageProcessorKwargs]) - added:
constructor Sam2Model(config: Sam2Config) - added:
constructor Sam2VisionModel(config: Sam2VisionConfig) - added:
field Sam2HieraDetModel.config_class - added:
field Sam2HieraDetModel.main_input_name - added:
field Sam2ImageProcessorFast.do_convert_rgb - added:
field Sam2ImageProcessorFast.do_normalize - added:
field Sam2ImageProcessorFast.do_pad - added:
field Sam2ImageProcessorFast.do_rescale - added:
field Sam2ImageProcessorFast.do_resize - added:
field Sam2ImageProcessorFast.image_mean - added:
field Sam2ImageProcessorFast.image_std - added:
field Sam2ImageProcessorFast.mask_pad_size - added:
field Sam2ImageProcessorFast.mask_size - added:
field Sam2ImageProcessorFast.pad_size - added:
field Sam2ImageProcessorFast.resample - added:
field Sam2ImageProcessorFast.size - added:
field Sam2ImageProcessorFast.valid_kwargs - added:
field Sam2PreTrainedModel.base_model_prefix - added:
field Sam2PreTrainedModel.config_class - added:
field Sam2PreTrainedModel.main_input_name - added:
field Sam2VisionModel.config_class - added:
field Sam2VisionModel.main_input_name - added:
method Sam2HieraDetModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Sam2HieraDetModelOutput] - added:
method Sam2HieraDetModel.get_input_embeddings() - added:
method Sam2ImageProcessorFast.pad_image() - added:
method Sam2ImageProcessorFast.post_process_masks(masks, original_sizes, mask_threshold=…, binarize=…, max_hole_area=…, max_sprinkle_area=…, apply_non_overlapping_constraints=…, **kwargs) - added:
method Sam2ImageProcessorFast.resize() - added:
method Sam2Model.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, target_embedding: Optiona… - added:
method Sam2Model.get_image_embeddings(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> list[torch.Tensor] - added:
method Sam2Model.get_image_features(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> tuple[list[torch.Tensor], list[torch.Tensor], Optional[tuple[torch.FloatTensor, ...]], Optional[tuple[torch.FloatTensor, ...]]] - added:
method Sam2Model.get_image_wide_positional_embeddings() -> torch.Tensor - added:
method Sam2VisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, Sam2VisionEncoderOutput] - added:
method Sam2VisionModel.get_input_embeddings()
- added:
class Sam2Processor(ProcessorMixin) - added:
constructor Sam2Processor(image_processor, target_size: Optional[int]=…, point_pad_value: int=…, **kwargs) - added:
field Sam2Processor.attributes - added:
field Sam2Processor.image_processor_class - added:
method Sam2Processor.post_process_masks(masks, original_sizes, mask_threshold=…, binarize=…, max_hole_area=…, max_sprinkle_area=…, apply_non_overlapping_constraints=…, **kwargs)
- added:
export define_import_structure
- added:
class Sam2VideoConfig(PretrainedConfig) - added:
class Sam2VideoMaskDecoderConfig(PretrainedConfig) - added:
class Sam2VideoPromptEncoderConfig(PretrainedConfig) - added:
constructor Sam2VideoConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, num_maskmem=…, image_size=…, sigmoid_scale_for_mem_enc=…, sigmoid_bias_for_mem_enc=…, enable_occlusion_spatial_embedding=…, multimask_output_in_sam=…, multimask_min_pt_num=…, multimask_max_pt_num=…, multimask_output_for_tracking=…, max_object_pointers_in_encoder=…, enable_temporal_pos… - added:
constructor Sam2VideoMaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, dynamic_multimask_via_stability=…, dynamic_multimask_stability_delta=…, dynamic_multimask_stability_thresh=…, **kwargs) - added:
constructor Sam2VideoPromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, scale=…, **kwargs) - added:
field Sam2VideoConfig.model_type - added:
field Sam2VideoConfig.sub_configs - added:
field Sam2VideoMaskDecoderConfig.base_config_key - added:
field Sam2VideoPromptEncoderConfig.base_config_key
- added:
class Sam2VideoInferenceSession - added:
class Sam2VideoModel(Sam2VideoPreTrainedModel) - added:
class Sam2VideoPreTrainedModel(PreTrainedModel) - added:
constructor Sam2VideoInferenceSession(video: Optional[torch.FloatTensor]=…, video_height: Optional[int]=…, video_width: Optional[int]=…, inference_device: Union[torch.device, str]=…, inference_state_device: Union[torch.device, str]=…, video_storage_device: Union[torch.device, str]=…, dtype: Union[torch.dtype, str]=…, max_vision_features_cache_size: int=…) - added:
constructor Sam2VideoModel(config: Sam2VideoConfig) - added:
field Sam2VideoPreTrainedModel.base_model_prefix - added:
field Sam2VideoPreTrainedModel.config_class - added:
field Sam2VideoPreTrainedModel.main_input_name - added:
method Sam2VideoInferenceSession.add_mask_inputs(obj_idx: int, frame_idx: int, inputs: torch.Tensor) - added:
method Sam2VideoInferenceSession.add_new_frame(pixel_values: torch.Tensor, frame_idx: Optional[int]=…) -> int - added:
method Sam2VideoInferenceSession.add_point_inputs(obj_idx: int, frame_idx: int, inputs: dict) - added:
method Sam2VideoInferenceSession.get_frame(frame_idx: int) -> torch.Tensor - added:
method Sam2VideoInferenceSession.get_obj_num() -> int - added:
method Sam2VideoInferenceSession.get_output(obj_idx: int, frame_idx: int, output_key: str, is_conditioning_frame: bool=…) - added:
method Sam2VideoInferenceSession.obj_id_to_idx(obj_id: int) -> int - added:
method Sam2VideoInferenceSession.obj_idx_to_id(obj_idx: int) -> int - added:
method Sam2VideoInferenceSession.remove_mask_inputs(obj_idx: int, frame_idx: int) - added:
method Sam2VideoInferenceSession.remove_point_inputs(obj_idx: int, frame_idx: int) - added:
method Sam2VideoInferenceSession.reset_inference_session() - added:
method Sam2VideoInferenceSession.reset_tracking_data() - added:
method Sam2VideoInferenceSession.store_output(obj_idx: int, frame_idx: int, output_key: Optional[str]=…, output_value: Optional[Union[torch.Tensor, dict]]=…, is_conditioning_frame: bool=…) - added:
method Sam2VideoModel.forward(inference_session: Sam2VideoInferenceSession, frame_idx: Optional[int]=…, frame: Optional[torch.Tensor]=…, reverse: bool=…) -> Sam2VideoSegmentationOutput - added:
method Sam2VideoModel.get_image_embeddings(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> list[torch.Tensor] - added:
method Sam2VideoModel.get_image_features(pixel_values: torch.FloatTensor, **kwargs: Unpack[TransformersKwargs]) -> tuple[list[torch.Tensor], list[torch.Tensor], Optional[tuple[torch.FloatTensor, ...]], Optional[tuple[torch.FloatTensor, ...]]] - added:
method Sam2VideoModel.get_image_wide_positional_embeddings() -> torch.Tensor - added:
method Sam2VideoModel.get_input_embeddings() - added:
method Sam2VideoModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Sam2VideoModel.propagate_in_video_iterator(inference_session: Sam2VideoInferenceSession, start_frame_idx: Optional[int]=…, max_frame_num_to_track: Optional[int]=…, reverse: bool=…) -> Iterator[Sam2VideoSegmentationOutput] - added:
property Sam2VideoInferenceSession.num_frames -> Optional[int]
- added:
class Sam2VideoConfig(PretrainedConfig) - added:
class Sam2VideoInferenceSession - added:
class Sam2VideoMaskDecoderConfig(Sam2MaskDecoderConfig) - added:
class Sam2VideoModel(Sam2Model) - added:
class Sam2VideoPreTrainedModel(PreTrainedModel) - added:
class Sam2VideoProcessor(Sam2Processor) - added:
class Sam2VideoPromptEncoderConfig(Sam2PromptEncoderConfig) - added:
constructor Sam2VideoConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, num_maskmem=…, image_size=…, sigmoid_scale_for_mem_enc=…, sigmoid_bias_for_mem_enc=…, enable_occlusion_spatial_embedding=…, multimask_output_in_sam=…, multimask_min_pt_num=…, multimask_max_pt_num=…, multimask_output_for_tracking=…, max_object_pointers_in_encoder=…, enable_temporal_pos… - added:
constructor Sam2VideoInferenceSession(video: Optional[torch.FloatTensor]=…, video_height: Optional[int]=…, video_width: Optional[int]=…, inference_device: Union[torch.device, str]=…, inference_state_device: Union[torch.device, str]=…, video_storage_device: Union[torch.device, str]=…, dtype: Union[torch.dtype, str]=…, max_vision_features_cache_size: int=…) - added:
constructor Sam2VideoModel(config: Sam2VideoConfig) - added:
constructor Sam2VideoProcessor(image_processor, video_processor, target_size: Optional[int]=…, point_pad_value: int=…, **kwargs) - added:
field Sam2VideoConfig.model_type - added:
field Sam2VideoConfig.sub_configs - added:
field Sam2VideoPreTrainedModel.base_model_prefix - added:
field Sam2VideoPreTrainedModel.config_class - added:
field Sam2VideoPreTrainedModel.main_input_name - added:
field Sam2VideoProcessor.attributes - added:
field Sam2VideoProcessor.image_processor_class - added:
field Sam2VideoProcessor.video_processor_class - added:
method Sam2VideoInferenceSession.add_mask_inputs(obj_idx: int, frame_idx: int, inputs: torch.Tensor) - added:
method Sam2VideoInferenceSession.add_new_frame(pixel_values: torch.Tensor, frame_idx: Optional[int]=…) -> int - added:
method Sam2VideoInferenceSession.add_point_inputs(obj_idx: int, frame_idx: int, inputs: dict) - added:
method Sam2VideoInferenceSession.get_frame(frame_idx: int) -> torch.Tensor - added:
method Sam2VideoInferenceSession.get_obj_num() -> int - added:
method Sam2VideoInferenceSession.get_output(obj_idx: int, frame_idx: int, output_key: str, is_conditioning_frame: bool=…) - added:
method Sam2VideoInferenceSession.obj_id_to_idx(obj_id: int) -> int - added:
method Sam2VideoInferenceSession.obj_idx_to_id(obj_idx: int) -> int - added:
method Sam2VideoInferenceSession.remove_mask_inputs(obj_idx: int, frame_idx: int) - added:
method Sam2VideoInferenceSession.remove_point_inputs(obj_idx: int, frame_idx: int) - added:
method Sam2VideoInferenceSession.reset_inference_session() - added:
method Sam2VideoInferenceSession.reset_tracking_data() - added:
method Sam2VideoInferenceSession.store_output(obj_idx: int, frame_idx: int, output_key: Optional[str]=…, output_value: Optional[Union[torch.Tensor, dict]]=…, is_conditioning_frame: bool=…) - added:
method Sam2VideoModel.forward(inference_session: Sam2VideoInferenceSession, frame_idx: Optional[int]=…, frame: Optional[torch.Tensor]=…, reverse: bool=…) -> Sam2VideoSegmentationOutput - added:
method Sam2VideoModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) -> tuple[torch.Tensor, torch.Tensor] - added:
method Sam2VideoModel.propagate_in_video_iterator(inference_session: Sam2VideoInferenceSession, start_frame_idx: Optional[int]=…, max_frame_num_to_track: Optional[int]=…, reverse: bool=…) -> Iterator[Sam2VideoSegmentationOutput] - added:
method Sam2VideoProcessor.add_inputs_to_inference_session(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: Union[list[int], int], input_points: Optional[Union[list[list[list[list[float]]]], torch.Tensor]]=…, input_labels: Optional[Union[list[list[list[int]]], torch.Tensor]]=…, input_boxes: Optional[Union[list[list[list[float]]], torch.Tensor]]=…, input_masks: Optional[Union[… - added:
method Sam2VideoProcessor.init_video_session(video: Optional[VideoInput]=…, inference_device: Union[str, torch.device]=…, inference_state_device: Union[str, torch.device]=…, processing_device: Union[str, torch.device]=…, video_storage_device: Union[str, torch.device]=…, max_vision_features_cache_size: int=…, dtype: torch.dtype=…) - added:
method Sam2VideoProcessor.process_new_mask_for_video_frame(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: list[int], input_masks: Union[np.ndarray, torch.Tensor, list[np.ndarray], list[torch.Tensor]]) - added:
method Sam2VideoProcessor.process_new_points_or_boxes_for_video_frame(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: list[int], input_points: Optional[Union[list[list[list[list[float]]]], torch.Tensor]]=…, input_labels: Optional[Union[list[list[list[int]]], torch.Tensor]]=…, input_boxes: Optional[Union[list[list[list[float]]], torch.Tensor]]=…, original_size: Optional[tupl… - added:
property Sam2VideoInferenceSession.num_frames -> Optional[int]
- added:
class Sam2VideoProcessor(ProcessorMixin) - added:
constructor Sam2VideoProcessor(image_processor, video_processor, target_size: Optional[int]=…, point_pad_value: int=…, **kwargs) - added:
field Sam2VideoProcessor.attributes - added:
field Sam2VideoProcessor.image_processor_class - added:
field Sam2VideoProcessor.video_processor_class - added:
method Sam2VideoProcessor.add_inputs_to_inference_session(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: Union[list[int], int], input_points: Optional[Union[list[list[list[list[float]]]], torch.Tensor]]=…, input_labels: Optional[Union[list[list[list[int]]], torch.Tensor]]=…, input_boxes: Optional[Union[list[list[list[float]]], torch.Tensor]]=…, input_masks: Optional[Union[… - added:
method Sam2VideoProcessor.init_video_session(video: Optional[VideoInput]=…, inference_device: Union[str, torch.device]=…, inference_state_device: Union[str, torch.device]=…, processing_device: Union[str, torch.device]=…, video_storage_device: Union[str, torch.device]=…, max_vision_features_cache_size: int=…, dtype: torch.dtype=…) - added:
method Sam2VideoProcessor.post_process_masks(masks, original_sizes, mask_threshold=…, binarize=…, max_hole_area=…, max_sprinkle_area=…, apply_non_overlapping_constraints=…, **kwargs) - added:
method Sam2VideoProcessor.process_new_mask_for_video_frame(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: list[int], input_masks: Union[np.ndarray, torch.Tensor, list[np.ndarray], list[torch.Tensor]]) - added:
method Sam2VideoProcessor.process_new_points_or_boxes_for_video_frame(inference_session: Sam2VideoInferenceSession, frame_idx: int, obj_ids: list[int], input_points: Optional[Union[list[list[list[list[float]]]], torch.Tensor]]=…, input_labels: Optional[Union[list[list[list[int]]], torch.Tensor]]=…, input_boxes: Optional[Union[list[list[list[float]]], torch.Tensor]]=…, original_size: Optional[tupl…
- added:
class Sam2VideoVideoProcessor(BaseVideoProcessor) - added:
field Sam2VideoVideoProcessor.do_convert_rgb - added:
field Sam2VideoVideoProcessor.do_normalize - added:
field Sam2VideoVideoProcessor.do_rescale - added:
field Sam2VideoVideoProcessor.do_resize - added:
field Sam2VideoVideoProcessor.image_mean - added:
field Sam2VideoVideoProcessor.image_std - added:
field Sam2VideoVideoProcessor.resample - added:
field Sam2VideoVideoProcessor.size - added:
method Sam2VideoVideoProcessor.post_process_masks(masks, original_sizes, reshaped_input_sizes, mask_threshold=…, binarize=…, pad_size=…)
- added:
export define_import_structure
- added:
class SamHQConfig(PretrainedConfig) - added:
class SamHQMaskDecoderConfig(PretrainedConfig) - added:
class SamHQPromptEncoderConfig(PretrainedConfig) - added:
class SamHQVisionConfig(PretrainedConfig) - added:
constructor SamHQConfig(vision_config=…, prompt_encoder_config=…, mask_decoder_config=…, initializer_range=…, **kwargs) - added:
constructor SamHQMaskDecoderConfig(hidden_size=…, hidden_act=…, mlp_dim=…, num_hidden_layers=…, num_attention_heads=…, attention_downsample_rate=…, num_multimask_outputs=…, iou_head_depth=…, iou_head_hidden_dim=…, layer_norm_eps=…, vit_dim=…, **kwargs) - added:
constructor SamHQPromptEncoderConfig(hidden_size=…, image_size=…, patch_size=…, mask_input_channels=…, num_point_embeddings=…, hidden_act=…, layer_norm_eps=…, **kwargs) - added:
constructor SamHQVisionConfig(hidden_size=…, output_channels=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, qkv_bias=…, mlp_ratio=…, use_abs_pos=…, use_rel_pos=…, window_size=…, global_attn_indexes=…, num_pos_feats=…, mlp_dim=…, **kwargs) - added:
field SamHQConfig.model_type - added:
field SamHQConfig.sub_configs - added:
field SamHQMaskDecoderConfig.base_config_key - added:
field SamHQPromptEncoderConfig.base_config_key - added:
field SamHQVisionConfig.base_config_key - added:
field SamHQVisionConfig.model_type
- added:
class SamHQModel(SamHQPreTrainedModel) - added:
class SamHQPreTrainedModel(PreTrainedModel) - added:
class SamHQVisionModel(SamHQPreTrainedModel) - added:
constructor SamHQModel(config) - added:
constructor SamHQVisionModel(config: SamHQVisionConfig) - added:
field SamHQPreTrainedModel.base_model_prefix - added:
field SamHQPreTrainedModel.config: SamHQConfig - added:
field SamHQPreTrainedModel.main_input_name - added:
field SamHQPreTrainedModel.supports_gradient_checkpointing - added:
field SamHQVisionModel.config: SamHQVisionConfig - added:
field SamHQVisionModel.main_input_name - added:
method SamHQModel.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, hq_token_only: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, t… - added:
method SamHQModel.get_image_embeddings(pixel_values) - added:
method SamHQModel.get_image_wide_positional_embeddings() - added:
method SamHQModel.get_input_embeddings() - added:
method SamHQModel.get_prompt_embeddings(input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…) - added:
method SamHQVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, SamHQVisionEncoderOutput] - added:
method SamHQVisionModel.get_input_embeddings() -> nn.Module
- added:
class SamHQConfig(SamConfig) - added:
class SamHQMaskDecoderConfig(SamMaskDecoderConfig) - added:
class SamHQModel(SamModel) - added:
class SamHQPreTrainedModel(SamPreTrainedModel) - added:
class SamHQPromptEncoderConfig(SamPromptEncoderConfig) - added:
class SamHQVisionConfig(SamVisionConfig) - added:
class SamHQVisionModel(SamVisionModel) - added:
constructor SamHQMaskDecoderConfig(vit_dim=…, **super_kwargs) - added:
constructor SamHQModel(config) - added:
method SamHQModel.forward(pixel_values: Optional[torch.FloatTensor]=…, input_points: Optional[torch.FloatTensor]=…, input_labels: Optional[torch.LongTensor]=…, input_boxes: Optional[torch.FloatTensor]=…, input_masks: Optional[torch.LongTensor]=…, image_embeddings: Optional[torch.FloatTensor]=…, multimask_output: bool=…, hq_token_only: bool=…, attention_similarity: Optional[torch.FloatTensor]=…, t… - added:
method SamHQModel.get_image_embeddings(pixel_values)
- added:
class SamHQProcessor(ProcessorMixin) - added:
constructor SamHQProcessor(image_processor) - added:
field SamHQProcessor.attributes - added:
field SamHQProcessor.image_processor_class - added:
method SamHQProcessor.post_process_masks(*args, **kwargs) - added:
property SamHQProcessor.model_input_names
- added:
export define_import_structure
- added:
class SeamlessM4TConfig(PretrainedConfig) - added:
constructor SeamlessM4TConfig(vocab_size=…, t2u_vocab_size=…, hidden_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, max_position_embeddings=…, is_encoder_decoder=…, encoder_layerdrop=…, decoder_layerdrop=…, activation_function=…, dropout=…, attention_dropout=…, activation_dropout=…, scale_embedding=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, d… - added:
field SeamlessM4TConfig.model_type
- added:
class SeamlessM4TFeatureExtractor(SequenceFeatureExtractor) - added:
constructor SeamlessM4TFeatureExtractor(feature_size=…, sampling_rate=…, num_mel_bins=…, padding_value=…, stride=…, **kwargs) - added:
field SeamlessM4TFeatureExtractor.model_input_names - added:
static method SeamlessM4TFeatureExtractor.zero_mean_unit_var_norm(input_values: list[np.ndarray], attention_mask: list[np.ndarray], padding_value: float=…) -> list[np.ndarray]
- added:
class SeamlessM4TCodeHifiGan(PreTrainedModel) - added:
class SeamlessM4TForSpeechToSpeech(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4TForSpeechToText(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4TForTextToSpeech(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4TForTextToText(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4THifiGan(nn.Module) - added:
class SeamlessM4TModel(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4TPreTrainedModel(PreTrainedModel) - added:
class SeamlessM4TTextToUnitForConditionalGeneration(SeamlessM4TPreTrainedModel, GenerationMixin) - added:
class SeamlessM4TTextToUnitModel(SeamlessM4TPreTrainedModel) - added:
constructor SeamlessM4TCodeHifiGan(config) - added:
constructor SeamlessM4TForSpeechToSpeech(config) - added:
constructor SeamlessM4TForSpeechToText(config: SeamlessM4TConfig) - added:
constructor SeamlessM4TForTextToSpeech(config: SeamlessM4TConfig) - added:
constructor SeamlessM4TForTextToText(config: SeamlessM4TConfig) - added:
constructor SeamlessM4THifiGan(config: SeamlessM4TConfig) - added:
constructor SeamlessM4TModel(config, current_modality=…) - added:
constructor SeamlessM4TTextToUnitForConditionalGeneration(config: SeamlessM4TConfig, embed_tokens_decoder: Optional[nn.Embedding]=…) - added:
constructor SeamlessM4TTextToUnitModel(config: SeamlessM4TConfig, embed_tokens_decoder: Optional[nn.Embedding]=…) - added:
field SeamlessM4TCodeHifiGan.config: SeamlessM4TConfig - added:
field SeamlessM4TCodeHifiGan.main_input_name - added:
field SeamlessM4TForSpeechToSpeech.main_input_name - added:
field SeamlessM4TForSpeechToText.main_input_name - added:
field SeamlessM4TForTextToSpeech.main_input_name - added:
field SeamlessM4TForTextToText.main_input_name - added:
field SeamlessM4TPreTrainedModel.base_model_prefix - added:
field SeamlessM4TPreTrainedModel.config: SeamlessM4TConfig - added:
field SeamlessM4TPreTrainedModel.supports_gradient_checkpointing - added:
method SeamlessM4TCodeHifiGan.apply_weight_norm() - added:
method SeamlessM4TCodeHifiGan.forward(input_ids: torch.LongTensor, spkr_id: torch.Tensor, lang_id: torch.Tensor) -> tuple[torch.Tensor] - added:
method SeamlessM4TCodeHifiGan.remove_weight_norm() - added:
method SeamlessM4TForSpeechToSpeech.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embed… - added:
method SeamlessM4TForSpeechToSpeech.generate(input_features: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, spkr_id: Optional[int]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4TGenerationOutput] - added:
method SeamlessM4TForSpeechToSpeech.get_decoder() - added:
method SeamlessM4TForSpeechToSpeech.get_encoder() - added:
method SeamlessM4TForSpeechToSpeech.get_input_embeddings() - added:
method SeamlessM4TForSpeechToSpeech.set_input_embeddings(value) - added:
method SeamlessM4TForSpeechToText.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds:… - added:
method SeamlessM4TForSpeechToText.generate(input_features=…, tgt_lang=…, generation_config=…, logits_processor=…, stopping_criteria=…, prefix_allowed_tokens_fn=…, synced_gpus=…, **kwargs) - added:
method SeamlessM4TForSpeechToText.get_decoder() - added:
method SeamlessM4TForSpeechToText.get_encoder() - added:
method SeamlessM4TForSpeechToText.get_input_embeddings() - added:
method SeamlessM4TForSpeechToText.set_input_embeddings(value) - added:
method SeamlessM4TForTextToSpeech.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Opti… - added:
method SeamlessM4TForTextToSpeech.generate(input_ids: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, spkr_id: Optional[int]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4TGenerationOutput] - added:
method SeamlessM4TForTextToSpeech.get_decoder() - added:
method SeamlessM4TForTextToSpeech.get_encoder() - added:
method SeamlessM4TForTextToSpeech.get_input_embeddings() - added:
method SeamlessM4TForTextToSpeech.set_input_embeddings(value) - added:
method SeamlessM4TForTextToText.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Option… - added:
method SeamlessM4TForTextToText.generate(input_ids=…, tgt_lang=…, generation_config=…, logits_processor=…, stopping_criteria=…, prefix_allowed_tokens_fn=…, synced_gpus=…, **kwargs) - added:
method SeamlessM4TForTextToText.get_decoder() - added:
method SeamlessM4TForTextToText.get_encoder() - added:
method SeamlessM4TForTextToText.get_input_embeddings() - added:
method SeamlessM4TForTextToText.set_input_embeddings(value) - added:
method SeamlessM4THifiGan.forward(input_embeds: torch.FloatTensor) -> torch.FloatTensor - added:
method SeamlessM4TModel.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatT… - added:
method SeamlessM4TModel.generate(input_ids: Optional[torch.Tensor]=…, input_features: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, spkr_id: Optional[int]=…, generate_speech: Optional[bool]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4TGenerationOutput] - added:
method SeamlessM4TModel.get_encoder() - added:
method SeamlessM4TModel.get_input_embeddings() - added:
method SeamlessM4TModel.set_input_embeddings(value) - added:
method SeamlessM4TModel.set_modality(modality=…) - added:
method SeamlessM4TPreTrainedModel.compute_last_hidden_states_per_sample(hidden_states: tuple[tuple[torch.Tensor]], beam_indices: Optional[torch.Tensor]=…) -> torch.Tensor - added:
method SeamlessM4TTextToUnitForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_… - added:
method SeamlessM4TTextToUnitForConditionalGeneration.get_decoder() - added:
method SeamlessM4TTextToUnitForConditionalGeneration.get_encoder() - added:
method SeamlessM4TTextToUnitForConditionalGeneration.get_input_embeddings() - added:
method SeamlessM4TTextToUnitForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method SeamlessM4TTextToUnitForConditionalGeneration.set_input_embeddings(value) - added:
method SeamlessM4TTextToUnitModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Opti…
- added:
class SeamlessM4TProcessor(ProcessorMixin) - added:
constructor SeamlessM4TProcessor(feature_extractor, tokenizer) - added:
field SeamlessM4TProcessor.feature_extractor_class - added:
field SeamlessM4TProcessor.tokenizer_class - added:
field SeamlessM4TProcessor.valid_processor_kwargs
- added:
class SeamlessM4TTokenizer(PreTrainedTokenizer) - added:
constructor SeamlessM4TTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, tokenizer_file=…, src_lang=…, tgt_lang=…, sp_model_kwargs: Optional[dict[str, Any]]=…, additional_special_tokens=…, add_prefix_space=…, **kwargs) - added:
field SeamlessM4TTokenizer.model_input_names - added:
field SeamlessM4TTokenizer.prefix_tokens: list[int] - added:
field SeamlessM4TTokenizer.suffix_tokens: list[int] - added:
field SeamlessM4TTokenizer.vocab_files_names - added:
method SeamlessM4TTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SeamlessM4TTokenizer.convert_tokens_to_string(tokens) - added:
method SeamlessM4TTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SeamlessM4TTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method SeamlessM4TTokenizer.get_spm_processor(from_slow=…) - added:
method SeamlessM4TTokenizer.get_vocab() - added:
method SeamlessM4TTokenizer.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method SeamlessM4TTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method SeamlessM4TTokenizer.set_src_lang_special_tokens(src_lang) -> None - added:
method SeamlessM4TTokenizer.set_tgt_lang_special_tokens(lang: str) -> None - added:
method SeamlessM4TTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
property SeamlessM4TTokenizer.src_lang -> str - added:
property SeamlessM4TTokenizer.tgt_lang -> str - added:
property SeamlessM4TTokenizer.unk_token_length - added:
property SeamlessM4TTokenizer.vocab_size
- added:
class SeamlessM4TTokenizerFast(PreTrainedTokenizerFast) - added:
constructor SeamlessM4TTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, src_lang=…, tgt_lang=…, additional_special_tokens=…, **kwargs) - added:
field SeamlessM4TTokenizerFast.model_input_names - added:
field SeamlessM4TTokenizerFast.prefix_tokens: list[int] - added:
field SeamlessM4TTokenizerFast.slow_tokenizer_class - added:
field SeamlessM4TTokenizerFast.suffix_tokens: list[int] - added:
field SeamlessM4TTokenizerFast.vocab_files_names - added:
method SeamlessM4TTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SeamlessM4TTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SeamlessM4TTokenizerFast.prepare_seq2seq_batch(src_texts: list[str], src_lang: str=…, tgt_texts: Optional[list[str]]=…, tgt_lang: str=…, **kwargs) -> BatchEncoding - added:
method SeamlessM4TTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method SeamlessM4TTokenizerFast.set_src_lang_special_tokens(src_lang) -> None - added:
method SeamlessM4TTokenizerFast.set_tgt_lang_special_tokens(lang: str) -> None - added:
property SeamlessM4TTokenizerFast.src_lang -> str - added:
property SeamlessM4TTokenizerFast.tgt_lang -> str
- added:
export define_import_structure
- added:
class SeamlessM4Tv2Config(PretrainedConfig) - added:
constructor SeamlessM4Tv2Config(vocab_size=…, t2u_vocab_size=…, char_vocab_size=…, hidden_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, max_position_embeddings=…, is_encoder_decoder=…, encoder_layerdrop=…, decoder_layerdrop=…, activation_function=…, dropout=…, attention_dropout=…, activation_dropout=…, scale_embedding=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…… - added:
field SeamlessM4Tv2Config.model_type
- added:
class SeamlessM4Tv2ForSpeechToSpeech(SeamlessM4Tv2PreTrainedModel, GenerationMixin) - added:
class SeamlessM4Tv2ForSpeechToText(SeamlessM4Tv2PreTrainedModel, GenerationMixin) - added:
class SeamlessM4Tv2ForTextToSpeech(SeamlessM4Tv2PreTrainedModel, GenerationMixin) - added:
class SeamlessM4Tv2ForTextToText(SeamlessM4Tv2PreTrainedModel, GenerationMixin) - added:
class SeamlessM4Tv2Model(SeamlessM4Tv2PreTrainedModel, GenerationMixin) - added:
class SeamlessM4Tv2PreTrainedModel(PreTrainedModel) - added:
constructor SeamlessM4Tv2ForSpeechToSpeech(config) - added:
constructor SeamlessM4Tv2ForSpeechToText(config: SeamlessM4Tv2Config) - added:
constructor SeamlessM4Tv2ForTextToSpeech(config: SeamlessM4Tv2Config) - added:
constructor SeamlessM4Tv2ForTextToText(config: SeamlessM4Tv2Config) - added:
constructor SeamlessM4Tv2Model(config, current_modality=…) - added:
field SeamlessM4Tv2ForSpeechToSpeech.main_input_name - added:
field SeamlessM4Tv2ForSpeechToText.main_input_name - added:
field SeamlessM4Tv2ForTextToSpeech.main_input_name - added:
field SeamlessM4Tv2ForTextToText.main_input_name - added:
field SeamlessM4Tv2PreTrainedModel.base_model_prefix - added:
field SeamlessM4Tv2PreTrainedModel.config: SeamlessM4Tv2Config - added:
field SeamlessM4Tv2PreTrainedModel.supports_gradient_checkpointing - added:
method SeamlessM4Tv2ForSpeechToSpeech.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_emb… - added:
method SeamlessM4Tv2ForSpeechToSpeech.generate(input_features: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, speaker_id: Optional[int]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4Tv2GenerationOutput] - added:
method SeamlessM4Tv2ForSpeechToSpeech.get_decoder() - added:
method SeamlessM4Tv2ForSpeechToSpeech.get_encoder() - added:
method SeamlessM4Tv2ForSpeechToSpeech.get_input_embeddings() - added:
method SeamlessM4Tv2ForSpeechToSpeech.set_input_embeddings(value) - added:
method SeamlessM4Tv2ForSpeechToText.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embed… - added:
method SeamlessM4Tv2ForSpeechToText.generate(input_features=…, tgt_lang=…, generation_config=…, logits_processor=…, stopping_criteria=…, prefix_allowed_tokens_fn=…, synced_gpus=…, **kwargs) - added:
method SeamlessM4Tv2ForSpeechToText.get_decoder() - added:
method SeamlessM4Tv2ForSpeechToText.get_encoder() - added:
method SeamlessM4Tv2ForSpeechToText.get_input_embeddings() - added:
method SeamlessM4Tv2ForSpeechToText.set_input_embeddings(value) - added:
method SeamlessM4Tv2ForTextToSpeech.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Op… - added:
method SeamlessM4Tv2ForTextToSpeech.generate(input_ids: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, speaker_id: Optional[int]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4Tv2GenerationOutput] - added:
method SeamlessM4Tv2ForTextToSpeech.get_decoder() - added:
method SeamlessM4Tv2ForTextToSpeech.get_encoder() - added:
method SeamlessM4Tv2ForTextToSpeech.get_input_embeddings() - added:
method SeamlessM4Tv2ForTextToSpeech.set_input_embeddings(value) - added:
method SeamlessM4Tv2ForTextToText.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Opti… - added:
method SeamlessM4Tv2ForTextToText.generate(input_ids=…, tgt_lang=…, generation_config=…, logits_processor=…, stopping_criteria=…, prefix_allowed_tokens_fn=…, synced_gpus=…, **kwargs) - added:
method SeamlessM4Tv2ForTextToText.get_decoder() - added:
method SeamlessM4Tv2ForTextToText.get_encoder() - added:
method SeamlessM4Tv2ForTextToText.get_input_embeddings() - added:
method SeamlessM4Tv2ForTextToText.set_input_embeddings(value) - added:
method SeamlessM4Tv2Model.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Floa… - added:
method SeamlessM4Tv2Model.generate(input_ids: Optional[torch.Tensor]=…, input_features: Optional[torch.Tensor]=…, return_intermediate_token_ids: Optional[bool]=…, tgt_lang: Optional[str]=…, speaker_id: Optional[int]=…, generate_speech: Optional[bool]=…, **kwargs) -> Union[torch.Tensor, SeamlessM4Tv2GenerationOutput] - added:
method SeamlessM4Tv2Model.get_encoder() - added:
method SeamlessM4Tv2Model.get_input_embeddings() - added:
method SeamlessM4Tv2Model.set_input_embeddings(value) - added:
method SeamlessM4Tv2Model.set_modality(modality=…)
- added:
export define_import_structure
- added:
class SeedOssConfig(PretrainedConfig) - added:
constructor SeedOssConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, pretraining_tp=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, attention_bias=…, attention_out_bias=…, attent… - added:
field SeedOssConfig.base_model_pp_plan - added:
field SeedOssConfig.base_model_tp_plan - added:
field SeedOssConfig.keys_to_ignore_at_inference - added:
field SeedOssConfig.model_type
- added:
class SeedOssForCausalLM(SeedOssPreTrainedModel, GenerationMixin) - added:
class SeedOssForQuestionAnswering(GenericForQuestionAnswering, SeedOssPreTrainedModel) - added:
class SeedOssForSequenceClassification(GenericForSequenceClassification, SeedOssPreTrainedModel) - added:
class SeedOssForTokenClassification(GenericForTokenClassification, SeedOssPreTrainedModel) - added:
class SeedOssModel(SeedOssPreTrainedModel) - added:
class SeedOssPreTrainedModel(PreTrainedModel) - added:
constructor SeedOssForCausalLM(config) - added:
constructor SeedOssModel(config: SeedOssConfig) - added:
field SeedOssForQuestionAnswering.base_model_prefix - added:
field SeedOssPreTrainedModel.base_model_prefix - added:
field SeedOssPreTrainedModel.config: SeedOssConfig - added:
field SeedOssPreTrainedModel.supports_gradient_checkpointing - added:
method SeedOssForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method SeedOssModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_position: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class SeedOssForCausalLM(LlamaForCausalLM) - added:
class SeedOssForQuestionAnswering(LlamaForQuestionAnswering) - added:
class SeedOssForSequenceClassification(LlamaForSequenceClassification) - added:
class SeedOssForTokenClassification(LlamaForTokenClassification) - added:
class SeedOssModel(LlamaModel) - added:
class SeedOssPreTrainedModel(LlamaPreTrainedModel) - added:
method SeedOssForCausalLM.forward(**super_kwargs: Unpack[TransformersKwargs]) -> CausalLMOutputWithPast
- added:
export define_import_structure
- added:
class SegformerConfig(PretrainedConfig) - added:
class SegformerOnnxConfig(OnnxConfig) - added:
constructor SegformerConfig(num_channels=…, num_encoder_blocks=…, depths=…, sr_ratios=…, hidden_sizes=…, patch_sizes=…, strides=…, num_attention_heads=…, mlp_ratios=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, classifier_dropout_prob=…, initializer_range=…, drop_path_rate=…, layer_norm_eps=…, decoder_hidden_size=…, semantic_loss_ignore_index=…, **kwargs) - added:
field SegformerConfig.model_type - added:
field SegformerOnnxConfig.torch_onnx_minimum_version - added:
property SegformerOnnxConfig.atol_for_validation -> float - added:
property SegformerOnnxConfig.default_onnx_opset -> int - added:
property SegformerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class SegformerFeatureExtractor(SegformerImageProcessor) - added:
constructor SegformerFeatureExtractor(*args, **kwargs) -> None
- added:
class SegformerImageProcessor(BaseImageProcessor) - added:
constructor SegformerImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_reduce_labels: bool=…, **kwargs) -> None - added:
field SegformerImageProcessor.model_input_names - added:
method SegformerImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method SegformerImageProcessor.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[flo… - added:
method SegformerImageProcessor.reduce_label(label: ImageInput) -> np.ndarray - added:
method SegformerImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class SegformerImageProcessorFast(BaseImageProcessorFast) - added:
constructor SegformerImageProcessorFast(**kwargs: Unpack[SegformerFastImageProcessorKwargs]) - added:
field SegformerImageProcessorFast.crop_size - added:
field SegformerImageProcessorFast.default_to_square - added:
field SegformerImageProcessorFast.do_center_crop - added:
field SegformerImageProcessorFast.do_normalize - added:
field SegformerImageProcessorFast.do_reduce_labels - added:
field SegformerImageProcessorFast.do_rescale - added:
field SegformerImageProcessorFast.do_resize - added:
field SegformerImageProcessorFast.image_mean - added:
field SegformerImageProcessorFast.image_std - added:
field SegformerImageProcessorFast.resample - added:
field SegformerImageProcessorFast.rescale_factor - added:
field SegformerImageProcessorFast.size - added:
field SegformerImageProcessorFast.valid_kwargs - added:
method SegformerImageProcessorFast.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple]]=…) - added:
method SegformerImageProcessorFast.preprocess(images: ImageInput, segmentation_maps: Optional[ImageInput]=…, **kwargs: Unpack[SegformerFastImageProcessorKwargs]) -> BatchFeature - added:
method SegformerImageProcessorFast.reduce_label(labels: list[torch.Tensor])
- added:
class SegformerDecodeHead(SegformerPreTrainedModel) - added:
class SegformerForImageClassification(SegformerPreTrainedModel) - added:
class SegformerForSemanticSegmentation(SegformerPreTrainedModel) - added:
class SegformerLayer(nn.Module) - added:
class SegformerModel(SegformerPreTrainedModel) - added:
class SegformerPreTrainedModel(PreTrainedModel) - added:
constructor SegformerDecodeHead(config) - added:
constructor SegformerForImageClassification(config) - added:
constructor SegformerForSemanticSegmentation(config) - added:
constructor SegformerLayer(config, hidden_size, num_attention_heads, drop_path, sequence_reduction_ratio, mlp_ratio) - added:
constructor SegformerModel(config) - added:
field SegformerPreTrainedModel.base_model_prefix - added:
field SegformerPreTrainedModel.config: SegformerConfig - added:
field SegformerPreTrainedModel.main_input_name - added:
method SegformerDecodeHead.forward(encoder_hidden_states: torch.FloatTensor) -> torch.Tensor - added:
method SegformerForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SegFormerImageClassifierOutput] - added:
method SegformerForSemanticSegmentation.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput] - added:
method SegformerLayer.forward(hidden_states, height, width, output_attentions=…) - added:
method SegformerModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
class TFSegformerDecodeHead(TFSegformerPreTrainedModel) - added:
class TFSegformerForImageClassification(TFSegformerPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFSegformerForSemanticSegmentation(TFSegformerPreTrainedModel) - added:
class TFSegformerModel(TFSegformerPreTrainedModel) - added:
class TFSegformerPreTrainedModel(TFPreTrainedModel) - added:
constructor TFSegformerDecodeHead(config: SegformerConfig, **kwargs) - added:
constructor TFSegformerForImageClassification(config: SegformerConfig, *inputs, **kwargs) - added:
constructor TFSegformerForSemanticSegmentation(config: SegformerConfig, **kwargs) - added:
constructor TFSegformerModel(config: SegformerConfig, *inputs, **kwargs) - added:
field TFSegformerPreTrainedModel.base_model_prefix - added:
field TFSegformerPreTrainedModel.config_class - added:
field TFSegformerPreTrainedModel.main_input_name - added:
method TFSegformerDecodeHead.build(input_shape=…) - added:
method TFSegformerDecodeHead.call(encoder_hidden_states: tf.Tensor, training: bool=…) -> tf.Tensor - added:
method TFSegformerForImageClassification.build(input_shape=…) - added:
method TFSegformerForImageClassification.call(pixel_values: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…) -> tuple | TFSequenceClassifierOutput - added:
method TFSegformerForSemanticSegmentation.build(input_shape=…) - added:
method TFSegformerForSemanticSegmentation.call(pixel_values: tf.Tensor, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…) -> tuple | TFSemanticSegmenterOutput - added:
method TFSegformerForSemanticSegmentation.hf_compute_loss(logits, labels) - added:
method TFSegformerModel.build(input_shape=…) - added:
method TFSegformerModel.call(pixel_values: tf.Tensor, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple | TFBaseModelOutput - added:
property TFSegformerPreTrainedModel.input_signature
- added:
class SegformerImageProcessorFast(BeitImageProcessorFast) - added:
field SegformerImageProcessorFast.crop_size - added:
field SegformerImageProcessorFast.do_center_crop - added:
field SegformerImageProcessorFast.do_normalize - added:
field SegformerImageProcessorFast.do_reduce_labels - added:
field SegformerImageProcessorFast.do_rescale - added:
field SegformerImageProcessorFast.do_resize - added:
field SegformerImageProcessorFast.image_mean - added:
field SegformerImageProcessorFast.image_std - added:
field SegformerImageProcessorFast.resample - added:
field SegformerImageProcessorFast.rescale_factor - added:
field SegformerImageProcessorFast.size
- added:
export define_import_structure
- added:
class SegGptConfig(PretrainedConfig) - added:
constructor SegGptConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, hidden_act=…, hidden_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, mlp_dim=…, drop_path_rate=…, pretrain_image_size=…, decoder_hidden_size=…, use_relative_position_embeddings=…, merge_index=…, intermediate_hidden_state_indices=…, beta=…, **kwargs) - added:
field SegGptConfig.model_type
- added:
class SegGptImageProcessor(BaseImageProcessor) - added:
constructor SegGptImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs) -> None - added:
field SegGptImageProcessor.model_input_names - added:
method SegGptImageProcessor.get_palette(num_labels: int) -> list[tuple[int, int]] - added:
method SegGptImageProcessor.mask_to_rgb(image: np.ndarray, palette: Optional[list[tuple[int, int]]]=…, data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method SegGptImageProcessor.post_process_semantic_segmentation(outputs, target_sizes: Optional[list[tuple[int, int]]]=…, num_labels: Optional[int]=…) - added:
method SegGptImageProcessor.preprocess(images: Optional[ImageInput]=…, prompt_images: Optional[ImageInput]=…, prompt_masks: Optional[ImageInput]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…… - added:
method SegGptImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class SegGptForImageSegmentation(SegGptPreTrainedModel) - added:
class SegGptModel(SegGptPreTrainedModel) - added:
class SegGptPreTrainedModel(PreTrainedModel) - added:
constructor SegGptForImageSegmentation(config: SegGptConfig) - added:
constructor SegGptModel(config: SegGptConfig) - added:
field SegGptPreTrainedModel.base_model_prefix - added:
field SegGptPreTrainedModel.config: SegGptConfig - added:
field SegGptPreTrainedModel.main_input_name - added:
field SegGptPreTrainedModel.supports_gradient_checkpointing - added:
method SegGptForImageSegmentation.forward(pixel_values: torch.Tensor, prompt_pixel_values: torch.Tensor, prompt_masks: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…, feature_ensemble: Optional[bool]=…, embedding_type: Optional[str]=…, labels: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) ->… - added:
method SegGptModel.forward(pixel_values: torch.Tensor, prompt_pixel_values: torch.Tensor, prompt_masks: torch.Tensor, bool_masked_pos: Optional[torch.BoolTensor]=…, feature_ensemble: Optional[bool]=…, embedding_type: Optional[str]=…, labels: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Se… - added:
method SegGptModel.get_input_embeddings() -> SegGptPatchEmbeddings
- added:
export define_import_structure
- added:
class SEWConfig(PretrainedConfig) - added:
constructor SEWConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, squeeze_factor=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel=…, conv_bias=… - added:
field SEWConfig.model_type - added:
property SEWConfig.inputs_to_logits_ratio
- added:
class SEWFeatureExtractor(SEWFeatureEncoder) - added:
constructor SEWFeatureExtractor(config)
- added:
class SEWForCTC(SEWPreTrainedModel) - added:
class SEWForSequenceClassification(SEWPreTrainedModel) - added:
class SEWModel(SEWPreTrainedModel) - added:
class SEWPreTrainedModel(PreTrainedModel) - added:
constructor SEWForCTC(config, target_lang: Optional[str]=…) - added:
constructor SEWForSequenceClassification(config) - added:
constructor SEWModel(config: SEWConfig) - added:
field SEWPreTrainedModel.base_model_prefix - added:
field SEWPreTrainedModel.config: SEWConfig - added:
field SEWPreTrainedModel.main_input_name - added:
field SEWPreTrainedModel.supports_gradient_checkpointing - added:
method SEWForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method SEWForCTC.freeze_base_model() - added:
method SEWForCTC.freeze_feature_encoder() - added:
method SEWForCTC.freeze_feature_extractor() - added:
method SEWForCTC.tie_weights() - added:
method SEWForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method SEWForSequenceClassification.freeze_base_model() - added:
method SEWForSequenceClassification.freeze_feature_encoder() - added:
method SEWForSequenceClassification.freeze_feature_extractor() - added:
method SEWModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
class SEWForCTC(Wav2Vec2ForCTC) - added:
class SEWForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class SEWModel(SEWPreTrainedModel) - added:
class SEWPreTrainedModel(PreTrainedModel) - added:
constructor SEWModel(config: SEWConfig) - added:
field SEWPreTrainedModel.base_model_prefix - added:
field SEWPreTrainedModel.config: SEWConfig - added:
field SEWPreTrainedModel.main_input_name - added:
field SEWPreTrainedModel.supports_gradient_checkpointing - added:
method SEWModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
export define_import_structure
- added:
class SEWDConfig(PretrainedConfig) - added:
constructor SEWDConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, squeeze_factor=…, max_position_embeddings=…, position_buckets=…, share_att_key=…, relative_attention=…, pos_att_type=…, norm_rel_ebd=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, final_dropout=…, initializer_range=…, layer_norm_e… - added:
field SEWDConfig.model_type - added:
method SEWDConfig.to_dict() - added:
property SEWDConfig.inputs_to_logits_ratio
- added:
class SEWDForCTC(SEWDPreTrainedModel) - added:
class SEWDForSequenceClassification(SEWDPreTrainedModel) - added:
class SEWDModel(SEWDPreTrainedModel) - added:
class SEWDPreTrainedModel(PreTrainedModel) - added:
constructor SEWDForCTC(config, target_lang: Optional[str]=…) - added:
constructor SEWDForSequenceClassification(config) - added:
constructor SEWDModel(config: SEWDConfig) - added:
field SEWDPreTrainedModel.base_model_prefix - added:
field SEWDPreTrainedModel.config: SEWDConfig - added:
field SEWDPreTrainedModel.main_input_name - added:
field SEWDPreTrainedModel.supports_gradient_checkpointing - added:
method SEWDForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method SEWDForCTC.freeze_base_model() - added:
method SEWDForCTC.freeze_feature_encoder() - added:
method SEWDForCTC.freeze_feature_extractor() - added:
method SEWDForCTC.tie_weights() - added:
method SEWDForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method SEWDForSequenceClassification.freeze_base_model() - added:
method SEWDForSequenceClassification.freeze_feature_encoder() - added:
method SEWDForSequenceClassification.freeze_feature_extractor() - added:
method SEWDModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput]
- added:
export define_import_structure
- added:
class ShieldGemma2Config(PretrainedConfig) - added:
constructor ShieldGemma2Config(text_config=…, vision_config=…, mm_tokens_per_image: int=…, boi_token_index: int=…, eoi_token_index: int=…, image_token_index: int=…, initializer_range: float=…, **kwargs) - added:
field ShieldGemma2Config.attribute_map - added:
field ShieldGemma2Config.model_type - added:
field ShieldGemma2Config.sub_configs
- added:
class ShieldGemma2ForImageClassification(PreTrainedModel) - added:
constructor ShieldGemma2ForImageClassification(config: ShieldGemma2Config) - added:
field ShieldGemma2ForImageClassification.config: ShieldGemma2Config - added:
method ShieldGemma2ForImageClassification.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, token_type_ids: Optional[torch.LongTensor]=…, cache_position: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, lab… - added:
method ShieldGemma2ForImageClassification.get_decoder() - added:
method ShieldGemma2ForImageClassification.get_input_embeddings() - added:
method ShieldGemma2ForImageClassification.get_output_embeddings() - added:
method ShieldGemma2ForImageClassification.set_decoder(decoder) - added:
method ShieldGemma2ForImageClassification.set_input_embeddings(value) - added:
method ShieldGemma2ForImageClassification.set_output_embeddings(new_embeddings) - added:
method ShieldGemma2ForImageClassification.tie_weights()
- added:
class ShieldGemma2Processor(Gemma3Processor) - added:
constructor ShieldGemma2Processor(image_processor, tokenizer, chat_template=…, image_seq_length=…, policy_definitions=…, **kwargs)
- added:
export define_import_structure
- added:
class SiglipConfig(PretrainedConfig) - added:
class SiglipTextConfig(PretrainedConfig) - added:
class SiglipVisionConfig(PretrainedConfig) - added:
constructor SiglipConfig(text_config=…, vision_config=…, **kwargs) - added:
constructor SiglipTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, projection_size=…, **kwargs) - added:
constructor SiglipVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, **kwargs) - added:
field SiglipConfig.model_type - added:
field SiglipConfig.sub_configs - added:
field SiglipTextConfig.base_config_key - added:
field SiglipTextConfig.model_type - added:
field SiglipVisionConfig.base_config_key - added:
field SiglipVisionConfig.model_type
- added:
class SiglipImageProcessor(BaseImageProcessor) - added:
constructor SiglipImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, **kwargs) -> None - added:
field SiglipImageProcessor.model_input_names - added:
method SiglipImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, T…
- added:
class SiglipImageProcessorFast(BaseImageProcessorFast) - added:
field SiglipImageProcessorFast.default_to_square - added:
field SiglipImageProcessorFast.do_normalize - added:
field SiglipImageProcessorFast.do_rescale - added:
field SiglipImageProcessorFast.do_resize - added:
field SiglipImageProcessorFast.image_mean - added:
field SiglipImageProcessorFast.image_std - added:
field SiglipImageProcessorFast.resample - added:
field SiglipImageProcessorFast.size
- added:
class SiglipForImageClassification(SiglipPreTrainedModel) - added:
class SiglipModel(SiglipPreTrainedModel) - added:
class SiglipPreTrainedModel(PreTrainedModel) - added:
class SiglipTextModel(SiglipPreTrainedModel) - added:
class SiglipVisionModel(SiglipPreTrainedModel) - added:
constructor SiglipForImageClassification(config: SiglipConfig) -> None - added:
constructor SiglipModel(config: SiglipConfig) - added:
constructor SiglipTextModel(config: SiglipTextConfig) - added:
constructor SiglipVisionModel(config: SiglipVisionConfig) - added:
field SiglipForImageClassification.main_input_name - added:
field SiglipModel.config: SiglipConfig - added:
field SiglipPreTrainedModel.base_model_prefix - added:
field SiglipPreTrainedModel.config: SiglipConfig - added:
field SiglipPreTrainedModel.supports_gradient_checkpointing - added:
field SiglipTextModel.config: SiglipTextConfig - added:
field SiglipVisionModel.config: SiglipVisionConfig - added:
field SiglipVisionModel.main_input_name - added:
method SiglipForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method SiglipModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> SiglipOutput - added:
method SiglipModel.get_image_features(pixel_values: torch.FloatTensor, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> torch.FloatTensor - added:
method SiglipModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method SiglipTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method SiglipTextModel.get_input_embeddings() -> nn.Module - added:
method SiglipTextModel.set_input_embeddings(value) - added:
method SiglipVisionModel.forward(pixel_values, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method SiglipVisionModel.get_input_embeddings() -> nn.Module
- added:
class SiglipProcessor(ProcessorMixin) - added:
constructor SiglipProcessor(image_processor, tokenizer) - added:
field SiglipProcessor.attributes - added:
field SiglipProcessor.image_processor_class - added:
field SiglipProcessor.tokenizer_class
- added:
class SiglipTokenizer(PreTrainedTokenizer) - added:
constructor SiglipTokenizer(vocab_file, eos_token=…, unk_token=…, pad_token=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, model_max_length=…, do_lower_case=…, **kwargs) -> None - added:
field SiglipTokenizer.model_input_names - added:
field SiglipTokenizer.vocab_files_names - added:
method SiglipTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SiglipTokenizer.canonicalize_text(text, *, keep_punctuation_exact_string=…) - added:
method SiglipTokenizer.convert_tokens_to_string(tokens) - added:
method SiglipTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SiglipTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method SiglipTokenizer.get_spm_processor() - added:
method SiglipTokenizer.get_vocab() - added:
method SiglipTokenizer.remove_punctuation(text: str) -> str - added:
method SiglipTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method SiglipTokenizer.tokenize(text: TextInput, add_special_tokens=…, **kwargs) -> list[str] - added:
property SiglipTokenizer.unk_token_length - added:
property SiglipTokenizer.vocab_size
- added:
export define_import_structure
- added:
class Siglip2Config(PretrainedConfig) - added:
class Siglip2TextConfig(PretrainedConfig) - added:
class Siglip2VisionConfig(PretrainedConfig) - added:
constructor Siglip2Config(text_config=…, vision_config=…, **kwargs) - added:
constructor Siglip2TextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, projection_size=…, **kwargs) - added:
constructor Siglip2VisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, num_patches=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, **kwargs) - added:
field Siglip2Config.model_type - added:
field Siglip2Config.sub_configs - added:
field Siglip2TextConfig.base_config_key - added:
field Siglip2TextConfig.model_type - added:
field Siglip2VisionConfig.base_config_key - added:
field Siglip2VisionConfig.model_type
- added:
class Siglip2ImageProcessor(BaseImageProcessor) - added:
constructor Siglip2ImageProcessor(do_resize: bool=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, patch_size: int=…, max_num_patches: int=…, **kwargs) - added:
field Siglip2ImageProcessor.model_input_names - added:
method Siglip2ImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, TensorType]]=…, input_data_format:…
- added:
class Siglip2ImageProcessorFast(BaseImageProcessorFast) - added:
constructor Siglip2ImageProcessorFast(**kwargs: Unpack[Siglip2FastImageProcessorKwargs]) - added:
field Siglip2ImageProcessorFast.do_normalize - added:
field Siglip2ImageProcessorFast.do_rescale - added:
field Siglip2ImageProcessorFast.do_resize - added:
field Siglip2ImageProcessorFast.image_mean - added:
field Siglip2ImageProcessorFast.image_std - added:
field Siglip2ImageProcessorFast.max_num_patches - added:
field Siglip2ImageProcessorFast.patch_size - added:
field Siglip2ImageProcessorFast.resample - added:
field Siglip2ImageProcessorFast.unused_kwargs - added:
field Siglip2ImageProcessorFast.valid_kwargs - added:
method Siglip2ImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Siglip2FastImageProcessorKwargs]) -> BatchFeature
- added:
class Siglip2ForImageClassification(Siglip2PreTrainedModel) - added:
class Siglip2Model(Siglip2PreTrainedModel) - added:
class Siglip2PreTrainedModel(PreTrainedModel) - added:
class Siglip2TextModel(Siglip2PreTrainedModel) - added:
class Siglip2VisionModel(Siglip2PreTrainedModel) - added:
constructor Siglip2ForImageClassification(config: Siglip2Config) -> None - added:
constructor Siglip2Model(config: Siglip2Config) - added:
constructor Siglip2TextModel(config: Siglip2TextConfig) - added:
constructor Siglip2VisionModel(config: Siglip2VisionConfig) - added:
field Siglip2ForImageClassification.main_input_name - added:
field Siglip2Model.config: Siglip2Config - added:
field Siglip2PreTrainedModel.base_model_prefix - added:
field Siglip2PreTrainedModel.config: Siglip2Config - added:
field Siglip2PreTrainedModel.supports_gradient_checkpointing - added:
field Siglip2TextModel.config: Siglip2TextConfig - added:
field Siglip2VisionModel.config: Siglip2VisionConfig - added:
field Siglip2VisionModel.main_input_name - added:
method Siglip2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> ImageClassifierOutput - added:
method Siglip2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=… - added:
method Siglip2Model.get_image_features(pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…) -> torch.FloatTensor - added:
method Siglip2Model.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method Siglip2TextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method Siglip2TextModel.get_input_embeddings() -> nn.Module - added:
method Siglip2TextModel.set_input_embeddings(value) - added:
method Siglip2VisionModel.forward(pixel_values: torch.FloatTensor, pixel_attention_mask: torch.Tensor, spatial_shapes: torch.LongTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPooling - added:
method Siglip2VisionModel.get_input_embeddings() -> nn.Module
- added:
class Siglip2Config(SiglipConfig) - added:
class Siglip2ForImageClassification(SiglipForImageClassification) - added:
class Siglip2Model(SiglipModel) - added:
class Siglip2PreTrainedModel(SiglipPreTrainedModel) - added:
class Siglip2TextConfig(SiglipTextConfig) - added:
class Siglip2TextModel(SiglipTextModel) - added:
class Siglip2VisionConfig(SiglipVisionConfig) - added:
class Siglip2VisionModel(SiglipVisionModel) - added:
constructor Siglip2VisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, num_patches=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, **kwargs) - added:
method Siglip2ForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> ImageClassifierOutput - added:
method Siglip2Model.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=… - added:
method Siglip2Model.get_image_features(pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.Tensor]=…, spatial_shapes: Optional[torch.LongTensor]=…) -> torch.FloatTensor - added:
method Siglip2VisionModel.forward(pixel_values: torch.FloatTensor, pixel_attention_mask: torch.Tensor, spatial_shapes: torch.LongTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> BaseModelOutputWithPooling
- added:
class Siglip2Processor(ProcessorMixin) - added:
constructor Siglip2Processor(image_processor, tokenizer) - added:
field Siglip2Processor.attributes - added:
field Siglip2Processor.image_processor_class - added:
field Siglip2Processor.tokenizer_class - added:
field Siglip2Processor.valid_processor_kwargs
- added:
export define_import_structure
- added:
class SmolLM3Config(PretrainedConfig) - added:
constructor SmolLM3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, rope_theta=…, rope_scaling=…, use_sliding_window=…, sliding_window=…, no_rope_layers=…, no_rope_layer_interval=…, lay… - added:
field SmolLM3Config.base_model_pp_plan - added:
field SmolLM3Config.base_model_tp_plan - added:
field SmolLM3Config.keys_to_ignore_at_inference - added:
field SmolLM3Config.model_type
- added:
class SmolLM3ForCausalLM(SmolLM3PreTrainedModel, GenerationMixin) - added:
class SmolLM3ForQuestionAnswering(GenericForQuestionAnswering, SmolLM3PreTrainedModel) - added:
class SmolLM3ForSequenceClassification(GenericForSequenceClassification, SmolLM3PreTrainedModel) - added:
class SmolLM3ForTokenClassification(GenericForTokenClassification, SmolLM3PreTrainedModel) - added:
class SmolLM3Model(SmolLM3PreTrainedModel) - added:
class SmolLM3PreTrainedModel(PreTrainedModel) - added:
constructor SmolLM3ForCausalLM(config) - added:
constructor SmolLM3Model(config: SmolLM3Config) - added:
field SmolLM3ForQuestionAnswering.base_model_prefix - added:
field SmolLM3PreTrainedModel.base_model_prefix - added:
field SmolLM3PreTrainedModel.config: SmolLM3Config - added:
field SmolLM3PreTrainedModel.supports_gradient_checkpointing - added:
method SmolLM3ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]=…,… - added:
method SmolLM3Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPast
- added:
class SmolLM3Config(PretrainedConfig) - added:
class SmolLM3ForCausalLM(LlamaForCausalLM) - added:
class SmolLM3ForQuestionAnswering(LlamaForQuestionAnswering) - added:
class SmolLM3ForSequenceClassification(LlamaForSequenceClassification) - added:
class SmolLM3ForTokenClassification(LlamaForTokenClassification) - added:
class SmolLM3Model(Qwen2Model) - added:
class SmolLM3PreTrainedModel(LlamaPreTrainedModel) - added:
constructor SmolLM3Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, rope_theta=…, rope_scaling=…, use_sliding_window=…, sliding_window=…, no_rope_layers=…, no_rope_layer_interval=…, lay… - added:
field SmolLM3Config.base_model_pp_plan - added:
field SmolLM3Config.base_model_tp_plan - added:
field SmolLM3Config.keys_to_ignore_at_inference - added:
field SmolLM3Config.model_type
- added:
export define_import_structure
- added:
class SmolVLMConfig(PretrainedConfig) - added:
class SmolVLMVisionConfig(PretrainedConfig) - added:
constructor SmolVLMConfig(use_cache=…, image_token_id=…, tie_word_embeddings=…, vision_config=…, text_config=…, scale_factor=…, pad_token_id=…, **kwargs) - added:
constructor SmolVLMVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, **kwargs) - added:
field SmolVLMConfig.model_type - added:
field SmolVLMConfig.sub_configs - added:
field SmolVLMVisionConfig.base_config_key - added:
field SmolVLMVisionConfig.model_type
- added:
class SmolVLMImageProcessor(BaseImageProcessor) - added:
constructor SmolVLMImageProcessor(do_convert_rgb: bool=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_image_splitting: bool=…, max_image_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: float=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…,… - added:
field SmolVLMImageProcessor.model_input_names - added:
method SmolVLMImageProcessor.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method SmolVLMImageProcessor.pad(images: list[list[np.ndarray]], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method SmolVLMImageProcessor.preprocess(images: ImageInput, do_convert_rgb: Optional[bool]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_image_splitting: Optional[bool]=…, do_rescale: Optional[bool]=…, max_image_size: Optional[dict[str, int]]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[… - added:
method SmolVLMImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method SmolVLMImageProcessor.resize_for_vision_encoder(image: np.ndarray, vision_encoder_max_size: int, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method SmolVLMImageProcessor.split_image(image, max_image_size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…)
- added:
class SmolVLMImageProcessorFast(BaseImageProcessorFast) - added:
field SmolVLMImageProcessorFast.do_convert_rgb - added:
field SmolVLMImageProcessorFast.do_image_splitting - added:
field SmolVLMImageProcessorFast.do_normalize - added:
field SmolVLMImageProcessorFast.do_pad - added:
field SmolVLMImageProcessorFast.do_rescale - added:
field SmolVLMImageProcessorFast.do_resize - added:
field SmolVLMImageProcessorFast.image_mean - added:
field SmolVLMImageProcessorFast.image_std - added:
field SmolVLMImageProcessorFast.max_image_size - added:
field SmolVLMImageProcessorFast.resample - added:
field SmolVLMImageProcessorFast.return_row_col_info - added:
field SmolVLMImageProcessorFast.size - added:
field SmolVLMImageProcessorFast.valid_kwargs - added:
method SmolVLMImageProcessorFast.get_number_of_image_patches(height: int, width: int, images_kwargs=…) - added:
method SmolVLMImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], fill: int=…, return_pixel_mask: bool=…) - added:
method SmolVLMImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[SmolVLMFastImageProcessorKwargs]) -> BatchFeature - added:
method SmolVLMImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method SmolVLMImageProcessorFast.resize_for_vision_encoder(image: torch.Tensor, vision_encoder_max_size: int, interpolation: Optional[F.InterpolationMode]=…) - added:
method SmolVLMImageProcessorFast.split_images(images: torch.Tensor, max_image_size: dict[str, int], interpolation: Optional[F.InterpolationMode]=…) - added:
method SmolVLMImageProcessorFast.to_dict()
- added:
class SmolVLMForConditionalGeneration(SmolVLMPreTrainedModel, GenerationMixin) - added:
class SmolVLMModel(SmolVLMPreTrainedModel) - added:
class SmolVLMPreTrainedModel(PreTrainedModel) - added:
class SmolVLMVisionTransformer(SmolVLMPreTrainedModel) - added:
constructor SmolVLMForConditionalGeneration(config) - added:
constructor SmolVLMModel(config: SmolVLMConfig) - added:
constructor SmolVLMVisionTransformer(config: SmolVLMVisionConfig) - added:
field SmolVLMPreTrainedModel.base_model_prefix - added:
field SmolVLMPreTrainedModel.config: SmolVLMConfig - added:
field SmolVLMPreTrainedModel.supports_gradient_checkpointing - added:
field SmolVLMVisionTransformer.config: SmolVLMVisionConfig - added:
method SmolVLMForConditionalGeneration.disable_input_require_grads() - added:
method SmolVLMForConditionalGeneration.enable_input_require_grads() - added:
method SmolVLMForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTenso… - added:
method SmolVLMForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method SmolVLMForConditionalGeneration.get_input_embeddings() - added:
method SmolVLMForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, pixel_values=…, pixel_attention_mask=…, image_hidden_states=…, logits_to_keep=…, **kwargs) - added:
method SmolVLMForConditionalGeneration.set_input_embeddings(value) - added:
method SmolVLMModel.disable_input_require_grads() - added:
method SmolVLMModel.enable_input_require_grads() - added:
method SmolVLMModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTensor]=…, use_cache: Op… - added:
method SmolVLMModel.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method SmolVLMModel.get_input_embeddings() - added:
method SmolVLMModel.inputs_merger(input_ids: torch.LongTensor, inputs_embeds: torch.Tensor, image_hidden_states: torch.Tensor) - added:
method SmolVLMModel.set_input_embeddings(value) - added:
method SmolVLMVisionTransformer.forward(pixel_values, patch_attention_mask: Optional[torch.BoolTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> Union[tuple, BaseModelOutput] - added:
method SmolVLMVisionTransformer.get_input_embeddings() - added:
method SmolVLMVisionTransformer.set_input_embeddings(value)
- added:
class SmolVLMConfig(Idefics3Config) - added:
class SmolVLMForConditionalGeneration(Idefics3ForConditionalGeneration) - added:
class SmolVLMImageProcessor(Idefics3ImageProcessor) - added:
class SmolVLMImageProcessorFast(Idefics3ImageProcessorFast) - added:
class SmolVLMModel(Idefics3Model) - added:
class SmolVLMPreTrainedModel(Idefics3PreTrainedModel) - added:
class SmolVLMVisionConfig(Idefics3VisionConfig) - added:
class SmolVLMVisionTransformer(Idefics3VisionTransformer) - added:
constructor SmolVLMForConditionalGeneration(config) - added:
field SmolVLMConfig.model_type - added:
field SmolVLMVisionConfig.model_type - added:
method SmolVLMForConditionalGeneration.forward(**super_kwargs) - added:
method SmolVLMModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_attention_mask: Optional[torch.BoolTensor]=…, image_hidden_states: Optional[torch.FloatTensor]=…, use_cache: Op… - added:
method SmolVLMModel.get_image_features(pixel_values: torch.FloatTensor, pixel_attention_mask: Optional[torch.LongTensor]=…) - added:
method SmolVLMModel.inputs_merger(input_ids: torch.LongTensor, inputs_embeds: torch.Tensor, image_hidden_states: torch.Tensor)
- added:
class SmolVLMProcessor(ProcessorMixin) - added:
constructor SmolVLMProcessor(image_processor, tokenizer, video_processor, image_seq_len: int=…, chat_template: Optional[str]=…, **kwargs) - added:
field SmolVLMProcessor.attributes - added:
field SmolVLMProcessor.image_processor_class - added:
field SmolVLMProcessor.tokenizer_class - added:
field SmolVLMProcessor.video_processor_class - added:
method SmolVLMProcessor.apply_chat_template(conversation: Union[list[dict[str, str]], list[list[dict[str, str]]]], chat_template: Optional[str]=…, **kwargs: Unpack[AllKwargsForChatTemplate]) -> str - added:
method SmolVLMProcessor.expand_text_with_image_tokens(text, image_rows, image_cols) - added:
method SmolVLMProcessor.expand_text_with_video_tokens(text, video_inputs)
- added:
class SmolVLMVideoProcessor(BaseVideoProcessor) - added:
constructor SmolVLMVideoProcessor(**kwargs: Unpack[SmolVLMVideoProcessorInitKwargs]) - added:
field SmolVLMVideoProcessor.do_convert_rgb - added:
field SmolVLMVideoProcessor.do_normalize - added:
field SmolVLMVideoProcessor.do_pad - added:
field SmolVLMVideoProcessor.do_rescale - added:
field SmolVLMVideoProcessor.do_resize - added:
field SmolVLMVideoProcessor.do_sample_frames - added:
field SmolVLMVideoProcessor.image_mean - added:
field SmolVLMVideoProcessor.image_std - added:
field SmolVLMVideoProcessor.max_image_size - added:
field SmolVLMVideoProcessor.model_input_names - added:
field SmolVLMVideoProcessor.resample - added:
field SmolVLMVideoProcessor.size - added:
field SmolVLMVideoProcessor.valid_kwargs - added:
method SmolVLMVideoProcessor.pad(video: torch.Tensor, padded_size: tuple[int, int], max_num_frames: int, fill: int=…, return_pixel_mask: bool=…) - added:
method SmolVLMVideoProcessor.resize(video: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor - added:
method SmolVLMVideoProcessor.sample_frames(metadata: VideoMetadata, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, skip_secs: Optional[int]=…, **kwargs)
- added:
export define_import_structure
- added:
class SpeechEncoderDecoderConfig(PretrainedConfig) - added:
constructor SpeechEncoderDecoderConfig(**kwargs) - added:
field SpeechEncoderDecoderConfig.has_no_defaults_at_init - added:
field SpeechEncoderDecoderConfig.model_type - added:
field SpeechEncoderDecoderConfig.sub_configs - added:
static method SpeechEncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs) -> PretrainedConfig
- added:
class FlaxSpeechEncoderDecoderModel(FlaxPreTrainedModel) - added:
constructor FlaxSpeechEncoderDecoderModel(config: SpeechEncoderDecoderConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxSpeechEncoderDecoderModel.base_model_prefix: str - added:
field FlaxSpeechEncoderDecoderModel.config_class - added:
field FlaxSpeechEncoderDecoderModel.module_class - added:
method FlaxSpeechEncoderDecoderModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional… - added:
method FlaxSpeechEncoderDecoderModel.encode(inputs: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, freeze_feature_encoder: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxSpeechEncoderDecoderModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxSpeechEncoderDecoderModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxSpeechEncoderDecoderModel.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxSpeechEncoderDecoderModel.update_inputs_for_generation(model_outputs, model_kwargs) - added:
static method FlaxSpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, decoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, *model_args, **kwargs) -> FlaxPreTrainedModel
- added:
class SpeechEncoderDecoderModel(PreTrainedModel, GenerationMixin) - added:
constructor SpeechEncoderDecoderModel(config: Optional[PretrainedConfig]=…, encoder: Optional[PreTrainedModel]=…, decoder: Optional[PreTrainedModel]=…) - added:
field SpeechEncoderDecoderModel.base_model_prefix - added:
field SpeechEncoderDecoderModel.config: SpeechEncoderDecoderConfig - added:
field SpeechEncoderDecoderModel.main_input_name - added:
field SpeechEncoderDecoderModel.supports_gradient_checkpointing - added:
method SpeechEncoderDecoderModel.forward(inputs: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.L… - added:
method SpeechEncoderDecoderModel.freeze_feature_encoder() - added:
method SpeechEncoderDecoderModel.get_encoder() - added:
method SpeechEncoderDecoderModel.get_input_embeddings() - added:
method SpeechEncoderDecoderModel.get_output_embeddings() - added:
method SpeechEncoderDecoderModel.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method SpeechEncoderDecoderModel.resize_token_embeddings(*args, **kwargs) - added:
method SpeechEncoderDecoderModel.set_output_embeddings(new_embeddings) - added:
static method SpeechEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[str]=…, decoder_pretrained_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel
- added:
export define_import_structure
- added:
class Speech2TextConfig(PretrainedConfig) - added:
constructor Speech2TextConfig(vocab_size=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, decoder_start_token_id=…, scale_embedding=…, pad_t… - added:
field Speech2TextConfig.attribute_map - added:
field Speech2TextConfig.keys_to_ignore_at_inference - added:
field Speech2TextConfig.model_type
- added:
class Speech2TextFeatureExtractor(SequenceFeatureExtractor) - added:
constructor Speech2TextFeatureExtractor(feature_size=…, sampling_rate=…, num_mel_bins=…, padding_value=…, dither=…, do_ceptral_normalize=…, normalize_means=…, normalize_vars=…, **kwargs) - added:
field Speech2TextFeatureExtractor.model_input_names - added:
method Speech2TextFeatureExtractor.normalize(input_features: list[np.ndarray], attention_mask: Optional[np.ndarray]=…) -> list[np.ndarray] - added:
static method Speech2TextFeatureExtractor.utterance_cmvn(x: np.ndarray, input_length: int, normalize_means: Optional[bool]=…, normalize_vars: Optional[bool]=…, padding_value: float=…) -> np.ndarray
- added:
class Speech2TextForConditionalGeneration(Speech2TextPreTrainedModel, GenerationMixin) - added:
class Speech2TextModel(Speech2TextPreTrainedModel) - added:
class Speech2TextPreTrainedModel(PreTrainedModel) - added:
constructor Speech2TextForConditionalGeneration(config: Speech2TextConfig) - added:
constructor Speech2TextModel(config: Speech2TextConfig) - added:
field Speech2TextForConditionalGeneration.base_model_prefix - added:
field Speech2TextPreTrainedModel.base_model_prefix - added:
field Speech2TextPreTrainedModel.config: Speech2TextConfig - added:
field Speech2TextPreTrainedModel.main_input_name - added:
field Speech2TextPreTrainedModel.supports_gradient_checkpointing - added:
method Speech2TextForConditionalGeneration.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[t… - added:
method Speech2TextForConditionalGeneration.get_decoder() - added:
method Speech2TextForConditionalGeneration.get_encoder() - added:
method Speech2TextModel.forward(input_features: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.Fl… - added:
method Speech2TextModel.get_encoder() - added:
method Speech2TextModel.get_input_embeddings() - added:
method Speech2TextModel.set_input_embeddings(value)
- added:
class TFSpeech2TextForConditionalGeneration(TFSpeech2TextPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFSpeech2TextModel(TFSpeech2TextPreTrainedModel) - added:
class TFSpeech2TextPreTrainedModel(TFPreTrainedModel) - added:
constructor TFSpeech2TextForConditionalGeneration(config: Speech2TextConfig) - added:
constructor TFSpeech2TextModel(config: Speech2TextConfig, *inputs, **kwargs) - added:
field TFSpeech2TextPreTrainedModel.base_model_prefix - added:
field TFSpeech2TextPreTrainedModel.config_class - added:
field TFSpeech2TextPreTrainedModel.main_input_name - added:
method TFSpeech2TextForConditionalGeneration.build(input_shape=…) - added:
method TFSpeech2TextForConditionalGeneration.call(input_features: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | None=… - added:
method TFSpeech2TextForConditionalGeneration.get_decoder() - added:
method TFSpeech2TextForConditionalGeneration.get_encoder() - added:
method TFSpeech2TextForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, attention_mask=…, head_mask=…, decoder_head_mask=…, cross_attn_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFSpeech2TextForConditionalGeneration.resize_token_embeddings(new_num_tokens: int) -> tf.Variable - added:
method TFSpeech2TextForConditionalGeneration.serving_output(output) - added:
method TFSpeech2TextForConditionalGeneration.tf_to_pt_weight_rename(tf_weight) - added:
method TFSpeech2TextModel.build(input_shape=…) - added:
method TFSpeech2TextModel.call(input_features: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs:… - added:
method TFSpeech2TextModel.get_decoder() - added:
method TFSpeech2TextModel.get_encoder() - added:
method TFSpeech2TextModel.serving_output(output) - added:
property TFSpeech2TextPreTrainedModel.input_signature
- added:
class Speech2TextProcessor(ProcessorMixin) - added:
constructor Speech2TextProcessor(feature_extractor, tokenizer) - added:
field Speech2TextProcessor.feature_extractor_class - added:
field Speech2TextProcessor.tokenizer_class - added:
method Speech2TextProcessor.as_target_processor()
- added:
class Speech2TextTokenizer(PreTrainedTokenizer) - added:
constructor Speech2TextTokenizer(vocab_file, spm_file, bos_token=…, eos_token=…, pad_token=…, unk_token=…, do_upper_case=…, do_lower_case=…, tgt_lang=…, lang_codes=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field Speech2TextTokenizer.model_input_names - added:
field Speech2TextTokenizer.prefix_tokens: list[int] - added:
field Speech2TextTokenizer.vocab_files_names - added:
method Speech2TextTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method Speech2TextTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method Speech2TextTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method Speech2TextTokenizer.get_vocab() -> dict - added:
method Speech2TextTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method Speech2TextTokenizer.set_tgt_lang_special_tokens(tgt_lang: str) -> None - added:
property Speech2TextTokenizer.tgt_lang -> str - added:
property Speech2TextTokenizer.vocab_size -> int
- added:
export define_import_structure
- added:
class SpeechT5Config(PretrainedConfig) - added:
class SpeechT5HifiGanConfig(PretrainedConfig) - added:
constructor SpeechT5Config(vocab_size=…, hidden_size=…, encoder_layers=…, encoder_attention_heads=…, encoder_ffn_dim=…, encoder_layerdrop=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, decoder_layerdrop=…, hidden_act=…, positional_dropout=…, hidden_dropout=…, attention_dropout=…, activation_dropout=…, initializer_range=…, layer_norm_eps=…, scale_embedding=…, feat_extract_norm=… - added:
constructor SpeechT5HifiGanConfig(model_in_dim=…, sampling_rate=…, upsample_initial_channel=…, upsample_rates=…, upsample_kernel_sizes=…, resblock_kernel_sizes=…, resblock_dilation_sizes=…, initializer_range=…, leaky_relu_slope=…, normalize_before=…, **kwargs) - added:
field SpeechT5Config.attribute_map - added:
field SpeechT5Config.model_type - added:
field SpeechT5HifiGanConfig.model_type - added:
method SpeechT5Config.inputs_to_logits_ratio()
- added:
class SpeechT5FeatureExtractor(SequenceFeatureExtractor) - added:
constructor SpeechT5FeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, do_normalize: bool=…, num_mel_bins: int=…, hop_length: int=…, win_length: int=…, win_function: str=…, frame_signal_scale: float=…, fmin: float=…, fmax: float=…, mel_floor: float=…, reduction_factor: int=…, return_attention_mask: bool=…, **kwargs) - added:
field SpeechT5FeatureExtractor.model_input_names - added:
method SpeechT5FeatureExtractor.to_dict() -> dict[str, Any] - added:
static method SpeechT5FeatureExtractor.zero_mean_unit_var_norm(input_values: list[np.ndarray], attention_mask: list[np.ndarray], padding_value: float=…) -> list[np.ndarray]
- added:
class SpeechT5ForSpeechToSpeech(SpeechT5PreTrainedModel) - added:
class SpeechT5ForSpeechToText(SpeechT5PreTrainedModel, GenerationMixin) - added:
class SpeechT5ForTextToSpeech(SpeechT5PreTrainedModel) - added:
class SpeechT5HifiGan(PreTrainedModel) - added:
class SpeechT5Model(SpeechT5PreTrainedModel) - added:
class SpeechT5PreTrainedModel(PreTrainedModel) - added:
constructor SpeechT5ForSpeechToSpeech(config: SpeechT5Config) - added:
constructor SpeechT5ForSpeechToText(config: SpeechT5Config) - added:
constructor SpeechT5ForTextToSpeech(config: SpeechT5Config) - added:
constructor SpeechT5HifiGan(config: SpeechT5HifiGanConfig) - added:
constructor SpeechT5Model(config: SpeechT5Config, encoder: Optional[nn.Module]=…, decoder: Optional[nn.Module]=…) - added:
field SpeechT5ForTextToSpeech.main_input_name - added:
field SpeechT5HifiGan.config: SpeechT5HifiGanConfig - added:
field SpeechT5HifiGan.main_input_name - added:
field SpeechT5PreTrainedModel.base_model_prefix - added:
field SpeechT5PreTrainedModel.config: SpeechT5Config - added:
field SpeechT5PreTrainedModel.main_input_name - added:
field SpeechT5PreTrainedModel.supports_gradient_checkpointing - added:
method SpeechT5ForSpeechToSpeech.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_values: Optional[torch.FloatTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Opt… - added:
method SpeechT5ForSpeechToSpeech.freeze_feature_encoder() - added:
method SpeechT5ForSpeechToSpeech.generate_speech(input_values: torch.FloatTensor, speaker_embeddings: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, threshold: float=…, minlenratio: float=…, maxlenratio: float=…, vocoder: Optional[nn.Module]=…, output_cross_attentions: bool=…, return_output_lengths: bool=…) -> torch.FloatTensor - added:
method SpeechT5ForSpeechToSpeech.get_decoder() - added:
method SpeechT5ForSpeechToSpeech.get_encoder() - added:
method SpeechT5ForSpeechToText.forward(input_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[… - added:
method SpeechT5ForSpeechToText.freeze_feature_encoder() - added:
method SpeechT5ForSpeechToText.get_decoder() - added:
method SpeechT5ForSpeechToText.get_encoder() - added:
method SpeechT5ForSpeechToText.get_output_embeddings() - added:
method SpeechT5ForSpeechToText.set_output_embeddings(new_embeddings) - added:
method SpeechT5ForTextToSpeech.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_values: Optional[torch.FloatTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[… - added:
method SpeechT5ForTextToSpeech.generate(input_ids: torch.LongTensor, attention_mask: Optional[torch.LongTensor]=…, speaker_embeddings: Optional[torch.FloatTensor]=…, threshold: float=…, minlenratio: float=…, maxlenratio: float=…, vocoder: Optional[nn.Module]=…, output_cross_attentions: bool=…, return_output_lengths: bool=…, **kwargs) -> Union[torch.FloatTensor, tuple[torch.FloatTensor, torch.Floa… - added:
method SpeechT5ForTextToSpeech.generate_speech(input_ids: torch.LongTensor, speaker_embeddings: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, threshold: float=…, minlenratio: float=…, maxlenratio: float=…, vocoder: Optional[nn.Module]=…, output_cross_attentions: bool=…, return_output_lengths: bool=…) -> Union[torch.FloatTensor, tuple[torch.FloatTensor, torch.FloatTe… - added:
method SpeechT5ForTextToSpeech.get_decoder() - added:
method SpeechT5ForTextToSpeech.get_encoder() - added:
method SpeechT5HifiGan.apply_weight_norm() - added:
method SpeechT5HifiGan.forward(spectrogram: torch.FloatTensor) -> torch.FloatTensor - added:
method SpeechT5HifiGan.remove_weight_norm() - added:
method SpeechT5Model.forward(input_values: Optional[torch.Tensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_values: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torc… - added:
method SpeechT5Model.freeze_feature_encoder() - added:
method SpeechT5Model.get_encoder() - added:
method SpeechT5Model.get_input_embeddings() - added:
method SpeechT5Model.set_input_embeddings(value) - added:
static method SpeechT5ForTextToSpeech.can_generate() -> bool
- added:
class EnglishNumberNormalizer - added:
constructor EnglishNumberNormalizer() - added:
method EnglishNumberNormalizer.convert(number) - added:
method EnglishNumberNormalizer.spell_number(num)
- added:
class SpeechT5Processor(ProcessorMixin) - added:
constructor SpeechT5Processor(feature_extractor, tokenizer) - added:
field SpeechT5Processor.feature_extractor_class - added:
field SpeechT5Processor.tokenizer_class - added:
method SpeechT5Processor.pad(*args, **kwargs)
- added:
class SpeechT5Tokenizer(PreTrainedTokenizer) - added:
constructor SpeechT5Tokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, normalize=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field SpeechT5Tokenizer.model_input_names - added:
field SpeechT5Tokenizer.vocab_files_names - added:
method SpeechT5Tokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method SpeechT5Tokenizer.convert_tokens_to_string(tokens) - added:
method SpeechT5Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method SpeechT5Tokenizer.get_vocab() - added:
method SpeechT5Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method SpeechT5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property SpeechT5Tokenizer.normalizer - added:
property SpeechT5Tokenizer.vocab_size
- added:
export define_import_structure
- added:
class SplinterConfig(PretrainedConfig) - added:
constructor SplinterConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, use_cache=…, pad_token_id=…, question_token_id=…, **kwargs) - added:
field SplinterConfig.model_type
- added:
class SplinterForPreTraining(SplinterPreTrainedModel) - added:
class SplinterForQuestionAnswering(SplinterPreTrainedModel) - added:
class SplinterLayer(GradientCheckpointingLayer) - added:
class SplinterModel(SplinterPreTrainedModel) - added:
class SplinterPreTrainedModel(PreTrainedModel) - added:
constructor SplinterForPreTraining(config) - added:
constructor SplinterForQuestionAnswering(config) - added:
constructor SplinterLayer(config) - added:
constructor SplinterModel(config) - added:
field SplinterPreTrainedModel.base_model_prefix - added:
field SplinterPreTrainedModel.config: SplinterConfig - added:
field SplinterPreTrainedModel.supports_gradient_checkpointing - added:
method SplinterForPreTraining.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optional[bo… - added:
method SplinterForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor]=…, output_attentions: Optio… - added:
method SplinterLayer.feed_forward_chunk(attention_output) - added:
method SplinterLayer.forward(hidden_states: torch.Tensor, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, **kwargs) -> tuple[torch.Tensor] - added:
method SplinterModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseMod… - added:
method SplinterModel.get_input_embeddings() - added:
method SplinterModel.set_input_embeddings(value)
- added:
class SplinterTokenizer(PreTrainedTokenizer) - added:
constructor SplinterTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, question_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field SplinterTokenizer.vocab_files_names - added:
method SplinterTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SplinterTokenizer.convert_tokens_to_string(tokens) - added:
method SplinterTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SplinterTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method SplinterTokenizer.get_vocab() - added:
method SplinterTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property SplinterTokenizer.do_lower_case - added:
property SplinterTokenizer.question_token_id - added:
property SplinterTokenizer.vocab_size
- added:
class SplinterTokenizerFast(PreTrainedTokenizerFast) - added:
constructor SplinterTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, question_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field SplinterTokenizerFast.slow_tokenizer_class - added:
field SplinterTokenizerFast.vocab_files_names - added:
method SplinterTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SplinterTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SplinterTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property SplinterTokenizerFast.question_token_id
- added:
export define_import_structure
- added:
class SqueezeBertConfig(PretrainedConfig) - added:
class SqueezeBertOnnxConfig(OnnxConfig) - added:
constructor SqueezeBertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, embedding_size=…, q_groups=…, k_groups=…, v_groups=…, post_attention_groups=…, intermediate_groups=…, out… - added:
field SqueezeBertConfig.model_type - added:
property SqueezeBertOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class SqueezeBertForMaskedLM(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForMultipleChoice(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForQuestionAnswering(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForSequenceClassification(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForTokenClassification(SqueezeBertPreTrainedModel) - added:
class SqueezeBertModel(SqueezeBertPreTrainedModel) - added:
class SqueezeBertModule(nn.Module) - added:
class SqueezeBertPreTrainedModel(PreTrainedModel) - added:
constructor SqueezeBertForMaskedLM(config) - added:
constructor SqueezeBertForMultipleChoice(config) - added:
constructor SqueezeBertForQuestionAnswering(config) - added:
constructor SqueezeBertForSequenceClassification(config) - added:
constructor SqueezeBertForTokenClassification(config) - added:
constructor SqueezeBertModel(config) - added:
constructor SqueezeBertModule(config) - added:
field SqueezeBertPreTrainedModel.base_model_prefix - added:
field SqueezeBertPreTrainedModel.config: SqueezeBertConfig - added:
method SqueezeBertForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict… - added:
method SqueezeBertForMaskedLM.get_output_embeddings() - added:
method SqueezeBertForMaskedLM.set_output_embeddings(new_embeddings) - added:
method SqueezeBertForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retur… - added:
method SqueezeBertForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[b… - added:
method SqueezeBertForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=… - added:
method SqueezeBertForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…,… - added:
method SqueezeBertModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple,… - added:
method SqueezeBertModel.get_input_embeddings() - added:
method SqueezeBertModel.set_input_embeddings(new_embeddings) - added:
method SqueezeBertModule.forward(hidden_states, attention_mask, output_attentions)
- added:
class SqueezeBertTokenizer(PreTrainedTokenizer) - added:
constructor SqueezeBertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, clean_up_tokenization_spaces=…, **kwargs) - added:
field SqueezeBertTokenizer.vocab_files_names - added:
method SqueezeBertTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method SqueezeBertTokenizer.convert_tokens_to_string(tokens) - added:
method SqueezeBertTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method SqueezeBertTokenizer.get_vocab() - added:
method SqueezeBertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property SqueezeBertTokenizer.do_lower_case - added:
property SqueezeBertTokenizer.vocab_size
- added:
class SqueezeBertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor SqueezeBertTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - added:
field SqueezeBertTokenizerFast.slow_tokenizer_class - added:
field SqueezeBertTokenizerFast.vocab_files_names - added:
method SqueezeBertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method SqueezeBertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class StableLmConfig(PretrainedConfig) - added:
constructor StableLmConfig(vocab_size=…, intermediate_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, layer_norm_eps=…, use_cache=…, tie_word_embeddings=…, rope_theta=…, rope_scaling=…, use_qkv_bias=…, qk_layernorm=…, use_parallel_residual=…, hidden_dropout=…, attention_dropout=…, partial_rotar… - added:
field StableLmConfig.keys_to_ignore_at_inference - added:
field StableLmConfig.model_type
- added:
class StableLmForCausalLM(StableLmPreTrainedModel, GenerationMixin) - added:
class StableLmForSequenceClassification(GenericForSequenceClassification, StableLmPreTrainedModel) - added:
class StableLmForTokenClassification(GenericForTokenClassification, StableLmPreTrainedModel) - added:
class StableLmModel(StableLmPreTrainedModel) - added:
class StableLmPreTrainedModel(PreTrainedModel) - added:
constructor StableLmForCausalLM(config) - added:
constructor StableLmModel(config: StableLmConfig) - added:
field StableLmPreTrainedModel.base_model_prefix - added:
field StableLmPreTrainedModel.config: StableLmConfig - added:
field StableLmPreTrainedModel.supports_gradient_checkpointing - added:
method StableLmForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_posit… - added:
method StableLmModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…) -> BaseMo…
- added:
export define_import_structure
- added:
class Starcoder2Config(PretrainedConfig) - added:
constructor Starcoder2Config(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, hidden_act=…, max_position_embeddings=…, initializer_range=…, norm_epsilon=…, use_cache=…, bos_token_id=…, eos_token_id=…, rope_theta=…, rope_scaling=…, sliding_window=…, attention_dropout=…, residual_dropout=…, embedding_dropout=…, use_bias=…, **kwargs) - added:
field Starcoder2Config.base_model_pp_plan - added:
field Starcoder2Config.base_model_tp_plan - added:
field Starcoder2Config.keys_to_ignore_at_inference - added:
field Starcoder2Config.model_type
- added:
class Starcoder2ForCausalLM(Starcoder2PreTrainedModel, GenerationMixin) - added:
class Starcoder2ForSequenceClassification(GenericForSequenceClassification, Starcoder2PreTrainedModel) - added:
class Starcoder2ForTokenClassification(GenericForTokenClassification, Starcoder2PreTrainedModel) - added:
class Starcoder2Model(Starcoder2PreTrainedModel) - added:
class Starcoder2PreTrainedModel(PreTrainedModel) - added:
constructor Starcoder2ForCausalLM(config) - added:
constructor Starcoder2Model(config: Starcoder2Config) - added:
field Starcoder2PreTrainedModel.base_model_prefix - added:
field Starcoder2PreTrainedModel.config: Starcoder2Config - added:
field Starcoder2PreTrainedModel.supports_gradient_checkpointing - added:
method Starcoder2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, logits_to_keep: Union[int, torch.Tensor]… - added:
method Starcoder2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOu…
- added:
class Starcoder2ForCausalLM(MistralForCausalLM) - added:
class Starcoder2ForSequenceClassification(MistralForSequenceClassification) - added:
class Starcoder2ForTokenClassification(MistralForTokenClassification) - added:
class Starcoder2Model(MistralModel) - added:
constructor Starcoder2Model(config: Starcoder2Config) - added:
method Starcoder2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOu…
- added:
export define_import_structure
- added:
class SuperGlueConfig(PretrainedConfig) - added:
constructor SuperGlueConfig(keypoint_detector_config: SuperPointConfig=…, hidden_size: int=…, keypoint_encoder_sizes: Optional[list[int]]=…, gnn_layers_types: Optional[list[str]]=…, num_attention_heads: int=…, sinkhorn_iterations: int=…, matching_threshold: float=…, initializer_range: float=…, **kwargs) - added:
field SuperGlueConfig.model_type - added:
property SuperGlueConfig.sub_configs
- added:
class SuperGlueImageProcessor(BaseImageProcessor) - added:
constructor SuperGlueImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_grayscale: bool=…, **kwargs) -> None - added:
field SuperGlueImageProcessor.model_input_names - added:
method SuperGlueImageProcessor.post_process_keypoint_matching(outputs: KeypointMatchingOutput, target_sizes: Union[TensorType, list[tuple]], threshold: float=…) -> list[dict[str, torch.Tensor]] - added:
method SuperGlueImageProcessor.preprocess(images, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_grayscale: Optional[bool]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, *… - added:
method SuperGlueImageProcessor.resize(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method SuperGlueImageProcessor.visualize_keypoint_matching(images: ImageInput, keypoint_matching_output: list[dict[str, torch.Tensor]]) -> list[Image.Image]
- added:
class SuperGlueForKeypointMatching(SuperGluePreTrainedModel) - added:
class SuperGluePreTrainedModel(PreTrainedModel) - added:
constructor SuperGlueForKeypointMatching(config: SuperGlueConfig) -> None - added:
field SuperGluePreTrainedModel.base_model_prefix - added:
field SuperGluePreTrainedModel.config: SuperGlueConfig - added:
field SuperGluePreTrainedModel.main_input_name - added:
method SuperGlueForKeypointMatching.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, KeypointMatchingOutput]
- added:
export define_import_structure
- added:
class SuperPointConfig(PretrainedConfig) - added:
constructor SuperPointConfig(encoder_hidden_sizes: list[int]=…, decoder_hidden_size: int=…, keypoint_decoder_dim: int=…, descriptor_decoder_dim: int=…, keypoint_threshold: float=…, max_keypoints: int=…, nms_radius: int=…, border_removal_distance: int=…, initializer_range=…, **kwargs) - added:
field SuperPointConfig.model_type
- added:
class SuperPointImageProcessor(BaseImageProcessor) - added:
constructor SuperPointImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: float=…, do_grayscale: bool=…, **kwargs) -> None - added:
field SuperPointImageProcessor.model_input_names - added:
method SuperPointImageProcessor.post_process_keypoint_detection(outputs: SuperPointKeypointDescriptionOutput, target_sizes: Union[TensorType, list[tuple]]) -> list[dict[str, torch.Tensor]] - added:
method SuperPointImageProcessor.preprocess(images, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_grayscale: Optional[bool]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: ChannelDimension=…, input_data_format: Optional[Union[str, ChannelDimension]]=…,… - added:
method SuperPointImageProcessor.resize(image: np.ndarray, size: dict[str, int], data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs)
- added:
class SuperPointImageProcessorFast(BaseImageProcessorFast) - added:
constructor SuperPointImageProcessorFast(**kwargs: Unpack[SuperPointFastImageProcessorKwargs]) - added:
field SuperPointImageProcessorFast.default_to_square - added:
field SuperPointImageProcessorFast.do_normalize - added:
field SuperPointImageProcessorFast.do_rescale - added:
field SuperPointImageProcessorFast.do_resize - added:
field SuperPointImageProcessorFast.resample - added:
field SuperPointImageProcessorFast.rescale_factor - added:
field SuperPointImageProcessorFast.size - added:
field SuperPointImageProcessorFast.valid_kwargs - added:
method SuperPointImageProcessorFast.post_process_keypoint_detection(outputs: SuperPointKeypointDescriptionOutput, target_sizes: Union[TensorType, list[tuple]]) -> list[dict[str, torch.Tensor]]
- added:
class SuperPointForKeypointDetection(SuperPointPreTrainedModel) - added:
class SuperPointPreTrainedModel(PreTrainedModel) - added:
constructor SuperPointForKeypointDetection(config: SuperPointConfig) -> None - added:
field SuperPointPreTrainedModel.base_model_prefix - added:
field SuperPointPreTrainedModel.config: SuperPointConfig - added:
field SuperPointPreTrainedModel.main_input_name - added:
field SuperPointPreTrainedModel.supports_gradient_checkpointing - added:
method SuperPointForKeypointDetection.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, SuperPointKeypointDescriptionOutput] - added:
method SuperPointPreTrainedModel.extract_one_channel_pixel_values(pixel_values: torch.FloatTensor) -> torch.FloatTensor
- added:
export define_import_structure
- added:
class SwiftFormerConfig(PretrainedConfig) - added:
class SwiftFormerOnnxConfig(OnnxConfig) - added:
constructor SwiftFormerConfig(image_size=…, num_channels=…, depths=…, embed_dims=…, mlp_ratio=…, downsamples=…, hidden_act=…, down_patch_size=…, down_stride=…, down_pad=…, drop_path_rate=…, drop_mlp_rate=…, drop_conv_encoder_rate=…, use_layer_scale=…, layer_scale_init_value=…, batch_norm_eps=…, **kwargs) - added:
field SwiftFormerConfig.model_type - added:
field SwiftFormerOnnxConfig.torch_onnx_minimum_version - added:
property SwiftFormerOnnxConfig.atol_for_validation -> float - added:
property SwiftFormerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class SwiftFormerForImageClassification(SwiftFormerPreTrainedModel) - added:
class SwiftFormerModel(SwiftFormerPreTrainedModel) - added:
class SwiftFormerPreTrainedModel(PreTrainedModel) - added:
constructor SwiftFormerForImageClassification(config: SwiftFormerConfig) -> None - added:
constructor SwiftFormerModel(config: SwiftFormerConfig) - added:
field SwiftFormerPreTrainedModel.base_model_prefix - added:
field SwiftFormerPreTrainedModel.config: SwiftFormerConfig - added:
field SwiftFormerPreTrainedModel.main_input_name - added:
field SwiftFormerPreTrainedModel.supports_gradient_checkpointing - added:
method SwiftFormerForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutputWithNoAttention] - added:
method SwiftFormerModel.forward(pixel_values: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithNoAttention]
- added:
class TFSwiftFormerForImageClassification(TFSwiftFormerPreTrainedModel) - added:
class TFSwiftFormerModel(TFSwiftFormerPreTrainedModel) - added:
class TFSwiftFormerPreTrainedModel(TFPreTrainedModel) - added:
constructor TFSwiftFormerForImageClassification(config: SwiftFormerConfig, **kwargs) -> None - added:
constructor TFSwiftFormerModel(config: SwiftFormerConfig, *inputs, **kwargs) - added:
field TFSwiftFormerPreTrainedModel.base_model_prefix - added:
field TFSwiftFormerPreTrainedModel.config_class - added:
field TFSwiftFormerPreTrainedModel.main_input_name - added:
method TFSwiftFormerForImageClassification.build(input_shape=…) - added:
method TFSwiftFormerForImageClassification.call(pixel_values: Optional[tf.Tensor]=…, labels: Optional[tf.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[tuple, TFImageClassifierOutputWithNoAttention] - added:
method TFSwiftFormerForImageClassification.hf_compute_loss(labels, logits) - added:
method TFSwiftFormerModel.build(input_shape=…) - added:
method TFSwiftFormerModel.call(pixel_values: Optional[tf.Tensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, training: bool=…) -> Union[TFBaseModelOutputWithNoAttention, tuple[tf.Tensor]]
- added:
export define_import_structure
- added:
class SwinConfig(BackboneConfigMixin, PretrainedConfig) - added:
class SwinOnnxConfig(OnnxConfig) - added:
constructor SwinConfig(image_size=…, patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, window_size=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, use_absolute_embeddings=…, initializer_range=…, layer_norm_eps=…, encoder_stride=…, out_features=…, out_indices=…, **kwargs) - added:
field SwinConfig.attribute_map - added:
field SwinConfig.model_type - added:
field SwinOnnxConfig.torch_onnx_minimum_version - added:
property SwinOnnxConfig.atol_for_validation -> float - added:
property SwinOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class SwinBackbone(SwinPreTrainedModel, BackboneMixin) - added:
class SwinForImageClassification(SwinPreTrainedModel) - added:
class SwinForMaskedImageModeling(SwinPreTrainedModel) - added:
class SwinModel(SwinPreTrainedModel) - added:
class SwinPreTrainedModel(PreTrainedModel) - added:
constructor SwinBackbone(config: SwinConfig) - added:
constructor SwinForImageClassification(config) - added:
constructor SwinForMaskedImageModeling(config) - added:
constructor SwinModel(config, add_pooling_layer=…, use_mask_token=…) - added:
field SwinPreTrainedModel.base_model_prefix - added:
field SwinPreTrainedModel.config: SwinConfig - added:
field SwinPreTrainedModel.main_input_name - added:
field SwinPreTrainedModel.supports_gradient_checkpointing - added:
method SwinBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method SwinBackbone.get_input_embeddings() - added:
method SwinForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, SwinImageClassifierOutput] - added:
method SwinForMaskedImageModeling.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, SwinMaskedImageModelingOutput] - added:
method SwinModel.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, SwinModelOutput] - added:
method SwinModel.get_input_embeddings()
- added:
class TFSwinForImageClassification(TFSwinPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFSwinForMaskedImageModeling(TFSwinPreTrainedModel) - added:
class TFSwinModel(TFSwinPreTrainedModel) - added:
class TFSwinPreTrainedModel(TFPreTrainedModel) - added:
constructor TFSwinForImageClassification(config: SwinConfig) - added:
constructor TFSwinForMaskedImageModeling(config: SwinConfig) - added:
constructor TFSwinModel(config: SwinConfig, add_pooling_layer: bool=…, use_mask_token: bool=…, **kwargs) -> None - added:
field TFSwinPreTrainedModel.base_model_prefix - added:
field TFSwinPreTrainedModel.config_class - added:
field TFSwinPreTrainedModel.main_input_name - added:
method TFSwinForImageClassification.build(input_shape=…) - added:
method TFSwinForImageClassification.call(pixel_values: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, labels: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple[tf.Tensor, ...] | TFSwinImageClassifierOutput - added:
method TFSwinForMaskedImageModeling.build(input_shape=…) - added:
method TFSwinForMaskedImageModeling.call(pixel_values: tf.Tensor | None=…, bool_masked_pos: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple | TFSwinMaskedImageModelingOutput - added:
method TFSwinModel.build(input_shape=…) - added:
method TFSwinModel.call(pixel_values: tf.Tensor | None=…, bool_masked_pos: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFSwinModelOutput | tuple[tf.Tensor, ...]
- added:
export define_import_structure
- added:
class Swin2SRConfig(PretrainedConfig) - added:
constructor Swin2SRConfig(image_size=…, patch_size=…, num_channels=…, num_channels_out=…, embed_dim=…, depths=…, num_heads=…, window_size=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, use_absolute_embeddings=…, initializer_range=…, layer_norm_eps=…, upscale=…, img_range=…, resi_connection=…, upsampler=…, **kwargs) - added:
field Swin2SRConfig.attribute_map - added:
field Swin2SRConfig.model_type
- added:
class Swin2SRImageProcessor(BaseImageProcessor) - added:
constructor Swin2SRImageProcessor(do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_pad: bool=…, size_divisor: int=…, **kwargs) -> None - added:
field Swin2SRImageProcessor.model_input_names - added:
method Swin2SRImageProcessor.pad(image: np.ndarray, size: int, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method Swin2SRImageProcessor.preprocess(images: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_pad: Optional[bool]=…, size_divisor: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Union[str, ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
property Swin2SRImageProcessor.pad_size
- added:
class Swin2SRImageProcessorFast(BaseImageProcessorFast) - added:
constructor Swin2SRImageProcessorFast(**kwargs: Unpack[Swin2SRFastImageProcessorKwargs]) - added:
field Swin2SRImageProcessorFast.do_pad - added:
field Swin2SRImageProcessorFast.do_rescale - added:
field Swin2SRImageProcessorFast.rescale_factor - added:
field Swin2SRImageProcessorFast.size_divisor - added:
field Swin2SRImageProcessorFast.valid_kwargs - added:
method Swin2SRImageProcessorFast.pad(images: torch.Tensor, size_divisor: int) -> torch.Tensor - added:
method Swin2SRImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[Swin2SRFastImageProcessorKwargs]) -> BatchFeature - added:
property Swin2SRImageProcessorFast.pad_size
- added:
class Swin2SRForImageSuperResolution(Swin2SRPreTrainedModel) - added:
class Swin2SRModel(Swin2SRPreTrainedModel) - added:
class Swin2SRPreTrainedModel(PreTrainedModel) - added:
constructor Swin2SRForImageSuperResolution(config) - added:
constructor Swin2SRModel(config) - added:
field Swin2SRPreTrainedModel.base_model_prefix - added:
field Swin2SRPreTrainedModel.config: Swin2SRConfig - added:
field Swin2SRPreTrainedModel.main_input_name - added:
field Swin2SRPreTrainedModel.supports_gradient_checkpointing - added:
method Swin2SRForImageSuperResolution.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageSuperResolutionOutput] - added:
method Swin2SRModel.forward(pixel_values: torch.FloatTensor, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method Swin2SRModel.get_input_embeddings() - added:
method Swin2SRModel.pad_and_normalize(pixel_values)
- added:
export define_import_structure
- added:
class Swinv2Config(BackboneConfigMixin, PretrainedConfig) - added:
constructor Swinv2Config(image_size=…, patch_size=…, num_channels=…, embed_dim=…, depths=…, num_heads=…, window_size=…, pretrained_window_sizes=…, mlp_ratio=…, qkv_bias=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, drop_path_rate=…, hidden_act=…, use_absolute_embeddings=…, initializer_range=…, layer_norm_eps=…, encoder_stride=…, out_features=…, out_indices=…, **kwargs) - added:
field Swinv2Config.attribute_map - added:
field Swinv2Config.model_type
- added:
class Swinv2Backbone(Swinv2PreTrainedModel, BackboneMixin) - added:
class Swinv2ForImageClassification(Swinv2PreTrainedModel) - added:
class Swinv2ForMaskedImageModeling(Swinv2PreTrainedModel) - added:
class Swinv2Model(Swinv2PreTrainedModel) - added:
class Swinv2PreTrainedModel(PreTrainedModel) - added:
constructor Swinv2Backbone(config) - added:
constructor Swinv2ForImageClassification(config) - added:
constructor Swinv2ForMaskedImageModeling(config) - added:
constructor Swinv2Model(config, add_pooling_layer=…, use_mask_token=…) - added:
field Swinv2PreTrainedModel.base_model_prefix - added:
field Swinv2PreTrainedModel.config: Swinv2Config - added:
field Swinv2PreTrainedModel.main_input_name - added:
field Swinv2PreTrainedModel.supports_gradient_checkpointing - added:
method Swinv2Backbone.forward(pixel_values: Tensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method Swinv2Backbone.get_input_embeddings() - added:
method Swinv2ForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, Swinv2ImageClassifierOutput] - added:
method Swinv2ForMaskedImageModeling.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, Swinv2MaskedImageModelingOutput] - added:
method Swinv2Model.forward(pixel_values: Optional[torch.FloatTensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, Swinv2ModelOutput] - added:
method Swinv2Model.get_input_embeddings()
- added:
export define_import_structure
- added:
class SwitchTransformersConfig(PretrainedConfig) - added:
constructor SwitchTransformersConfig(vocab_size=…, d_model=…, d_kv=…, d_ff=…, expert_capacity=…, num_layers=…, num_sparse_encoder_layers=…, num_decoder_layers=…, num_sparse_decoder_layers=…, num_heads=…, num_experts=…, router_bias=…, router_jitter_noise=…, router_dtype=…, router_ignore_padding_tokens=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_nor… - added:
field SwitchTransformersConfig.attribute_map - added:
field SwitchTransformersConfig.keys_to_ignore_at_inference - added:
field SwitchTransformersConfig.model_type
- added:
class SwitchTransformersEncoderModel(SwitchTransformersPreTrainedModel) - added:
class SwitchTransformersForConditionalGeneration(SwitchTransformersPreTrainedModel, GenerationMixin) - added:
class SwitchTransformersModel(SwitchTransformersPreTrainedModel) - added:
class SwitchTransformersPreTrainedModel(PreTrainedModel) - added:
class SwitchTransformersSparseMLP(nn.Module) - added:
class SwitchTransformersTop1Router(nn.Module) - added:
constructor SwitchTransformersEncoderModel(config: SwitchTransformersConfig) - added:
constructor SwitchTransformersForConditionalGeneration(config: SwitchTransformersConfig) - added:
constructor SwitchTransformersModel(config: SwitchTransformersConfig) - added:
constructor SwitchTransformersSparseMLP(config: SwitchTransformersConfig, expert_class: nn.Module=…) - added:
constructor SwitchTransformersTop1Router(config: SwitchTransformersConfig) - added:
field SwitchTransformersPreTrainedModel.base_model_prefix - added:
field SwitchTransformersPreTrainedModel.config: SwitchTransformersConfig - added:
field SwitchTransformersPreTrainedModel.supports_gradient_checkpointing - added:
method SwitchTransformersEncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, output_router_logits: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTen… - added:
method SwitchTransformersEncoderModel.get_encoder() - added:
method SwitchTransformersEncoderModel.get_input_embeddings() - added:
method SwitchTransformersEncoderModel.set_input_embeddings(new_embeddings) - added:
method SwitchTransformersForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_ou… - added:
method SwitchTransformersForConditionalGeneration.get_encoder() - added:
method SwitchTransformersForConditionalGeneration.get_input_embeddings() - added:
method SwitchTransformersForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method SwitchTransformersForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method SwitchTransformersModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tup… - added:
method SwitchTransformersModel.get_encoder() - added:
method SwitchTransformersModel.get_input_embeddings() - added:
method SwitchTransformersModel.set_input_embeddings(new_embeddings) - added:
method SwitchTransformersSparseMLP.forward(hidden_states) - added:
method SwitchTransformersTop1Router.forward(hidden_states: torch.Tensor) -> tuple - added:
property SwitchTransformersPreTrainedModel.dummy_inputs
- added:
export define_import_structure
- added:
class T5Config(PretrainedConfig) - added:
class T5OnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor T5Config(vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, use_cache=…, pad_token_id=…, eos_token_id=…, classifier_dropout=…, **kwargs) - added:
field T5Config.attribute_map - added:
field T5Config.keys_to_ignore_at_inference - added:
field T5Config.model_type - added:
property T5OnnxConfig.default_onnx_opset -> int - added:
property T5OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxT5EncoderModel(FlaxT5PreTrainedModel) - added:
class FlaxT5ForConditionalGeneration(FlaxT5PreTrainedModel) - added:
class FlaxT5Model(FlaxT5PreTrainedModel) - added:
class FlaxT5PreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxT5PreTrainedModel(config: T5Config, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxT5EncoderModel.module_class - added:
field FlaxT5ForConditionalGeneration.module_class - added:
field FlaxT5Model.module_class - added:
field FlaxT5PreTrainedModel.base_model_prefix - added:
field FlaxT5PreTrainedModel.config_class - added:
field FlaxT5PreTrainedModel.module_class: nn.Module - added:
method FlaxT5ForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxT5ForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxT5ForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxT5PreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxT5PreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxT5PreTrainedModel.encode(input_ids: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxT5PreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxT5PreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class T5EncoderModel(T5PreTrainedModel) - added:
class T5ForConditionalGeneration(T5PreTrainedModel, GenerationMixin) - added:
class T5ForQuestionAnswering(T5PreTrainedModel) - added:
class T5ForSequenceClassification(T5PreTrainedModel) - added:
class T5ForTokenClassification(T5PreTrainedModel) - added:
class T5Model(T5PreTrainedModel) - added:
class T5PreTrainedModel(PreTrainedModel) - added:
constructor T5EncoderModel(config: T5Config) - added:
constructor T5ForConditionalGeneration(config: T5Config) - added:
constructor T5ForQuestionAnswering(config: T5Config) - added:
constructor T5ForSequenceClassification(config: T5Config) - added:
constructor T5ForTokenClassification(config: T5Config) - added:
constructor T5Model(config: T5Config) - added:
field T5PreTrainedModel.base_model_prefix - added:
field T5PreTrainedModel.config: T5Config - added:
field T5PreTrainedModel.is_parallelizable - added:
field T5PreTrainedModel.load_tf_weights - added:
field T5PreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_t5(model, config, tf_checkpoint_path) - added:
method T5EncoderModel.deparallelize() - added:
method T5EncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method T5EncoderModel.get_encoder() - added:
method T5EncoderModel.get_input_embeddings() - added:
method T5EncoderModel.parallelize(device_map=…) - added:
method T5EncoderModel.set_input_embeddings(new_embeddings) - added:
method T5ForConditionalGeneration.deparallelize() - added:
method T5ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[… - added:
method T5ForConditionalGeneration.get_encoder() - added:
method T5ForConditionalGeneration.get_input_embeddings() - added:
method T5ForConditionalGeneration.parallelize(device_map=…) - added:
method T5ForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method T5ForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method T5ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tupl… - added:
method T5ForQuestionAnswering.get_encoder() - added:
method T5ForQuestionAnswering.get_input_embeddings() - added:
method T5ForQuestionAnswering.set_input_embeddings(new_embeddings) - added:
method T5ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.Flo… - added:
method T5ForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], TokenClassifierOutput] - added:
method T5Model.deparallelize() - added:
method T5Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.F… - added:
method T5Model.get_encoder() - added:
method T5Model.get_input_embeddings() - added:
method T5Model.parallelize(device_map=…) - added:
method T5Model.set_input_embeddings(new_embeddings) - added:
property T5PreTrainedModel.dummy_inputs
- added:
class TFT5EncoderModel(TFT5PreTrainedModel) - added:
class TFT5ForConditionalGeneration(TFT5PreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFT5Model(TFT5PreTrainedModel) - added:
class TFT5PreTrainedModel(TFPreTrainedModel) - added:
constructor TFT5EncoderModel(config, *inputs, **kwargs) - added:
constructor TFT5ForConditionalGeneration(config, *inputs, **kwargs) - added:
constructor TFT5Model(config, *inputs, **kwargs) - added:
field TFT5PreTrainedModel.base_model_prefix - added:
field TFT5PreTrainedModel.config_class - added:
method TFT5EncoderModel.build(input_shape=…) - added:
method TFT5EncoderModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> tuple | TFBaseModelOutput - added:
method TFT5EncoderModel.get_encoder() - added:
method TFT5ForConditionalGeneration.build(input_shape=…) - added:
method TFT5ForConditionalGeneration.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, past_key_values:… - added:
method TFT5ForConditionalGeneration.get_encoder() - added:
method TFT5ForConditionalGeneration.get_output_embeddings() - added:
method TFT5ForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFT5ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, decoder_attention_mask=…, head_mask=…, decoder_head_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFT5ForConditionalGeneration.serving_output(output) - added:
method TFT5ForConditionalGeneration.set_output_embeddings(value) - added:
method TFT5Model.build(input_shape=…) - added:
method TFT5Model.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[np.nda… - added:
method TFT5Model.get_encoder() - added:
method TFT5PreTrainedModel.get_input_embeddings() - added:
method TFT5PreTrainedModel.set_input_embeddings(value)
- added:
class T5Tokenizer(PreTrainedTokenizer) - added:
constructor T5Tokenizer(vocab_file, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, legacy=…, add_prefix_space=…, **kwargs) -> None - added:
field T5Tokenizer.model_input_names - added:
field T5Tokenizer.vocab_files_names - added:
method T5Tokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method T5Tokenizer.convert_tokens_to_string(tokens) - added:
method T5Tokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method T5Tokenizer.get_sentinel_token_ids() - added:
method T5Tokenizer.get_sentinel_tokens() - added:
method T5Tokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method T5Tokenizer.get_spm_processor(from_slow=…) - added:
method T5Tokenizer.get_vocab() - added:
method T5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method T5Tokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
property T5Tokenizer.unk_token_length - added:
property T5Tokenizer.vocab_size
- added:
class T5TokenizerFast(PreTrainedTokenizerFast) - added:
constructor T5TokenizerFast(vocab_file=…, tokenizer_file=…, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, add_prefix_space=…, **kwargs) - added:
field T5TokenizerFast.model_input_names - added:
field T5TokenizerFast.prefix_tokens: list[int] - added:
field T5TokenizerFast.slow_tokenizer_class - added:
field T5TokenizerFast.vocab_files_names - added:
method T5TokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method T5TokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method T5TokenizerFast.get_sentinel_token_ids() - added:
method T5TokenizerFast.get_sentinel_tokens() - added:
method T5TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class T5GemmaConfig(PretrainedConfig) - added:
class T5GemmaModuleConfig(PretrainedConfig) - added:
constructor T5GemmaConfig(encoder: Optional[Union[T5GemmaModuleConfig, dict[Any, Any]]]=…, decoder: Optional[Union[T5GemmaModuleConfig, dict[Any, Any]]]=…, is_encoder_decoder: bool=…, dropout_rate: float=…, classifier_dropout_rate: float=…, attention_dropout: float=…, tie_word_embeddings: bool=…, vocab_size: int=…, **kwargs) - added:
constructor T5GemmaModuleConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_s… - added:
field T5GemmaConfig.base_model_pp_plan - added:
field T5GemmaConfig.base_model_tp_plan - added:
field T5GemmaConfig.keys_to_ignore_at_inference - added:
field T5GemmaConfig.model_type - added:
field T5GemmaModuleConfig.base_model_pp_plan - added:
field T5GemmaModuleConfig.base_model_tp_plan - added:
field T5GemmaModuleConfig.keys_to_ignore_at_inference - added:
field T5GemmaModuleConfig.model_type
- added:
class T5GemmaEncoderModel(T5GemmaPreTrainedModel) - added:
class T5GemmaForConditionalGeneration(T5GemmaPreTrainedModel, GenerationMixin) - added:
class T5GemmaForSequenceClassification(T5GemmaPreTrainedModel) - added:
class T5GemmaForTokenClassification(T5GemmaPreTrainedModel) - added:
class T5GemmaModel(T5GemmaPreTrainedModel) - added:
class T5GemmaPreTrainedModel(PreTrainedModel) - added:
constructor T5GemmaEncoderModel(config: T5GemmaConfig) - added:
constructor T5GemmaForConditionalGeneration(config: T5GemmaConfig) - added:
constructor T5GemmaForSequenceClassification(config: T5GemmaConfig, is_encoder_decoder: Optional[bool]=…) - added:
constructor T5GemmaForTokenClassification(config: T5GemmaConfig, is_encoder_decoder: Optional[bool]=…) - added:
constructor T5GemmaModel(config: T5GemmaConfig) - added:
field T5GemmaPreTrainedModel.base_model_prefix - added:
field T5GemmaPreTrainedModel.config: T5GemmaConfig - added:
field T5GemmaPreTrainedModel.supports_gradient_checkpointing - added:
method T5GemmaEncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method T5GemmaEncoderModel.get_input_embeddings() - added:
method T5GemmaEncoderModel.set_input_embeddings(new_embeddings) - added:
method T5GemmaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, past_key_values: Op… - added:
method T5GemmaForConditionalGeneration.get_decoder() - added:
method T5GemmaForConditionalGeneration.get_encoder() - added:
method T5GemmaForConditionalGeneration.get_output_embeddings() - added:
method T5GemmaForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method T5GemmaForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method T5GemmaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, inputs_embeds: Optional[tor… - added:
method T5GemmaForSequenceClassification.get_input_embeddings() - added:
method T5GemmaForSequenceClassification.set_input_embeddings(value) - added:
method T5GemmaForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, inputs_embeds: Optional[torch.… - added:
method T5GemmaForTokenClassification.get_input_embeddings() - added:
method T5GemmaForTokenClassification.set_input_embeddings(value) - added:
method T5GemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, past_key_values: Optional[EncoderDecod… - added:
method T5GemmaModel.get_encoder() - added:
method T5GemmaModel.get_input_embeddings() - added:
method T5GemmaModel.set_input_embeddings(new_embeddings)
- added:
class T5GemmaConfig(PretrainedConfig) - added:
class T5GemmaEncoderModel(T5GemmaPreTrainedModel) - added:
class T5GemmaForConditionalGeneration(T5GemmaPreTrainedModel, GenerationMixin) - added:
class T5GemmaForSequenceClassification(T5GemmaPreTrainedModel) - added:
class T5GemmaForTokenClassification(T5GemmaPreTrainedModel) - added:
class T5GemmaModel(T5GemmaPreTrainedModel) - added:
class T5GemmaModuleConfig(Gemma2Config) - added:
class T5GemmaPreTrainedModel(Gemma2PreTrainedModel) - added:
constructor T5GemmaConfig(encoder: Optional[Union[T5GemmaModuleConfig, dict[Any, Any]]]=…, decoder: Optional[Union[T5GemmaModuleConfig, dict[Any, Any]]]=…, is_encoder_decoder: bool=…, dropout_rate: float=…, classifier_dropout_rate: float=…, attention_dropout: float=…, tie_word_embeddings: bool=…, vocab_size: int=…, **kwargs) - added:
constructor T5GemmaEncoderModel(config: T5GemmaConfig) - added:
constructor T5GemmaForConditionalGeneration(config: T5GemmaConfig) - added:
constructor T5GemmaForSequenceClassification(config: T5GemmaConfig, is_encoder_decoder: Optional[bool]=…) - added:
constructor T5GemmaForTokenClassification(config: T5GemmaConfig, is_encoder_decoder: Optional[bool]=…) - added:
constructor T5GemmaModel(config: T5GemmaConfig) - added:
field T5GemmaConfig.base_model_pp_plan - added:
field T5GemmaConfig.base_model_tp_plan - added:
field T5GemmaConfig.keys_to_ignore_at_inference - added:
field T5GemmaConfig.model_type - added:
field T5GemmaPreTrainedModel.base_model_prefix - added:
field T5GemmaPreTrainedModel.config: T5GemmaConfig - added:
field T5GemmaPreTrainedModel.supports_gradient_checkpointing - added:
method T5GemmaEncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method T5GemmaEncoderModel.get_input_embeddings() - added:
method T5GemmaEncoderModel.set_input_embeddings(new_embeddings) - added:
method T5GemmaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, past_key_values: Op… - added:
method T5GemmaForConditionalGeneration.get_decoder() - added:
method T5GemmaForConditionalGeneration.get_encoder() - added:
method T5GemmaForConditionalGeneration.get_output_embeddings() - added:
method T5GemmaForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method T5GemmaForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method T5GemmaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, inputs_embeds: Optional[tor… - added:
method T5GemmaForSequenceClassification.get_input_embeddings() - added:
method T5GemmaForSequenceClassification.set_input_embeddings(value) - added:
method T5GemmaForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.Tensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, inputs_embeds: Optional[torch.… - added:
method T5GemmaForTokenClassification.get_input_embeddings() - added:
method T5GemmaForTokenClassification.set_input_embeddings(value) - added:
method T5GemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, position_ids: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, decoder_position_ids: Optional[torch.LongTensor]=…, encoder_outputs: Optional[BaseModelOutput]=…, past_key_values: Optional[EncoderDecod… - added:
method T5GemmaModel.get_encoder() - added:
method T5GemmaModel.get_input_embeddings() - added:
method T5GemmaModel.set_input_embeddings(new_embeddings)
- added:
export define_import_structure
- added:
class TableTransformerConfig(PretrainedConfig) - added:
class TableTransformerOnnxConfig(OnnxConfig) - added:
constructor TableTransformerConfig(use_timm_backbone=…, backbone_config=…, num_channels=…, num_queries=…, encoder_layers=…, encoder_ffn_dim=…, encoder_attention_heads=…, decoder_layers=…, decoder_ffn_dim=…, decoder_attention_heads=…, encoder_layerdrop=…, decoder_layerdrop=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, i… - added:
field TableTransformerConfig.attribute_map - added:
field TableTransformerConfig.keys_to_ignore_at_inference - added:
field TableTransformerConfig.model_type - added:
field TableTransformerOnnxConfig.torch_onnx_minimum_version - added:
property TableTransformerConfig.hidden_size -> int - added:
property TableTransformerConfig.num_attention_heads -> int - added:
property TableTransformerConfig.sub_configs - added:
property TableTransformerOnnxConfig.atol_for_validation -> float - added:
property TableTransformerOnnxConfig.default_onnx_opset -> int - added:
property TableTransformerOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class TableTransformerForObjectDetection(TableTransformerPreTrainedModel) - added:
class TableTransformerModel(TableTransformerPreTrainedModel) - added:
class TableTransformerPreTrainedModel(PreTrainedModel) - added:
constructor TableTransformerForObjectDetection(config: TableTransformerConfig) - added:
constructor TableTransformerModel(config: TableTransformerConfig) - added:
field TableTransformerPreTrainedModel.base_model_prefix - added:
field TableTransformerPreTrainedModel.config: TableTransformerConfig - added:
field TableTransformerPreTrainedModel.main_input_name - added:
method TableTransformerForObjectDetection.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[list[dict]]=…, output_attentions: Optional[bool]=…, o… - added:
method TableTransformerModel.forward(pixel_values: torch.FloatTensor, pixel_mask: Optional[torch.FloatTensor]=…, decoder_attention_mask: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return… - added:
method TableTransformerModel.freeze_backbone() - added:
method TableTransformerModel.get_encoder() - added:
method TableTransformerModel.unfreeze_backbone()
- added:
export define_import_structure
- added:
class TapasConfig(PretrainedConfig) - added:
constructor TapasConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_sizes=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, positive_label_weight=…, num_aggregation_labels=…, aggregation_loss_weight=…, use_answer_as_supervision=…, ans… - added:
field TapasConfig.model_type
- added:
class TapasForMaskedLM(TapasPreTrainedModel) - added:
class TapasForQuestionAnswering(TapasPreTrainedModel) - added:
class TapasForSequenceClassification(TapasPreTrainedModel) - added:
class TapasModel(TapasPreTrainedModel) - added:
class TapasPreTrainedModel(PreTrainedModel) - added:
constructor TapasForMaskedLM(config) - added:
constructor TapasForQuestionAnswering(config: TapasConfig) - added:
constructor TapasForSequenceClassification(config) - added:
constructor TapasModel(config, add_pooling_layer=…) - added:
field TapasForMaskedLM.base_model_prefix - added:
field TapasForMaskedLM.config: TapasConfig - added:
field TapasPreTrainedModel.base_model_prefix - added:
field TapasPreTrainedModel.config: TapasConfig - added:
field TapasPreTrainedModel.supports_gradient_checkpointing - added:
function load_tf_weights_in_tapas(model, config, tf_checkpoint_path) - added:
method TapasForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTen… - added:
method TapasForMaskedLM.get_output_embeddings() - added:
method TapasForMaskedLM.set_output_embeddings(new_embeddings) - added:
method TapasForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, table_mask: Optional[torch.LongTensor]=…, labels: Optional[torch.LongTensor]=…, aggregation_… - added:
method TapasForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidde… - added:
method TapasModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.FloatTensor]=…… - added:
method TapasModel.get_input_embeddings() - added:
method TapasModel.set_input_embeddings(value)
- added:
class TFTapasForMaskedLM(TFTapasPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFTapasForQuestionAnswering(TFTapasPreTrainedModel) - added:
class TFTapasForSequenceClassification(TFTapasPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFTapasModel(TFTapasPreTrainedModel) - added:
class TFTapasPreTrainedModel(TFPreTrainedModel) - added:
constructor TFTapasForMaskedLM(config: TapasConfig, *inputs, **kwargs) - added:
constructor TFTapasForQuestionAnswering(config: TapasConfig, *inputs, **kwargs) - added:
constructor TFTapasForSequenceClassification(config: TapasConfig, *inputs, **kwargs) - added:
constructor TFTapasModel(config: TapasConfig, *inputs, **kwargs) - added:
field TFTapasPreTrainedModel.base_model_prefix - added:
field TFTapasPreTrainedModel.config_class - added:
method TFTapasForMaskedLM.build(input_shape=…) - added:
method TFTapasForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | No… - added:
method TFTapasForMaskedLM.get_lm_head() -> keras.layers.Layer - added:
method TFTapasForQuestionAnswering.build(input_shape=…) - added:
method TFTapasForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, table_mask: np.ndarray | tf.Tensor | None=…, aggregation_labels: np.ndarray | tf.Ten… - added:
method TFTapasForSequenceClassification.build(input_shape=…) - added:
method TFTapasForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_d… - added:
method TFTapasModel.build(input_shape=…) - added:
method TFTapasModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…,… - added:
property TFTapasPreTrainedModel.input_signature
- added:
class TapasTokenizer(PreTrainedTokenizer) - added:
constructor TapasTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, empty_token=…, tokenize_chinese_chars=…, strip_accents=…, cell_trim_length: int=…, max_column_id: Optional[int]=…, max_row_id: Optional[int]=…, strip_column_names: bool=…, update_answer_coordinates: bool=…, min_question_length=…, max_questio… - added:
field TapasTokenizer.vocab_files_names - added:
method TapasTokenizer.batch_encode_plus(table: pd.DataFrame, queries: Optional[Union[list[TextInput], list[PreTokenizedInput], list[EncodedInput]]]=…, answer_coordinates: Optional[list[list[tuple]]]=…, answer_text: Optional[list[list[TextInput]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TapasTruncationStrategy]=…, max_length: Option… - added:
method TapasTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method TapasTokenizer.convert_logits_to_predictions(data, logits, logits_agg=…, cell_classification_threshold=…) - added:
method TapasTokenizer.convert_tokens_to_string(tokens) - added:
method TapasTokenizer.create_attention_mask_from_sequences(query_ids: list[int], table_values: list[TableValue]) -> list[int] - added:
method TapasTokenizer.create_column_token_type_ids_from_sequences(query_ids: list[int], table_values: list[TableValue]) -> list[int] - added:
method TapasTokenizer.create_row_token_type_ids_from_sequences(query_ids: list[int], table_values: list[TableValue]) -> list[int] - added:
method TapasTokenizer.create_segment_token_type_ids_from_sequences(query_ids: list[int], table_values: list[TableValue]) -> list[int] - added:
method TapasTokenizer.encode(table: pd.DataFrame, query: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TapasTruncationStrategy]=…, max_length: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, **kwargs) -> list[int] - added:
method TapasTokenizer.encode_plus(table: pd.DataFrame, query: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, answer_coordinates: Optional[list[tuple]]=…, answer_text: Optional[list[TextInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TapasTruncationStrategy]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optiona… - added:
method TapasTokenizer.get_answer_ids(column_ids, row_ids, tokenized_table, answer_texts_question, answer_coordinates_question) - added:
method TapasTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method TapasTokenizer.get_vocab() - added:
method TapasTokenizer.prepare_for_model(raw_table: pd.DataFrame, raw_query: Union[TextInput, PreTokenizedInput, EncodedInput], tokenized_table: Optional[TokenizedTable]=…, query_tokens: Optional[TokenizedTable]=…, answer_coordinates: Optional[list[tuple]]=…, answer_text: Optional[list[TextInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool,… - added:
method TapasTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property TapasTokenizer.do_lower_case - added:
property TapasTokenizer.vocab_size
- added:
export define_import_structure
- added:
class TextNetConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor TextNetConfig(stem_kernel_size=…, stem_stride=…, stem_num_channels=…, stem_out_channels=…, stem_act_func=…, image_size=…, conv_layer_kernel_sizes=…, conv_layer_strides=…, hidden_sizes=…, batch_norm_eps=…, initializer_range=…, out_features=…, out_indices=…, **kwargs) - added:
field TextNetConfig.model_type
- added:
class TextNetImageProcessor(BaseImageProcessor) - added:
constructor TextNetImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, size_divisor: int=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb:… - added:
field TextNetImageProcessor.model_input_names - added:
method TextNetImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, size_divisor: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float… - added:
method TextNetImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class TextNetImageProcessorFast(BaseImageProcessorFast) - added:
constructor TextNetImageProcessorFast(**kwargs: Unpack[TextNetFastImageProcessorKwargs]) -> None - added:
field TextNetImageProcessorFast.crop_size - added:
field TextNetImageProcessorFast.default_to_square - added:
field TextNetImageProcessorFast.do_center_crop - added:
field TextNetImageProcessorFast.do_convert_rgb - added:
field TextNetImageProcessorFast.do_normalize - added:
field TextNetImageProcessorFast.do_rescale - added:
field TextNetImageProcessorFast.do_resize - added:
field TextNetImageProcessorFast.image_mean - added:
field TextNetImageProcessorFast.image_std - added:
field TextNetImageProcessorFast.resample - added:
field TextNetImageProcessorFast.size - added:
field TextNetImageProcessorFast.size_divisor - added:
field TextNetImageProcessorFast.valid_kwargs - added:
method TextNetImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[TextNetFastImageProcessorKwargs]) -> BatchFeature - added:
method TextNetImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, size_divisor: int=…, **kwargs) -> torch.Tensor
- added:
class TextNetBackbone(TextNetPreTrainedModel, BackboneMixin) - added:
class TextNetForImageClassification(TextNetPreTrainedModel) - added:
class TextNetModel(TextNetPreTrainedModel) - added:
class TextNetPreTrainedModel(PreTrainedModel) - added:
constructor TextNetBackbone(config) - added:
constructor TextNetForImageClassification(config) - added:
constructor TextNetModel(config) - added:
field TextNetBackbone.has_attentions - added:
field TextNetPreTrainedModel.base_model_prefix - added:
field TextNetPreTrainedModel.config: TextNetConfig - added:
field TextNetPreTrainedModel.main_input_name - added:
method TextNetBackbone.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[tuple], BackboneOutput] - added:
method TextNetForImageClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> ImageClassifierOutputWithNoAttention - added:
method TextNetModel.forward(pixel_values: Tensor, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[Any, list[Any]], tuple[Any], BaseModelOutputWithPoolingAndNoAttention]
- added:
export define_import_structure
- added:
class TimeSeriesTransformerConfig(PretrainedConfig) - added:
constructor TimeSeriesTransformerConfig(prediction_length: Optional[int]=…, context_length: Optional[int]=…, distribution_output: str=…, loss: str=…, input_size: int=…, lags_sequence: list[int]=…, scaling: Optional[Union[str, bool]]=…, num_dynamic_real_features: int=…, num_static_categorical_features: int=…, num_static_real_features: int=…, num_time_features: int=…, cardinality: Optional[list[int… - added:
field TimeSeriesTransformerConfig.attribute_map - added:
field TimeSeriesTransformerConfig.model_type
- added:
class TimeSeriesTransformerForPrediction(TimeSeriesTransformerPreTrainedModel) - added:
class TimeSeriesTransformerModel(TimeSeriesTransformerPreTrainedModel) - added:
class TimeSeriesTransformerPreTrainedModel(PreTrainedModel) - added:
constructor TimeSeriesTransformerForPrediction(config: TimeSeriesTransformerConfig) - added:
constructor TimeSeriesTransformerModel(config: TimeSeriesTransformerConfig) - added:
field TimeSeriesTransformerPreTrainedModel.base_model_prefix - added:
field TimeSeriesTransformerPreTrainedModel.config: TimeSeriesTransformerConfig - added:
field TimeSeriesTransformerPreTrainedModel.main_input_name - added:
field TimeSeriesTransformerPreTrainedModel.supports_gradient_checkpointing - added:
method TimeSeriesTransformerForPrediction.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, future_observed_mask: Optional[torch.Tensor]=…, decoder_attent… - added:
method TimeSeriesTransformerForPrediction.generate(past_values: torch.Tensor, past_time_features: torch.Tensor, future_time_features: torch.Tensor, past_observed_mask: Optional[torch.Tensor]=…, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> SampleTSPredictionOutp… - added:
method TimeSeriesTransformerForPrediction.get_decoder() - added:
method TimeSeriesTransformerForPrediction.get_encoder() - added:
method TimeSeriesTransformerForPrediction.output_distribution(params, loc=…, scale=…, trailing_n=…) -> torch.distributions.Distribution - added:
method TimeSeriesTransformerForPrediction.output_params(dec_output) - added:
method TimeSeriesTransformerModel.create_network_inputs(past_values: torch.Tensor, past_time_features: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, past_observed_mask: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…) - added:
method TimeSeriesTransformerModel.forward(past_values: torch.Tensor, past_time_features: torch.Tensor, past_observed_mask: torch.Tensor, static_categorical_features: Optional[torch.Tensor]=…, static_real_features: Optional[torch.Tensor]=…, future_values: Optional[torch.Tensor]=…, future_time_features: Optional[torch.Tensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optio… - added:
method TimeSeriesTransformerModel.get_encoder() - added:
method TimeSeriesTransformerModel.get_lagged_subsequences(sequence: torch.Tensor, subsequences_length: int, shift: int=…) -> torch.Tensor
- added:
export define_import_structure
- added:
class TimesFmConfig(PretrainedConfig) - added:
constructor TimesFmConfig(patch_length: int=…, context_length: int=…, horizon_length: int=…, freq_size: int=…, num_hidden_layers: int=…, hidden_size: int=…, intermediate_size: int=…, head_dim: int=…, num_attention_heads: int=…, tolerance: float=…, rms_norm_eps: float=…, quantiles: list[float]=…, pad_val: float=…, attention_dropout: float=…, use_positional_embedding: bool=…, initializer_range: flo… - added:
field TimesFmConfig.is_encoder_decoder - added:
field TimesFmConfig.keys_to_ignore_at_inference - added:
field TimesFmConfig.model_type
- added:
class TimesFmModel(TimesFmPreTrainedModel) - added:
class TimesFmModelForPrediction(TimesFmPreTrainedModel) - added:
class TimesFmPreTrainedModel(PreTrainedModel) - added:
constructor TimesFmModel(config: TimesFmConfig) - added:
constructor TimesFmModelForPrediction(config: TimesFmConfig) - added:
field TimesFmPreTrainedModel.base_model_prefix - added:
field TimesFmPreTrainedModel.config: TimesFmConfig - added:
field TimesFmPreTrainedModel.main_input_name - added:
method TimesFmModel.forward(past_values: torch.Tensor, past_values_padding: torch.LongTensor, freq: torch.Tensor, output_attentions: bool=…, output_hidden_states: bool=…) -> TimesFmOutput - added:
method TimesFmModelForPrediction.forward(past_values: Sequence[torch.Tensor], freq: Optional[Sequence[Union[torch.Tensor, int]]]=…, window_size: Optional[int]=…, future_values: Optional[torch.Tensor]=…, forecast_context_len: Optional[int]=…, return_forecast_on_context: bool=…, truncate_negative: bool=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> TimesFmOutputF…
- added:
class TimesFmModel(TimesFmPreTrainedModel) - added:
class TimesFmModelForPrediction(TimesFmPreTrainedModel) - added:
class TimesFmPreTrainedModel(PreTrainedModel) - added:
constructor TimesFmModel(config: TimesFmConfig) - added:
constructor TimesFmModelForPrediction(config: TimesFmConfig) - added:
field TimesFmPreTrainedModel.base_model_prefix - added:
field TimesFmPreTrainedModel.config: TimesFmConfig - added:
field TimesFmPreTrainedModel.main_input_name - added:
method TimesFmModel.forward(past_values: torch.Tensor, past_values_padding: torch.LongTensor, freq: torch.Tensor, output_attentions: bool=…, output_hidden_states: bool=…) -> TimesFmOutput - added:
method TimesFmModelForPrediction.forward(past_values: Sequence[torch.Tensor], freq: Optional[Sequence[Union[torch.Tensor, int]]]=…, window_size: Optional[int]=…, future_values: Optional[torch.Tensor]=…, forecast_context_len: Optional[int]=…, return_forecast_on_context: bool=…, truncate_negative: bool=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> TimesFmOutputF…
- added:
export define_import_structure
- added:
class TimesformerConfig(PretrainedConfig) - added:
constructor TimesformerConfig(image_size=…, patch_size=…, num_channels=…, num_frames=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, attention_type=…, drop_path_rate=…, **kwargs) - added:
field TimesformerConfig.model_type
- added:
class TimesformerForVideoClassification(TimesformerPreTrainedModel) - added:
class TimesformerModel(TimesformerPreTrainedModel) - added:
class TimesformerPreTrainedModel(PreTrainedModel) - added:
constructor TimesformerForVideoClassification(config) - added:
constructor TimesformerModel(config) - added:
field TimesformerPreTrainedModel.base_model_prefix - added:
field TimesformerPreTrainedModel.config: TimesformerConfig - added:
field TimesformerPreTrainedModel.main_input_name - added:
field TimesformerPreTrainedModel.supports_gradient_checkpointing - added:
method TimesformerForVideoClassification.forward(pixel_values: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method TimesformerModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method TimesformerModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class TimmBackboneConfig(PretrainedConfig) - added:
constructor TimmBackboneConfig(backbone=…, num_channels=…, features_only=…, use_pretrained_backbone=…, out_indices=…, freeze_batch_norm_2d=…, **kwargs) - added:
field TimmBackboneConfig.model_type
- added:
class TimmBackbone(PreTrainedModel, BackboneMixin) - added:
constructor TimmBackbone(config, **kwargs) - added:
field TimmBackbone.config: TimmBackboneConfig - added:
field TimmBackbone.main_input_name - added:
field TimmBackbone.supports_gradient_checkpointing - added:
method TimmBackbone.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, **kwargs) -> Union[BackboneOutput, tuple[Tensor, ...]] - added:
method TimmBackbone.freeze_batch_norm_2d() - added:
method TimmBackbone.unfreeze_batch_norm_2d() - added:
static method TimmBackbone.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
- added:
export define_import_structure
- added:
class TimmWrapperConfig(PretrainedConfig) - added:
constructor TimmWrapperConfig(architecture: str=…, initializer_range: float=…, do_pooling: bool=…, model_args: Optional[dict[str, Any]]=…, **kwargs) - added:
field TimmWrapperConfig.model_type - added:
method TimmWrapperConfig.to_dict() -> dict[str, Any] - added:
static method TimmWrapperConfig.from_dict(config_dict: dict[str, Any], **kwargs)
- added:
class TimmWrapperImageProcessor(BaseImageProcessor) - added:
constructor TimmWrapperImageProcessor(pretrained_cfg: dict[str, Any], architecture: Optional[str]=…, **kwargs) - added:
field TimmWrapperImageProcessor.main_input_name - added:
method TimmWrapperImageProcessor.preprocess(images: ImageInput, return_tensors: Optional[Union[str, TensorType]]=…) -> BatchFeature - added:
method TimmWrapperImageProcessor.save_pretrained(*args, **kwargs) - added:
method TimmWrapperImageProcessor.to_dict() -> dict[str, Any] - added:
static method TimmWrapperImageProcessor.get_image_processor_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]]
- added:
class TimmWrapperForImageClassification(TimmWrapperPreTrainedModel) - added:
class TimmWrapperModel(TimmWrapperPreTrainedModel) - added:
class TimmWrapperPreTrainedModel(PreTrainedModel) - added:
constructor TimmWrapperForImageClassification(config: TimmWrapperConfig) - added:
constructor TimmWrapperModel(config: TimmWrapperConfig) - added:
constructor TimmWrapperPreTrainedModel(*args, **kwargs) - added:
field TimmWrapperPreTrainedModel.accepts_loss_kwargs - added:
field TimmWrapperPreTrainedModel.config: TimmWrapperConfig - added:
field TimmWrapperPreTrainedModel.main_input_name - added:
field TimmWrapperPreTrainedModel.model_tags - added:
method TimmWrapperForImageClassification.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[Union[bool, list[int]]]=…, return_dict: Optional[bool]=…, **kwargs) -> Union[ImageClassifierOutput, tuple[Tensor, ...]] - added:
method TimmWrapperModel.forward(pixel_values: torch.FloatTensor, output_attentions: Optional[bool]=…, output_hidden_states: Optional[Union[bool, list[int]]]=…, return_dict: Optional[bool]=…, do_pooling: Optional[bool]=…, **kwargs) -> Union[TimmWrapperModelOutput, tuple[Tensor, ...]] - added:
method TimmWrapperPreTrainedModel.load_state_dict(state_dict, *args, **kwargs) - added:
method TimmWrapperPreTrainedModel.post_init()
- added:
export define_import_structure
- added:
class TrOCRConfig(PretrainedConfig) - added:
constructor TrOCRConfig(vocab_size=…, d_model=…, decoder_layers=…, decoder_attention_heads=…, decoder_ffn_dim=…, activation_function=…, max_position_embeddings=…, dropout=…, attention_dropout=…, activation_dropout=…, decoder_start_token_id=…, init_std=…, decoder_layerdrop=…, use_cache=…, scale_embedding=…, use_learned_position_embeddings=…, layernorm_embedding=…, pad_token_id=…, bos_token_id=…, e… - added:
field TrOCRConfig.attribute_map - added:
field TrOCRConfig.keys_to_ignore_at_inference - added:
field TrOCRConfig.model_type
- added:
class TrOCRForCausalLM(TrOCRPreTrainedModel, GenerationMixin) - added:
class TrOCRPreTrainedModel(PreTrainedModel) - added:
constructor TrOCRForCausalLM(config) - added:
field TrOCRPreTrainedModel.base_model_prefix - added:
field TrOCRPreTrainedModel.config: TrOCRConfig - added:
field TrOCRPreTrainedModel.supports_gradient_checkpointing - added:
method TrOCRForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: O… - added:
method TrOCRForCausalLM.get_decoder() - added:
method TrOCRForCausalLM.get_input_embeddings() - added:
method TrOCRForCausalLM.get_output_embeddings() - added:
method TrOCRForCausalLM.set_decoder(decoder) - added:
method TrOCRForCausalLM.set_input_embeddings(value) - added:
method TrOCRForCausalLM.set_output_embeddings(new_embeddings)
- added:
class TrOCRProcessor(ProcessorMixin) - added:
constructor TrOCRProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field TrOCRProcessor.attributes - added:
field TrOCRProcessor.image_processor_class - added:
field TrOCRProcessor.tokenizer_class - added:
method TrOCRProcessor.as_target_processor() - added:
property TrOCRProcessor.feature_extractor - added:
property TrOCRProcessor.feature_extractor_class - added:
property TrOCRProcessor.model_input_names
- added:
export define_import_structure
- added:
class TvpConfig(PretrainedConfig) - added:
constructor TvpConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, distance_loss_weight=…, duration_loss_weight=…, visual_prompter_type=…, visual_prompter_apply=…, visual_prompt_size=…, max_img_size=…, num_frames=…, vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, max_grid_… - added:
field TvpConfig.model_type - added:
method TvpConfig.to_dict() - added:
property TvpConfig.sub_configs - added:
static method TvpConfig.from_backbone_config(backbone_config: PretrainedConfig, **kwargs)
- added:
class TvpImageProcessor(BaseImageProcessor) - added:
constructor TvpImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_pad: bool=…, pad_size: Optional[dict[str, int]]=…, constant_values: Union[float, Iterable[float]]=…, pad_mode: PaddingMode=…, do_normalize: bool=…, do_flip_chann… - added:
field TvpImageProcessor.model_input_names - added:
method TvpImageProcessor.pad_image(image: np.ndarray, pad_size: Optional[dict[str, int]]=…, constant_values: Union[float, Iterable[float]]=…, pad_mode: PaddingMode=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method TvpImageProcessor.preprocess(videos: Union[ImageInput, list[ImageInput], list[list[ImageInput]]], do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_pad: Optional[bool]=…, pad_size: Optional[dict[s… - added:
method TvpImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class TvpImageProcessorFast(BaseImageProcessorFast) - added:
constructor TvpImageProcessorFast(**kwargs: Unpack[TvpFastImageProcessorKwargs]) - added:
field TvpImageProcessorFast.constant_values - added:
field TvpImageProcessorFast.crop_size - added:
field TvpImageProcessorFast.default_to_square - added:
field TvpImageProcessorFast.do_center_crop - added:
field TvpImageProcessorFast.do_flip_channel_order - added:
field TvpImageProcessorFast.do_normalize - added:
field TvpImageProcessorFast.do_pad - added:
field TvpImageProcessorFast.do_rescale - added:
field TvpImageProcessorFast.do_resize - added:
field TvpImageProcessorFast.image_mean - added:
field TvpImageProcessorFast.image_std - added:
field TvpImageProcessorFast.pad_mode - added:
field TvpImageProcessorFast.pad_size - added:
field TvpImageProcessorFast.resample - added:
field TvpImageProcessorFast.rescale_factor - added:
field TvpImageProcessorFast.size - added:
field TvpImageProcessorFast.valid_kwargs - added:
method TvpImageProcessorFast.preprocess(videos: Union[ImageInput, list[ImageInput], list[list[ImageInput]]], **kwargs: Unpack[TvpFastImageProcessorKwargs]) -> BatchFeature - added:
method TvpImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, antialias: bool=…, **kwargs) -> torch.Tensor
- added:
class TvpForVideoGrounding(TvpPreTrainedModel) - added:
class TvpModel(TvpPreTrainedModel) - added:
class TvpPreTrainedModel(PreTrainedModel) - added:
constructor TvpForVideoGrounding(config) - added:
constructor TvpModel(config) - added:
field TvpPreTrainedModel.base_model_prefix - added:
field TvpPreTrainedModel.config: TvpConfig - added:
field TvpPreTrainedModel.supports_gradient_checkpointing - added:
method TvpForVideoGrounding.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, labels: Optional[tuple[torch.Tensor]]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, interpolate_pos_encoding: bool=…) - added:
method TvpModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, interpolate_pos_encoding: bool=…) - added:
method TvpModel.get_input_embeddings() - added:
method TvpModel.set_input_embeddings(value)
- added:
class TvpProcessor(ProcessorMixin) - added:
constructor TvpProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field TvpProcessor.attributes - added:
field TvpProcessor.image_processor_class - added:
field TvpProcessor.tokenizer_class - added:
method TvpProcessor.post_process_video_grounding(logits, video_durations)
- added:
export define_import_structure
- added:
class UdopConfig(PretrainedConfig) - added:
constructor UdopConfig(vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, relative_bias_args=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, use_cache=…, pad_token_id=…, eos_token_id=…, max_2d_position_embeddings=…, image_size=…, pat… - added:
field UdopConfig.attribute_map - added:
field UdopConfig.keys_to_ignore_at_inference - added:
field UdopConfig.model_type
- added:
class UdopEncoderModel(UdopPreTrainedModel) - added:
class UdopForConditionalGeneration(UdopPreTrainedModel, GenerationMixin) - added:
class UdopModel(UdopPreTrainedModel) - added:
class UdopPreTrainedModel(PreTrainedModel) - added:
constructor UdopEncoderModel(config: UdopConfig) - added:
constructor UdopForConditionalGeneration(config) - added:
constructor UdopModel(config) - added:
field UdopPreTrainedModel.base_model_prefix - added:
field UdopPreTrainedModel.config: UdopConfig - added:
field UdopPreTrainedModel.supports_gradient_checkpointing - added:
method UdopEncoderModel.forward(input_ids: Optional[Tensor]=…, bbox: Optional[dict[str, Any]]=…, attention_mask: Optional[Tensor]=…, pixel_values: Optional[Tensor]=…, visual_bbox: Optional[dict[str, Any]]=…, head_mask: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[to… - added:
method UdopEncoderModel.get_encoder() - added:
method UdopEncoderModel.get_input_embeddings() - added:
method UdopEncoderModel.set_input_embeddings(new_embeddings) - added:
method UdopForConditionalGeneration.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, bbox: Optional[dict[str, Any]]=…, pixel_values: Optional[Tensor]=…, visual_bbox: Optional[dict[str, Any]]=…, decoder_input_ids: Optional[Tensor]=…, decoder_attention_mask: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, encoder_outputs: Optional[Tensor]=…, past_key_values: Optiona… - added:
method UdopForConditionalGeneration.get_encoder() - added:
method UdopForConditionalGeneration.get_input_embeddings() - added:
method UdopForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method UdopModel.forward(input_ids: Optional[Tensor]=…, attention_mask: Optional[Tensor]=…, bbox: Optional[dict[str, Any]]=…, pixel_values: Optional[Tensor]=…, visual_bbox: Optional[dict[str, Any]]=…, decoder_input_ids: Optional[Tensor]=…, decoder_attention_mask: Optional[Tensor]=…, inputs_embeds: Optional[Tensor]=…, encoder_outputs: Optional[Tensor]=…, past_key_values: Optional[Cache]=…, head_ma… - added:
method UdopModel.get_encoder() - added:
method UdopModel.get_input_embeddings() - added:
method UdopModel.set_input_embeddings(new_embeddings)
- added:
class UdopProcessor(ProcessorMixin) - added:
constructor UdopProcessor(image_processor, tokenizer) - added:
field UdopProcessor.attributes - added:
field UdopProcessor.image_processor_class - added:
field UdopProcessor.tokenizer_class - added:
method UdopProcessor.get_overflowing_images(images, overflow_to_sample_mapping) - added:
property UdopProcessor.model_input_names
- added:
class UdopTokenizer(PreTrainedTokenizer) - added:
constructor UdopTokenizer(vocab_file, eos_token=…, unk_token=…, sep_token=…, pad_token=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, legacy=…, add_prefix_space=…, **kwargs) -> None - added:
field UdopTokenizer.model_input_names - added:
field UdopTokenizer.vocab_files_names - added:
method UdopTokenizer.batch_encode_plus_boxes(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[… - added:
method UdopTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method UdopTokenizer.call_boxes(text: Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]], text_pair: Optional[Union[PreTokenizedInput, list[PreTokenizedInput]]]=…, boxes: Optional[Union[list[list[int]], list[list[list[int]]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncati… - added:
method UdopTokenizer.convert_tokens_to_string(tokens) - added:
method UdopTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method UdopTokenizer.encode_boxes(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, s… - added:
method UdopTokenizer.encode_plus_boxes(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, is_split_into_words: bool=…… - added:
method UdopTokenizer.get_sentinel_token_ids() - added:
method UdopTokenizer.get_sentinel_tokens() - added:
method UdopTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method UdopTokenizer.get_vocab() - added:
method UdopTokenizer.prepare_for_model_boxes(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optiona… - added:
method UdopTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method UdopTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
method UdopTokenizer.truncate_sequences(ids: list[int], token_boxes: list[list[int]], pair_ids: Optional[list[int]]=…, pair_token_boxes: Optional[list[list[int]]]=…, labels: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[list[int], list[int], list[int]] - added:
property UdopTokenizer.vocab_size
- added:
class UdopTokenizerFast(PreTrainedTokenizerFast) - added:
constructor UdopTokenizerFast(vocab_file=…, tokenizer_file=…, eos_token=…, sep_token=…, unk_token=…, pad_token=…, sep_token_box=…, pad_token_box=…, pad_token_label=…, only_label_first_subword=…, additional_special_tokens=…, **kwargs) - added:
field UdopTokenizerFast.model_input_names - added:
field UdopTokenizerFast.slow_tokenizer_class - added:
field UdopTokenizerFast.vocab_files_names - added:
method UdopTokenizerFast.batch_encode_plus_boxes(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput]], is_pair: Optional[bool]=…, boxes: Optional[list[list[list[int]]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optio… - added:
method UdopTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method UdopTokenizerFast.call_boxes(text: Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]], text_pair: Optional[Union[PreTokenizedInput, list[PreTokenizedInput]]]=…, boxes: Optional[Union[list[list[int]], list[list[list[int]]]]]=…, word_labels: Optional[Union[list[int], list[list[int]]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, trun… - added:
method UdopTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method UdopTokenizerFast.encode_boxes(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=… - added:
method UdopTokenizerFast.encode_plus_boxes(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[PreTokenizedInput]=…, boxes: Optional[list[list[int]]]=…, word_labels: Optional[list[list[int]]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, is_split_into_words: bo… - added:
method UdopTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method UdopTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str]
- added:
export define_import_structure
- added:
class UMT5Config(PretrainedConfig) - added:
class UMT5OnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor UMT5Config(vocab_size=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_decoder_layers=…, num_heads=…, relative_attention_num_buckets=…, relative_attention_max_distance=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, feed_forward_proj=…, is_encoder_decoder=…, use_cache=…, tokenizer_class=…, tie_word_embeddings=…, pad_token_id=…, eos_token_id=…, decoder_start_token_id=…, c… - added:
field UMT5Config.attribute_map - added:
field UMT5Config.keys_to_ignore_at_inference - added:
field UMT5Config.model_type - added:
property UMT5OnnxConfig.atol_for_validation -> float - added:
property UMT5OnnxConfig.default_onnx_opset -> int - added:
property UMT5OnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class UMT5EncoderModel(UMT5PreTrainedModel) - added:
class UMT5ForConditionalGeneration(UMT5PreTrainedModel, GenerationMixin) - added:
class UMT5ForQuestionAnswering(UMT5PreTrainedModel) - added:
class UMT5ForSequenceClassification(UMT5PreTrainedModel) - added:
class UMT5ForTokenClassification(UMT5PreTrainedModel) - added:
class UMT5Model(UMT5PreTrainedModel) - added:
class UMT5PreTrainedModel(PreTrainedModel) - added:
constructor UMT5EncoderModel(config) - added:
constructor UMT5ForConditionalGeneration(config) - added:
constructor UMT5ForQuestionAnswering(config) - added:
constructor UMT5ForSequenceClassification(config: UMT5Config) - added:
constructor UMT5ForTokenClassification(config: UMT5Config) - added:
constructor UMT5Model(config) - added:
field UMT5EncoderModel.model_type - added:
field UMT5ForConditionalGeneration.model_type - added:
field UMT5Model.config: UMT5Config - added:
field UMT5Model.model_type - added:
field UMT5PreTrainedModel.base_model_prefix - added:
field UMT5PreTrainedModel.config: UMT5Config - added:
field UMT5PreTrainedModel.supports_gradient_checkpointing - added:
method UMT5EncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], BaseModelOutput] - added:
method UMT5EncoderModel.get_encoder() - added:
method UMT5EncoderModel.get_input_embeddings() - added:
method UMT5EncoderModel.set_input_embeddings(new_embeddings) - added:
method UMT5ForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optiona… - added:
method UMT5ForConditionalGeneration.get_encoder() - added:
method UMT5ForConditionalGeneration.get_input_embeddings() - added:
method UMT5ForConditionalGeneration.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method UMT5ForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method UMT5ForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tu… - added:
method UMT5ForQuestionAnswering.get_encoder() - added:
method UMT5ForQuestionAnswering.get_input_embeddings() - added:
method UMT5ForQuestionAnswering.set_input_embeddings(new_embeddings) - added:
method UMT5ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[list[torch.F… - added:
method UMT5ForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], TokenClassifierOutput] - added:
method UMT5Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.FloatTensor]=…, decoder_head_mask: Optional[torch.FloatTensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch… - added:
method UMT5Model.get_encoder() - added:
method UMT5Model.get_input_embeddings() - added:
method UMT5Model.set_input_embeddings(new_embeddings) - added:
property UMT5PreTrainedModel.dummy_inputs
- added:
export define_import_structure
- added:
class UniSpeechConfig(PretrainedConfig) - added:
constructor UniSpeechConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, feat_quantizer_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel… - added:
field UniSpeechConfig.model_type - added:
property UniSpeechConfig.inputs_to_logits_ratio
- added:
class UniSpeechForCTC(UniSpeechPreTrainedModel) - added:
class UniSpeechForPreTraining(UniSpeechPreTrainedModel) - added:
class UniSpeechForSequenceClassification(UniSpeechPreTrainedModel) - added:
class UniSpeechModel(UniSpeechPreTrainedModel) - added:
class UniSpeechPreTrainedModel(PreTrainedModel) - added:
constructor UniSpeechForCTC(config, target_lang: Optional[str]=…) - added:
constructor UniSpeechForPreTraining(config: UniSpeechConfig) - added:
constructor UniSpeechForSequenceClassification(config) - added:
constructor UniSpeechModel(config: UniSpeechConfig) - added:
field UniSpeechPreTrainedModel.base_model_prefix - added:
field UniSpeechPreTrainedModel.config: UniSpeechConfig - added:
field UniSpeechPreTrainedModel.main_input_name - added:
field UniSpeechPreTrainedModel.supports_gradient_checkpointing - added:
method UniSpeechForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method UniSpeechForCTC.freeze_base_model() - added:
method UniSpeechForCTC.freeze_feature_encoder() - added:
method UniSpeechForCTC.freeze_feature_extractor() - added:
method UniSpeechForCTC.tie_weights() - added:
method UniSpeechForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechForPreTrainingOutput] - added:
method UniSpeechForPreTraining.freeze_feature_encoder() - added:
method UniSpeechForPreTraining.freeze_feature_extractor() - added:
method UniSpeechForPreTraining.set_gumbel_temperature(temperature: int) - added:
method UniSpeechForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method UniSpeechForSequenceClassification.freeze_base_model() - added:
method UniSpeechForSequenceClassification.freeze_feature_encoder() - added:
method UniSpeechForSequenceClassification.freeze_feature_extractor() - added:
method UniSpeechModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechBaseModelOutput] - added:
static method UniSpeechForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
class UniSpeechForCTC(Wav2Vec2ForCTC) - added:
class UniSpeechForPreTraining(UniSpeechPreTrainedModel) - added:
class UniSpeechForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class UniSpeechModel(UniSpeechPreTrainedModel, Wav2Vec2Model) - added:
class UniSpeechPreTrainedModel(PreTrainedModel) - added:
constructor UniSpeechForPreTraining(config: UniSpeechConfig) - added:
constructor UniSpeechModel(config: UniSpeechConfig) - added:
field UniSpeechPreTrainedModel.base_model_prefix - added:
field UniSpeechPreTrainedModel.config: UniSpeechConfig - added:
field UniSpeechPreTrainedModel.main_input_name - added:
field UniSpeechPreTrainedModel.supports_gradient_checkpointing - added:
method UniSpeechForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechForPreTrainingOutput] - added:
method UniSpeechForPreTraining.freeze_feature_encoder() - added:
method UniSpeechForPreTraining.freeze_feature_extractor() - added:
method UniSpeechForPreTraining.set_gumbel_temperature(temperature: int) - added:
method UniSpeechModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechBaseModelOutput] - added:
method UniSpeechModel.freeze_feature_encoder() - added:
method UniSpeechModel.freeze_feature_extractor() - added:
static method UniSpeechForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
export define_import_structure
- added:
class UniSpeechSatConfig(PretrainedConfig) - added:
constructor UniSpeechSatConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, feat_quantizer_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_ker… - added:
field UniSpeechSatConfig.model_type - added:
property UniSpeechSatConfig.inputs_to_logits_ratio
- added:
class UniSpeechSatForAudioFrameClassification(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatForCTC(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatForPreTraining(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatForSequenceClassification(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatForXVector(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatModel(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatPreTrainedModel(PreTrainedModel) - added:
constructor UniSpeechSatForAudioFrameClassification(config) - added:
constructor UniSpeechSatForCTC(config, target_lang: Optional[str]=…) - added:
constructor UniSpeechSatForPreTraining(config: UniSpeechSatConfig) - added:
constructor UniSpeechSatForSequenceClassification(config) - added:
constructor UniSpeechSatForXVector(config) - added:
constructor UniSpeechSatModel(config: UniSpeechSatConfig) - added:
field UniSpeechSatPreTrainedModel.base_model_prefix - added:
field UniSpeechSatPreTrainedModel.config: UniSpeechSatConfig - added:
field UniSpeechSatPreTrainedModel.main_input_name - added:
field UniSpeechSatPreTrainedModel.supports_gradient_checkpointing - added:
method UniSpeechSatForAudioFrameClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method UniSpeechSatForAudioFrameClassification.freeze_base_model() - added:
method UniSpeechSatForAudioFrameClassification.freeze_feature_encoder() - added:
method UniSpeechSatForAudioFrameClassification.freeze_feature_extractor() - added:
method UniSpeechSatForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method UniSpeechSatForCTC.freeze_base_model() - added:
method UniSpeechSatForCTC.freeze_feature_encoder() - added:
method UniSpeechSatForCTC.freeze_feature_extractor() - added:
method UniSpeechSatForCTC.tie_weights() - added:
method UniSpeechSatForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechSatForPreTrainingOutput] - added:
method UniSpeechSatForPreTraining.freeze_feature_encoder() - added:
method UniSpeechSatForPreTraining.freeze_feature_extractor() - added:
method UniSpeechSatForPreTraining.set_gumbel_temperature(temperature: int) - added:
method UniSpeechSatForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method UniSpeechSatForSequenceClassification.freeze_base_model() - added:
method UniSpeechSatForSequenceClassification.freeze_feature_encoder() - added:
method UniSpeechSatForSequenceClassification.freeze_feature_extractor() - added:
method UniSpeechSatForXVector.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method UniSpeechSatForXVector.freeze_base_model() - added:
method UniSpeechSatForXVector.freeze_feature_encoder() - added:
method UniSpeechSatForXVector.freeze_feature_extractor() - added:
method UniSpeechSatModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechSatBaseModelOutput] - added:
static method UniSpeechSatForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
class UniSpeechSatForAudioFrameClassification(Wav2Vec2ForAudioFrameClassification) - added:
class UniSpeechSatForCTC(Wav2Vec2ForCTC) - added:
class UniSpeechSatForPreTraining(UniSpeechSatPreTrainedModel) - added:
class UniSpeechSatForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class UniSpeechSatForXVector(Wav2Vec2ForXVector) - added:
class UniSpeechSatModel(UniSpeechSatPreTrainedModel, Wav2Vec2Model) - added:
class UniSpeechSatPreTrainedModel(PreTrainedModel) - added:
constructor UniSpeechSatForPreTraining(config: UniSpeechSatConfig) - added:
constructor UniSpeechSatModel(config: UniSpeechSatConfig) - added:
field UniSpeechSatPreTrainedModel.base_model_prefix - added:
field UniSpeechSatPreTrainedModel.config: UniSpeechSatConfig - added:
field UniSpeechSatPreTrainedModel.main_input_name - added:
field UniSpeechSatPreTrainedModel.supports_gradient_checkpointing - added:
method UniSpeechSatForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechSatForPreTrainingOutput] - added:
method UniSpeechSatForPreTraining.freeze_feature_encoder() - added:
method UniSpeechSatForPreTraining.freeze_feature_extractor() - added:
method UniSpeechSatForPreTraining.set_gumbel_temperature(temperature: int) - added:
method UniSpeechSatModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, UniSpeechSatBaseModelOutput] - added:
method UniSpeechSatModel.freeze_feature_encoder() - added:
method UniSpeechSatModel.freeze_feature_extractor() - added:
static method UniSpeechSatForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
export define_import_structure
- added:
class UnivNetConfig(PretrainedConfig) - added:
constructor UnivNetConfig(model_in_channels=…, model_hidden_channels=…, num_mel_bins=…, resblock_kernel_sizes=…, resblock_stride_sizes=…, resblock_dilation_sizes=…, kernel_predictor_num_blocks=…, kernel_predictor_hidden_channels=…, kernel_predictor_conv_size=…, kernel_predictor_dropout=…, initializer_range=…, leaky_relu_slope=…, **kwargs) - added:
field UnivNetConfig.model_type
- added:
class UnivNetFeatureExtractor(SequenceFeatureExtractor) - added:
constructor UnivNetFeatureExtractor(feature_size: int=…, sampling_rate: int=…, padding_value: float=…, do_normalize: bool=…, num_mel_bins: int=…, hop_length: int=…, win_length: int=…, win_function: str=…, filter_length: Optional[int]=…, max_length_s: int=…, fmin: float=…, fmax: Optional[float]=…, mel_floor: float=…, center: bool=…, compression_factor: float=…, compression_clip_val: float=…, norma… - added:
field UnivNetFeatureExtractor.model_input_names - added:
method UnivNetFeatureExtractor.batch_decode(waveforms, waveform_lengths=…) -> list[np.ndarray] - added:
method UnivNetFeatureExtractor.denormalize(spectrogram) - added:
method UnivNetFeatureExtractor.generate_noise(noise_length: int, generator: Optional[np.random.Generator]=…) -> np.ndarray - added:
method UnivNetFeatureExtractor.mel_spectrogram(waveform: np.ndarray) -> np.ndarray - added:
method UnivNetFeatureExtractor.normalize(spectrogram) - added:
method UnivNetFeatureExtractor.to_dict() -> dict[str, Any]
- added:
class UnivNetModel(PreTrainedModel) - added:
constructor UnivNetModel(config: UnivNetConfig) - added:
field UnivNetModel.config: UnivNetConfig - added:
field UnivNetModel.main_input_name - added:
method UnivNetModel.apply_weight_norm() - added:
method UnivNetModel.forward(input_features: torch.FloatTensor, noise_sequence: Optional[torch.FloatTensor]=…, padding_mask: Optional[torch.FloatTensor]=…, generator: Optional[torch.Generator]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.FloatTensor], UnivNetModelOutput] - added:
method UnivNetModel.remove_weight_norm()
- added:
export define_import_structure
- added:
class UperNetConfig(PretrainedConfig) - added:
constructor UperNetConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, hidden_size=…, initializer_range=…, pool_scales=…, use_auxiliary_head=…, auxiliary_loss_weight=…, auxiliary_in_channels=…, auxiliary_channels=…, auxiliary_num_convs=…, auxiliary_concat_input=…, loss_ignore_index=…, **kwargs) - added:
field UperNetConfig.model_type - added:
property UperNetConfig.sub_configs
- added:
class UperNetForSemanticSegmentation(UperNetPreTrainedModel) - added:
class UperNetPreTrainedModel(PreTrainedModel) - added:
constructor UperNetForSemanticSegmentation(config) - added:
field UperNetPreTrainedModel.config: UperNetConfig - added:
field UperNetPreTrainedModel.main_input_name - added:
method UperNetForSemanticSegmentation.forward(pixel_values: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…) -> Union[tuple, SemanticSegmenterOutput]
- added:
export define_import_structure
- added:
class VaultGemmaConfig(PretrainedConfig) - added:
constructor VaultGemmaConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, num_key_value_heads=…, head_dim=…, hidden_activation=…, max_position_embeddings=…, initializer_range=…, rms_norm_eps=…, use_cache=…, pad_token_id=…, eos_token_id=…, bos_token_id=…, tie_word_embeddings=…, rope_theta=…, attention_bias=…, attention_dropout=…, query_pre_attn_scal… - added:
field VaultGemmaConfig.base_model_pp_plan - added:
field VaultGemmaConfig.base_model_tp_plan - added:
field VaultGemmaConfig.keys_to_ignore_at_inference - added:
field VaultGemmaConfig.model_type
- added:
class VaultGemmaForCausalLM(VaultGemmaPreTrainedModel, GenerationMixin) - added:
class VaultGemmaModel(VaultGemmaPreTrainedModel) - added:
class VaultGemmaPreTrainedModel(PreTrainedModel) - added:
constructor VaultGemmaForCausalLM(config) - added:
constructor VaultGemmaModel(config: VaultGemmaConfig) - added:
field VaultGemmaPreTrainedModel.base_model_prefix - added:
field VaultGemmaPreTrainedModel.config: VaultGemmaConfig - added:
field VaultGemmaPreTrainedModel.supports_gradient_checkpointing - added:
method VaultGemmaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_pos… - added:
method VaultGemmaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, cache_position: Optional[torch.LongTensor]=…, **kwarg…
- added:
class VaultGemmaConfig(Gemma2Config) - added:
class VaultGemmaForCausalLM(Gemma2ForCausalLM)
- added:
export define_import_structure
- added:
class VideoLlavaConfig(PretrainedConfig) - added:
constructor VideoLlavaConfig(vision_config=…, text_config=…, image_token_index=…, video_token_index=…, projector_hidden_act=…, vision_feature_select_strategy=…, vision_feature_layer=…, image_seq_length=…, video_seq_length=…, multimodal_projector_bias=…, **kwargs) - added:
field VideoLlavaConfig.attribute_map - added:
field VideoLlavaConfig.model_type - added:
field VideoLlavaConfig.sub_configs
- added:
class VideoLlavaImageProcessor(BaseImageProcessor) - added:
constructor VideoLlavaImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: bool=…, **kwargs)… - added:
field VideoLlavaImageProcessor.model_input_names - added:
method VideoLlavaImageProcessor.preprocess(images: Optional[list[ImageInput]]=…, videos: Optional[list[VideoInput]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[int]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Opti… - added:
method VideoLlavaImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class VideoLlavaForConditionalGeneration(VideoLlavaPreTrainedModel, GenerationMixin) - added:
class VideoLlavaModel(VideoLlavaPreTrainedModel) - added:
class VideoLlavaPreTrainedModel(PreTrainedModel) - added:
constructor VideoLlavaForConditionalGeneration(config: VideoLlavaConfig) - added:
constructor VideoLlavaModel(config: VideoLlavaConfig) - added:
field VideoLlavaPreTrainedModel.base_model_prefix - added:
field VideoLlavaPreTrainedModel.config: VideoLlavaConfig - added:
field VideoLlavaPreTrainedModel.supports_gradient_checkpointing - added:
method VideoLlavaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values_images: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union… - added:
method VideoLlavaForConditionalGeneration.get_decoder() - added:
method VideoLlavaForConditionalGeneration.get_image_features(pixel_values_images: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method VideoLlavaForConditionalGeneration.get_input_embeddings() - added:
method VideoLlavaForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method VideoLlavaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values_images=…, pixel_values_videos=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method VideoLlavaForConditionalGeneration.set_decoder(decoder) - added:
method VideoLlavaForConditionalGeneration.set_input_embeddings(value) - added:
method VideoLlavaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values_images: Optional[torch.FloatTensor]=…, pixel_values_videos: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layer: Optional[Union[int, list[int]]]=…… - added:
method VideoLlavaModel.get_decoder() - added:
method VideoLlavaModel.get_image_features(pixel_values_images: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…, vision_feature_select_strategy: Optional[str]=…) - added:
method VideoLlavaModel.get_input_embeddings() - added:
method VideoLlavaModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: Optional[torch.FloatTensor]=…, video_features: Optional[torch.FloatTensor]=…) - added:
method VideoLlavaModel.get_video_features(pixel_values_videos: torch.FloatTensor, vision_feature_layer: Optional[Union[int, list[int]]]=…) - added:
method VideoLlavaModel.set_decoder(decoder) - added:
method VideoLlavaModel.set_input_embeddings(value) - added:
property VideoLlavaForConditionalGeneration.image_tower - added:
property VideoLlavaForConditionalGeneration.language_model - added:
property VideoLlavaForConditionalGeneration.multi_modal_projector - added:
property VideoLlavaForConditionalGeneration.video_tower
- added:
class VideoLlavaProcessor(ProcessorMixin) - added:
constructor VideoLlavaProcessor(image_processor=…, video_processor=…, tokenizer=…, patch_size=…, vision_feature_select_strategy=…, image_token=…, video_token=…, chat_template=…, num_additional_image_tokens=…, **kwargs) - added:
field VideoLlavaProcessor.attributes - added:
field VideoLlavaProcessor.image_processor_class - added:
field VideoLlavaProcessor.tokenizer_class - added:
field VideoLlavaProcessor.video_processor_class
- added:
class VideoLlavaVideoProcessor(BaseVideoProcessor) - added:
constructor VideoLlavaVideoProcessor(**kwargs: Unpack[VideoLlavaFastVideoProcessorInitKwargs]) - added:
field VideoLlavaVideoProcessor.crop_size - added:
field VideoLlavaVideoProcessor.default_to_square - added:
field VideoLlavaVideoProcessor.do_center_crop - added:
field VideoLlavaVideoProcessor.do_convert_rgb - added:
field VideoLlavaVideoProcessor.do_normalize - added:
field VideoLlavaVideoProcessor.do_rescale - added:
field VideoLlavaVideoProcessor.do_resize - added:
field VideoLlavaVideoProcessor.do_sample_frames - added:
field VideoLlavaVideoProcessor.image_mean - added:
field VideoLlavaVideoProcessor.image_std - added:
field VideoLlavaVideoProcessor.model_input_names - added:
field VideoLlavaVideoProcessor.resample - added:
field VideoLlavaVideoProcessor.size - added:
field VideoLlavaVideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class VideoMAEConfig(PretrainedConfig) - added:
constructor VideoMAEConfig(image_size=…, patch_size=…, num_channels=…, num_frames=…, tubelet_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, use_mean_pooling=…, decoder_num_attention_heads=…, decoder_hidden_size=…, decoder_num_hidden_layer… - added:
field VideoMAEConfig.model_type
- added:
class VideoMAEFeatureExtractor(VideoMAEImageProcessor) - added:
constructor VideoMAEFeatureExtractor(*args, **kwargs) -> None
- added:
class VideoMAEImageProcessor(BaseImageProcessor) - added:
constructor VideoMAEImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field VideoMAEImageProcessor.model_input_names - added:
method VideoMAEImageProcessor.preprocess(videos: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: O… - added:
method VideoMAEImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class VideoMAEForPreTraining(VideoMAEPreTrainedModel) - added:
class VideoMAEForVideoClassification(VideoMAEPreTrainedModel) - added:
class VideoMAEModel(VideoMAEPreTrainedModel) - added:
class VideoMAEPreTrainedModel(PreTrainedModel) - added:
constructor VideoMAEForPreTraining(config) - added:
constructor VideoMAEForVideoClassification(config) - added:
constructor VideoMAEModel(config) - added:
field VideoMAEPreTrainedModel.base_model_prefix - added:
field VideoMAEPreTrainedModel.config: VideoMAEConfig - added:
field VideoMAEPreTrainedModel.main_input_name - added:
field VideoMAEPreTrainedModel.supports_gradient_checkpointing - added:
method VideoMAEForPreTraining.forward(pixel_values: torch.FloatTensor, bool_masked_pos: torch.BoolTensor, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> VideoMAEForPreTrainingOutput - added:
method VideoMAEForVideoClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method VideoMAEModel.forward(pixel_values: torch.FloatTensor, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method VideoMAEModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class ViltConfig(PretrainedConfig) - added:
constructor ViltConfig(vocab_size=…, type_vocab_size=…, modality_type_vocab_size=…, max_position_embeddings=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, max_image_length=…, tie_word_embeddings=…, n… - added:
field ViltConfig.model_type
- added:
class ViltFeatureExtractor(ViltImageProcessor) - added:
constructor ViltFeatureExtractor(*args, **kwargs) -> None
- added:
class ViltImageProcessor(BaseImageProcessor) - added:
constructor ViltImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, size_divisor: int=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, **kwargs) -> None - added:
field ViltImageProcessor.model_input_names - added:
method ViltImageProcessor.pad(images: list[np.ndarray], constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> BatchFeature - added:
method ViltImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, size_divisor: Optional[int]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad:… - added:
method ViltImageProcessor.resize(image: np.ndarray, size: dict[str, int], size_divisor: int=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
static method ViltImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class ViltImageProcessorFast(BaseImageProcessorFast) - added:
field ViltImageProcessorFast.default_to_square - added:
field ViltImageProcessorFast.do_normalize - added:
field ViltImageProcessorFast.do_pad - added:
field ViltImageProcessorFast.do_rescale - added:
field ViltImageProcessorFast.do_resize - added:
field ViltImageProcessorFast.image_mean - added:
field ViltImageProcessorFast.image_std - added:
field ViltImageProcessorFast.model_input_names - added:
field ViltImageProcessorFast.resample - added:
field ViltImageProcessorFast.size - added:
field ViltImageProcessorFast.size_divisor - added:
field ViltImageProcessorFast.valid_kwargs - added:
method ViltImageProcessorFast.resize(images: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, size_divisor: Optional[int]=…) -> torch.Tensor
- added:
class ViltForImageAndTextRetrieval(ViltPreTrainedModel) - added:
class ViltForImagesAndTextClassification(ViltPreTrainedModel) - added:
class ViltForMaskedLM(ViltPreTrainedModel) - added:
class ViltForQuestionAnswering(ViltPreTrainedModel) - added:
class ViltForTokenClassification(ViltPreTrainedModel) - added:
class ViltLayer(GradientCheckpointingLayer) - added:
class ViltModel(ViltPreTrainedModel) - added:
class ViltPreTrainedModel(PreTrainedModel) - added:
constructor ViltForImageAndTextRetrieval(config) - added:
constructor ViltForImagesAndTextClassification(config) - added:
constructor ViltForMaskedLM(config) - added:
constructor ViltForQuestionAnswering(config) - added:
constructor ViltForTokenClassification(config) - added:
constructor ViltLayer(config) - added:
constructor ViltModel(config, add_pooling_layer=…) - added:
field ViltPreTrainedModel.base_model_prefix - added:
field ViltPreTrainedModel.config: ViltConfig - added:
field ViltPreTrainedModel.supports_gradient_checkpointing - added:
method ViltForImageAndTextRetrieval.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, l… - added:
method ViltForImagesAndTextClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor… - added:
method ViltForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, labels: Option… - added:
method ViltForMaskedLM.get_output_embeddings() - added:
method ViltForMaskedLM.set_output_embeddings(new_embeddings) - added:
method ViltForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, label… - added:
method ViltForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, lab… - added:
method ViltLayer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…) - added:
method ViltModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, pixel_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, image_embeds: Optional[torch.FloatTensor]=…, image_token_type_idx… - added:
method ViltModel.get_input_embeddings() - added:
method ViltModel.set_input_embeddings(value)
- added:
class ViltProcessor(ProcessorMixin) - added:
constructor ViltProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field ViltProcessor.attributes - added:
field ViltProcessor.image_processor_class - added:
field ViltProcessor.tokenizer_class - added:
field ViltProcessor.valid_processor_kwargs - added:
property ViltProcessor.feature_extractor - added:
property ViltProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class VipLlavaConfig(PretrainedConfig) - added:
constructor VipLlavaConfig(vision_config=…, text_config=…, image_token_index=…, projector_hidden_act=…, projector_layernorm_eps=…, vision_feature_layers=…, image_seq_length=…, **kwargs) - added:
field VipLlavaConfig.attribute_map - added:
field VipLlavaConfig.model_type - added:
field VipLlavaConfig.sub_configs
- added:
class VipLlavaForConditionalGeneration(VipLlavaPreTrainedModel, GenerationMixin) - added:
class VipLlavaModel(VipLlavaPreTrainedModel) - added:
class VipLlavaPreTrainedModel(PreTrainedModel) - added:
constructor VipLlavaForConditionalGeneration(config: VipLlavaConfig) - added:
constructor VipLlavaModel(config: VipLlavaConfig) - added:
field VipLlavaPreTrainedModel.base_model_prefix - added:
field VipLlavaPreTrainedModel.config: VipLlavaConfig - added:
field VipLlavaPreTrainedModel.supports_gradient_checkpointing - added:
method VipLlavaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layers: Optional[Union[int, list[int]]]=…, labels: Optional[torch.LongTensor]=…, u… - added:
method VipLlavaForConditionalGeneration.get_decoder() - added:
method VipLlavaForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layers: Optional[Union[int, list[int]]]=…) - added:
method VipLlavaForConditionalGeneration.get_input_embeddings() - added:
method VipLlavaForConditionalGeneration.get_output_embeddings() -> nn.Module - added:
method VipLlavaForConditionalGeneration.prepare_inputs_for_generation(input_ids, past_key_values=…, inputs_embeds=…, pixel_values=…, attention_mask=…, cache_position=…, logits_to_keep=…, **kwargs) - added:
method VipLlavaForConditionalGeneration.set_decoder(decoder) - added:
method VipLlavaForConditionalGeneration.set_input_embeddings(value) - added:
method VipLlavaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layers: Optional[Union[int, list[int]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[b… - added:
method VipLlavaModel.get_decoder() - added:
method VipLlavaModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layers: Optional[Union[int, list[int]]]=…) - added:
method VipLlavaModel.get_input_embeddings() - added:
method VipLlavaModel.get_placeholder_mask(input_ids: torch.LongTensor, inputs_embeds: torch.FloatTensor, image_features: torch.FloatTensor) - added:
method VipLlavaModel.set_decoder(decoder) - added:
method VipLlavaModel.set_input_embeddings(value) - added:
property VipLlavaForConditionalGeneration.language_model - added:
property VipLlavaForConditionalGeneration.multi_modal_projector - added:
property VipLlavaForConditionalGeneration.vision_tower
- added:
class VipLlavaForConditionalGeneration(LlavaForConditionalGeneration) - added:
class VipLlavaModel(LlavaModel) - added:
class VipLlavaPreTrainedModel(LlavaPreTrainedModel) - added:
method VipLlavaForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layers: Optional[Union[int, list[int]]]=…, labels: Optional[torch.LongTensor]=…, u… - added:
method VipLlavaForConditionalGeneration.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layers: Optional[Union[int, list[int]]]=…) - added:
method VipLlavaModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, vision_feature_layers: Optional[Union[int, list[int]]]=…, use_cache: Optional[bool]=…, output_attentions: Optional[b… - added:
method VipLlavaModel.get_image_features(pixel_values: torch.FloatTensor, vision_feature_layers: Optional[Union[int, list[int]]]=…)
- added:
export define_import_structure
- added:
class VisionEncoderDecoderConfig(PretrainedConfig) - added:
class VisionEncoderDecoderOnnxConfig(OnnxConfig) - added:
constructor VisionEncoderDecoderConfig(**kwargs) - added:
field VisionEncoderDecoderConfig.has_no_defaults_at_init - added:
field VisionEncoderDecoderConfig.model_type - added:
field VisionEncoderDecoderConfig.sub_configs - added:
method VisionEncoderDecoderOnnxConfig.get_decoder_config(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, feature: str=…) -> OnnxConfig - added:
method VisionEncoderDecoderOnnxConfig.get_encoder_config(encoder_config: PretrainedConfig) -> OnnxConfig - added:
property VisionEncoderDecoderOnnxConfig.inputs -> None - added:
static method VisionEncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs) -> PretrainedConfig
- added:
class FlaxVisionEncoderDecoderModel(FlaxPreTrainedModel) - added:
constructor FlaxVisionEncoderDecoderModel(config: VisionEncoderDecoderConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxVisionEncoderDecoderModel.base_model_prefix - added:
field FlaxVisionEncoderDecoderModel.config_class - added:
field FlaxVisionEncoderDecoderModel.main_input_name - added:
field FlaxVisionEncoderDecoderModel.module_class - added:
method FlaxVisionEncoderDecoderModel.decode(decoder_input_ids, encoder_outputs, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxVisionEncoderDecoderModel.encode(pixel_values: jnp.ndarray, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…) - added:
method FlaxVisionEncoderDecoderModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxVisionEncoderDecoderModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxVisionEncoderDecoderModel.prepare_inputs_for_generation(decoder_input_ids, max_length, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxVisionEncoderDecoderModel.update_inputs_for_generation(model_outputs, model_kwargs) - added:
static method FlaxVisionEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, decoder_pretrained_model_name_or_path: Optional[Union[str, os.PathLike]]=…, *model_args, **kwargs) -> FlaxPreTrainedModel
- added:
class TFVisionEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss) - added:
constructor TFVisionEncoderDecoderModel(config: PretrainedConfig | None=…, encoder: TFPreTrainedModel | None=…, decoder: TFPreTrainedModel | None=…) - added:
field TFVisionEncoderDecoderModel.base_model_prefix - added:
field TFVisionEncoderDecoderModel.config_class - added:
field TFVisionEncoderDecoderModel.load_weight_prefix - added:
field TFVisionEncoderDecoderModel.main_input_name - added:
method TFVisionEncoderDecoderModel.build(input_shape=…) - added:
method TFVisionEncoderDecoderModel.call(pixel_values: np.ndarray | tf.Tensor | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, encoder_outputs: tuple | TFBaseModelOutput | None=…, past_key_values: tuple[tuple[np.ndarray | tf.Tensor]] | None=…, decoder_inputs_embeds: np.ndarray | tf.Tensor | None=…, labels: np.ndarray | tf.Tensor… - added:
method TFVisionEncoderDecoderModel.get_encoder() - added:
method TFVisionEncoderDecoderModel.get_input_embeddings() - added:
method TFVisionEncoderDecoderModel.get_output_embeddings() - added:
method TFVisionEncoderDecoderModel.prepare_decoder_input_ids_from_labels(labels: tf.Tensor) - added:
method TFVisionEncoderDecoderModel.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method TFVisionEncoderDecoderModel.resize_token_embeddings(*args, **kwargs) - added:
method TFVisionEncoderDecoderModel.serving_output(output) - added:
method TFVisionEncoderDecoderModel.set_output_embeddings(new_embeddings) - added:
method TFVisionEncoderDecoderModel.tf_to_pt_weight_rename(tf_weight) - added:
property TFVisionEncoderDecoderModel.input_signature - added:
static method TFVisionEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: str | None=…, decoder_pretrained_model_name_or_path: str | None=…, *model_args, **kwargs) -> TFPreTrainedModel
- added:
class VisionEncoderDecoderModel(PreTrainedModel, GenerationMixin) - added:
constructor VisionEncoderDecoderModel(config: Optional[PretrainedConfig]=…, encoder: Optional[PreTrainedModel]=…, decoder: Optional[PreTrainedModel]=…) - added:
field VisionEncoderDecoderModel.base_model_prefix - added:
field VisionEncoderDecoderModel.config: VisionEncoderDecoderConfig - added:
field VisionEncoderDecoderModel.main_input_name - added:
field VisionEncoderDecoderModel.supports_gradient_checkpointing - added:
method VisionEncoderDecoderModel.forward(pixel_values: Optional[torch.FloatTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.BoolTensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, past_key_values: Optional[Cache]=…, decoder_inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…… - added:
method VisionEncoderDecoderModel.get_encoder() - added:
method VisionEncoderDecoderModel.get_input_embeddings() - added:
method VisionEncoderDecoderModel.get_output_embeddings() - added:
method VisionEncoderDecoderModel.prepare_decoder_input_ids_from_labels(labels: torch.Tensor) - added:
method VisionEncoderDecoderModel.set_output_embeddings(new_embeddings) - added:
static method VisionEncoderDecoderModel.from_encoder_decoder_pretrained(encoder_pretrained_model_name_or_path: Optional[str]=…, decoder_pretrained_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel - added:
static method VisionEncoderDecoderModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
- added:
export define_import_structure
- added:
class VisionTextDualEncoderConfig(PretrainedConfig) - added:
constructor VisionTextDualEncoderConfig(projection_dim=…, logit_scale_init_value=…, **kwargs) - added:
field VisionTextDualEncoderConfig.has_no_defaults_at_init - added:
field VisionTextDualEncoderConfig.model_type - added:
field VisionTextDualEncoderConfig.sub_configs - added:
static method VisionTextDualEncoderConfig.from_vision_text_configs(vision_config: PretrainedConfig, text_config: PretrainedConfig, **kwargs)
- added:
class FlaxVisionTextDualEncoderModel(FlaxPreTrainedModel) - added:
constructor FlaxVisionTextDualEncoderModel(config: VisionTextDualEncoderConfig, input_shape: Optional[tuple]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxVisionTextDualEncoderModel.config_class - added:
field FlaxVisionTextDualEncoderModel.module_class - added:
method FlaxVisionTextDualEncoderModel.get_image_features(pixel_values, params: Optional[dict]=…, dropout_rng: jax.random.PRNGKey=…, train=…) - added:
method FlaxVisionTextDualEncoderModel.get_text_features(input_ids, attention_mask=…, position_ids=…, token_type_ids=…, params: Optional[dict]=…, dropout_rng: jax.random.PRNGKey=…, train=…) - added:
method FlaxVisionTextDualEncoderModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
static method FlaxVisionTextDualEncoderModel.from_vision_text_pretrained(vision_model_name_or_path: Optional[str]=…, text_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> FlaxPreTrainedModel
- added:
class TFVisionTextDualEncoderModel(TFPreTrainedModel) - added:
constructor TFVisionTextDualEncoderModel(config: VisionTextDualEncoderConfig | None=…, vision_model: TFPreTrainedModel | None=…, text_model: TFPreTrainedModel | None=…) - added:
field TFVisionTextDualEncoderModel.base_model_prefix - added:
field TFVisionTextDualEncoderModel.config_class - added:
field TFVisionTextDualEncoderModel.load_weight_prefix - added:
method TFVisionTextDualEncoderModel.build(input_shape=…) - added:
method TFVisionTextDualEncoderModel.call(input_ids: tf.Tensor | None=…, pixel_values: tf.Tensor | None=…, attention_mask: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, return_loss: bool | None=…, token_type_ids: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> tuple[tf.Tensor] | TFCLIPOutput - added:
method TFVisionTextDualEncoderModel.get_image_features(pixel_values=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method TFVisionTextDualEncoderModel.get_text_features(input_ids=…, attention_mask=…, position_ids=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method TFVisionTextDualEncoderModel.tf_to_pt_weight_rename(tf_weight) - added:
property TFVisionTextDualEncoderModel.dummy_inputs - added:
static method TFVisionTextDualEncoderModel.from_vision_text_pretrained(vision_model_name_or_path: str | None=…, text_model_name_or_path: str | None=…, *model_args, **kwargs) -> TFPreTrainedModel
- added:
class VisionTextDualEncoderModel(PreTrainedModel) - added:
constructor VisionTextDualEncoderModel(config: Optional[VisionTextDualEncoderConfig]=…, vision_model: Optional[PreTrainedModel]=…, text_model: Optional[PreTrainedModel]=…) - added:
field VisionTextDualEncoderModel.base_model_prefix - added:
field VisionTextDualEncoderModel.config: VisionTextDualEncoderConfig - added:
method VisionTextDualEncoderModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, token_type_ids: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…)… - added:
method VisionTextDualEncoderModel.get_image_features(pixel_values: torch.Tensor) -> torch.FloatTensor - added:
method VisionTextDualEncoderModel.get_text_features(input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
static method VisionTextDualEncoderModel.from_vision_text_pretrained(vision_model_name_or_path: Optional[str]=…, text_model_name_or_path: Optional[str]=…, *model_args, **kwargs) -> PreTrainedModel
- added:
class VisionTextDualEncoderProcessor(ProcessorMixin) - added:
constructor VisionTextDualEncoderProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field VisionTextDualEncoderProcessor.attributes - added:
field VisionTextDualEncoderProcessor.image_processor_class - added:
field VisionTextDualEncoderProcessor.tokenizer_class - added:
property VisionTextDualEncoderProcessor.feature_extractor - added:
property VisionTextDualEncoderProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class VisualBertConfig(PretrainedConfig) - added:
constructor VisualBertConfig(vocab_size=…, hidden_size=…, visual_embedding_dim=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, bypass_transformer=…, special_visual_initialize=…, pad_token_id=…, bos_token_id=…, eos_token_id=…,… - added:
field VisualBertConfig.model_type
- added:
class VisualBertForMultipleChoice(VisualBertPreTrainedModel) - added:
class VisualBertForPreTraining(VisualBertPreTrainedModel) - added:
class VisualBertForQuestionAnswering(VisualBertPreTrainedModel) - added:
class VisualBertForRegionToPhraseAlignment(VisualBertPreTrainedModel) - added:
class VisualBertForVisualReasoning(VisualBertPreTrainedModel) - added:
class VisualBertLayer(GradientCheckpointingLayer) - added:
class VisualBertModel(VisualBertPreTrainedModel) - added:
class VisualBertPreTrainedModel(PreTrainedModel) - added:
constructor VisualBertForMultipleChoice(config) - added:
constructor VisualBertForPreTraining(config) - added:
constructor VisualBertForQuestionAnswering(config) - added:
constructor VisualBertForRegionToPhraseAlignment(config) - added:
constructor VisualBertForVisualReasoning(config) - added:
constructor VisualBertLayer(config) - added:
constructor VisualBertModel(config, add_pooling_layer=…) - added:
field VisualBertPreTrainedModel.base_model_prefix - added:
field VisualBertPreTrainedModel.config: VisualBertConfig - added:
field VisualBertPreTrainedModel.supports_gradient_checkpointing - added:
method VisualBertForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.LongTens… - added:
method VisualBertForPreTraining.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.LongTensor]… - added:
method VisualBertForPreTraining.get_output_embeddings() - added:
method VisualBertForPreTraining.set_output_embeddings(new_embeddings) - added:
method VisualBertForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.LongT… - added:
method VisualBertForRegionToPhraseAlignment.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch… - added:
method VisualBertForVisualReasoning.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.LongTen… - added:
method VisualBertLayer.feed_forward_chunk(attention_output) - added:
method VisualBertLayer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…) - added:
method VisualBertModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, visual_embeds: Optional[torch.FloatTensor]=…, visual_attention_mask: Optional[torch.LongTensor]=…, visua… - added:
method VisualBertModel.get_input_embeddings() - added:
method VisualBertModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class ViTConfig(PretrainedConfig) - added:
class ViTOnnxConfig(OnnxConfig) - added:
constructor ViTConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, encoder_stride=…, pooler_output_size=…, pooler_act=…, **kwargs) - added:
field ViTConfig.model_type - added:
field ViTOnnxConfig.torch_onnx_minimum_version - added:
property ViTOnnxConfig.atol_for_validation -> float - added:
property ViTOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class ViTFeatureExtractor(ViTImageProcessor) - added:
constructor ViTFeatureExtractor(*args, **kwargs) -> None
- added:
class ViTImageProcessor(BaseImageProcessor) - added:
constructor ViTImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_rgb: Optional[bool]=…, **kwargs) -> None - added:
field ViTImageProcessor.model_input_names - added:
method ViTImageProcessor.preprocess(images: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optional[Union[str, Tens… - added:
method ViTImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class ViTImageProcessorFast(BaseImageProcessorFast) - added:
field ViTImageProcessorFast.do_normalize - added:
field ViTImageProcessorFast.do_rescale - added:
field ViTImageProcessorFast.do_resize - added:
field ViTImageProcessorFast.image_mean - added:
field ViTImageProcessorFast.image_std - added:
field ViTImageProcessorFast.resample - added:
field ViTImageProcessorFast.size
- added:
class FlaxViTForImageClassification(FlaxViTPreTrainedModel) - added:
class FlaxViTModel(FlaxViTPreTrainedModel) - added:
class FlaxViTPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxViTPreTrainedModel(config: ViTConfig, input_shape=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxViTForImageClassification.module_class - added:
field FlaxViTModel.module_class - added:
field FlaxViTPreTrainedModel.base_model_prefix - added:
field FlaxViTPreTrainedModel.config_class - added:
field FlaxViTPreTrainedModel.main_input_name - added:
field FlaxViTPreTrainedModel.module_class: nn.Module - added:
method FlaxViTPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFViTForImageClassification(TFViTPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFViTModel(TFViTPreTrainedModel) - added:
class TFViTPreTrainedModel(TFPreTrainedModel) - added:
constructor TFViTForImageClassification(config: ViTConfig, *inputs, **kwargs) - added:
constructor TFViTModel(config: ViTConfig, *inputs, add_pooling_layer=…, **kwargs) - added:
field TFViTPreTrainedModel.base_model_prefix - added:
field TFViTPreTrainedModel.config_class - added:
field TFViTPreTrainedModel.main_input_name - added:
method TFViTForImageClassification.build(input_shape=…) - added:
method TFViTForImageClassification.call(pixel_values: TFModelInputType | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, interpolate_pos_encoding: bool | None=…, return_dict: bool | None=…, labels: np.ndarray | tf.Tensor | None=…, training: bool | None=…) -> TFSequenceClassifierOutput | tuple[tf.Tensor] - added:
method TFViTModel.build(input_shape=…) - added:
method TFViTModel.call(pixel_values: TFModelInputType | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, interpolate_pos_encoding: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutputWithPooling | tuple[tf.Tensor]
- added:
class ViTForImageClassification(ViTPreTrainedModel) - added:
class ViTForMaskedImageModeling(ViTPreTrainedModel) - added:
class ViTModel(ViTPreTrainedModel) - added:
class ViTPreTrainedModel(PreTrainedModel) - added:
constructor ViTForImageClassification(config: ViTConfig) - added:
constructor ViTForMaskedImageModeling(config: ViTConfig) - added:
constructor ViTModel(config: ViTConfig, add_pooling_layer: bool=…, use_mask_token: bool=…) - added:
field ViTPreTrainedModel.base_model_prefix - added:
field ViTPreTrainedModel.config: ViTConfig - added:
field ViTPreTrainedModel.main_input_name - added:
field ViTPreTrainedModel.supports_gradient_checkpointing - added:
method ViTForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method ViTForMaskedImageModeling.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> MaskedImageModelingOutput - added:
method ViTModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method ViTModel.get_input_embeddings() -> ViTPatchEmbeddings
- added:
export define_import_structure
- added:
class ViTMAEConfig(PretrainedConfig) - added:
constructor ViTMAEConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, decoder_num_attention_heads=…, decoder_hidden_size=…, decoder_num_hidden_layers=…, decoder_intermediate_size=…, mask_ratio=…, norm… - added:
field ViTMAEConfig.model_type
- added:
class TFViTMAEForPreTraining(TFViTMAEPreTrainedModel) - added:
class TFViTMAEModel(TFViTMAEPreTrainedModel) - added:
class TFViTMAEPreTrainedModel(TFPreTrainedModel) - added:
constructor TFViTMAEForPreTraining(config) - added:
constructor TFViTMAEModel(config: ViTMAEConfig, *inputs, **kwargs) - added:
field TFViTMAEPreTrainedModel.base_model_prefix - added:
field TFViTMAEPreTrainedModel.config_class - added:
field TFViTMAEPreTrainedModel.main_input_name - added:
method TFViTMAEForPreTraining.build(input_shape=…) - added:
method TFViTMAEForPreTraining.call(pixel_values: TFModelInputType | None=…, noise: tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…, interpolate_pos_encoding: bool=…) -> TFViTMAEForPreTrainingOutput | tuple[tf.Tensor] - added:
method TFViTMAEForPreTraining.forward_loss(pixel_values, pred, mask, interpolate_pos_encoding: bool=…) - added:
method TFViTMAEForPreTraining.get_input_embeddings() - added:
method TFViTMAEForPreTraining.patchify(pixel_values, interpolate_pos_encoding: bool=…) - added:
method TFViTMAEForPreTraining.unpatchify(patchified_pixel_values, original_image_size: tuple[int, int] | None=…) - added:
method TFViTMAEModel.build(input_shape=…) - added:
method TFViTMAEModel.call(pixel_values: TFModelInputType | None=…, noise: tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…, interpolate_pos_encoding: bool=…) -> TFViTMAEModelOutput | tuple[tf.Tensor] - added:
method TFViTMAEModel.get_input_embeddings()
- added:
class ViTMAEForPreTraining(ViTMAEPreTrainedModel) - added:
class ViTMAELayer(GradientCheckpointingLayer) - added:
class ViTMAEModel(ViTMAEPreTrainedModel) - added:
class ViTMAEPreTrainedModel(PreTrainedModel) - added:
constructor ViTMAEForPreTraining(config: ViTMAEConfig) - added:
constructor ViTMAELayer(config: ViTMAEConfig) - added:
constructor ViTMAEModel(config) - added:
field ViTMAEPreTrainedModel.base_model_prefix - added:
field ViTMAEPreTrainedModel.config: ViTMAEConfig - added:
field ViTMAEPreTrainedModel.main_input_name - added:
field ViTMAEPreTrainedModel.supports_gradient_checkpointing - added:
method ViTMAEForPreTraining.forward(pixel_values: Optional[torch.FloatTensor]=…, noise: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> ViTMAEForPreTrainingOutput - added:
method ViTMAEForPreTraining.forward_loss(pixel_values, pred, mask, interpolate_pos_encoding: bool=…) - added:
method ViTMAEForPreTraining.get_input_embeddings() - added:
method ViTMAEForPreTraining.patchify(pixel_values, interpolate_pos_encoding: bool=…) - added:
method ViTMAEForPreTraining.unpatchify(patchified_pixel_values, original_image_size: Optional[tuple[int, int]]=…) - added:
method ViTMAELayer.forward(hidden_states: torch.Tensor, head_mask: Optional[torch.Tensor]=…) -> torch.Tensor - added:
method ViTMAEModel.forward(pixel_values: Optional[torch.FloatTensor]=…, noise: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> ViTMAEModelOutput - added:
method ViTMAEModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class ViTMSNConfig(PretrainedConfig) - added:
constructor ViTMSNConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, **kwargs) - added:
field ViTMSNConfig.model_type
- added:
class ViTMSNForImageClassification(ViTMSNPreTrainedModel) - added:
class ViTMSNModel(ViTMSNPreTrainedModel) - added:
class ViTMSNPreTrainedModel(PreTrainedModel) - added:
constructor ViTMSNForImageClassification(config: ViTMSNConfig) -> None - added:
constructor ViTMSNModel(config: ViTMSNConfig, use_mask_token: bool=…) - added:
field ViTMSNPreTrainedModel.base_model_prefix - added:
field ViTMSNPreTrainedModel.config: ViTMSNConfig - added:
field ViTMSNPreTrainedModel.main_input_name - added:
field ViTMSNPreTrainedModel.supports_gradient_checkpointing - added:
method ViTMSNForImageClassification.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method ViTMSNModel.forward(pixel_values: Optional[torch.Tensor]=…, bool_masked_pos: Optional[torch.BoolTensor]=…, head_mask: Optional[torch.Tensor]=…, interpolate_pos_encoding: Optional[bool]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutput - added:
method ViTMSNModel.get_input_embeddings() -> ViTMSNPatchEmbeddings
- added:
export define_import_structure
- added:
class VitDetConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor VitDetConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, hidden_act=…, dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, pretrain_image_size=…, patch_size=…, num_channels=…, qkv_bias=…, drop_path_rate=…, window_block_indices=…, residual_block_indices=…, use_absolute_position_embeddings=…, use_relative_position_embeddings=…, window_size=…,… - added:
field VitDetConfig.model_type
- added:
class VitDetBackbone(VitDetPreTrainedModel, BackboneMixin) - added:
class VitDetModel(VitDetPreTrainedModel) - added:
class VitDetPreTrainedModel(PreTrainedModel) - added:
constructor VitDetBackbone(config) - added:
constructor VitDetModel(config: VitDetConfig) - added:
field VitDetPreTrainedModel.base_model_prefix - added:
field VitDetPreTrainedModel.config: VitDetConfig - added:
field VitDetPreTrainedModel.main_input_name - added:
field VitDetPreTrainedModel.supports_gradient_checkpointing - added:
method VitDetBackbone.forward(pixel_values: torch.Tensor, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) -> BackboneOutput - added:
method VitDetBackbone.get_input_embeddings() -> VitDetEmbeddings - added:
method VitDetModel.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutput] - added:
method VitDetModel.get_input_embeddings() -> VitDetEmbeddings
- added:
export define_import_structure
- added:
class VitMatteConfig(PretrainedConfig) - added:
constructor VitMatteConfig(backbone_config: Optional[PretrainedConfig]=…, backbone=…, use_pretrained_backbone=…, use_timm_backbone=…, backbone_kwargs=…, hidden_size: int=…, batch_norm_eps: float=…, initializer_range: float=…, convstream_hidden_sizes: list[int]=…, fusion_hidden_sizes: list[int]=…, **kwargs) - added:
field VitMatteConfig.model_type - added:
method VitMatteConfig.to_dict() - added:
property VitMatteConfig.sub_configs
- added:
class VitMatteImageProcessor(BaseImageProcessor) - added:
constructor VitMatteImageProcessor(do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: bool=…, size_divisor: int=…, **kwargs) -> None - added:
field VitMatteImageProcessor.model_input_names - added:
method VitMatteImageProcessor.pad_image(image: np.ndarray, size_divisibility: int=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method VitMatteImageProcessor.preprocess(images: ImageInput, trimaps: ImageInput, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_pad: Optional[bool]=…, size_divisor: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, data_forma… - added:
property VitMatteImageProcessor.size_divisibility
- added:
class VitMatteImageProcessorFast(BaseImageProcessorFast) - added:
constructor VitMatteImageProcessorFast(**kwargs: Unpack[VitMatteFastImageProcessorKwargs]) -> None - added:
field VitMatteImageProcessorFast.do_normalize: bool - added:
field VitMatteImageProcessorFast.do_pad: bool - added:
field VitMatteImageProcessorFast.do_rescale: bool - added:
field VitMatteImageProcessorFast.image_mean: Optional[Union[float, list[float]]] - added:
field VitMatteImageProcessorFast.image_std: Optional[Union[float, list[float]]] - added:
field VitMatteImageProcessorFast.rescale_factor: Union[int, float] - added:
field VitMatteImageProcessorFast.size_divisor: int - added:
field VitMatteImageProcessorFast.valid_kwargs - added:
method VitMatteImageProcessorFast.preprocess(images: list[torch.Tensor], trimaps: list[torch.Tensor], **kwargs: Unpack[VitMatteFastImageProcessorKwargs]) -> BatchFeature - added:
property VitMatteImageProcessorFast.size_divisibility
- added:
class VitMatteForImageMatting(VitMattePreTrainedModel) - added:
class VitMattePreTrainedModel(PreTrainedModel) - added:
constructor VitMatteForImageMatting(config) - added:
field VitMattePreTrainedModel.config: VitMatteConfig - added:
field VitMattePreTrainedModel.main_input_name - added:
field VitMattePreTrainedModel.supports_gradient_checkpointing - added:
method VitMatteForImageMatting.forward(pixel_values: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, labels: Optional[torch.Tensor]=…, return_dict: Optional[bool]=…)
- added:
export define_import_structure
- added:
class VitPoseConfig(PretrainedConfig) - added:
constructor VitPoseConfig(backbone_config: Optional[PretrainedConfig]=…, backbone: Optional[str]=…, use_pretrained_backbone: bool=…, use_timm_backbone: bool=…, backbone_kwargs: Optional[dict]=…, initializer_range: float=…, scale_factor: int=…, use_simple_decoder: bool=…, **kwargs) - added:
field VitPoseConfig.model_type - added:
property VitPoseConfig.sub_configs
- added:
class VitPoseImageProcessor(BaseImageProcessor) - added:
constructor VitPoseImageProcessor(do_affine_transform: bool=…, size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) - added:
field VitPoseImageProcessor.model_input_names - added:
method VitPoseImageProcessor.affine_transform(image: np.ndarray, center: tuple[float], scale: tuple[float], rotation: float, size: dict[str, int], data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method VitPoseImageProcessor.keypoints_from_heatmaps(heatmaps: np.ndarray, center: np.ndarray, scale: np.ndarray, kernel: int=…) - added:
method VitPoseImageProcessor.post_process_pose_estimation(outputs: VitPoseEstimatorOutput, boxes: Union[list[list[list[float]]], np.ndarray], kernel_size: int=…, threshold: Optional[float]=…, target_sizes: Union[TensorType, list[tuple]]=…) - added:
method VitPoseImageProcessor.preprocess(images: ImageInput, boxes: Union[list[list[float]], np.ndarray], do_affine_transform: Optional[bool]=…, size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, return_tensors: Optiona…
- added:
class VitPoseForPoseEstimation(VitPosePreTrainedModel) - added:
class VitPosePreTrainedModel(PreTrainedModel) - added:
constructor VitPoseForPoseEstimation(config: VitPoseConfig) - added:
field VitPosePreTrainedModel.base_model_prefix - added:
field VitPosePreTrainedModel.config: VitPoseConfig - added:
field VitPosePreTrainedModel.main_input_name - added:
field VitPosePreTrainedModel.supports_gradient_checkpointing - added:
method VitPoseForPoseEstimation.forward(pixel_values: torch.Tensor, dataset_index: Optional[torch.Tensor]=…, flip_pairs: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> VitPoseEstimatorOutput
- added:
export define_import_structure
- added:
class VitPoseBackboneConfig(BackboneConfigMixin, PretrainedConfig) - added:
constructor VitPoseBackboneConfig(image_size=…, patch_size=…, num_channels=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, mlp_ratio=…, num_experts=…, part_features=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, out_features=…, out_indices=…, **kwargs) - added:
field VitPoseBackboneConfig.model_type
- added:
class VitPoseBackbone(VitPoseBackbonePreTrainedModel, BackboneMixin) - added:
class VitPoseBackbonePreTrainedModel(PreTrainedModel) - added:
constructor VitPoseBackbone(config: VitPoseBackboneConfig) - added:
field VitPoseBackbonePreTrainedModel.base_model_prefix - added:
field VitPoseBackbonePreTrainedModel.config: VitPoseBackboneConfig - added:
field VitPoseBackbonePreTrainedModel.main_input_name - added:
field VitPoseBackbonePreTrainedModel.supports_gradient_checkpointing - added:
method VitPoseBackbone.forward(pixel_values: torch.Tensor, dataset_index: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, output_hidden_states: Optional[bool]=…, **kwargs)
- added:
export define_import_structure
- added:
class VitsConfig(PretrainedConfig) - added:
constructor VitsConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, window_size=…, use_bias=…, ffn_dim=…, layerdrop=…, ffn_kernel_size=…, flow_size=…, spectrogram_bins=…, hidden_act=…, hidden_dropout=…, attention_dropout=…, activation_dropout=…, initializer_range=…, layer_norm_eps=…, use_stochastic_duration_prediction=…, num_speakers=…, speaker_embedding_size=…, upsamp… - added:
field VitsConfig.model_type
- added:
class VitsModel(VitsPreTrainedModel) - added:
class VitsPreTrainedModel(PreTrainedModel) - added:
constructor VitsModel(config: VitsConfig) - added:
field VitsPreTrainedModel.base_model_prefix - added:
field VitsPreTrainedModel.config: VitsConfig - added:
field VitsPreTrainedModel.main_input_name - added:
field VitsPreTrainedModel.supports_gradient_checkpointing - added:
method VitsModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, speaker_id: Optional[int]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.FloatTensor]=…) -> Union[tuple[Any], VitsModelOutput] - added:
method VitsModel.get_encoder()
- added:
class VitsTokenizer(PreTrainedTokenizer) - added:
constructor VitsTokenizer(vocab_file, pad_token=…, unk_token=…, language=…, add_blank=…, normalize=…, phonemize=…, is_uroman=…, **kwargs) -> None - added:
field VitsTokenizer.model_input_names - added:
field VitsTokenizer.vocab_files_names - added:
method VitsTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method VitsTokenizer.get_vocab() - added:
method VitsTokenizer.normalize_text(input_string) - added:
method VitsTokenizer.prepare_for_tokenization(text: str, is_split_into_words: bool=…, normalize: Optional[bool]=…, **kwargs) -> tuple[str, dict[str, Any]] - added:
method VitsTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Union[tuple[str], None] - added:
property VitsTokenizer.vocab_size
- added:
export define_import_structure
- added:
class VivitConfig(PretrainedConfig) - added:
constructor VivitConfig(image_size=…, num_frames=…, tubelet_size=…, num_channels=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, qkv_bias=…, **kwargs) - added:
field VivitConfig.model_type
- added:
class VivitImageProcessor(BaseImageProcessor) - added:
constructor VivitImageProcessor(do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_center_crop: bool=…, crop_size: Optional[dict[str, int]]=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, offset: bool=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, **kwargs) -> None - added:
field VivitImageProcessor.model_input_names - added:
method VivitImageProcessor.preprocess(videos: ImageInput, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample: Optional[PILImageResampling]=…, do_center_crop: Optional[bool]=…, crop_size: Optional[dict[str, int]]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, offset: Optional[bool]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[f… - added:
method VivitImageProcessor.rescale(image: np.ndarray, scale: Union[int, float], offset: bool=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) - added:
method VivitImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray
- added:
class VivitForVideoClassification(VivitPreTrainedModel) - added:
class VivitModel(VivitPreTrainedModel) - added:
class VivitPreTrainedModel(PreTrainedModel) - added:
constructor VivitForVideoClassification(config: VivitConfig) - added:
constructor VivitModel(config: VivitConfig, add_pooling_layer: bool=…) - added:
field VivitPreTrainedModel.base_model_prefix - added:
field VivitPreTrainedModel.config: VivitConfig - added:
field VivitPreTrainedModel.main_input_name - added:
field VivitPreTrainedModel.supports_gradient_checkpointing - added:
method VivitForVideoClassification.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> ImageClassifierOutput - added:
method VivitModel.forward(pixel_values: Optional[torch.FloatTensor]=…, head_mask: Optional[torch.FloatTensor]=…, interpolate_pos_encoding: bool=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method VivitModel.get_input_embeddings()
- added:
export define_import_structure
- added:
class VJEPA2Config(PretrainedConfig) - added:
constructor VJEPA2Config(patch_size=…, crop_size=…, frames_per_clip=…, tubelet_size=…, hidden_size=…, in_chans=…, num_attention_heads=…, num_hidden_layers=…, drop_path_rate=…, mlp_ratio=…, layer_norm_eps=…, qkv_bias=…, attention_probs_dropout_prob=…, hidden_act=…, initializer_range=…, attention_dropout=…, num_pooler_layers=…, pred_hidden_size=…, pred_num_attention_heads=…, pred_num_hidden_layers=… - added:
field VJEPA2Config.model_type
- added:
class VJEPA2ForVideoClassification(VJEPA2PreTrainedModel) - added:
class VJEPA2Model(VJEPA2PreTrainedModel) - added:
class VJEPA2PreTrainedModel(PreTrainedModel) - added:
constructor VJEPA2ForVideoClassification(config: VJEPA2Config) - added:
constructor VJEPA2Model(config: VJEPA2Config) - added:
field VJEPA2PreTrainedModel.base_model_prefix - added:
field VJEPA2PreTrainedModel.config: VJEPA2Config - added:
field VJEPA2PreTrainedModel.main_input_name - added:
field VJEPA2PreTrainedModel.supports_gradient_checkpointing - added:
method VJEPA2ForVideoClassification.forward(pixel_values_videos: torch.Tensor, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…) -> Union[tuple, ImageClassifierOutput] - added:
method VJEPA2Model.forward(pixel_values_videos: torch.Tensor, context_head_mask: Optional[torch.Tensor]=…, context_mask: Optional[list[torch.Tensor]]=…, target_head_mask: Optional[torch.Tensor]=…, target_mask: Optional[list[torch.Tensor]]=…, skip_predictor: bool=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> VJEPA2WithMaskedInputModelOutput - added:
method VJEPA2Model.get_input_embeddings() -> VJEPA2PatchEmbeddings3D - added:
method VJEPA2Model.get_vision_features(pixel_values_videos) -> torch.Tensor
- added:
class VJEPA2VideoProcessor(BaseVideoProcessor) - added:
constructor VJEPA2VideoProcessor(**kwargs: Unpack[VJEPA2VideoProcessorInitKwargs]) - added:
field VJEPA2VideoProcessor.crop_size - added:
field VJEPA2VideoProcessor.do_center_crop - added:
field VJEPA2VideoProcessor.do_normalize - added:
field VJEPA2VideoProcessor.do_rescale - added:
field VJEPA2VideoProcessor.do_resize - added:
field VJEPA2VideoProcessor.image_mean - added:
field VJEPA2VideoProcessor.image_std - added:
field VJEPA2VideoProcessor.model_input_names - added:
field VJEPA2VideoProcessor.resample - added:
field VJEPA2VideoProcessor.size - added:
field VJEPA2VideoProcessor.valid_kwargs
- added:
export define_import_structure
- added:
class VoxtralConfig(PretrainedConfig) - added:
class VoxtralEncoderConfig(PretrainedConfig) - added:
constructor VoxtralConfig(audio_config=…, text_config=…, audio_token_id=…, projector_hidden_act=…, **kwargs) - added:
constructor VoxtralEncoderConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, scale_embedding=…, activation_function=…, num_mel_bins=…, max_source_positions=…, initializer_range=…, attention_dropout=…, **kwargs) - added:
field VoxtralConfig.model_type - added:
field VoxtralConfig.sub_configs - added:
field VoxtralEncoderConfig.attribute_map - added:
field VoxtralEncoderConfig.model_type
- added:
class VoxtralEncoder(VoxtralPreTrainedModel) - added:
class VoxtralForConditionalGeneration(VoxtralPreTrainedModel, GenerationMixin) - added:
class VoxtralPreTrainedModel(PreTrainedModel) - added:
constructor VoxtralEncoder(config: VoxtralEncoderConfig) - added:
constructor VoxtralForConditionalGeneration(config) - added:
field VoxtralEncoder.config: VoxtralEncoderConfig - added:
field VoxtralEncoder.main_input_name - added:
field VoxtralPreTrainedModel.base_model_prefix - added:
field VoxtralPreTrainedModel.config: VoxtralConfig - added:
field VoxtralPreTrainedModel.supports_gradient_checkpointing - added:
method VoxtralEncoder.forward(input_features, attention_mask=…, **kwargs: Unpack[TransformersKwargs]) - added:
method VoxtralEncoder.get_input_embeddings() -> nn.Module - added:
method VoxtralEncoder.set_input_embeddings(value: nn.Module) - added:
method VoxtralForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torc… - added:
method VoxtralForConditionalGeneration.get_audio_embeds(input_features: torch.FloatTensor) - added:
method VoxtralForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor) - added:
method VoxtralForConditionalGeneration.get_decoder() - added:
method VoxtralForConditionalGeneration.get_input_embeddings() - added:
method VoxtralForConditionalGeneration.get_output_embeddings() - added:
method VoxtralForConditionalGeneration.prepare_inputs_for_generation(*args, **kwargs) - added:
method VoxtralForConditionalGeneration.set_decoder(decoder) - added:
method VoxtralForConditionalGeneration.set_input_embeddings(value) - added:
method VoxtralForConditionalGeneration.set_output_embeddings(new_embeddings)
- added:
class VoxtralEncoder(Qwen2AudioEncoder) - added:
class VoxtralForConditionalGeneration(VoxtralPreTrainedModel, GenerationMixin) - added:
class VoxtralPreTrainedModel(Qwen2AudioPreTrainedModel) - added:
constructor VoxtralForConditionalGeneration(config) - added:
method VoxtralEncoder.forward(input_features, attention_mask=…, **kwargs: Unpack[TransformersKwargs]) - added:
method VoxtralForConditionalGeneration.forward(input_ids: Optional[torch.LongTensor]=…, input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, cache_position: Optional[torc… - added:
method VoxtralForConditionalGeneration.get_audio_embeds(input_features: torch.FloatTensor) - added:
method VoxtralForConditionalGeneration.get_audio_features(input_features: torch.FloatTensor) - added:
method VoxtralForConditionalGeneration.get_decoder() - added:
method VoxtralForConditionalGeneration.get_input_embeddings() - added:
method VoxtralForConditionalGeneration.get_output_embeddings() - added:
method VoxtralForConditionalGeneration.prepare_inputs_for_generation(*args, **kwargs) - added:
method VoxtralForConditionalGeneration.set_decoder(decoder) - added:
method VoxtralForConditionalGeneration.set_input_embeddings(value) - added:
method VoxtralForConditionalGeneration.set_output_embeddings(new_embeddings)
- added:
class VoxtralProcessor(ProcessorMixin) - added:
constructor VoxtralProcessor(feature_extractor, tokenizer) - added:
field VoxtralProcessor.attributes - added:
field VoxtralProcessor.feature_extractor_class - added:
field VoxtralProcessor.tokenizer_class - added:
method VoxtralProcessor.apply_chat_template(conversation: Union[list[dict[str, str]], list[list[dict[str, str]]]], **kwargs: Unpack[AllKwargsForChatTemplate]) -> str - added:
method VoxtralProcessor.apply_transcription_request(language: Union[str, list[str]], audio: Union[str, list[str], AudioInput], model_id: str, sampling_rate: Optional[int]=…, format: Optional[Union[str, list[str]]]=…, **kwargs: Unpack[VoxtralProcessorKwargs])
- added:
export define_import_structure
- added:
class Wav2Vec2Config(PretrainedConfig) - added:
constructor Wav2Vec2Config(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, feat_quantizer_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel=… - added:
field Wav2Vec2Config.model_type - added:
property Wav2Vec2Config.inputs_to_logits_ratio
- added:
class Wav2Vec2FeatureExtractor(SequenceFeatureExtractor) - added:
constructor Wav2Vec2FeatureExtractor(feature_size=…, sampling_rate=…, padding_value=…, return_attention_mask=…, do_normalize=…, **kwargs) - added:
field Wav2Vec2FeatureExtractor.model_input_names - added:
static method Wav2Vec2FeatureExtractor.zero_mean_unit_var_norm(input_values: list[np.ndarray], attention_mask: list[np.ndarray], padding_value: float=…) -> list[np.ndarray]
- added:
class FlaxWav2Vec2ForCTC(FlaxWav2Vec2PreTrainedModel) - added:
class FlaxWav2Vec2ForPreTraining(FlaxWav2Vec2PreTrainedModel) - added:
class FlaxWav2Vec2Model(FlaxWav2Vec2PreTrainedModel) - added:
class FlaxWav2Vec2PreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxWav2Vec2PreTrainedModel(config: Wav2Vec2Config, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxWav2Vec2ForCTC.module_class - added:
field FlaxWav2Vec2ForPreTraining.module_class - added:
field FlaxWav2Vec2Model.module_class - added:
field FlaxWav2Vec2PreTrainedModel.base_model_prefix: str - added:
field FlaxWav2Vec2PreTrainedModel.config_class - added:
field FlaxWav2Vec2PreTrainedModel.main_input_name - added:
field FlaxWav2Vec2PreTrainedModel.module_class: nn.Module - added:
method FlaxWav2Vec2PreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFWav2Vec2ForCTC(TFWav2Vec2PreTrainedModel) - added:
class TFWav2Vec2ForSequenceClassification(TFWav2Vec2PreTrainedModel) - added:
class TFWav2Vec2Model(TFWav2Vec2PreTrainedModel) - added:
class TFWav2Vec2PreTrainedModel(TFPreTrainedModel) - added:
constructor TFWav2Vec2ForCTC(config: Wav2Vec2Config, *inputs, **kwargs) - added:
constructor TFWav2Vec2ForSequenceClassification(config) - added:
constructor TFWav2Vec2Model(config: Wav2Vec2Config, *inputs, **kwargs) - added:
constructor TFWav2Vec2PreTrainedModel(config, *inputs, **kwargs) - added:
field TFWav2Vec2PreTrainedModel.base_model_prefix - added:
field TFWav2Vec2PreTrainedModel.config_class - added:
field TFWav2Vec2PreTrainedModel.main_input_name - added:
method TFWav2Vec2ForCTC.build(input_shape=…) - added:
method TFWav2Vec2ForCTC.call(input_values: tf.Tensor, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, labels: tf.Tensor | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool | None=…) -> TFCausalLMOutput |… - added:
method TFWav2Vec2ForCTC.freeze_feature_encoder() - added:
method TFWav2Vec2ForCTC.freeze_feature_extractor() - added:
method TFWav2Vec2ForSequenceClassification.build(input_shape=…) - added:
method TFWav2Vec2ForSequenceClassification.call(input_values: tf.Tensor, attention_mask: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, labels: tf.Tensor | None=…, training: bool=…) -> TFSequenceClassifierOutput | tuple[tf.Tensor] - added:
method TFWav2Vec2ForSequenceClassification.freeze_base_model() - added:
method TFWav2Vec2ForSequenceClassification.freeze_feature_encoder() - added:
method TFWav2Vec2ForSequenceClassification.freeze_feature_extractor() - added:
method TFWav2Vec2Model.build(input_shape=…) - added:
method TFWav2Vec2Model.call(input_values: tf.Tensor, attention_mask: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool | None=…, training: bool=…) -> TFBaseModelOutput | tuple[tf.Tensor] - added:
property TFWav2Vec2PreTrainedModel.dummy_inputs - added:
property TFWav2Vec2PreTrainedModel.input_signature
- added:
class Wav2Vec2ForAudioFrameClassification(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2ForCTC(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2ForMaskedLM(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2ForPreTraining(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2ForSequenceClassification(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2ForXVector(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2Model(Wav2Vec2PreTrainedModel) - added:
class Wav2Vec2PreTrainedModel(PreTrainedModel) - added:
constructor Wav2Vec2ForAudioFrameClassification(config) - added:
constructor Wav2Vec2ForCTC(config, target_lang: Optional[str]=…) - added:
constructor Wav2Vec2ForMaskedLM(config) - added:
constructor Wav2Vec2ForPreTraining(config: Wav2Vec2Config) - added:
constructor Wav2Vec2ForSequenceClassification(config) - added:
constructor Wav2Vec2ForXVector(config) - added:
constructor Wav2Vec2Model(config: Wav2Vec2Config) - added:
field Wav2Vec2PreTrainedModel.base_model_prefix - added:
field Wav2Vec2PreTrainedModel.config: Wav2Vec2Config - added:
field Wav2Vec2PreTrainedModel.main_input_name - added:
field Wav2Vec2PreTrainedModel.supports_gradient_checkpointing - added:
method Wav2Vec2ForAudioFrameClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method Wav2Vec2ForAudioFrameClassification.freeze_base_model() - added:
method Wav2Vec2ForAudioFrameClassification.freeze_feature_encoder() - added:
method Wav2Vec2ForAudioFrameClassification.freeze_feature_extractor() - added:
method Wav2Vec2ForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method Wav2Vec2ForCTC.freeze_base_model() - added:
method Wav2Vec2ForCTC.freeze_feature_encoder() - added:
method Wav2Vec2ForCTC.freeze_feature_extractor() - added:
method Wav2Vec2ForCTC.tie_weights() - added:
method Wav2Vec2ForMaskedLM.forward(input_values: torch.FloatTensor, attention_mask: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, MaskedLMOutput] - added:
method Wav2Vec2ForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.BoolTensor]=…, sampled_negative_indices: Optional[torch.BoolTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2ForPreTrainingOutput] - added:
method Wav2Vec2ForPreTraining.freeze_feature_encoder() - added:
method Wav2Vec2ForPreTraining.freeze_feature_extractor() - added:
method Wav2Vec2ForPreTraining.set_gumbel_temperature(temperature: int) - added:
method Wav2Vec2ForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method Wav2Vec2ForSequenceClassification.freeze_base_model() - added:
method Wav2Vec2ForSequenceClassification.freeze_feature_encoder() - added:
method Wav2Vec2ForSequenceClassification.freeze_feature_extractor() - added:
method Wav2Vec2ForXVector.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method Wav2Vec2ForXVector.freeze_base_model() - added:
method Wav2Vec2ForXVector.freeze_feature_encoder() - added:
method Wav2Vec2ForXVector.freeze_feature_extractor() - added:
method Wav2Vec2Model.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2BaseModelOutput] - added:
method Wav2Vec2Model.freeze_feature_encoder() - added:
method Wav2Vec2Model.freeze_feature_extractor() - added:
method Wav2Vec2PreTrainedModel.init_adapter_layers() - added:
method Wav2Vec2PreTrainedModel.load_adapter(target_lang: str, force_load=…, **kwargs) - added:
static method Wav2Vec2ForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
class Wav2Vec2Processor(ProcessorMixin) - added:
constructor Wav2Vec2Processor(feature_extractor, tokenizer) - added:
field Wav2Vec2Processor.feature_extractor_class - added:
field Wav2Vec2Processor.tokenizer_class - added:
method Wav2Vec2Processor.as_target_processor() - added:
method Wav2Vec2Processor.pad(*args, **kwargs) - added:
property Wav2Vec2Processor.model_input_names - added:
static method Wav2Vec2Processor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
class Wav2Vec2CTCTokenizer(PreTrainedTokenizer) - added:
class Wav2Vec2Tokenizer(PreTrainedTokenizer) - added:
constructor Wav2Vec2CTCTokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, word_delimiter_token=…, replace_word_delimiter_char=…, do_lower_case=…, target_lang=…, **kwargs) - added:
constructor Wav2Vec2Tokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, word_delimiter_token=…, do_lower_case=…, do_normalize=…, return_attention_mask=…, **kwargs) - added:
field Wav2Vec2CTCTokenizer.model_input_names - added:
field Wav2Vec2CTCTokenizer.vocab_files_names - added:
field Wav2Vec2Tokenizer.model_input_names - added:
field Wav2Vec2Tokenizer.pretrained_vocab_files_map - added:
field Wav2Vec2Tokenizer.vocab_files_names - added:
method Wav2Vec2CTCTokenizer.batch_decode(sequences: Union[list[int], list[list[int]], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_char_offsets: bool=…, output_word_offsets: bool=…, **kwargs) -> list[str] - added:
method Wav2Vec2CTCTokenizer.convert_tokens_to_string(tokens: list[str], group_tokens: bool=…, spaces_between_special_tokens: bool=…, output_char_offsets: bool=…, output_word_offsets: bool=…) -> dict[str, Union[str, float]] - added:
method Wav2Vec2CTCTokenizer.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_char_offsets: bool=…, output_word_offsets: bool=…, **kwargs) -> str - added:
method Wav2Vec2CTCTokenizer.get_vocab() -> dict - added:
method Wav2Vec2CTCTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method Wav2Vec2CTCTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method Wav2Vec2CTCTokenizer.set_target_lang(target_lang: str) - added:
method Wav2Vec2Tokenizer.convert_tokens_to_string(tokens: list[str]) -> str - added:
method Wav2Vec2Tokenizer.get_vocab() -> dict - added:
method Wav2Vec2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property Wav2Vec2CTCTokenizer.vocab_size -> int - added:
property Wav2Vec2CTCTokenizer.word_delimiter_token -> str - added:
property Wav2Vec2CTCTokenizer.word_delimiter_token_id -> Optional[int] - added:
property Wav2Vec2Tokenizer.vocab_size -> int - added:
property Wav2Vec2Tokenizer.word_delimiter_token -> str - added:
property Wav2Vec2Tokenizer.word_delimiter_token_id -> Optional[int]
- added:
export define_import_structure
- added:
class Wav2Vec2BertConfig(PretrainedConfig) - added:
constructor Wav2Vec2BertConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, feature_projection_input_dim=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, apply_spec_augment=…, mask_time_prob=…, mask_time_length=…, mask_time_min_mas… - added:
field Wav2Vec2BertConfig.model_type - added:
property Wav2Vec2BertConfig.inputs_to_logits_ratio
- added:
class Wav2Vec2BertForAudioFrameClassification(Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertForCTC(Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertForSequenceClassification(Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertForXVector(Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertModel(Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertPreTrainedModel(PreTrainedModel) - added:
constructor Wav2Vec2BertForAudioFrameClassification(config) - added:
constructor Wav2Vec2BertForCTC(config, target_lang: Optional[str]=…) - added:
constructor Wav2Vec2BertForSequenceClassification(config) - added:
constructor Wav2Vec2BertForXVector(config) - added:
constructor Wav2Vec2BertModel(config: Wav2Vec2BertConfig) - added:
field Wav2Vec2BertPreTrainedModel.base_model_prefix - added:
field Wav2Vec2BertPreTrainedModel.config: Wav2Vec2BertConfig - added:
field Wav2Vec2BertPreTrainedModel.main_input_name - added:
field Wav2Vec2BertPreTrainedModel.supports_gradient_checkpointing - added:
method Wav2Vec2BertForAudioFrameClassification.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method Wav2Vec2BertForAudioFrameClassification.freeze_base_model() - added:
method Wav2Vec2BertForCTC.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method Wav2Vec2BertForSequenceClassification.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method Wav2Vec2BertForSequenceClassification.freeze_base_model() - added:
method Wav2Vec2BertForXVector.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method Wav2Vec2BertForXVector.freeze_base_model() - added:
method Wav2Vec2BertModel.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2BertBaseModelOutput]
- added:
class Wav2Vec2BertForAudioFrameClassification(Wav2Vec2ConformerForAudioFrameClassification) - added:
class Wav2Vec2BertForCTC(Wav2Vec2ConformerForCTC) - added:
class Wav2Vec2BertForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class Wav2Vec2BertForXVector(Wav2Vec2ConformerForXVector) - added:
class Wav2Vec2BertModel(Wav2Vec2Model, Wav2Vec2BertPreTrainedModel) - added:
class Wav2Vec2BertPreTrainedModel(PreTrainedModel) - added:
constructor Wav2Vec2BertForAudioFrameClassification(config) - added:
constructor Wav2Vec2BertForCTC(config, target_lang: Optional[str]=…) - added:
constructor Wav2Vec2BertForSequenceClassification(config) - added:
constructor Wav2Vec2BertForXVector(config) - added:
constructor Wav2Vec2BertModel(config: Wav2Vec2BertConfig) - added:
field Wav2Vec2BertPreTrainedModel.base_model_prefix - added:
field Wav2Vec2BertPreTrainedModel.config: Wav2Vec2BertConfig - added:
field Wav2Vec2BertPreTrainedModel.main_input_name - added:
field Wav2Vec2BertPreTrainedModel.supports_gradient_checkpointing - added:
method Wav2Vec2BertForAudioFrameClassification.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method Wav2Vec2BertForAudioFrameClassification.freeze_feature_encoder() - added:
method Wav2Vec2BertForCTC.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method Wav2Vec2BertForCTC.freeze_feature_encoder() - added:
method Wav2Vec2BertForSequenceClassification.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method Wav2Vec2BertForSequenceClassification.freeze_base_model() - added:
method Wav2Vec2BertForSequenceClassification.freeze_feature_encoder() - added:
method Wav2Vec2BertForSequenceClassification.freeze_feature_extractor() - added:
method Wav2Vec2BertForXVector.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method Wav2Vec2BertForXVector.freeze_feature_encoder() - added:
method Wav2Vec2BertModel.forward(input_features: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2BertBaseModelOutput] - added:
method Wav2Vec2BertModel.freeze_feature_encoder() - added:
method Wav2Vec2BertModel.freeze_feature_extractor()
- added:
class Wav2Vec2BertProcessor(ProcessorMixin) - added:
constructor Wav2Vec2BertProcessor(feature_extractor, tokenizer) - added:
field Wav2Vec2BertProcessor.feature_extractor_class - added:
field Wav2Vec2BertProcessor.tokenizer_class - added:
method Wav2Vec2BertProcessor.pad(input_features=…, labels=…, **kwargs) - added:
property Wav2Vec2BertProcessor.model_input_names - added:
static method Wav2Vec2BertProcessor.from_pretrained(pretrained_model_name_or_path, **kwargs)
- added:
export define_import_structure
- added:
class Wav2Vec2ConformerConfig(PretrainedConfig) - added:
constructor Wav2Vec2ConformerConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, feat_quantizer_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, con… - added:
field Wav2Vec2ConformerConfig.model_type - added:
property Wav2Vec2ConformerConfig.inputs_to_logits_ratio
- added:
class Wav2Vec2ConformerForAudioFrameClassification(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerForCTC(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerForPreTraining(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerForSequenceClassification(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerForXVector(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerModel(Wav2Vec2ConformerPreTrainedModel) - added:
class Wav2Vec2ConformerPreTrainedModel(PreTrainedModel) - added:
constructor Wav2Vec2ConformerForAudioFrameClassification(config) - added:
constructor Wav2Vec2ConformerForCTC(config, target_lang: Optional[str]=…) - added:
constructor Wav2Vec2ConformerForPreTraining(config: Wav2Vec2ConformerConfig) - added:
constructor Wav2Vec2ConformerForSequenceClassification(config) - added:
constructor Wav2Vec2ConformerForXVector(config) - added:
constructor Wav2Vec2ConformerModel(config: Wav2Vec2ConformerConfig) - added:
field Wav2Vec2ConformerPreTrainedModel.base_model_prefix - added:
field Wav2Vec2ConformerPreTrainedModel.config: Wav2Vec2ConformerConfig - added:
field Wav2Vec2ConformerPreTrainedModel.main_input_name - added:
field Wav2Vec2ConformerPreTrainedModel.supports_gradient_checkpointing - added:
method Wav2Vec2ConformerForAudioFrameClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method Wav2Vec2ConformerForAudioFrameClassification.freeze_base_model() - added:
method Wav2Vec2ConformerForAudioFrameClassification.freeze_feature_encoder() - added:
method Wav2Vec2ConformerForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method Wav2Vec2ConformerForCTC.freeze_feature_encoder() - added:
method Wav2Vec2ConformerForPreTraining.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.BoolTensor]=…, sampled_negative_indices: Optional[torch.BoolTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2ConformerForPreTrainingOutput] - added:
method Wav2Vec2ConformerForPreTraining.freeze_feature_encoder() - added:
method Wav2Vec2ConformerForPreTraining.set_gumbel_temperature(temperature: int) - added:
method Wav2Vec2ConformerForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method Wav2Vec2ConformerForSequenceClassification.freeze_base_model() - added:
method Wav2Vec2ConformerForSequenceClassification.freeze_feature_encoder() - added:
method Wav2Vec2ConformerForXVector.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method Wav2Vec2ConformerForXVector.freeze_base_model() - added:
method Wav2Vec2ConformerForXVector.freeze_feature_encoder() - added:
method Wav2Vec2ConformerModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, Wav2Vec2ConformerBaseModelOutput] - added:
method Wav2Vec2ConformerModel.freeze_feature_encoder() - added:
static method Wav2Vec2ConformerForPreTraining.compute_contrastive_logits(target_features: torch.FloatTensor, negative_features: torch.FloatTensor, predicted_features: torch.FloatTensor, temperature: int=…)
- added:
class Wav2Vec2ConformerForAudioFrameClassification(Wav2Vec2ForAudioFrameClassification) - added:
class Wav2Vec2ConformerForCTC(Wav2Vec2ForCTC) - added:
class Wav2Vec2ConformerForPreTraining(Wav2Vec2ForPreTraining) - added:
class Wav2Vec2ConformerForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class Wav2Vec2ConformerForXVector(Wav2Vec2ForXVector) - added:
class Wav2Vec2ConformerModel(Wav2Vec2ConformerPreTrainedModel, Wav2Vec2Model) - added:
class Wav2Vec2ConformerPreTrainedModel(PreTrainedModel) - added:
constructor Wav2Vec2ConformerForAudioFrameClassification(config) - added:
constructor Wav2Vec2ConformerForCTC(config, target_lang: Optional[str]=…) - added:
constructor Wav2Vec2ConformerForPreTraining(config: Wav2Vec2ConformerConfig) - added:
constructor Wav2Vec2ConformerForSequenceClassification(config) - added:
constructor Wav2Vec2ConformerForXVector(config) - added:
constructor Wav2Vec2ConformerModel(config: Wav2Vec2ConformerConfig) - added:
field Wav2Vec2ConformerPreTrainedModel.base_model_prefix - added:
field Wav2Vec2ConformerPreTrainedModel.config: Wav2Vec2ConformerConfig - added:
field Wav2Vec2ConformerPreTrainedModel.main_input_name - added:
field Wav2Vec2ConformerPreTrainedModel.supports_gradient_checkpointing - added:
method Wav2Vec2ConformerForAudioFrameClassification.freeze_feature_extractor() - added:
method Wav2Vec2ConformerForCTC.freeze_base_model() - added:
method Wav2Vec2ConformerForCTC.freeze_feature_extractor() - added:
method Wav2Vec2ConformerForCTC.tie_weights() - added:
method Wav2Vec2ConformerForPreTraining.freeze_feature_extractor() - added:
method Wav2Vec2ConformerForSequenceClassification.freeze_feature_extractor() - added:
method Wav2Vec2ConformerForXVector.freeze_feature_extractor() - added:
method Wav2Vec2ConformerModel.freeze_feature_extractor()
- added:
export define_import_structure
- added:
class Wav2Vec2PhonemeCTCTokenizer(PreTrainedTokenizer) - added:
constructor Wav2Vec2PhonemeCTCTokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, phone_delimiter_token=…, word_delimiter_token=…, do_phonemize=…, phonemizer_lang=…, phonemizer_backend=…, **kwargs) - added:
field Wav2Vec2PhonemeCTCTokenizer.model_input_names - added:
field Wav2Vec2PhonemeCTCTokenizer.vocab_files_names - added:
method Wav2Vec2PhonemeCTCTokenizer.batch_decode(sequences: Union[list[int], list[list[int]], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_char_offsets: bool=…, **kwargs) -> list[str] - added:
method Wav2Vec2PhonemeCTCTokenizer.convert_tokens_to_string(tokens: list[str], group_tokens: bool=…, spaces_between_special_tokens: bool=…, filter_word_delimiter_token: bool=…, output_char_offsets: bool=…) -> str - added:
method Wav2Vec2PhonemeCTCTokenizer.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_char_offsets: bool=…, **kwargs) -> str - added:
method Wav2Vec2PhonemeCTCTokenizer.get_vocab() -> dict - added:
method Wav2Vec2PhonemeCTCTokenizer.init_backend(phonemizer_lang: str) - added:
method Wav2Vec2PhonemeCTCTokenizer.phonemize(text: str, phonemizer_lang: Optional[str]=…) -> str - added:
method Wav2Vec2PhonemeCTCTokenizer.prepare_for_tokenization(text: str, is_split_into_words: bool=…, phonemizer_lang: Optional[str]=…, do_phonemize: Optional[bool]=…) -> tuple[str, dict[str, Any]] - added:
method Wav2Vec2PhonemeCTCTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property Wav2Vec2PhonemeCTCTokenizer.phone_delimiter_token -> str - added:
property Wav2Vec2PhonemeCTCTokenizer.phone_delimiter_token_id -> Optional[int] - added:
property Wav2Vec2PhonemeCTCTokenizer.vocab_size -> int - added:
property Wav2Vec2PhonemeCTCTokenizer.word_delimiter_token -> str - added:
property Wav2Vec2PhonemeCTCTokenizer.word_delimiter_token_id -> Optional[int]
- added:
export define_import_structure
- added:
class Wav2Vec2ProcessorWithLM(ProcessorMixin) - added:
constructor Wav2Vec2ProcessorWithLM(feature_extractor: FeatureExtractionMixin, tokenizer: PreTrainedTokenizerBase, decoder: BeamSearchDecoderCTC) - added:
field Wav2Vec2ProcessorWithLM.feature_extractor_class - added:
field Wav2Vec2ProcessorWithLM.tokenizer_class - added:
method Wav2Vec2ProcessorWithLM.as_target_processor() - added:
method Wav2Vec2ProcessorWithLM.batch_decode(logits: np.ndarray, pool: Optional[Pool]=…, num_processes: Optional[int]=…, beam_width: Optional[int]=…, beam_prune_logp: Optional[float]=…, token_min_logp: Optional[float]=…, hotwords: Optional[Iterable[str]]=…, hotword_weight: Optional[float]=…, alpha: Optional[float]=…, beta: Optional[float]=…, unk_score_offset: Optional[float]=…, lm_score_boundary:… - added:
method Wav2Vec2ProcessorWithLM.decode(logits: np.ndarray, beam_width: Optional[int]=…, beam_prune_logp: Optional[float]=…, token_min_logp: Optional[float]=…, hotwords: Optional[Iterable[str]]=…, hotword_weight: Optional[float]=…, alpha: Optional[float]=…, beta: Optional[float]=…, unk_score_offset: Optional[float]=…, lm_score_boundary: Optional[bool]=…, output_word_offsets: bool=…, n_best: int=…) - added:
method Wav2Vec2ProcessorWithLM.pad(*args, **kwargs) - added:
method Wav2Vec2ProcessorWithLM.save_pretrained(save_directory) - added:
property Wav2Vec2ProcessorWithLM.language_model - added:
static method Wav2Vec2ProcessorWithLM.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
static method Wav2Vec2ProcessorWithLM.get_missing_alphabet_tokens(decoder, tokenizer)
- added:
export define_import_structure
- added:
class WavLMConfig(PretrainedConfig) - added:
constructor WavLMConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout=…, activation_dropout=…, attention_dropout=…, feat_proj_dropout=…, final_dropout=…, layerdrop=…, initializer_range=…, layer_norm_eps=…, feat_extract_norm=…, feat_extract_activation=…, conv_dim=…, conv_stride=…, conv_kernel=…, conv_bias=…, num_conv_pos_… - added:
field WavLMConfig.model_type - added:
property WavLMConfig.inputs_to_logits_ratio
- added:
class WavLMForAudioFrameClassification(WavLMPreTrainedModel) - added:
class WavLMForCTC(WavLMPreTrainedModel) - added:
class WavLMForSequenceClassification(WavLMPreTrainedModel) - added:
class WavLMForXVector(WavLMPreTrainedModel) - added:
class WavLMModel(WavLMPreTrainedModel) - added:
class WavLMPreTrainedModel(PreTrainedModel) - added:
constructor WavLMForAudioFrameClassification(config) - added:
constructor WavLMForCTC(config, target_lang: Optional[str]=…) - added:
constructor WavLMForSequenceClassification(config) - added:
constructor WavLMForXVector(config) - added:
constructor WavLMModel(config: WavLMConfig) - added:
field WavLMPreTrainedModel.base_model_prefix - added:
field WavLMPreTrainedModel.config: WavLMConfig - added:
field WavLMPreTrainedModel.main_input_name - added:
field WavLMPreTrainedModel.supports_gradient_checkpointing - added:
method WavLMForAudioFrameClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, TokenClassifierOutput] - added:
method WavLMForAudioFrameClassification.freeze_base_model() - added:
method WavLMForAudioFrameClassification.freeze_feature_encoder() - added:
method WavLMForAudioFrameClassification.freeze_feature_extractor() - added:
method WavLMForCTC.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, CausalLMOutput] - added:
method WavLMForCTC.freeze_base_model() - added:
method WavLMForCTC.freeze_feature_encoder() - added:
method WavLMForCTC.freeze_feature_extractor() - added:
method WavLMForCTC.tie_weights() - added:
method WavLMForSequenceClassification.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, SequenceClassifierOutput] - added:
method WavLMForSequenceClassification.freeze_base_model() - added:
method WavLMForSequenceClassification.freeze_feature_encoder() - added:
method WavLMForSequenceClassification.freeze_feature_extractor() - added:
method WavLMForXVector.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, labels: Optional[torch.Tensor]=…) -> Union[tuple, XVectorOutput] - added:
method WavLMForXVector.freeze_base_model() - added:
method WavLMForXVector.freeze_feature_encoder() - added:
method WavLMForXVector.freeze_feature_extractor() - added:
method WavLMModel.forward(input_values: Optional[torch.Tensor], attention_mask: Optional[torch.Tensor]=…, mask_time_indices: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, WavLMBaseModelOutput] - added:
method WavLMModel.freeze_feature_encoder() - added:
method WavLMModel.freeze_feature_extractor()
- added:
class WavLMForAudioFrameClassification(Wav2Vec2ForAudioFrameClassification) - added:
class WavLMForCTC(Wav2Vec2ForCTC) - added:
class WavLMForSequenceClassification(Wav2Vec2ForSequenceClassification) - added:
class WavLMForXVector(Wav2Vec2ForXVector) - added:
class WavLMModel(Wav2Vec2Model) - added:
class WavLMPreTrainedModel(PreTrainedModel, Wav2Vec2PreTrainedModel) - added:
field WavLMPreTrainedModel.base_model_prefix - added:
field WavLMPreTrainedModel.config: WavLMConfig - added:
field WavLMPreTrainedModel.main_input_name - added:
field WavLMPreTrainedModel.supports_gradient_checkpointing - added:
method WavLMPreTrainedModel.init_adapter_layers() - added:
method WavLMPreTrainedModel.load_adapter()
- added:
export define_import_structure
- added:
class WhisperConfig(PretrainedConfig) - added:
class WhisperOnnxConfig(OnnxSeq2SeqConfigWithPast) - added:
constructor WhisperConfig(vocab_size=…, num_mel_bins=…, encoder_layers=…, encoder_attention_heads=…, decoder_layers=…, decoder_attention_heads=…, decoder_ffn_dim=…, encoder_ffn_dim=…, encoder_layerdrop=…, decoder_layerdrop=…, decoder_start_token_id=…, use_cache=…, is_encoder_decoder=…, activation_function=…, d_model=…, dropout=…, attention_dropout=…, activation_dropout=…, init_std=…, scale_embedd… - added:
field WhisperConfig.attribute_map - added:
field WhisperConfig.keys_to_ignore_at_inference - added:
field WhisperConfig.model_type - added:
method WhisperOnnxConfig.generate_dummy_inputs(preprocessor: Union[PreTrainedTokenizerBase, FeatureExtractionMixin], batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, sampling_rate: int=…, time_duration: float=…, frequency: int=…) -> Mapping[str, Any] - added:
property WhisperOnnxConfig.atol_for_validation -> float - added:
property WhisperOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class BasicTextNormalizer - added:
class EnglishNumberNormalizer - added:
class EnglishSpellingNormalizer - added:
class EnglishTextNormalizer - added:
constructor BasicTextNormalizer(remove_diacritics: bool=…, split_letters: bool=…) - added:
constructor EnglishNumberNormalizer() - added:
constructor EnglishSpellingNormalizer(english_spelling_mapping) - added:
constructor EnglishTextNormalizer(english_spelling_mapping) - added:
function remove_symbols(s: str) - added:
function remove_symbols_and_diacritics(s: str, keep=…) - added:
method EnglishNumberNormalizer.postprocess(s: str) - added:
method EnglishNumberNormalizer.preprocess(s: str) - added:
method EnglishNumberNormalizer.process_words(words: list[str]) -> Iterator[str] - added:
variable ADDITIONAL_DIACRITICS
- added:
class WhisperFeatureExtractor(SequenceFeatureExtractor) - added:
constructor WhisperFeatureExtractor(feature_size=…, sampling_rate=…, hop_length=…, chunk_length=…, n_fft=…, padding_value=…, dither=…, return_attention_mask=…, **kwargs) - added:
field WhisperFeatureExtractor.model_input_names - added:
static method WhisperFeatureExtractor.zero_mean_unit_var_norm(input_values: list[np.ndarray], attention_mask: list[np.ndarray], padding_value: float=…) -> list[np.ndarray]
- added:
class WhisperGenerationMixin(GenerationMixin) - added:
method WhisperGenerationMixin.detect_language(input_features: Optional[torch.FloatTensor]=…, encoder_outputs: Optional[Union[torch.FloatTensor, BaseModelOutput]]=…, generation_config: Optional[GenerationConfig]=…, num_segment_frames: int=…) -> torch.Tensor - added:
method WhisperGenerationMixin.generate(input_features: Optional[torch.Tensor]=…, generation_config: Optional[GenerationConfig]=…, logits_processor: Optional[LogitsProcessorList]=…, stopping_criteria: Optional[StoppingCriteriaList]=…, prefix_allowed_tokens_fn: Optional[Callable[[int, torch.Tensor], list[int]]]=…, synced_gpus: bool=…, return_timestamps: Optional[bool]=…, task: Optional[str]=…, lang… - added:
method WhisperGenerationMixin.generate_with_fallback(segment_input, decoder_input_ids, cur_bsz, seek, batch_idx_map, temperatures, generation_config, logits_processor, stopping_criteria, prefix_allowed_tokens_fn, synced_gpus, return_token_timestamps, do_condition_on_prev_tokens, is_shortform, batch_size, attention_mask, kwargs) - added:
variable logger
- added:
class FlaxWhisperForAudioClassification(FlaxWhisperPreTrainedModel) - added:
class FlaxWhisperForConditionalGeneration(FlaxWhisperPreTrainedModel) - added:
class FlaxWhisperModel(FlaxWhisperPreTrainedModel) - added:
class FlaxWhisperPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxWhisperPreTrainedModel(config: WhisperConfig, input_shape: Optional[tuple[int]]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxWhisperForAudioClassification.dtype: jnp.dtype - added:
field FlaxWhisperForAudioClassification.module_class - added:
field FlaxWhisperForConditionalGeneration.dtype: jnp.dtype - added:
field FlaxWhisperForConditionalGeneration.module_class - added:
field FlaxWhisperModel.config: WhisperConfig - added:
field FlaxWhisperModel.dtype: jnp.dtype - added:
field FlaxWhisperModel.module_class - added:
field FlaxWhisperPreTrainedModel.base_model_prefix: str - added:
field FlaxWhisperPreTrainedModel.config_class - added:
field FlaxWhisperPreTrainedModel.main_input_name - added:
field FlaxWhisperPreTrainedModel.module_class: nn.Module - added:
method FlaxWhisperForAudioClassification.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict - added:
method FlaxWhisperForConditionalGeneration.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Op… - added:
method FlaxWhisperForConditionalGeneration.generate(input_features, generation_config=…, logits_processor=…, return_timestamps=…, task=…, language=…, is_multilingual=…, **kwargs) - added:
method FlaxWhisperForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, max_length, attention_mask: Optional[jax.Array]=…, decoder_attention_mask: Optional[jax.Array]=…, encoder_outputs=…, **kwargs) - added:
method FlaxWhisperForConditionalGeneration.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxWhisperPreTrainedModel.decode(decoder_input_ids, encoder_outputs, encoder_attention_mask: Optional[jnp.ndarray]=…, decoder_attention_mask: Optional[jnp.ndarray]=…, decoder_position_ids: Optional[jnp.ndarray]=…, past_key_values: Optional[dict]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[di… - added:
method FlaxWhisperPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxWhisperPreTrainedModel.encode(input_features: jnp.ndarray, attention_mask: Optional[jnp.ndarray]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, train: bool=…, params: Optional[dict]=…, dropout_rng: PRNGKey=…, **kwargs) - added:
method FlaxWhisperPreTrainedModel.init_cache(batch_size, max_length, encoder_outputs) - added:
method FlaxWhisperPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFWhisperForConditionalGeneration(TFWhisperPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFWhisperModel(TFWhisperPreTrainedModel) - added:
class TFWhisperPreTrainedModel(TFPreTrainedModel) - added:
constructor TFWhisperForConditionalGeneration(config: WhisperConfig, **kwargs) - added:
constructor TFWhisperModel(config: WhisperConfig, **kwargs) - added:
field TFWhisperForConditionalGeneration.base_model_prefix - added:
field TFWhisperPreTrainedModel.base_model_prefix - added:
field TFWhisperPreTrainedModel.config_class - added:
field TFWhisperPreTrainedModel.main_input_name - added:
method TFWhisperForConditionalGeneration.build(input_shape=…) - added:
method TFWhisperForConditionalGeneration.call(input_features: TFModelInputType | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | Non… - added:
method TFWhisperForConditionalGeneration.generate(inputs: tf.Tensor | None=…, generation_config: GenerationConfig | None=…, logits_processor: TFLogitsProcessorList | None=…, seed: list[int] | None=…, return_timestamps: bool | None=…, task: str | None=…, language: str | None=…, is_multilingual: bool | None=…, prompt_ids: tf.Tensor | None=…, return_token_timestamps=…, **kwargs) - added:
method TFWhisperForConditionalGeneration.get_decoder() - added:
method TFWhisperForConditionalGeneration.get_encoder() - added:
method TFWhisperForConditionalGeneration.get_output_embeddings() - added:
method TFWhisperForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past_key_values=…, use_cache=…, encoder_outputs=…, attention_mask=…, decoder_attention_mask=…, **kwargs) - added:
method TFWhisperForConditionalGeneration.resize_token_embeddings(new_num_tokens: int) -> keras.layers.Embedding - added:
method TFWhisperForConditionalGeneration.serving_output(output) - added:
method TFWhisperForConditionalGeneration.set_output_embeddings(value) - added:
method TFWhisperModel.build(input_shape=…) - added:
method TFWhisperModel.call(input_features: TFModelInputType | None=…, decoder_input_ids: np.ndarray | tf.Tensor | None=…, decoder_attention_mask: np.ndarray | tf.Tensor | None=…, decoder_position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, decoder_head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | None=…, encoder_output… - added:
method TFWhisperModel.decoder() - added:
method TFWhisperModel.encoder() - added:
method TFWhisperModel.get_decoder() - added:
method TFWhisperModel.get_encoder() - added:
method TFWhisperModel.get_input_embeddings() - added:
method TFWhisperModel.serving_output(output) - added:
method TFWhisperModel.set_input_embeddings(value) - added:
property TFWhisperPreTrainedModel.dummy_inputs -> dict[str, tf.Tensor] - added:
property TFWhisperPreTrainedModel.input_signature
- added:
class WhisperForAudioClassification(WhisperPreTrainedModel) - added:
class WhisperForCausalLM(WhisperPreTrainedModel, GenerationMixin) - added:
class WhisperForConditionalGeneration(WhisperGenerationMixin, WhisperPreTrainedModel) - added:
class WhisperModel(WhisperPreTrainedModel) - added:
class WhisperPreTrainedModel(PreTrainedModel) - added:
constructor WhisperForAudioClassification(config) - added:
constructor WhisperForCausalLM(config) - added:
constructor WhisperForConditionalGeneration(config: WhisperConfig) - added:
constructor WhisperModel(config: WhisperConfig) - added:
field WhisperForCausalLM.main_input_name - added:
field WhisperForConditionalGeneration.base_model_prefix - added:
field WhisperPreTrainedModel.base_model_prefix - added:
field WhisperPreTrainedModel.config: WhisperConfig - added:
field WhisperPreTrainedModel.main_input_name - added:
field WhisperPreTrainedModel.supports_gradient_checkpointing - added:
method WhisperForAudioClassification.forward(input_features: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.FloatTensor]]]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], SequenceClassifierOutput] - added:
method WhisperForAudioClassification.freeze_encoder() - added:
method WhisperForAudioClassification.get_input_embeddings() -> nn.Module - added:
method WhisperForAudioClassification.set_input_embeddings(value: nn.Module) - added:
method WhisperForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[torch.FloatTensor]]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bo… - added:
method WhisperForCausalLM.get_decoder() - added:
method WhisperForCausalLM.get_input_embeddings() -> nn.Module - added:
method WhisperForCausalLM.get_output_embeddings() - added:
method WhisperForCausalLM.set_decoder(decoder) - added:
method WhisperForCausalLM.set_input_embeddings(value) - added:
method WhisperForCausalLM.set_output_embeddings(new_embeddings) - added:
method WhisperForConditionalGeneration.forward(input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[… - added:
method WhisperForConditionalGeneration.freeze_encoder() - added:
method WhisperForConditionalGeneration.get_decoder() - added:
method WhisperForConditionalGeneration.get_encoder() - added:
method WhisperForConditionalGeneration.get_input_embeddings() -> nn.Module - added:
method WhisperForConditionalGeneration.get_output_embeddings() - added:
method WhisperForConditionalGeneration.set_output_embeddings(new_embeddings) - added:
method WhisperModel.forward(input_features: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.LongTensor]=…, decoder_input_ids: Optional[torch.LongTensor]=…, decoder_attention_mask: Optional[torch.LongTensor]=…, head_mask: Optional[torch.Tensor]=…, decoder_head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, encoder_outputs: Optional[tuple[tuple[torch.F… - added:
method WhisperModel.freeze_encoder() - added:
method WhisperModel.get_encoder() - added:
method WhisperModel.get_input_embeddings() - added:
method WhisperModel.set_input_embeddings(value)
- added:
class WhisperProcessor(ProcessorMixin) - added:
constructor WhisperProcessor(feature_extractor, tokenizer) - added:
field WhisperProcessor.feature_extractor_class - added:
field WhisperProcessor.tokenizer_class - added:
method WhisperProcessor.get_decoder_prompt_ids(task=…, language=…, no_timestamps=…) - added:
method WhisperProcessor.get_prompt_ids(text: str, return_tensors=…)
- added:
class WhisperTokenizer(PreTrainedTokenizer) - added:
constructor WhisperTokenizer(vocab_file, merges_file, normalizer_file=…, errors=…, unk_token=…, bos_token=…, eos_token=…, pad_token=…, add_prefix_space=…, language=…, task=…, predict_timestamps=…, **kwargs) - added:
field WhisperTokenizer.model_input_names - added:
field WhisperTokenizer.vocab_files_names - added:
method WhisperTokenizer.bpe(token) - added:
method WhisperTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method WhisperTokenizer.convert_tokens_to_string(tokens) - added:
method WhisperTokenizer.decode(token_ids, skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_offsets: bool=…, time_precision: float=…, decode_with_timestamps: bool=…, normalize: bool=…, basic_normalize: bool=…, remove_diacritics: bool=…, **kwargs) -> str - added:
method WhisperTokenizer.get_decoder_prompt_ids(task=…, language=…, no_timestamps=…) - added:
method WhisperTokenizer.get_prompt_ids(text: str, return_tensors=…) - added:
method WhisperTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method WhisperTokenizer.get_vocab() - added:
method WhisperTokenizer.normalize(text) - added:
method WhisperTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - added:
method WhisperTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method WhisperTokenizer.set_prefix_tokens(language: Optional[str]=…, task: Optional[str]=…, predict_timestamps: Optional[bool]=…) - added:
method WhisperTokenizer.timestamp_ids(time_precision=…) - added:
property WhisperTokenizer.prefix_tokens -> list[int] - added:
property WhisperTokenizer.vocab_size -> int - added:
static method WhisperTokenizer.basic_normalize(text, remove_diacritics=…)
- added:
class WhisperTokenizerFast(PreTrainedTokenizerFast) - added:
constructor WhisperTokenizerFast(vocab_file=…, merges_file=…, normalizer_file=…, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, language=…, task=…, predict_timestamps=…, **kwargs) - added:
field WhisperTokenizerFast.model_input_names - added:
field WhisperTokenizerFast.slow_tokenizer_class - added:
field WhisperTokenizerFast.vocab_files_names - added:
method WhisperTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> list[int] - added:
method WhisperTokenizerFast.decode(token_ids, skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, output_offsets: bool=…, time_precision: float=…, decode_with_timestamps: bool=…, normalize: bool=…, basic_normalize: bool=…, remove_diacritics: bool=…, **kwargs) -> str - added:
method WhisperTokenizerFast.get_decoder_prompt_ids(task=…, language=…, no_timestamps=…) - added:
method WhisperTokenizerFast.get_prompt_ids(text: str, return_tensors=…) - added:
method WhisperTokenizerFast.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method WhisperTokenizerFast.normalize(text) - added:
method WhisperTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
method WhisperTokenizerFast.set_prefix_tokens(language: Optional[str]=…, task: Optional[str]=…, predict_timestamps: Optional[bool]=…) - added:
method WhisperTokenizerFast.timestamp_ids(time_precision=…) - added:
property WhisperTokenizerFast.prefix_tokens -> list[int] - added:
static method WhisperTokenizerFast.basic_normalize(text, remove_diacritics=…)
- added:
export define_import_structure
- added:
class XCLIPConfig(PretrainedConfig) - added:
class XCLIPTextConfig(PretrainedConfig) - added:
class XCLIPVisionConfig(PretrainedConfig) - added:
constructor XCLIPConfig(text_config=…, vision_config=…, projection_dim=…, prompt_layers=…, prompt_alpha=…, prompt_hidden_act=…, prompt_num_attention_heads=…, prompt_attention_dropout=…, prompt_projection_dropout=…, logit_scale_init_value=…, **kwargs) - added:
constructor XCLIPTextConfig(vocab_size=…, hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, max_position_embeddings=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
constructor XCLIPVisionConfig(hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, mit_hidden_size=…, mit_intermediate_size=…, mit_num_hidden_layers=…, mit_num_attention_heads=…, num_channels=…, image_size=…, patch_size=…, num_frames=…, hidden_act=…, layer_norm_eps=…, attention_dropout=…, initializer_range=…, initializer_factor=…, drop_path_rate=…, **kwargs) - added:
field XCLIPConfig.model_type - added:
field XCLIPConfig.sub_configs - added:
field XCLIPTextConfig.base_config_key - added:
field XCLIPTextConfig.model_type - added:
field XCLIPVisionConfig.base_config_key - added:
field XCLIPVisionConfig.model_type
- added:
class XCLIPModel(XCLIPPreTrainedModel) - added:
class XCLIPPreTrainedModel(PreTrainedModel) - added:
class XCLIPTextModel(XCLIPPreTrainedModel) - added:
class XCLIPVisionModel(XCLIPPreTrainedModel) - added:
constructor XCLIPModel(config: XCLIPConfig) - added:
constructor XCLIPTextModel(config: XCLIPTextConfig) - added:
constructor XCLIPVisionModel(config: XCLIPVisionConfig) - added:
field XCLIPModel.config: XCLIPConfig - added:
field XCLIPPreTrainedModel.base_model_prefix - added:
field XCLIPPreTrainedModel.config: XCLIPConfig - added:
field XCLIPPreTrainedModel.supports_gradient_checkpointing - added:
field XCLIPTextModel.config: XCLIPTextConfig - added:
field XCLIPVisionModel.config: XCLIPVisionConfig - added:
field XCLIPVisionModel.main_input_name - added:
method XCLIPModel.forward(input_ids: Optional[torch.LongTensor]=…, pixel_values: Optional[torch.FloatTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, return_loss: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, interpolate_pos_encoding: bool=…, return_dict: Optional[bool]=…) -> Union[tuple, XCLIPOutput] - added:
method XCLIPModel.get_text_features(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…) -> torch.FloatTensor - added:
method XCLIPModel.get_video_features(pixel_values: torch.Tensor) -> torch.FloatTensor - added:
method XCLIPTextModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method XCLIPTextModel.get_input_embeddings() -> nn.Module - added:
method XCLIPTextModel.set_input_embeddings(value) - added:
method XCLIPVisionModel.forward(pixel_values: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOutputWithPooling] - added:
method XCLIPVisionModel.get_input_embeddings() -> nn.Module
- added:
class XCLIPProcessor(ProcessorMixin) - added:
constructor XCLIPProcessor(image_processor=…, tokenizer=…, **kwargs) - added:
field XCLIPProcessor.attributes - added:
field XCLIPProcessor.image_processor_class - added:
field XCLIPProcessor.tokenizer_class - added:
property XCLIPProcessor.feature_extractor - added:
property XCLIPProcessor.feature_extractor_class
- added:
export define_import_structure
- added:
class XcodecConfig(PretrainedConfig) - added:
constructor XcodecConfig(target_bandwidths: Optional[list[float]]=…, sample_rate: int=…, kernel_size: int=…, channel_ratios: list[float]=…, strides: list[int]=…, block_dilations: list[int]=…, unit_kernel_size: int=…, codebook_size: int=…, codebook_dim: Optional[int]=…, initializer_range: float=…, acoustic_model_config: Union[dict, DacConfig]=…, semantic_model_config: Union[dict, HubertConfig]=…,… - added:
field XcodecConfig.model_type - added:
field XcodecConfig.sub_configs - added:
property XcodecConfig.codebook_nbits -> int - added:
property XcodecConfig.frame_rate -> int - added:
property XcodecConfig.hidden_size -> int - added:
property XcodecConfig.hop_length -> int - added:
property XcodecConfig.num_quantizers -> int - added:
property XcodecConfig.semantic_hidden_size -> int
- added:
class XcodecModel(XcodecPreTrainedModel) - added:
class XcodecPreTrainedModel(PreTrainedAudioTokenizerBase) - added:
constructor XcodecModel(config) - added:
field XcodecPreTrainedModel.base_model_prefix - added:
field XcodecPreTrainedModel.config_class - added:
field XcodecPreTrainedModel.main_input_name - added:
method XcodecModel.decode(audio_codes: torch.Tensor, return_dict: Optional[bool]=…) -> Union[torch.Tensor, XcodecDecoderOutput] - added:
method XcodecModel.encode(input_values: torch.Tensor, bandwidth: Optional[float]=…, return_dict: Optional[bool]=…) -> Union[torch.Tensor, XcodecEncoderOutput] - added:
method XcodecModel.forward(input_values: torch.Tensor, audio_codes: Optional[torch.Tensor]=…, bandwidth: Optional[float]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor, torch.Tensor], XcodecOutput] - added:
method XcodecPreTrainedModel.apply_weight_norm() - added:
method XcodecPreTrainedModel.remove_weight_norm()
- added:
export define_import_structure
- added:
class XGLMConfig(PretrainedConfig) - added:
constructor XGLMConfig(vocab_size=…, max_position_embeddings=…, d_model=…, ffn_dim=…, num_layers=…, attention_heads=…, activation_function=…, dropout=…, attention_dropout=…, activation_dropout=…, layerdrop=…, init_std=…, scale_embedding=…, use_cache=…, decoder_start_token_id=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, **kwargs) - added:
field XGLMConfig.attribute_map - added:
field XGLMConfig.keys_to_ignore_at_inference - added:
field XGLMConfig.model_type
- added:
class FlaxXGLMForCausalLM(FlaxXGLMPreTrainedModel) - added:
class FlaxXGLMModel(FlaxXGLMPreTrainedModel) - added:
class FlaxXGLMPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxXGLMPreTrainedModel(config: XGLMConfig, input_shape: tuple[int]=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, **kwargs) - added:
field FlaxXGLMForCausalLM.module_class - added:
field FlaxXGLMModel.module_class - added:
field FlaxXGLMPreTrainedModel.base_model_prefix: str - added:
field FlaxXGLMPreTrainedModel.config_class - added:
field FlaxXGLMPreTrainedModel.module_class: nn.Module - added:
method FlaxXGLMForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxXGLMForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxXGLMPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxXGLMPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFXGLMForCausalLM(TFXGLMPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFXGLMModel(TFXGLMPreTrainedModel) - added:
class TFXGLMPreTrainedModel(TFPreTrainedModel) - added:
constructor TFXGLMForCausalLM(config: XGLMConfig, embed_tokens: TFSharedEmbeddings | None=…, *inputs: Any, **kwargs: Any) -> None - added:
constructor TFXGLMModel(config: XGLMConfig, embed_tokens: TFSharedEmbeddings | None=…, *inputs: Any, **kwargs: Any) -> None - added:
field TFXGLMForCausalLM.base_model_prefix - added:
field TFXGLMPreTrainedModel.base_model_prefix - added:
field TFXGLMPreTrainedModel.config_class - added:
method TFXGLMForCausalLM.build(input_shape=…) - added:
method TFXGLMForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[… - added:
method TFXGLMForCausalLM.prepare_inputs_for_generation(inputs, past_key_values=…, use_cache=…, **kwargs) - added:
method TFXGLMForCausalLM.tf_to_pt_weight_rename(tf_weight) - added:
method TFXGLMModel.build(input_shape=…) - added:
method TFXGLMModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, cross_attn_head_mask: np.ndarray | tf.Tensor | None=…, past_key_values: tuple[tuple[…
- added:
class XGLMForCausalLM(XGLMPreTrainedModel, GenerationMixin) - added:
class XGLMModel(XGLMPreTrainedModel) - added:
class XGLMPreTrainedModel(PreTrainedModel) - added:
constructor XGLMForCausalLM(config) - added:
constructor XGLMModel(config: XGLMConfig, embed_tokens: Optional[nn.Embedding]=…) - added:
field XGLMForCausalLM.base_model_prefix - added:
field XGLMPreTrainedModel.base_model_prefix - added:
field XGLMPreTrainedModel.config: XGLMConfig - added:
field XGLMPreTrainedModel.supports_gradient_checkpointing - added:
method XGLMForCausalLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch… - added:
method XGLMModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, cross_attn_head_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cache]=…, inputs_embeds: Optional[torch.Tenso…
- added:
class XGLMTokenizer(PreTrainedTokenizer) - added:
constructor XGLMTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field XGLMTokenizer.model_input_names - added:
field XGLMTokenizer.vocab_files_names - added:
method XGLMTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XGLMTokenizer.convert_tokens_to_string(tokens) - added:
method XGLMTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XGLMTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method XGLMTokenizer.get_vocab() - added:
method XGLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property XGLMTokenizer.vocab_size
- added:
class XGLMTokenizerFast(PreTrainedTokenizerFast) - added:
constructor XGLMTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, **kwargs) - added:
field XGLMTokenizerFast.model_input_names - added:
field XGLMTokenizerFast.slow_tokenizer_class - added:
field XGLMTokenizerFast.vocab_files_names - added:
method XGLMTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XGLMTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XGLMTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class XLMConfig(PretrainedConfig) - added:
class XLMOnnxConfig(OnnxConfig) - added:
constructor XLMConfig(vocab_size=…, emb_dim=…, n_layers=…, n_heads=…, dropout=…, attention_dropout=…, gelu_activation=…, sinusoidal_embeddings=…, causal=…, asm=…, n_langs=…, use_lang_emb=…, max_position_embeddings=…, embed_init_std=…, layer_norm_eps=…, init_std=…, bos_index=…, eos_index=…, pad_index=…, unk_index=…, mask_index=…, is_encoder=…, summary_type=…, summary_use_proj=…, summary_activation… - added:
field XLMConfig.attribute_map - added:
field XLMConfig.model_type - added:
property XLMOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class TFXLMForMultipleChoice(TFXLMPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFXLMForQuestionAnsweringSimple(TFXLMPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFXLMForSequenceClassification(TFXLMPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFXLMForTokenClassification(TFXLMPreTrainedModel, TFTokenClassificationLoss) - added:
class TFXLMMainLayer(keras.layers.Layer) - added:
class TFXLMModel(TFXLMPreTrainedModel) - added:
class TFXLMPreTrainedModel(TFPreTrainedModel) - added:
class TFXLMWithLMHeadModel(TFXLMPreTrainedModel) - added:
constructor TFXLMForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFXLMForQuestionAnsweringSimple(config, *inputs, **kwargs) - added:
constructor TFXLMForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFXLMForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFXLMMainLayer(config, **kwargs) - added:
constructor TFXLMModel(config, *inputs, **kwargs) - added:
constructor TFXLMWithLMHeadModel(config, *inputs, **kwargs) - added:
field TFXLMMainLayer.config_class - added:
field TFXLMPreTrainedModel.base_model_prefix - added:
field TFXLMPreTrainedModel.config_class - added:
method TFXLMForMultipleChoice.build(input_shape=…) - added:
method TFXLMForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np… - added:
method TFXLMForQuestionAnsweringSimple.build(input_shape=…) - added:
method TFXLMForQuestionAnsweringSimple.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_e… - added:
method TFXLMForSequenceClassification.build(input_shape=…) - added:
method TFXLMForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_em… - added:
method TFXLMForTokenClassification.build(input_shape=…) - added:
method TFXLMForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embed… - added:
method TFXLMMainLayer.build(input_shape=…) - added:
method TFXLMMainLayer.call(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) -> TFBaseModelOutput | tuple[tf.Tensor] - added:
method TFXLMMainLayer.get_input_embeddings() - added:
method TFXLMMainLayer.set_input_embeddings(value) - added:
method TFXLMModel.build(input_shape=…) - added:
method TFXLMModel.call(input_ids: TFModelInputType | None=…, attention_mask: tf.Tensor | None=…, langs: tf.Tensor | None=…, token_type_ids: tf.Tensor | None=…, position_ids: tf.Tensor | None=…, lengths: tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: tf.Tensor | None=…, inputs_embeds: tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, r… - added:
method TFXLMWithLMHeadModel.build(input_shape=…) - added:
method TFXLMWithLMHeadModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, langs: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, lengths: np.ndarray | tf.Tensor | None=…, cache: dict[str, tf.Tensor] | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.n… - added:
method TFXLMWithLMHeadModel.get_lm_head() - added:
method TFXLMWithLMHeadModel.get_prefix_bias_name() - added:
method TFXLMWithLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - added:
property TFXLMForMultipleChoice.dummy_inputs - added:
property TFXLMPreTrainedModel.dummy_inputs
- added:
class XLMForMultipleChoice(XLMPreTrainedModel) - added:
class XLMForQuestionAnswering(XLMPreTrainedModel) - added:
class XLMForQuestionAnsweringSimple(XLMPreTrainedModel) - added:
class XLMForSequenceClassification(XLMPreTrainedModel) - added:
class XLMForTokenClassification(XLMPreTrainedModel) - added:
class XLMModel(XLMPreTrainedModel) - added:
class XLMPreTrainedModel(PreTrainedModel) - added:
class XLMWithLMHeadModel(XLMPreTrainedModel, GenerationMixin) - added:
constructor XLMForMultipleChoice(config, *inputs, **kwargs) - added:
constructor XLMForQuestionAnswering(config) - added:
constructor XLMForQuestionAnsweringSimple(config) - added:
constructor XLMForSequenceClassification(config) - added:
constructor XLMForTokenClassification(config) - added:
constructor XLMModel(config) - added:
constructor XLMPreTrainedModel(*inputs, **kwargs) - added:
constructor XLMWithLMHeadModel(config) - added:
field XLMPreTrainedModel.base_model_prefix - added:
field XLMPreTrainedModel.config: XLMConfig - added:
field XLMPreTrainedModel.load_tf_weights - added:
method XLMForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Opti… - added:
method XLMForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_pos… - added:
method XLMForQuestionAnsweringSimple.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, sta… - added:
method XLMForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labe… - added:
method XLMForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels:… - added:
method XLMModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optio… - added:
method XLMModel.get_input_embeddings() - added:
method XLMModel.set_input_embeddings(new_embeddings) - added:
method XLMWithLMHeadModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, langs: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, lengths: Optional[torch.Tensor]=…, cache: Optional[dict[str, torch.Tensor]]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Option… - added:
method XLMWithLMHeadModel.get_output_embeddings() - added:
method XLMWithLMHeadModel.prepare_inputs_for_generation(input_ids, **kwargs) - added:
method XLMWithLMHeadModel.set_output_embeddings(new_embeddings) - added:
property XLMPreTrainedModel.dummy_inputs
- added:
class XLMTokenizer(PreTrainedTokenizer) - added:
constructor XLMTokenizer(vocab_file, merges_file, unk_token=…, bos_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, lang2id=…, id2lang=…, do_lowercase_and_remove_accent=…, **kwargs) - added:
field XLMTokenizer.vocab_files_names - added:
method XLMTokenizer.bpe(token) - added:
method XLMTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMTokenizer.convert_tokens_to_string(tokens) - added:
method XLMTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method XLMTokenizer.get_vocab() - added:
method XLMTokenizer.ja_tokenize(text) - added:
method XLMTokenizer.moses_pipeline(text, lang) - added:
method XLMTokenizer.moses_punct_norm(text, lang) - added:
method XLMTokenizer.moses_tokenize(text, lang) - added:
method XLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property XLMTokenizer.do_lower_case - added:
property XLMTokenizer.vocab_size
- added:
export define_import_structure
- added:
class XLMRobertaConfig(PretrainedConfig) - added:
class XLMRobertaOnnxConfig(OnnxConfig) - added:
constructor XLMRobertaConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field XLMRobertaConfig.model_type - added:
property XLMRobertaOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class FlaxXLMRobertaForCausalLM(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaForMaskedLM(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaForMultipleChoice(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaForQuestionAnswering(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaForSequenceClassification(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaForTokenClassification(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaModel(FlaxXLMRobertaPreTrainedModel) - added:
class FlaxXLMRobertaPreTrainedModel(FlaxPreTrainedModel) - added:
constructor FlaxXLMRobertaPreTrainedModel(config: XLMRobertaConfig, input_shape: tuple=…, seed: int=…, dtype: jnp.dtype=…, _do_init: bool=…, gradient_checkpointing: bool=…, **kwargs) - added:
field FlaxXLMRobertaForCausalLM.module_class - added:
field FlaxXLMRobertaForMaskedLM.module_class - added:
field FlaxXLMRobertaForMultipleChoice.module_class - added:
field FlaxXLMRobertaForQuestionAnswering.module_class - added:
field FlaxXLMRobertaForSequenceClassification.module_class - added:
field FlaxXLMRobertaForTokenClassification.module_class - added:
field FlaxXLMRobertaModel.module_class - added:
field FlaxXLMRobertaPreTrainedModel.base_model_prefix - added:
field FlaxXLMRobertaPreTrainedModel.config_class - added:
field FlaxXLMRobertaPreTrainedModel.module_class: nn.Module - added:
method FlaxXLMRobertaForCausalLM.prepare_inputs_for_generation(input_ids, max_length, attention_mask: Optional[jax.Array]=…) - added:
method FlaxXLMRobertaForCausalLM.update_inputs_for_generation(model_outputs, model_kwargs) - added:
method FlaxXLMRobertaPreTrainedModel.enable_gradient_checkpointing() - added:
method FlaxXLMRobertaPreTrainedModel.init_cache(batch_size, max_length) - added:
method FlaxXLMRobertaPreTrainedModel.init_weights(rng: jax.random.PRNGKey, input_shape: tuple, params: FrozenDict=…) -> FrozenDict
- added:
class TFXLMRobertaForCausalLM(TFXLMRobertaPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFXLMRobertaForMaskedLM(TFXLMRobertaPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFXLMRobertaForMultipleChoice(TFXLMRobertaPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFXLMRobertaForQuestionAnswering(TFXLMRobertaPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFXLMRobertaForSequenceClassification(TFXLMRobertaPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFXLMRobertaForTokenClassification(TFXLMRobertaPreTrainedModel, TFTokenClassificationLoss) - added:
class TFXLMRobertaModel(TFXLMRobertaPreTrainedModel) - added:
class TFXLMRobertaPreTrainedModel(TFPreTrainedModel) - added:
constructor TFXLMRobertaForCausalLM(config: XLMRobertaConfig, *inputs, **kwargs) - added:
constructor TFXLMRobertaForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFXLMRobertaForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFXLMRobertaForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFXLMRobertaForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFXLMRobertaForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFXLMRobertaModel(config, *inputs, **kwargs) - added:
field TFXLMRobertaPreTrainedModel.base_model_prefix - added:
field TFXLMRobertaPreTrainedModel.config_class - added:
method TFXLMRobertaForCausalLM.build(input_shape=…) - added:
method TFXLMRobertaForCausalLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarr… - added:
method TFXLMRobertaForCausalLM.get_lm_head() - added:
method TFXLMRobertaForCausalLM.get_prefix_bias_name() - added:
method TFXLMRobertaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method TFXLMRobertaForMaskedLM.build(input_shape=…) - added:
method TFXLMRobertaForMaskedLM.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict: bool… - added:
method TFXLMRobertaForMaskedLM.get_lm_head() - added:
method TFXLMRobertaForMaskedLM.get_prefix_bias_name() - added:
method TFXLMRobertaForMultipleChoice.build(input_shape=…) - added:
method TFXLMRobertaForMultipleChoice.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_dict… - added:
method TFXLMRobertaForQuestionAnswering.build(input_shape=…) - added:
method TFXLMRobertaForQuestionAnswering.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return_d… - added:
method TFXLMRobertaForSequenceClassification.build(input_shape=…) - added:
method TFXLMRobertaForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, ret… - added:
method TFXLMRobertaForTokenClassification.build(input_shape=…) - added:
method TFXLMRobertaForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, output_attentions: bool | None=…, output_hidden_states: bool | None=…, return… - added:
method TFXLMRobertaModel.build(input_shape=…) - added:
method TFXLMRobertaModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, position_ids: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np.ndarray | tf.Tensor | None=…, encoder_hidden_states: np.ndarray | tf.Tensor | None=…, encoder_attention_mask: np.ndarray | t…
- added:
class XLMRobertaForCausalLM(XLMRobertaPreTrainedModel, GenerationMixin) - added:
class XLMRobertaForMaskedLM(XLMRobertaPreTrainedModel) - added:
class XLMRobertaForMultipleChoice(XLMRobertaPreTrainedModel) - added:
class XLMRobertaForQuestionAnswering(XLMRobertaPreTrainedModel) - added:
class XLMRobertaForSequenceClassification(XLMRobertaPreTrainedModel) - added:
class XLMRobertaForTokenClassification(XLMRobertaPreTrainedModel) - added:
class XLMRobertaModel(XLMRobertaPreTrainedModel) - added:
class XLMRobertaPreTrainedModel(PreTrainedModel) - added:
constructor XLMRobertaForCausalLM(config) - added:
constructor XLMRobertaForMaskedLM(config) - added:
constructor XLMRobertaForMultipleChoice(config) - added:
constructor XLMRobertaForQuestionAnswering(config) - added:
constructor XLMRobertaForSequenceClassification(config) - added:
constructor XLMRobertaForTokenClassification(config) - added:
constructor XLMRobertaModel(config, add_pooling_layer=…) - added:
field XLMRobertaPreTrainedModel.base_model_prefix - added:
field XLMRobertaPreTrainedModel.config: XLMRobertaConfig - added:
field XLMRobertaPreTrainedModel.supports_gradient_checkpointing - added:
method XLMRobertaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.Flo… - added:
method XLMRobertaForCausalLM.get_output_embeddings() - added:
method XLMRobertaForCausalLM.set_output_embeddings(new_embeddings) - added:
method XLMRobertaForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.Flo… - added:
method XLMRobertaForMaskedLM.get_output_embeddings() - added:
method XLMRobertaForMaskedLM.set_output_embeddings(new_embeddings) - added:
method XLMRobertaForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden_s… - added:
method XLMRobertaForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTensor… - added:
method XLMRobertaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_… - added:
method XLMRobertaForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hid… - added:
method XLMRobertaModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Cach… - added:
method XLMRobertaModel.get_input_embeddings() - added:
method XLMRobertaModel.set_input_embeddings(value)
- added:
class XLMRobertaTokenizer(PreTrainedTokenizer) - added:
constructor XLMRobertaTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field XLMRobertaTokenizer.model_input_names - added:
field XLMRobertaTokenizer.vocab_files_names - added:
method XLMRobertaTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMRobertaTokenizer.convert_tokens_to_string(tokens) - added:
method XLMRobertaTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMRobertaTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method XLMRobertaTokenizer.get_vocab() - added:
method XLMRobertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property XLMRobertaTokenizer.vocab_size
- added:
class XLMRobertaTokenizerFast(PreTrainedTokenizerFast) - added:
constructor XLMRobertaTokenizerFast(vocab_file=…, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
field XLMRobertaTokenizerFast.model_input_names - added:
field XLMRobertaTokenizerFast.slow_tokenizer_class - added:
field XLMRobertaTokenizerFast.vocab_files_names - added:
method XLMRobertaTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMRobertaTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLMRobertaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export define_import_structure
- added:
class XLMRobertaXLConfig(PretrainedConfig) - added:
class XLMRobertaXLOnnxConfig(OnnxConfig) - added:
constructor XLMRobertaXLConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, **kwargs) - added:
field XLMRobertaXLConfig.model_type - added:
property XLMRobertaXLOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class XLMRobertaXLForCausalLM(XLMRobertaXLPreTrainedModel, GenerationMixin) - added:
class XLMRobertaXLForMaskedLM(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLForMultipleChoice(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLForQuestionAnswering(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLForSequenceClassification(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLForTokenClassification(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLModel(XLMRobertaXLPreTrainedModel) - added:
class XLMRobertaXLPreTrainedModel(PreTrainedModel) - added:
constructor XLMRobertaXLForCausalLM(config) - added:
constructor XLMRobertaXLForMaskedLM(config) - added:
constructor XLMRobertaXLForMultipleChoice(config) - added:
constructor XLMRobertaXLForQuestionAnswering(config) - added:
constructor XLMRobertaXLForSequenceClassification(config) - added:
constructor XLMRobertaXLForTokenClassification(config) - added:
constructor XLMRobertaXLModel(config, add_pooling_layer=…) - added:
field XLMRobertaXLPreTrainedModel.base_model_prefix - added:
field XLMRobertaXLPreTrainedModel.config: XLMRobertaXLConfig - added:
method XLMRobertaXLForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_attention_mask: Optional[torch.F… - added:
method XLMRobertaXLForCausalLM.get_output_embeddings() - added:
method XLMRobertaXLForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, **model_kwargs) - added:
method XLMRobertaXLForCausalLM.set_output_embeddings(new_embeddings) - added:
method XLMRobertaXLForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.FloatT… - added:
method XLMRobertaXLForMaskedLM.get_output_embeddings() - added:
method XLMRobertaXLForMaskedLM.set_output_embeddings(new_embeddings) - added:
method XLMRobertaXLForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions: Optional[bool]=…, output_hidden… - added:
method XLMRobertaXLForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_positions: Optional[torch.LongTens… - added:
method XLMRobertaXLForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, outpu… - added:
method XLMRobertaXLForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_h… - added:
method XLMRobertaXLModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]=…, past_key_values: Optional[Ca… - added:
method XLMRobertaXLModel.get_input_embeddings() - added:
method XLMRobertaXLModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class XLNetConfig(PretrainedConfig) - added:
constructor XLNetConfig(vocab_size=…, d_model=…, n_layer=…, n_head=…, d_inner=…, ff_activation=…, untie_r=…, attn_type=…, initializer_range=…, layer_norm_eps=…, dropout=…, mem_len=…, reuse_len=…, use_mems_eval=…, use_mems_train=…, bi_data=…, clamp_len=…, same_length=…, summary_type=…, summary_use_proj=…, summary_activation=…, summary_last_dropout=…, start_n_top=…, end_n_top=…, pad_token_id=…, bos… - added:
field XLNetConfig.attribute_map - added:
field XLNetConfig.keys_to_ignore_at_inference - added:
field XLNetConfig.model_type - added:
property XLNetConfig.max_position_embeddings
- added:
class TFXLNetForMultipleChoice(TFXLNetPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFXLNetForQuestionAnsweringSimple(TFXLNetPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFXLNetForSequenceClassification(TFXLNetPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFXLNetForTokenClassification(TFXLNetPreTrainedModel, TFTokenClassificationLoss) - added:
class TFXLNetLMHeadModel(TFXLNetPreTrainedModel, TFCausalLanguageModelingLoss) - added:
class TFXLNetMainLayer(keras.layers.Layer) - added:
class TFXLNetModel(TFXLNetPreTrainedModel) - added:
class TFXLNetPreTrainedModel(TFPreTrainedModel) - added:
constructor TFXLNetForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFXLNetForQuestionAnsweringSimple(config, *inputs, **kwargs) - added:
constructor TFXLNetForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFXLNetForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFXLNetLMHeadModel(config, *inputs, **kwargs) - added:
constructor TFXLNetMainLayer(config, **kwargs) - added:
constructor TFXLNetModel(config, *inputs, **kwargs) - added:
field TFXLNetMainLayer.config_class - added:
field TFXLNetPreTrainedModel.base_model_prefix - added:
field TFXLNetPreTrainedModel.config_class - added:
method TFXLNetForMultipleChoice.build(input_shape=…) - added:
method TFXLNetForMultipleChoice.call(input_ids: TFModelInputType | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, input… - added:
method TFXLNetForQuestionAnsweringSimple.build(input_shape=…) - added:
method TFXLNetForQuestionAnsweringSimple.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None… - added:
method TFXLNetForSequenceClassification.build(input_shape=…) - added:
method TFXLNetForSequenceClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=… - added:
method TFXLNetForTokenClassification.build(input_shape=…) - added:
method TFXLNetForTokenClassification.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…,… - added:
method TFXLNetLMHeadModel.build(input_shape=…) - added:
method TFXLNetLMHeadModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embe… - added:
method TFXLNetLMHeadModel.get_lm_head() - added:
method TFXLNetLMHeadModel.get_prefix_bias_name() - added:
method TFXLNetLMHeadModel.prepare_inputs_for_generation(inputs, past_key_values=…, use_mems=…, **kwargs) - added:
method TFXLNetMainLayer.build(input_shape=…) - added:
method TFXLNetMainLayer.cache_mem(curr_out, prev_mem) - added:
method TFXLNetMainLayer.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds… - added:
method TFXLNetMainLayer.create_mask(qlen, mlen) - added:
method TFXLNetMainLayer.get_input_embeddings() - added:
method TFXLNetMainLayer.relative_positional_encoding(qlen, klen, bsz=…) - added:
method TFXLNetMainLayer.set_input_embeddings(value) - added:
method TFXLNetModel.build(input_shape=…) - added:
method TFXLNetModel.call(input_ids: TFModelInputType | None=…, attention_mask: np.ndarray | tf.Tensor | None=…, mems: np.ndarray | tf.Tensor | None=…, perm_mask: np.ndarray | tf.Tensor | None=…, target_mapping: np.ndarray | tf.Tensor | None=…, token_type_ids: np.ndarray | tf.Tensor | None=…, input_mask: np.ndarray | tf.Tensor | None=…, head_mask: np.ndarray | tf.Tensor | None=…, inputs_embeds: np… - added:
static method TFXLNetMainLayer.positional_embedding(pos_seq, inv_freq, bsz=…)
- added:
class XLNetForMultipleChoice(XLNetPreTrainedModel) - added:
class XLNetForQuestionAnswering(XLNetPreTrainedModel) - added:
class XLNetForQuestionAnsweringSimple(XLNetPreTrainedModel) - added:
class XLNetForSequenceClassification(XLNetPreTrainedModel) - added:
class XLNetForTokenClassification(XLNetPreTrainedModel) - added:
class XLNetLMHeadModel(XLNetPreTrainedModel, GenerationMixin) - added:
class XLNetModel(XLNetPreTrainedModel) - added:
class XLNetPreTrainedModel(PreTrainedModel) - added:
constructor XLNetForMultipleChoice(config) - added:
constructor XLNetForQuestionAnswering(config) - added:
constructor XLNetForQuestionAnsweringSimple(config) - added:
constructor XLNetForSequenceClassification(config) - added:
constructor XLNetForTokenClassification(config) - added:
constructor XLNetLMHeadModel(config) - added:
constructor XLNetModel(config) - added:
field XLNetPreTrainedModel.base_model_prefix - added:
field XLNetPreTrainedModel.config: XLNetConfig - added:
field XLNetPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_xlnet(model, config, tf_path) - added:
method XLNetForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optio… - added:
method XLNetForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_posi… - added:
method XLNetForQuestionAnsweringSimple.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, star… - added:
method XLNetForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, label… - added:
method XLNetForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels:… - added:
method XLNetLMHeadModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[to… - added:
method XLNetLMHeadModel.get_output_embeddings() - added:
method XLNetLMHeadModel.prepare_inputs_for_generation(input_ids, past_key_values=…, use_mems=…, **kwargs) - added:
method XLNetLMHeadModel.set_output_embeddings(new_embeddings) - added:
method XLNetModel.cache_mem(curr_out, prev_mem) - added:
method XLNetModel.create_mask(qlen, mlen) - added:
method XLNetModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, mems: Optional[torch.Tensor]=…, perm_mask: Optional[torch.Tensor]=…, target_mapping: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, input_mask: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, use_mems: Optional[bool]=… - added:
method XLNetModel.get_input_embeddings() - added:
method XLNetModel.relative_positional_encoding(qlen, klen, bsz=…) - added:
method XLNetModel.set_input_embeddings(new_embeddings) - added:
static method XLNetModel.positional_embedding(pos_seq, inv_freq, bsz=…)
- added:
class XLNetTokenizer(PreTrainedTokenizer) - added:
constructor XLNetTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, sp_model_kwargs: Optional[dict[str, Any]]=…, **kwargs) -> None - added:
field XLNetTokenizer.padding_side - added:
field XLNetTokenizer.vocab_files_names - added:
method XLNetTokenizer.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLNetTokenizer.convert_tokens_to_string(tokens) - added:
method XLNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLNetTokenizer.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - added:
method XLNetTokenizer.get_vocab() - added:
method XLNetTokenizer.preprocess_text(inputs) - added:
method XLNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str] - added:
property XLNetTokenizer.vocab_size
- added:
class XLNetTokenizerFast(PreTrainedTokenizerFast) - added:
constructor XLNetTokenizerFast(vocab_file=…, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - added:
field XLNetTokenizerFast.padding_side - added:
field XLNetTokenizerFast.slow_tokenizer_class - added:
field XLNetTokenizerFast.vocab_files_names - added:
method XLNetTokenizerFast.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLNetTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - added:
method XLNetTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str]
- added:
export OptionalDependencyNotAvailable - added:
export define_import_structure - added:
export is_torch_available
- added:
class xLSTMConfig(PretrainedConfig) - added:
constructor xLSTMConfig(vocab_size: int=…, hidden_size: int=…, embedding_dim: Optional[int]=…, num_hidden_layers: Optional[int]=…, num_blocks: Optional[int]=…, num_heads: int=…, use_bias: bool=…, norm_reduction_force_float32: bool=…, tie_word_embeddings: bool=…, add_out_norm: bool=…, norm_eps: float=…, qk_dim_factor: float=…, v_dim_factor: float=…, chunkwise_kernel: ChunkwiseKernelType=…, sequenc… - added:
field xLSTMConfig.model_type - added:
method xLSTMConfig.to_xlstm_block_config() - added:
property xLSTMConfig.qk_dim - added:
property xLSTMConfig.qk_head_dim - added:
property xLSTMConfig.v_dim - added:
property xLSTMConfig.v_head_dim
- added:
class xLSTMForCausalLM(xLSTMPreTrainedModel, GenerationMixin) - added:
class xLSTMModel(xLSTMPreTrainedModel) - added:
class xLSTMPreTrainedModel(PreTrainedModel) - added:
constructor xLSTMForCausalLM(config) - added:
constructor xLSTMModel(config) - added:
field xLSTMPreTrainedModel.base_model_prefix - added:
field xLSTMPreTrainedModel.config_class - added:
field xLSTMPreTrainedModel.supports_gradient_checkpointing - added:
method xLSTMForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, cache_params: Optional[xLSTMCache]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> Union[tuple, xLSTMCausalLMOutput] - added:
method xLSTMForCausalLM.get_input_embeddings() - added:
method xLSTMForCausalLM.get_output_embeddings() - added:
method xLSTMForCausalLM.prepare_inputs_for_generation(input_ids, attention_mask=…, inputs_embeds=…, use_cache=…, cache_params: Optional[xLSTMCache]=…, **kwargs) - added:
method xLSTMForCausalLM.set_input_embeddings(new_embeddings) - added:
method xLSTMForCausalLM.set_output_embeddings(new_embeddings) - added:
method xLSTMModel.forward(input_ids: Optional[torch.LongTensor]=…, inputs_embeds: Optional[torch.LongTensor]=…, cache_params: Optional[xLSTMCache]=…, use_cache: Optional[bool]=…, output_hidden_states: Optional[bool]=…, **kwargs) -> Union[tuple, xLSTMOutput] - added:
method xLSTMModel.get_input_embeddings() - added:
method xLSTMModel.set_input_embeddings(new_embedding)
- added:
export define_import_structure
- added:
class XmodConfig(PretrainedConfig) - added:
class XmodOnnxConfig(OnnxConfig) - added:
constructor XmodConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, position_embedding_type=…, use_cache=…, classifier_dropout=…, pre_norm=…, adapte… - added:
field XmodConfig.model_type - added:
property XmodOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class XmodForCausalLM(XmodPreTrainedModel, GenerationMixin) - added:
class XmodForMaskedLM(XmodPreTrainedModel) - added:
class XmodForMultipleChoice(XmodPreTrainedModel) - added:
class XmodForQuestionAnswering(XmodPreTrainedModel) - added:
class XmodForSequenceClassification(XmodPreTrainedModel) - added:
class XmodForTokenClassification(XmodPreTrainedModel) - added:
class XmodModel(XmodPreTrainedModel) - added:
class XmodPreTrainedModel(PreTrainedModel) - added:
constructor XmodForCausalLM(config) - added:
constructor XmodForMaskedLM(config) - added:
constructor XmodForMultipleChoice(config) - added:
constructor XmodForQuestionAnswering(config) - added:
constructor XmodForSequenceClassification(config) - added:
constructor XmodForTokenClassification(config) - added:
constructor XmodModel(config, add_pooling_layer=…) - added:
field XmodPreTrainedModel.base_model_prefix - added:
field XmodPreTrainedModel.config: XmodConfig - added:
field XmodPreTrainedModel.supports_gradient_checkpointing - added:
method XmodForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_… - added:
method XmodForCausalLM.get_output_embeddings() - added:
method XmodForCausalLM.set_output_embeddings(new_embeddings) - added:
method XmodForMaskedLM.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, encoder_hidden_states: Optional[torch.FloatTensor]=…, encoder_… - added:
method XmodForMaskedLM.get_output_embeddings() - added:
method XmodForMaskedLM.set_output_embeddings(new_embeddings) - added:
method XmodForMultipleChoice.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, output_attentions:… - added:
method XmodForQuestionAnswering.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, end_po… - added:
method XmodForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_att… - added:
method XmodForTokenClassification.forward(input_ids: Optional[torch.LongTensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.FloatTensor]=…, token_type_ids: Optional[torch.LongTensor]=…, position_ids: Optional[torch.LongTensor]=…, head_mask: Optional[torch.FloatTensor]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, output_attent… - added:
method XmodModel.forward(input_ids: Optional[torch.Tensor]=…, lang_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, encoder_hidden_states: Optional[torch.Tensor]=…, encoder_attention_mask: Optional[torch.Tensor]… - added:
method XmodModel.get_input_embeddings() - added:
method XmodModel.set_input_embeddings(value) - added:
method XmodPreTrainedModel.freeze_embeddings_and_language_adapters() - added:
method XmodPreTrainedModel.set_default_language(language: str)
- added:
export define_import_structure
- added:
class YolosConfig(PretrainedConfig) - added:
class YolosOnnxConfig(OnnxConfig) - added:
constructor YolosConfig(hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, initializer_range=…, layer_norm_eps=…, image_size=…, patch_size=…, num_channels=…, qkv_bias=…, num_detection_tokens=…, use_mid_position_embeddings=…, auxiliary_loss=…, class_cost=…, bbox_cost=…, giou_cost=…, bbox_loss_coeffici… - added:
field YolosConfig.model_type - added:
field YolosOnnxConfig.torch_onnx_minimum_version - added:
property YolosOnnxConfig.atol_for_validation -> float - added:
property YolosOnnxConfig.default_onnx_opset -> int - added:
property YolosOnnxConfig.inputs -> Mapping[str, Mapping[int, str]]
- added:
class YolosFeatureExtractor(YolosImageProcessor) - added:
constructor YolosFeatureExtractor(*args, **kwargs) -> None
- added:
class YolosImageProcessor(BaseImageProcessor) - added:
constructor YolosImageProcessor(format: Union[str, AnnotationFormat]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_convert_annotations: Optional[bool]=…, do_pad: bool=…, pad_… - added:
field YolosImageProcessor.model_input_names - added:
method YolosImageProcessor.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method YolosImageProcessor.pad(images: list[np.ndarray], annotations: Optional[list[dict[str, Any]]]=…, constant_values: Union[float, Iterable[float]]=…, return_pixel_mask: bool=…, return_tensors: Optional[Union[str, TensorType]]=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, update_bboxes: bool=…, pad_size: Optional[dict[str, int]]=…) -… - added:
method YolosImageProcessor.post_process(outputs, target_sizes) - added:
method YolosImageProcessor.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…) - added:
method YolosImageProcessor.prepare_annotation(image: np.ndarray, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method YolosImageProcessor.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, do_resize: Optional[bool]=…, size: Optional[dict[str, int]]=…, resample=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[Union[int, float]]=…, do_normalize: Optional[b… - added:
method YolosImageProcessor.rescale(image: np.ndarray, rescale_factor: float, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
method YolosImageProcessor.resize(image: np.ndarray, size: dict[str, int], resample: PILImageResampling=…, data_format: Optional[ChannelDimension]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…, **kwargs) -> np.ndarray - added:
method YolosImageProcessor.resize_annotation(annotation, orig_size, size, resample: PILImageResampling=…) -> dict - added:
static method YolosImageProcessor.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class YolosImageProcessorFast(BaseImageProcessorFast) - added:
constructor YolosImageProcessorFast(**kwargs: Unpack[YolosFastImageProcessorKwargs]) -> None - added:
field YolosImageProcessorFast.default_to_square - added:
field YolosImageProcessorFast.do_normalize - added:
field YolosImageProcessorFast.do_pad - added:
field YolosImageProcessorFast.do_rescale - added:
field YolosImageProcessorFast.do_resize - added:
field YolosImageProcessorFast.format - added:
field YolosImageProcessorFast.image_mean - added:
field YolosImageProcessorFast.image_std - added:
field YolosImageProcessorFast.model_input_names - added:
field YolosImageProcessorFast.resample - added:
field YolosImageProcessorFast.size - added:
field YolosImageProcessorFast.valid_kwargs - added:
method YolosImageProcessorFast.normalize_annotation(annotation: dict, image_size: tuple[int, int]) -> dict - added:
method YolosImageProcessorFast.pad(image: torch.Tensor, padded_size: tuple[int, int], annotation: Optional[dict[str, Any]]=…, update_bboxes: bool=…, fill: int=…) - added:
method YolosImageProcessorFast.post_process(outputs, target_sizes) - added:
method YolosImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method YolosImageProcessorFast.prepare_annotation(image: torch.Tensor, target: dict, format: Optional[AnnotationFormat]=…, return_segmentation_masks: Optional[bool]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> dict - added:
method YolosImageProcessorFast.preprocess(images: ImageInput, annotations: Optional[Union[AnnotationType, list[AnnotationType]]]=…, masks_path: Optional[Union[str, pathlib.Path]]=…, **kwargs: Unpack[YolosFastImageProcessorKwargs]) -> BatchFeature - added:
method YolosImageProcessorFast.resize(image: torch.Tensor, size: SizeDict, interpolation: Optional[F.InterpolationMode]=…, **kwargs) -> torch.Tensor - added:
method YolosImageProcessorFast.resize_annotation(annotation: dict[str, Any], orig_size: tuple[int, int], target_size: tuple[int, int], threshold: float=…, interpolation: Optional[F.InterpolationMode]=…) - added:
static method YolosImageProcessorFast.from_dict(image_processor_dict: dict[str, Any], **kwargs)
- added:
class YolosForObjectDetection(YolosPreTrainedModel) - added:
class YolosModel(YolosPreTrainedModel) - added:
class YolosPreTrainedModel(PreTrainedModel) - added:
constructor YolosForObjectDetection(config: YolosConfig) - added:
constructor YolosModel(config: YolosConfig, add_pooling_layer: bool=…) - added:
field YolosPreTrainedModel.base_model_prefix - added:
field YolosPreTrainedModel.config: YolosConfig - added:
field YolosPreTrainedModel.main_input_name - added:
field YolosPreTrainedModel.supports_gradient_checkpointing - added:
method YolosForObjectDetection.forward(pixel_values: torch.FloatTensor, labels: Optional[list[dict]]=…, **kwargs: Unpack[TransformersKwargs]) -> YolosObjectDetectionOutput - added:
method YolosModel.forward(pixel_values: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, **kwargs: Unpack[TransformersKwargs]) -> BaseModelOutputWithPooling - added:
method YolosModel.get_input_embeddings() -> YolosPatchEmbeddings
- added:
class YolosImageProcessorFast(DetrImageProcessorFast) - added:
method YolosImageProcessorFast.post_process(outputs, target_sizes) - added:
method YolosImageProcessorFast.post_process_instance() - added:
method YolosImageProcessorFast.post_process_instance_segmentation() - added:
method YolosImageProcessorFast.post_process_object_detection(outputs, threshold: float=…, target_sizes: Union[TensorType, list[tuple]]=…, top_k: int=…) - added:
method YolosImageProcessorFast.post_process_panoptic() - added:
method YolosImageProcessorFast.post_process_panoptic_segmentation() - added:
method YolosImageProcessorFast.post_process_segmentation() - added:
method YolosImageProcessorFast.post_process_semantic_segmentation()
- added:
export define_import_structure
- added:
class YosoConfig(PretrainedConfig) - added:
constructor YosoConfig(vocab_size=…, hidden_size=…, num_hidden_layers=…, num_attention_heads=…, intermediate_size=…, hidden_act=…, hidden_dropout_prob=…, attention_probs_dropout_prob=…, max_position_embeddings=…, type_vocab_size=…, initializer_range=…, layer_norm_eps=…, position_embedding_type=…, use_expectation=…, hash_code_len=…, num_hash=…, conv_window=…, use_fast_hash=…, lsh_backward=…, pad_t… - added:
field YosoConfig.model_type
- added:
class YosoForMaskedLM(YosoPreTrainedModel) - added:
class YosoForMultipleChoice(YosoPreTrainedModel) - added:
class YosoForQuestionAnswering(YosoPreTrainedModel) - added:
class YosoForSequenceClassification(YosoPreTrainedModel) - added:
class YosoForTokenClassification(YosoPreTrainedModel) - added:
class YosoLayer(GradientCheckpointingLayer) - added:
class YosoModel(YosoPreTrainedModel) - added:
class YosoPreTrainedModel(PreTrainedModel) - added:
constructor YosoForMaskedLM(config) - added:
constructor YosoForMultipleChoice(config) - added:
constructor YosoForQuestionAnswering(config) - added:
constructor YosoForSequenceClassification(config) - added:
constructor YosoForTokenClassification(config) - added:
constructor YosoLayer(config) - added:
constructor YosoModel(config) - added:
field YosoPreTrainedModel.base_model_prefix - added:
field YosoPreTrainedModel.config: YosoConfig - added:
field YosoPreTrainedModel.supports_gradient_checkpointing - added:
method YosoForMaskedLM.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optio… - added:
method YosoForMaskedLM.get_output_embeddings() - added:
method YosoForMaskedLM.set_output_embeddings(new_embeddings) - added:
method YosoForMultipleChoice.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict:… - added:
method YosoForQuestionAnswering.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, start_positions: Optional[torch.Tensor]=…, end_positions: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…,… - added:
method YosoForSequenceClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, retu… - added:
method YosoForTokenClassification.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, labels: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_… - added:
method YosoLayer.feed_forward_chunk(attention_output) - added:
method YosoLayer.forward(hidden_states, attention_mask=…, output_attentions=…) - added:
method YosoModel.forward(input_ids: Optional[torch.Tensor]=…, attention_mask: Optional[torch.Tensor]=…, token_type_ids: Optional[torch.Tensor]=…, position_ids: Optional[torch.Tensor]=…, head_mask: Optional[torch.Tensor]=…, inputs_embeds: Optional[torch.Tensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple, BaseModelOu… - added:
method YosoModel.get_input_embeddings() - added:
method YosoModel.set_input_embeddings(value)
- added:
export define_import_structure
- added:
class ZambaConfig(PretrainedConfig) - added:
constructor ZambaConfig(vocab_size=…, tie_word_embeddings=…, hidden_size=…, attention_hidden_size=…, intermediate_size=…, num_hidden_layers=…, num_attention_heads=…, attention_head_dim=…, num_key_value_heads=…, n_mamba_heads=…, hidden_act=…, hidden_mamba_act=…, initializer_range=…, rms_norm_eps=…, use_cache=…, num_logits_to_keep=…, pad_token_id=…, bos_token_id=…, eos_token_id=…, max_position_embe… - added:
field ZambaConfig.keys_to_ignore_at_inference - added:
field ZambaConfig.model_type
- added:
class ZambaForCausalLM(ZambaPreTrainedModel, GenerationMixin) - added:
class ZambaForSequenceClassification(ZambaPreTrainedModel) - added:
class ZambaModel(ZambaPreTrainedModel) - added:
class ZambaPreTrainedModel(PreTrainedModel) - added:
constructor ZambaForCausalLM(config: ZambaConfig) - added:
constructor ZambaForSequenceClassification(config) - added:
constructor ZambaModel(config: ZambaConfig) - added:
field ZambaPreTrainedModel.base_model_prefix - added:
field ZambaPreTrainedModel.config: ZambaConfig - added:
field ZambaPreTrainedModel.supports_gradient_checkpointing - added:
method ZambaForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[ZambaHybridDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]… - added:
method ZambaForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method ZambaForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_h… - added:
method ZambaModel.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[ZambaHybridDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_pos…
- added:
export define_import_structure
- added:
class Zamba2Config(PretrainedConfig) - added:
constructor Zamba2Config(vocab_size=…, max_position_embeddings=…, hidden_size=…, num_hidden_layers=…, layers_block_type=…, mamba_d_state=…, mamba_d_conv=…, mamba_expand=…, mamba_ngroups=…, time_step_min=…, time_step_max=…, time_step_floor=…, time_step_limit=…, n_mamba_heads=…, use_conv_bias=…, chunk_size=…, use_mem_eff_path=…, add_bias_linear=…, intermediate_size=…, hidden_act=…, num_attention_he… - added:
field Zamba2Config.attribute_map - added:
field Zamba2Config.keys_to_ignore_at_inference - added:
field Zamba2Config.model_type
- added:
class Zamba2ForCausalLM(Zamba2PreTrainedModel, GenerationMixin) - added:
class Zamba2ForSequenceClassification(Zamba2PreTrainedModel) - added:
class Zamba2Model(Zamba2PreTrainedModel) - added:
class Zamba2PreTrainedModel(PreTrainedModel) - added:
constructor Zamba2ForCausalLM(config: Zamba2Config) - added:
constructor Zamba2ForSequenceClassification(config) - added:
constructor Zamba2Model(config: Zamba2Config) - added:
field Zamba2PreTrainedModel.base_model_prefix - added:
field Zamba2PreTrainedModel.config: Zamba2Config - added:
field Zamba2PreTrainedModel.supports_gradient_checkpointing - added:
method Zamba2ForCausalLM.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Zamba2HybridDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[boo… - added:
method Zamba2ForCausalLM.prepare_inputs_for_generation(input_ids, past_key_values=…, attention_mask=…, inputs_embeds=…, cache_position=…, position_ids=…, use_cache=…, **kwargs) - added:
method Zamba2ForSequenceClassification.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Union[Cache, list[torch.FloatTensor]]]=…, inputs_embeds: Optional[torch.FloatTensor]=…, labels: Optional[torch.LongTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_… - added:
method Zamba2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Zamba2HybridDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_p… - added:
method Zamba2Model.get_layers(blocks, linear_layers, mamba_layers)
- added:
class Zamba2ForCausalLM(ZambaForCausalLM) - added:
class Zamba2ForSequenceClassification(ZambaForSequenceClassification) - added:
class Zamba2Model(ZambaModel, Zamba2PreTrainedModel) - added:
class Zamba2PreTrainedModel(PreTrainedModel) - added:
constructor Zamba2Model(config: Zamba2Config) - added:
field Zamba2PreTrainedModel.base_model_prefix - added:
field Zamba2PreTrainedModel.config: Zamba2Config - added:
field Zamba2PreTrainedModel.supports_gradient_checkpointing - added:
method Zamba2Model.forward(input_ids: Optional[torch.LongTensor]=…, attention_mask: Optional[torch.Tensor]=…, position_ids: Optional[torch.LongTensor]=…, past_key_values: Optional[Zamba2HybridDynamicCache]=…, inputs_embeds: Optional[torch.FloatTensor]=…, use_cache: Optional[bool]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…, cache_p… - added:
method Zamba2Model.get_layers(blocks, linear_layers, mamba_layers)
- added:
export define_import_structure
- added:
class ZoeDepthConfig(PretrainedConfig) - added:
constructor ZoeDepthConfig(backbone_config=…, backbone=…, use_pretrained_backbone=…, backbone_kwargs=…, hidden_act=…, initializer_range=…, batch_norm_eps=…, readout_type=…, reassemble_factors=…, neck_hidden_sizes=…, fusion_hidden_size=…, head_in_index=…, use_batch_norm_in_fusion_residual=…, use_bias_in_fusion_residual=…, num_relative_features=…, add_projection=…, bottleneck_features=…, num_attrac… - added:
field ZoeDepthConfig.model_type - added:
property ZoeDepthConfig.sub_configs - added:
variable ZOEDEPTH_PRETRAINED_CONFIG_ARCHIVE_MAP
- added:
class ZoeDepthImageProcessor(BaseImageProcessor) - added:
constructor ZoeDepthImageProcessor(do_pad: bool=…, do_rescale: bool=…, rescale_factor: Union[int, float]=…, do_normalize: bool=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_resize: bool=…, size: Optional[dict[str, int]]=…, resample: PILImageResampling=…, keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, **kwargs) -> None - added:
field ZoeDepthImageProcessor.model_input_names - added:
method ZoeDepthImageProcessor.pad_image(image: np.ndarray, mode: PaddingMode=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) - added:
method ZoeDepthImageProcessor.post_process_depth_estimation(outputs: ZoeDepthDepthEstimatorOutput, source_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…, outputs_flipped: Optional[Union[ZoeDepthDepthEstimatorOutput, None]]=…, do_remove_padding: Optional[Union[bool, None]]=…) -> list[dict[str, TensorType]] - added:
method ZoeDepthImageProcessor.preprocess(images: ImageInput, do_pad: Optional[bool]=…, do_rescale: Optional[bool]=…, rescale_factor: Optional[float]=…, do_normalize: Optional[bool]=…, image_mean: Optional[Union[float, list[float]]]=…, image_std: Optional[Union[float, list[float]]]=…, do_resize: Optional[bool]=…, size: Optional[int]=…, keep_aspect_ratio: Optional[bool]=…, ensure_multiple_of: Optio… - added:
method ZoeDepthImageProcessor.resize(image: np.ndarray, size: dict[str, int], keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, resample: PILImageResampling=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray
- added:
class ZoeDepthImageProcessorFast(BaseImageProcessorFast) - added:
constructor ZoeDepthImageProcessorFast(**kwargs: Unpack[ZoeDepthFastImageProcessorKwargs]) -> None - added:
field ZoeDepthImageProcessorFast.do_normalize - added:
field ZoeDepthImageProcessorFast.do_pad - added:
field ZoeDepthImageProcessorFast.do_rescale - added:
field ZoeDepthImageProcessorFast.do_resize - added:
field ZoeDepthImageProcessorFast.ensure_multiple_of - added:
field ZoeDepthImageProcessorFast.image_mean - added:
field ZoeDepthImageProcessorFast.image_std - added:
field ZoeDepthImageProcessorFast.keep_aspect_ratio - added:
field ZoeDepthImageProcessorFast.resample - added:
field ZoeDepthImageProcessorFast.size - added:
field ZoeDepthImageProcessorFast.valid_kwargs - added:
method ZoeDepthImageProcessorFast.post_process_depth_estimation(outputs: ZoeDepthDepthEstimatorOutput, source_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…, target_sizes: Optional[Union[TensorType, list[tuple[int, int]], None]]=…, outputs_flipped: Optional[Union[ZoeDepthDepthEstimatorOutput, None]]=…, do_remove_padding: Optional[Union[bool, None]]=…) -> list[dict[str, TensorTy… - added:
method ZoeDepthImageProcessorFast.preprocess(images: ImageInput, **kwargs: Unpack[ZoeDepthFastImageProcessorKwargs]) -> BatchFeature - added:
method ZoeDepthImageProcessorFast.resize(images: torch.Tensor, size: SizeDict, keep_aspect_ratio: bool=…, ensure_multiple_of: int=…, interpolation: Optional[F.InterpolationMode]=…) -> torch.Tensor
- added:
class ZoeDepthForDepthEstimation(ZoeDepthPreTrainedModel) - added:
class ZoeDepthPreTrainedModel(PreTrainedModel) - added:
constructor ZoeDepthForDepthEstimation(config) - added:
field ZoeDepthPreTrainedModel.base_model_prefix - added:
field ZoeDepthPreTrainedModel.config: ZoeDepthConfig - added:
field ZoeDepthPreTrainedModel.main_input_name - added:
field ZoeDepthPreTrainedModel.supports_gradient_checkpointing - added:
method ZoeDepthForDepthEstimation.forward(pixel_values: torch.FloatTensor, labels: Optional[torch.LongTensor]=…, output_attentions: Optional[bool]=…, output_hidden_states: Optional[bool]=…, return_dict: Optional[bool]=…) -> Union[tuple[torch.Tensor], DepthEstimatorOutput]
- added:
export EXTERNAL_DATA_FORMAT_SIZE_LIMIT - added:
export FeaturesManager - added:
export OnnxConfig - added:
export OnnxConfigWithPast - added:
export OnnxSeq2SeqConfigWithPast - added:
export ParameterFormat - added:
export PatchingSpec - added:
export compute_serialized_parameters_size - added:
export export - added:
export validate_model_outputs
- added:
class OnnxConfig(ABC) - added:
class OnnxConfigWithPast(OnnxConfig, ABC) - added:
class OnnxSeq2SeqConfigWithPast(OnnxConfigWithPast) - added:
class PatchingSpec - added:
constructor OnnxConfig(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…) - added:
constructor OnnxConfigWithPast(config: PretrainedConfig, task: str=…, patching_specs: Optional[list[PatchingSpec]]=…, use_past: bool=…) - added:
field OnnxConfig.default_fixed_batch - added:
field OnnxConfig.default_fixed_num_choices - added:
field OnnxConfig.default_fixed_sequence - added:
field OnnxConfig.torch_onnx_minimum_version - added:
field PatchingSpec.custom_op: Callable - added:
field PatchingSpec.name: str - added:
field PatchingSpec.o: Any - added:
field PatchingSpec.op_wrapper: Optional[Callable] - added:
field PatchingSpec.orig_op: Optional[Callable] - added:
method OnnxConfig.generate_dummy_inputs(preprocessor: Union[PreTrainedTokenizerBase, FeatureExtractionMixin, ImageProcessingMixin], batch_size: int=…, seq_length: int=…, num_choices: int=…, is_pair: bool=…, framework: Optional[TensorType]=…, num_channels: int=…, image_width: int=…, image_height: int=…, sampling_rate: int=…, time_duration: float=…, frequency: int=…, tokenizer: Optional[PreTrainedT… - added:
method OnnxConfig.generate_dummy_inputs_onnxruntime(reference_model_inputs: Mapping[str, Any]) -> Mapping[str, Any] - added:
method OnnxConfig.patch_ops() - added:
method OnnxConfig.restore_ops() - added:
method OnnxConfigWithPast.fill_with_past_key_values_(inputs_or_outputs: Mapping[str, Mapping[int, str]], direction: str, inverted_values_shape: bool=…) - added:
method OnnxConfigWithPast.flatten_output_collection_property(name: str, field: Iterable[Any]) -> dict[str, Any] - added:
method OnnxConfigWithPast.generate_dummy_inputs(tokenizer: PreTrainedTokenizerBase, batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
method OnnxSeq2SeqConfigWithPast.fill_with_past_key_values_(inputs_or_outputs: Mapping[str, Mapping[int, str]], direction: str) - added:
method OnnxSeq2SeqConfigWithPast.generate_dummy_inputs(tokenizer: Optional[PreTrainedTokenizerBase], batch_size: int=…, seq_length: int=…, is_pair: bool=…, framework: Optional[TensorType]=…) -> Mapping[str, Any] - added:
property OnnxConfig.atol_for_validation -> float - added:
property OnnxConfig.default_batch_size -> int - added:
property OnnxConfig.default_num_choices -> int - added:
property OnnxConfig.default_onnx_opset -> int - added:
property OnnxConfig.default_sequence_length -> int - added:
property OnnxConfig.inputs -> Mapping[str, Mapping[int, str]] - added:
property OnnxConfig.is_torch_support_available -> bool - added:
property OnnxConfig.outputs -> Mapping[str, Mapping[int, str]] - added:
property OnnxConfig.values_override -> Optional[Mapping[str, Any]] - added:
property OnnxConfigWithPast.num_attention_heads -> int - added:
property OnnxConfigWithPast.num_layers -> int - added:
property OnnxConfigWithPast.outputs -> Mapping[str, Mapping[int, str]] - added:
property OnnxConfigWithPast.values_override -> Optional[Mapping[str, Any]] - added:
property OnnxSeq2SeqConfigWithPast.num_attention_heads -> tuple[int, ...] - added:
property OnnxSeq2SeqConfigWithPast.num_layers -> tuple[int, ...] - added:
property OnnxSeq2SeqConfigWithPast.outputs -> Mapping[str, Mapping[int, str]] - added:
static method OnnxConfig.flatten_output_collection_property(name: str, field: Iterable[Any]) -> dict[str, Any] - added:
static method OnnxConfig.from_model_config(config: PretrainedConfig, task: str=…) -> OnnxConfig - added:
static method OnnxConfig.use_external_data_format(num_parameters: int) -> bool - added:
static method OnnxConfigWithPast.with_past(config: PretrainedConfig, task: str=…) -> OnnxConfigWithPast - added:
variable DEFAULT_ONNX_OPSET - added:
variable EXTERNAL_DATA_FORMAT_SIZE_LIMIT - added:
variable logger
- added:
function check_onnxruntime_requirements(minimum_version: Version) - added:
function ensure_model_and_config_inputs_match(model: Union[PreTrainedModel, TFPreTrainedModel], model_inputs: Iterable[str]) -> tuple[bool, list[str]] - added:
function export(preprocessor: Union[PreTrainedTokenizer, FeatureExtractionMixin, ProcessorMixin], model: Union[PreTrainedModel, TFPreTrainedModel], config: OnnxConfig, opset: int, output: Path, tokenizer: Optional[PreTrainedTokenizer]=…, device: str=…) -> tuple[list[str], list[str]] - added:
function export_pytorch(preprocessor: Union[PreTrainedTokenizer, FeatureExtractionMixin, ProcessorMixin], model: PreTrainedModel, config: OnnxConfig, opset: int, output: Path, tokenizer: Optional[PreTrainedTokenizer]=…, device: str=…) -> tuple[list[str], list[str]] - added:
function export_tensorflow(preprocessor: Union[PreTrainedTokenizer, FeatureExtractionMixin], model: TFPreTrainedModel, config: OnnxConfig, opset: int, output: Path, tokenizer: Optional[PreTrainedTokenizer]=…) -> tuple[list[str], list[str]] - added:
function validate_model_outputs(config: OnnxConfig, preprocessor: Union[PreTrainedTokenizer, FeatureExtractionMixin, ProcessorMixin], reference_model: Union[PreTrainedModel, TFPreTrainedModel], onnx_model: Path, onnx_named_outputs: list[str], atol: float, tokenizer: Optional[PreTrainedTokenizer]=…) - added:
variable ORT_QUANTIZE_MINIMUM_VERSION - added:
variable logger
- added:
class FeaturesManager - added:
field FeaturesManager.AVAILABLE_FEATURES - added:
function supported_features_mapping(*supported_features: str, onnx_config_cls: Optional[str]=…) -> dict[str, Callable[[PretrainedConfig], OnnxConfig]] - added:
method FeaturesManager.get_config(feature: str) -> OnnxConfig - added:
static method FeaturesManager.check_supported_model_or_raise(model: Union[PreTrainedModel, TFPreTrainedModel], feature: str=…) -> tuple[str, Callable] - added:
static method FeaturesManager.determine_framework(model: str, framework: Optional[str]=…) -> str - added:
static method FeaturesManager.feature_to_task(feature: str) -> str - added:
static method FeaturesManager.get_model_class_for_feature(feature: str, framework: str=…) -> type - added:
static method FeaturesManager.get_model_from_feature(feature: str, model: str, framework: Optional[str]=…, cache_dir: Optional[str]=…) -> Union[PreTrainedModel, TFPreTrainedModel] - added:
static method FeaturesManager.get_supported_features_for_model_type(model_type: str, model_name: Optional[str]=…) -> dict[str, Callable[[PretrainedConfig], OnnxConfig]] - added:
variable logger
- added:
class ParameterFormat(Enum) - added:
field ParameterFormat.Float - added:
function compute_effective_axis_dimension(dimension: int, fixed_dimension: int, num_token_to_add: int=…) -> int - added:
function compute_serialized_parameters_size(num_parameters: int, dtype: ParameterFormat) -> int - added:
function get_preprocessor(model_name: str) -> Optional[Union[AutoTokenizer, AutoFeatureExtractor, AutoProcessor]] - added:
property ParameterFormat.size -> int
- removed:
class AdamW(Optimizer) - removed:
constructor AdamW(params: Iterable[torch.nn.parameter.Parameter], lr: float=…, betas: Tuple[float, float]=…, eps: float=…, weight_decay: float=…, correct_bias: bool=…) - removed:
method AdamW.step(closure: Callable=…) - added:
class AdafactorSchedule(LambdaLR) - added:
constructor AdafactorSchedule(optimizer, initial_lr=…) - added:
function get_adafactor_schedule(optimizer, initial_lr=…) - added:
function get_cosine_with_min_lr_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: float=…, last_epoch: int=…, min_lr: Optional[float]=…, min_lr_rate: Optional[float]=…) - added:
function get_cosine_with_min_lr_schedule_with_warmup_lr_rate(optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: float=…, last_epoch: int=…, min_lr: Optional[float]=…, min_lr_rate: Optional[float]=…, warmup_lr_rate: Optional[float]=…) - added:
function get_inverse_sqrt_schedule(optimizer: Optimizer, num_warmup_steps: int, timescale: Optional[int]=…, last_epoch: int=…) - added:
function get_reduce_on_plateau_schedule(optimizer: Optimizer, **kwargs) - added:
function get_scheduler(name: Union[str, SchedulerType], optimizer: Optimizer, num_warmup_steps: Optional[int]=…, num_training_steps: Optional[int]=…, scheduler_specific_kwargs: Optional[dict]=…) - added:
function get_wsd_schedule(optimizer: Optimizer, num_warmup_steps: int, num_decay_steps: int, num_training_steps: Optional[int]=…, num_stable_steps: Optional[int]=…, warmup_type: str=…, decay_type: str=…, min_lr_ratio: float=…, num_cycles: float=…, last_epoch: int=…) - added:
method AdafactorSchedule.get_lr() - added:
variable TYPE_TO_SCHEDULER_FUNCTION
- changed:
class AdamWeightDecay(tf.keras.optimizers.Adam) → class AdamWeightDecay(Adam) - changed:
class GradientAccumulator(object) → class GradientAccumulator - changed:
class WarmUp(tf.keras.optimizers.schedules.LearningRateSchedule) → class WarmUp(schedules.LearningRateSchedule) - changed:
constructor AdamWeightDecay(learning_rate: Union[float, tf.keras.optimizers.schedules.LearningRateSchedule]=…, beta_1: float=…, beta_2: float=…, epsilon: float=…, amsgrad: bool=…, weight_decay_rate: float=…, include_in_weight_decay: Optional[List[str]]=…, exclude_from_weight_decay: Optional[List[str]]=…, name: str=…, **kwargs) → constructor AdamWeightDecay(learning_rate: Union[float, schedules.LearningRateSchedule]=…, beta_1: float=…, beta_2: float=…, epsilon: float=…, amsgrad: bool=…, weight_decay_rate: float=…, include_in_weight_decay: Optional[list[str]]=…, exclude_from_weight_decay: Optional[list[str]]=…, name: str=…, **kwargs) - changed:
constructor WarmUp(initial_learning_rate: float, decay_schedule_fn: Callable, warmup_steps: int, power: float=…, name: str=…) → constructor WarmUp(initial_learning_rate: float, decay_schedule_fn: Callable, warmup_steps: int, power: float=…, name: Optional[str]=…) - changed:
function create_optimizer(init_lr: float, num_train_steps: int, num_warmup_steps: int, min_lr_ratio: float=…, adam_beta1: float=…, adam_beta2: float=…, adam_epsilon: float=…, weight_decay_rate: float=…, power: float=…, include_in_weight_decay: Optional[List[str]]=…) → function create_optimizer(init_lr: float, num_train_steps: int, num_warmup_steps: int, min_lr_ratio: float=…, adam_beta1: float=…, adam_beta2: float=…, adam_epsilon: float=…, adam_clipnorm: Optional[float]=…, adam_global_clipnorm: Optional[float]=…, weight_decay_rate: float=…, power: float=…, include_in_weight_decay: Optional[list[str]]=…) - added:
variable schedules
- removed:
class ArgumentHandler(ABC) - removed:
class Conversation - removed:
class ConversationalPipeline(Pipeline) - removed:
class CsvPipelineDataFormat(PipelineDataFormat) - removed:
class FeatureExtractionPipeline(Pipeline) - removed:
class FillMaskPipeline(Pipeline) - removed:
class JsonPipelineDataFormat(PipelineDataFormat) - removed:
class PipedPipelineDataFormat(PipelineDataFormat) - removed:
class Pipeline(_ScikitCompat) - removed:
class PipelineDataFormat - removed:
class PipelineException(Exception) - removed:
class QuestionAnsweringArgumentHandler(ArgumentHandler) - removed:
class QuestionAnsweringPipeline(Pipeline) - removed:
class SummarizationPipeline(Pipeline) - removed:
class Text2TextGenerationPipeline(Pipeline) - removed:
class TextClassificationPipeline(Pipeline) - removed:
class TextGenerationPipeline(Pipeline) - removed:
class TokenClassificationArgumentHandler(ArgumentHandler) - removed:
class TokenClassificationPipeline(Pipeline) - removed:
class TranslationPipeline(Pipeline) - removed:
class ZeroShotClassificationArgumentHandler(ArgumentHandler) - removed:
class ZeroShotClassificationPipeline(Pipeline) - removed:
constructor Conversation(text: str=…, conversation_id: UUID=…) - removed:
constructor ConversationalPipeline(min_length_for_response=…, *args, **kwargs) - removed:
constructor CsvPipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) - removed:
constructor FeatureExtractionPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, args_parser: ArgumentHandler=…, device: int=…, task: str=…) - removed:
constructor FillMaskPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, args_parser: ArgumentHandler=…, device: int=…, top_k=…, task: str=…, **kwargs) - removed:
constructor JsonPipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) - removed:
constructor Pipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, task: str=…, args_parser: ArgumentHandler=…, device: int=…, binary_output: bool=…) - removed:
constructor PipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite: bool=…) - removed:
constructor PipelineException(task: str, model: str, reason: str) - removed:
constructor QuestionAnsweringPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, device: int=…, task: str=…, **kwargs) - removed:
constructor SummarizationPipeline(*args, **kwargs) - removed:
constructor Text2TextGenerationPipeline(*args, **kwargs) - removed:
constructor TextClassificationPipeline(return_all_scores: bool=…, **kwargs) - removed:
constructor TextGenerationPipeline(*args, **kwargs) - removed:
constructor TokenClassificationPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, args_parser: ArgumentHandler=…, device: int=…, binary_output: bool=…, ignore_labels=…, task: str=…, grouped_entities: bool=…, ignore_subwords: bool=…) - removed:
constructor TranslationPipeline(*args, **kwargs) - removed:
constructor ZeroShotClassificationPipeline(args_parser=…, *args, **kwargs) - removed:
field Pipeline.default_input_names - removed:
field PipelineDataFormat.SUPPORTED_FORMATS - removed:
field QuestionAnsweringPipeline.default_input_names - removed:
field TextGenerationPipeline.ALLOWED_MODELS - removed:
field TextGenerationPipeline.XL_PREFIX - removed:
field TokenClassificationPipeline.default_input_names - removed:
function get_default_model(targeted_task: Dict, framework: Optional[str], task_options: Optional[Any]) -> str - removed:
function get_framework(model, revision: Optional[str]=…) - removed:
method Conversation.add_user_input(text: str, overwrite: bool=…) - removed:
method Conversation.append_response(response: str) - removed:
method Conversation.mark_processed() - removed:
method Conversation.set_history(history: List[int]) - removed:
method CsvPipelineDataFormat.save(data: List[dict]) - removed:
method FillMaskPipeline.ensure_exactly_one_mask_token(masked_index: np.ndarray) - removed:
method JsonPipelineDataFormat.save(data: dict) - removed:
method PipedPipelineDataFormat.save(data: dict) - removed:
method PipedPipelineDataFormat.save_binary(data: Union[dict, List[dict]]) -> str - removed:
method Pipeline.check_model_type(supported_models: Union[List[str], dict]) - removed:
method Pipeline.device_placement() - removed:
method Pipeline.ensure_tensor_on_device(**inputs) - removed:
method Pipeline.predict(X) - removed:
method Pipeline.save_pretrained(save_directory: str) - removed:
method Pipeline.transform(X) - removed:
method PipelineDataFormat.save(data: Union[dict, List[dict]]) - removed:
method PipelineDataFormat.save_binary(data: Union[dict, List[dict]]) -> str - removed:
method QuestionAnsweringArgumentHandler.normalize(item) - removed:
method QuestionAnsweringPipeline.decode(start: np.ndarray, end: np.ndarray, topk: int, max_answer_len: int) -> Tuple - removed:
method QuestionAnsweringPipeline.span_to_answer(text: str, start: int, end: int) -> Dict[str, Union[str, int]] - removed:
method TokenClassificationPipeline.group_entities(entities: List[dict]) -> List[dict] - removed:
method TokenClassificationPipeline.group_sub_entities(entities: List[dict]) -> dict - removed:
property ZeroShotClassificationPipeline.entailment_id - removed:
static method PipelineDataFormat.from_str(format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) -> PipelineDataFormat - removed:
static method QuestionAnsweringPipeline.create_sample(question: Union[str, List[str]], context: Union[str, List[str]]) -> Union[SquadExample, List[SquadExample]] - removed:
variable NerPipeline - removed:
variable PIPELINE_INIT_ARGS - changed:
function check_task(task: str) -> Tuple[Dict, Any] → function check_task(task: str) -> tuple[str, dict, Any] - changed:
function pipeline(task: str, model: Optional=…, config: Optional[Union[str, PretrainedConfig]]=…, tokenizer: Optional[Union[str, PreTrainedTokenizer]]=…, framework: Optional[str]=…, revision: Optional[str]=…, use_fast: bool=…, **kwargs) -> Pipeline → function pipeline(task: Literal[None], model: Optional[Union[str, PreTrainedModel, TFPreTrainedModel]]=…, config: Optional[Union[str, PretrainedConfig]]=…, tokenizer: Optional[Union[str, PreTrainedTokenizer, PreTrainedTokenizerFast]]=…, feature_extractor: Optional[Union[str, PreTrainedFeatureExtractor]]=…, image_processor: Optional[Union[str, BaseImageProcessor]]=…, processor: Optional[Union[str,… - added:
export AggregationStrategy - added:
export ArgumentHandler - added:
export AudioClassificationPipeline - added:
export AutoConfig - added:
export AutoFeatureExtractor - added:
export AutoImageProcessor - added:
export AutoModel - added:
export AutoModelForAudioClassification - added:
export AutoModelForCTC - added:
export AutoModelForCausalLM - added:
export AutoModelForDepthEstimation - added:
export AutoModelForDocumentQuestionAnswering - added:
export AutoModelForImageClassification - added:
export AutoModelForImageSegmentation - added:
export AutoModelForImageTextToText - added:
export AutoModelForImageToImage - added:
export AutoModelForKeypointMatching - added:
export AutoModelForMaskGeneration - added:
export AutoModelForMaskedLM - added:
export AutoModelForObjectDetection - added:
export AutoModelForQuestionAnswering - added:
export AutoModelForSemanticSegmentation - added:
export AutoModelForSeq2SeqLM - added:
export AutoModelForSequenceClassification - added:
export AutoModelForSpeechSeq2Seq - added:
export AutoModelForTableQuestionAnswering - added:
export AutoModelForTextToSpectrogram - added:
export AutoModelForTextToWaveform - added:
export AutoModelForTokenClassification - added:
export AutoModelForVideoClassification - added:
export AutoModelForVision2Seq - added:
export AutoModelForVisualQuestionAnswering - added:
export AutoModelForZeroShotImageClassification - added:
export AutoModelForZeroShotObjectDetection - added:
export AutoProcessor - added:
export AutoTokenizer - added:
export AutomaticSpeechRecognitionPipeline - added:
export BaseImageProcessor - added:
export CONFIG_NAME - added:
export CsvPipelineDataFormat - added:
export DepthEstimationPipeline - added:
export DocumentQuestionAnsweringPipeline - added:
export FEATURE_EXTRACTOR_MAPPING - added:
export FeatureExtractionPipeline - added:
export FillMaskPipeline - added:
export HUGGINGFACE_CO_RESOLVE_ENDPOINT - added:
export IMAGE_PROCESSOR_MAPPING - added:
export ImageClassificationPipeline - added:
export ImageFeatureExtractionPipeline - added:
export ImageSegmentationPipeline - added:
export ImageTextToTextPipeline - added:
export ImageToImagePipeline - added:
export ImageToTextPipeline - added:
export JsonPipelineDataFormat - added:
export KeypointMatchingPipeline - added:
export MaskGenerationPipeline - added:
export NerPipeline - added:
export ObjectDetectionPipeline - added:
export PROCESSOR_MAPPING - added:
export PipedPipelineDataFormat - added:
export Pipeline - added:
export PipelineDataFormat - added:
export PipelineException - added:
export PipelineRegistry - added:
export PreTrainedFeatureExtractor - added:
export PreTrainedModel - added:
export PreTrainedTokenizer - added:
export PreTrainedTokenizerFast - added:
export PretrainedConfig - added:
export ProcessorMixin - added:
export QuestionAnsweringArgumentHandler - added:
export QuestionAnsweringPipeline - added:
export SummarizationPipeline - added:
export TFAutoModel - added:
export TFAutoModelForCausalLM - added:
export TFAutoModelForImageClassification - added:
export TFAutoModelForMaskedLM - added:
export TFAutoModelForQuestionAnswering - added:
export TFAutoModelForSeq2SeqLM - added:
export TFAutoModelForSequenceClassification - added:
export TFAutoModelForTableQuestionAnswering - added:
export TFAutoModelForTokenClassification - added:
export TFAutoModelForVision2Seq - added:
export TFAutoModelForZeroShotImageClassification - added:
export TFPreTrainedModel - added:
export TOKENIZER_MAPPING - added:
export TableQuestionAnsweringArgumentHandler - added:
export TableQuestionAnsweringPipeline - added:
export Text2TextGenerationPipeline - added:
export TextClassificationPipeline - added:
export TextGenerationPipeline - added:
export TextToAudioPipeline - added:
export TokenClassificationArgumentHandler - added:
export TokenClassificationPipeline - added:
export TranslationPipeline - added:
export VideoClassificationPipeline - added:
export VisualQuestionAnsweringPipeline - added:
export ZeroShotAudioClassificationPipeline - added:
export ZeroShotClassificationArgumentHandler - added:
export ZeroShotClassificationPipeline - added:
export ZeroShotImageClassificationPipeline - added:
export ZeroShotObjectDetectionPipeline - added:
export cached_file - added:
export extract_commit_hash - added:
export find_adapter_config_file - added:
export get_class_from_dynamic_module - added:
export get_default_model_and_revision - added:
export infer_framework_load_model - added:
export is_kenlm_available - added:
export is_offline_mode - added:
export is_peft_available - added:
export is_pyctcdecode_available - added:
export is_tf_available - added:
export is_torch_available - added:
export logging - added:
function clean_custom_task(task_info) - added:
function get_supported_tasks() -> list[str] - added:
function get_task(model: str, token: Optional[str]=…, **deprecated_kwargs) -> str - added:
variable PIPELINE_REGISTRY - added:
variable TASK_ALIASES
- added:
class AudioClassificationPipeline(Pipeline) - added:
constructor AudioClassificationPipeline(*args, **kwargs) - added:
function ffmpeg_read(bpayload: bytes, sampling_rate: int) -> np.ndarray - added:
method AudioClassificationPipeline.postprocess(model_outputs, top_k=…, function_to_apply=…) - added:
method AudioClassificationPipeline.preprocess(inputs) - added:
variable logger
- added:
function chunk_bytes_iter(iterator, chunk_len: int, stride: tuple[int, int], stream: bool=…) - added:
function ffmpeg_microphone(sampling_rate: int, chunk_length_s: float, format_for_conversion: str=…, ffmpeg_input_device: Optional[str]=…, ffmpeg_additional_args: Optional[list[str]]=…) - added:
function ffmpeg_microphone_live(sampling_rate: int, chunk_length_s: float, stream_chunk_s: Optional[int]=…, stride_length_s: Optional[Union[tuple[float, float], float]]=…, format_for_conversion: str=…, ffmpeg_input_device: Optional[str]=…, ffmpeg_additional_args: Optional[list[str]]=…) - added:
function ffmpeg_read(bpayload: bytes, sampling_rate: int) -> np.ndarray
- added:
class AutomaticSpeechRecognitionPipeline(ChunkPipeline) - added:
constructor AutomaticSpeechRecognitionPipeline(model: PreTrainedModel, feature_extractor: Optional[Union[SequenceFeatureExtractor, str]]=…, tokenizer: Optional[PreTrainedTokenizer]=…, decoder: Optional[Union[BeamSearchDecoderCTC, str]]=…, device: Optional[Union[int, torch.device]]=…, **kwargs) - added:
function chunk_iter(inputs, feature_extractor, chunk_len, stride_left, stride_right, dtype=…) - added:
function rescale_stride(stride, ratio) - added:
method AutomaticSpeechRecognitionPipeline.postprocess(model_outputs, decoder_kwargs: Optional[dict]=…, return_timestamps=…, return_language=…) - added:
method AutomaticSpeechRecognitionPipeline.preprocess(inputs, chunk_length_s=…, stride_length_s=…) - added:
variable logger
- added:
class ArgumentHandler(ABC) - added:
class ChunkPipeline(Pipeline) - added:
class CsvPipelineDataFormat(PipelineDataFormat) - added:
class JsonPipelineDataFormat(PipelineDataFormat) - added:
class PipedPipelineDataFormat(PipelineDataFormat) - added:
class Pipeline(_ScikitCompat, PushToHubMixin) - added:
class PipelineDataFormat - added:
class PipelineException(Exception) - added:
class PipelineRegistry - added:
constructor CsvPipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) - added:
constructor JsonPipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) - added:
constructor Pipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: Optional[PreTrainedTokenizer]=…, feature_extractor: Optional[PreTrainedFeatureExtractor]=…, image_processor: Optional[BaseImageProcessor]=…, processor: Optional[ProcessorMixin]=…, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, task: str=…, device: Optional[Union[int, torch.device]]=…, binary_output: b… - added:
constructor PipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite: bool=…) - added:
constructor PipelineException(task: str, model: str, reason: str) - added:
constructor PipelineRegistry(supported_tasks: dict[str, Any], task_aliases: dict[str, str]) -> None - added:
field Pipeline.default_input_names - added:
field PipelineDataFormat.SUPPORTED_FORMATS - added:
function build_pipeline_init_args(has_tokenizer: bool=…, has_feature_extractor: bool=…, has_image_processor: bool=…, has_processor: bool=…, supports_binary_output: bool=…) -> str - added:
function get_default_model_and_revision(targeted_task: dict, framework: Optional[str], task_options: Optional[Any]) -> tuple[str, str] - added:
function get_framework(model, revision: Optional[str]=…) - added:
function infer_framework_from_model(model, model_classes: Optional[dict[str, tuple[type]]]=…, task: Optional[str]=…, framework: Optional[str]=…, **model_kwargs) - added:
function infer_framework_load_model(model, config: AutoConfig, model_classes: Optional[dict[str, tuple[type]]]=…, task: Optional[str]=…, framework: Optional[str]=…, **model_kwargs) - added:
function load_assistant_model(model: PreTrainedModel, assistant_model: Optional[Union[str, PreTrainedModel]], assistant_tokenizer: Optional[PreTrainedTokenizer]) -> tuple[Optional[PreTrainedModel], Optional[PreTrainedTokenizer]] - added:
function no_collate_fn(items) - added:
function pad_collate_fn(tokenizer, feature_extractor) - added:
method ChunkPipeline.get_iterator(inputs, num_workers: int, batch_size: int, preprocess_params, forward_params, postprocess_params) - added:
method ChunkPipeline.run_single(inputs, preprocess_params, forward_params, postprocess_params) - added:
method CsvPipelineDataFormat.save(data: list[dict]) - added:
method JsonPipelineDataFormat.save(data: dict) - added:
method PipedPipelineDataFormat.save(data: dict) - added:
method PipedPipelineDataFormat.save_binary(data: Union[dict, list[dict]]) -> str - added:
method Pipeline.check_model_type(supported_models: Union[list[str], dict]) - added:
method Pipeline.device_placement() - added:
method Pipeline.ensure_tensor_on_device(**inputs) - added:
method Pipeline.forward(model_inputs, **forward_params) - added:
method Pipeline.get_inference_context() - added:
method Pipeline.get_iterator(inputs, num_workers: int, batch_size: int, preprocess_params, forward_params, postprocess_params) - added:
method Pipeline.iterate(inputs, preprocess_params, forward_params, postprocess_params) - added:
method Pipeline.postprocess(model_outputs: ModelOutput, **postprocess_parameters: dict) -> Any - added:
method Pipeline.predict(X) - added:
method Pipeline.preprocess(input_: Any, **preprocess_parameters: dict) -> dict[str, GenericTensor] - added:
method Pipeline.run_multi(inputs, preprocess_params, forward_params, postprocess_params) - added:
method Pipeline.run_single(inputs, preprocess_params, forward_params, postprocess_params) - added:
method Pipeline.save_pretrained(save_directory: Union[str, os.PathLike], safe_serialization: bool=…, **kwargs: Any) - added:
method Pipeline.transform(X) - added:
method PipelineDataFormat.save(data: Union[dict, list[dict]]) - added:
method PipelineDataFormat.save_binary(data: Union[dict, list[dict]]) -> str - added:
method PipelineRegistry.check_task(task: str) -> tuple[str, dict, Any] - added:
method PipelineRegistry.get_supported_tasks() -> list[str] - added:
method PipelineRegistry.register_pipeline(task: str, pipeline_class: type, pt_model: Optional[Union[type, tuple[type]]]=…, tf_model: Optional[Union[type, tuple[type]]]=…, default: Optional[dict]=…, type: Optional[str]=…) -> None - added:
method PipelineRegistry.to_dict() - added:
property Pipeline.dtype -> Optional[torch.dtype] - added:
property Pipeline.torch_dtype -> Optional[torch.dtype] - added:
static method PipelineDataFormat.from_str(format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) -> PipelineDataFormat - added:
variable Dataset - added:
variable GenericTensor - added:
variable KeyDataset - added:
variable PIPELINE_INIT_ARGS - added:
variable SUPPORTED_PEFT_TASKS - added:
variable logger
- added:
class DepthEstimationPipeline(Pipeline) - added:
constructor DepthEstimationPipeline(*args, **kwargs) - added:
method DepthEstimationPipeline.postprocess(model_outputs) - added:
method DepthEstimationPipeline.preprocess(image, timeout=…) - added:
variable logger
- added:
class DocumentQuestionAnsweringPipeline(ChunkPipeline) - added:
class ModelType(ExplicitEnum) - added:
constructor DocumentQuestionAnsweringPipeline(*args, **kwargs) - added:
field ModelType.LayoutLM - added:
field ModelType.LayoutLMv2andv3 - added:
field ModelType.VisionEncoderDecoder - added:
function apply_tesseract(image: Image.Image, lang: Optional[str], tesseract_config: Optional[str]) - added:
function normalize_box(box, width, height) - added:
method DocumentQuestionAnsweringPipeline.postprocess(model_outputs, top_k=…, **kwargs) - added:
method DocumentQuestionAnsweringPipeline.postprocess_encoder_decoder_single(model_outputs, **kwargs) - added:
method DocumentQuestionAnsweringPipeline.postprocess_extractive_qa(model_outputs, top_k=…, handle_impossible_answer=…, max_answer_len=…, **kwargs) - added:
method DocumentQuestionAnsweringPipeline.preprocess(input, padding=…, doc_stride=…, max_seq_len=…, word_boxes: Optional[tuple[str, list[float]]]=…, lang=…, tesseract_config=…, timeout=…) - added:
variable TESSERACT_LOADED - added:
variable logger
- added:
class FeatureExtractionPipeline(Pipeline) - added:
method FeatureExtractionPipeline.postprocess(model_outputs, return_tensors=…) - added:
method FeatureExtractionPipeline.preprocess(inputs, **tokenize_kwargs) -> dict[str, GenericTensor]
- added:
class FillMaskPipeline(Pipeline) - added:
method FillMaskPipeline.ensure_exactly_one_mask_token(model_inputs: GenericTensor) - added:
method FillMaskPipeline.get_masked_index(input_ids: GenericTensor) -> np.ndarray - added:
method FillMaskPipeline.get_target_ids(targets) - added:
method FillMaskPipeline.postprocess(model_outputs, top_k=…, target_ids=…) - added:
method FillMaskPipeline.preprocess(inputs, return_tensors=…, tokenizer_kwargs=…, **preprocess_parameters) -> dict[str, GenericTensor] - added:
variable logger
- added:
class ClassificationFunction(ExplicitEnum) - added:
class ImageClassificationPipeline(Pipeline) - added:
constructor ImageClassificationPipeline(*args, **kwargs) - added:
field ClassificationFunction.NONE - added:
field ClassificationFunction.SIGMOID - added:
field ClassificationFunction.SOFTMAX - added:
field ImageClassificationPipeline.function_to_apply: ClassificationFunction - added:
function sigmoid(_outputs) - added:
function softmax(_outputs) - added:
method ImageClassificationPipeline.postprocess(model_outputs, function_to_apply=…, top_k=…) - added:
method ImageClassificationPipeline.preprocess(image, timeout=…) - added:
variable logger
- added:
class ImageFeatureExtractionPipeline(Pipeline) - added:
method ImageFeatureExtractionPipeline.postprocess(model_outputs, pool=…, return_tensors=…) - added:
method ImageFeatureExtractionPipeline.preprocess(image, timeout=…, **image_processor_kwargs) -> dict[str, GenericTensor]
- added:
class ImageSegmentationPipeline(Pipeline) - added:
constructor ImageSegmentationPipeline(*args, **kwargs) - added:
method ImageSegmentationPipeline.postprocess(model_outputs, subtask=…, threshold=…, mask_threshold=…, overlap_mask_area_threshold=…) - added:
method ImageSegmentationPipeline.preprocess(image, subtask=…, timeout=…) - added:
variable logger
- added:
class Chat - added:
class ImageTextToTextPipeline(Pipeline) - added:
class ReturnType(enum.Enum) - added:
constructor Chat(messages: dict, images: Optional[Union[str, list[str], Image.Image, list[Image.Image]]]=…) - added:
constructor ImageTextToTextPipeline(*args, **kwargs) - added:
field ReturnType.FULL_TEXT - added:
field ReturnType.NEW_TEXT - added:
field ReturnType.TENSORS - added:
function add_images_to_messages(messages: dict, images: Optional[Union[str, list[str], Image.Image, list[Image.Image]]]) - added:
method ImageTextToTextPipeline.postprocess(model_outputs, return_type=…, continue_final_message=…, skip_special_tokens=…, **postprocess_kwargs) - added:
method ImageTextToTextPipeline.preprocess(inputs=…, timeout=…, continue_final_message=…, **processing_kwargs) - added:
variable IMAGE_TOKEN - added:
variable logger
- added:
class ImageToImagePipeline(Pipeline) - added:
constructor ImageToImagePipeline(*args, **kwargs) - added:
method ImageToImagePipeline.postprocess(model_outputs) - added:
method ImageToImagePipeline.preprocess(image, timeout=…) - added:
variable logger
- added:
class ImageToTextPipeline(Pipeline) - added:
constructor ImageToTextPipeline(*args, **kwargs) - added:
method ImageToTextPipeline.postprocess(model_outputs) - added:
method ImageToTextPipeline.preprocess(image, prompt=…, timeout=…) - added:
variable logger
- added:
class Keypoint(TypedDict) - added:
class KeypointMatchingPipeline(Pipeline) - added:
class Match(TypedDict) - added:
constructor KeypointMatchingPipeline(*args, **kwargs) - added:
field Keypoint.x: float - added:
field Keypoint.y: float - added:
field Match.keypoint_image_0: Keypoint - added:
field Match.keypoint_image_1: Keypoint - added:
field Match.score: float - added:
function validate_image_pairs(images: Any) -> Sequence[Sequence[ImagePair]] - added:
method KeypointMatchingPipeline.postprocess(forward_outputs, threshold=…) -> list[Match] - added:
method KeypointMatchingPipeline.preprocess(images, timeout=…) - added:
variable ImagePair: TypeAlias
- added:
class MaskGenerationPipeline(ChunkPipeline) - added:
constructor MaskGenerationPipeline(**kwargs) - added:
method MaskGenerationPipeline.postprocess(model_outputs, output_rle_mask=…, output_bboxes_mask=…, crops_nms_thresh=…) - added:
method MaskGenerationPipeline.preprocess(image, points_per_batch=…, crops_n_layers: int=…, crop_overlap_ratio: float=…, points_per_crop: int=…, crop_n_points_downscale_factor: int=…, timeout: Optional[float]=…) - added:
variable logger
- added:
class ObjectDetectionPipeline(Pipeline) - added:
constructor ObjectDetectionPipeline(*args, **kwargs) - added:
method ObjectDetectionPipeline.postprocess(model_outputs, threshold=…) - added:
method ObjectDetectionPipeline.preprocess(image, timeout=…) - added:
variable logger
- added:
class KeyDataset(Dataset) - added:
class KeyPairDataset(Dataset) - added:
class PipelineChunkIterator(PipelineIterator) - added:
class PipelineDataset(Dataset) - added:
class PipelineIterator(IterableDataset) - added:
class PipelinePackIterator(PipelineIterator) - added:
constructor KeyDataset(dataset: Dataset, key: str) - added:
constructor KeyPairDataset(dataset: Dataset, key1: str, key2: str) - added:
constructor PipelineChunkIterator(loader, infer, params, loader_batch_size=…) - added:
constructor PipelineDataset(dataset, process, params) - added:
constructor PipelineIterator(loader, infer, params, loader_batch_size=…) - added:
method PipelineIterator.loader_batch_item()
- added:
class QuestionAnsweringArgumentHandler(ArgumentHandler) - added:
class QuestionAnsweringPipeline(ChunkPipeline) - added:
constructor QuestionAnsweringPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, task: str=…, **kwargs) - added:
field QuestionAnsweringPipeline.default_input_names - added:
field QuestionAnsweringPipeline.handle_impossible_answer - added:
function decode_spans(start: np.ndarray, end: np.ndarray, topk: int, max_answer_len: int, undesired_tokens: np.ndarray) -> tuple - added:
function select_starts_ends(start, end, p_mask, attention_mask, min_null_score=…, top_k=…, handle_impossible_answer=…, max_answer_len=…) - added:
method QuestionAnsweringArgumentHandler.normalize(item) - added:
method QuestionAnsweringPipeline.get_answer(answers: list[dict], target: str) -> Optional[dict] - added:
method QuestionAnsweringPipeline.get_indices(enc: tokenizers.Encoding, s: int, e: int, sequence_index: int, align_to_words: bool) -> tuple[int, int] - added:
method QuestionAnsweringPipeline.postprocess(model_outputs, top_k=…, handle_impossible_answer=…, max_answer_len=…, align_to_words=…) - added:
method QuestionAnsweringPipeline.preprocess(example, padding=…, doc_stride=…, max_question_len=…, max_seq_len=…) - added:
method QuestionAnsweringPipeline.span_to_answer(text: str, start: int, end: int) -> dict[str, Union[str, int]] - added:
static method QuestionAnsweringPipeline.create_sample(question: Union[str, list[str]], context: Union[str, list[str]]) -> Union[SquadExample, list[SquadExample]] - added:
variable Dataset - added:
variable logger
- added:
class TableQuestionAnsweringArgumentHandler(ArgumentHandler) - added:
class TableQuestionAnsweringPipeline(Pipeline) - added:
constructor TableQuestionAnsweringPipeline(args_parser=…, **kwargs) - added:
field TableQuestionAnsweringPipeline.default_input_names - added:
method TableQuestionAnsweringPipeline.batch_inference(**inputs) - added:
method TableQuestionAnsweringPipeline.postprocess(model_outputs) - added:
method TableQuestionAnsweringPipeline.preprocess(pipeline_input, padding=…, truncation=…) - added:
method TableQuestionAnsweringPipeline.sequential_inference(**inputs)
- added:
class ReturnType(enum.Enum) - added:
class SummarizationPipeline(Text2TextGenerationPipeline) - added:
class Text2TextGenerationPipeline(Pipeline) - added:
class TranslationPipeline(Text2TextGenerationPipeline) - added:
constructor Text2TextGenerationPipeline(*args, **kwargs) - added:
field ReturnType.TENSORS - added:
field ReturnType.TEXT - added:
field SummarizationPipeline.return_name - added:
field Text2TextGenerationPipeline.return_name - added:
field TranslationPipeline.return_name - added:
method SummarizationPipeline.check_inputs(input_length: int, min_length: int, max_length: int) -> bool - added:
method Text2TextGenerationPipeline.check_inputs(input_length: int, min_length: int, max_length: int) - added:
method Text2TextGenerationPipeline.postprocess(model_outputs, return_type=…, clean_up_tokenization_spaces=…) - added:
method Text2TextGenerationPipeline.preprocess(inputs, truncation=…, **kwargs) - added:
method TranslationPipeline.check_inputs(input_length: int, min_length: int, max_length: int) - added:
method TranslationPipeline.preprocess(*args, truncation=…, src_lang=…, tgt_lang=…) - added:
variable logger
- added:
class ClassificationFunction(ExplicitEnum) - added:
class TextClassificationPipeline(Pipeline) - added:
constructor TextClassificationPipeline(**kwargs) - added:
field ClassificationFunction.NONE - added:
field ClassificationFunction.SIGMOID - added:
field ClassificationFunction.SOFTMAX - added:
field TextClassificationPipeline.function_to_apply - added:
field TextClassificationPipeline.return_all_scores - added:
function sigmoid(_outputs) - added:
function softmax(_outputs) - added:
method TextClassificationPipeline.postprocess(model_outputs, function_to_apply=…, top_k=…, _legacy=…) - added:
method TextClassificationPipeline.preprocess(inputs, **tokenizer_kwargs) -> dict[str, GenericTensor]
- added:
class Chat - added:
class ReturnType(enum.Enum) - added:
class TextGenerationPipeline(Pipeline) - added:
constructor Chat(messages: dict) - added:
constructor TextGenerationPipeline(*args, **kwargs) - added:
field ReturnType.FULL_TEXT - added:
field ReturnType.NEW_TEXT - added:
field ReturnType.TENSORS - added:
field TextGenerationPipeline.XL_PREFIX - added:
method TextGenerationPipeline.postprocess(model_outputs, return_type=…, clean_up_tokenization_spaces=…, continue_final_message=…, skip_special_tokens=…) - added:
method TextGenerationPipeline.preprocess(prompt_text, prefix=…, handle_long_generation=…, add_special_tokens=…, truncation=…, padding=…, max_length=…, continue_final_message=…, tokenizer_encode_kwargs=…, **generate_kwargs) - added:
variable ChatType
- added:
class TextToAudioPipeline(Pipeline) - added:
constructor TextToAudioPipeline(*args, vocoder=…, sampling_rate=…, no_processor=…, **kwargs) - added:
method TextToAudioPipeline.postprocess(audio) - added:
method TextToAudioPipeline.preprocess(text, **kwargs) - added:
variable DEFAULT_VOCODER_ID
- added:
class AggregationStrategy(ExplicitEnum) - added:
class TokenClassificationArgumentHandler(ArgumentHandler) - added:
class TokenClassificationPipeline(ChunkPipeline) - added:
constructor TokenClassificationPipeline(args_parser=…, **kwargs) - added:
field AggregationStrategy.AVERAGE - added:
field AggregationStrategy.FIRST - added:
field AggregationStrategy.MAX - added:
field AggregationStrategy.NONE - added:
field AggregationStrategy.SIMPLE - added:
field TokenClassificationPipeline.default_input_names - added:
method TokenClassificationPipeline.aggregate(pre_entities: list[dict], aggregation_strategy: AggregationStrategy) -> list[dict] - added:
method TokenClassificationPipeline.aggregate_overlapping_entities(entities) - added:
method TokenClassificationPipeline.aggregate_word(entities: list[dict], aggregation_strategy: AggregationStrategy) -> dict - added:
method TokenClassificationPipeline.aggregate_words(entities: list[dict], aggregation_strategy: AggregationStrategy) -> list[dict] - added:
method TokenClassificationPipeline.gather_pre_entities(sentence: str, input_ids: np.ndarray, scores: np.ndarray, offset_mapping: Optional[list[tuple[int, int]]], special_tokens_mask: np.ndarray, aggregation_strategy: AggregationStrategy, word_ids: Optional[list[Optional[int]]]=…, word_to_chars_map: Optional[list[tuple[int, int]]]=…) -> list[dict] - added:
method TokenClassificationPipeline.get_tag(entity_name: str) -> tuple[str, str] - added:
method TokenClassificationPipeline.group_entities(entities: list[dict]) -> list[dict] - added:
method TokenClassificationPipeline.group_sub_entities(entities: list[dict]) -> dict - added:
method TokenClassificationPipeline.postprocess(all_outputs, aggregation_strategy=…, ignore_labels=…) - added:
method TokenClassificationPipeline.preprocess(sentence, offset_mapping=…, **preprocess_params) - added:
variable NerPipeline
- added:
class VideoClassificationPipeline(Pipeline) - added:
constructor VideoClassificationPipeline(*args, **kwargs) - added:
function read_video_pyav(container, indices) - added:
method VideoClassificationPipeline.postprocess(model_outputs, top_k=…, function_to_apply=…) - added:
method VideoClassificationPipeline.preprocess(video, num_frames=…, frame_sampling_rate=…) - added:
variable logger
- added:
class VisualQuestionAnsweringPipeline(Pipeline) - added:
constructor VisualQuestionAnsweringPipeline(*args, **kwargs) - added:
method VisualQuestionAnsweringPipeline.postprocess(model_outputs, top_k=…) - added:
method VisualQuestionAnsweringPipeline.preprocess(inputs, padding=…, truncation=…, timeout=…) - added:
variable logger
- added:
class ZeroShotAudioClassificationPipeline(Pipeline) - added:
constructor ZeroShotAudioClassificationPipeline(**kwargs) - added:
method ZeroShotAudioClassificationPipeline.postprocess(model_outputs) - added:
method ZeroShotAudioClassificationPipeline.preprocess(audio, candidate_labels=…, hypothesis_template=…) - added:
variable logger
- added:
class ZeroShotClassificationArgumentHandler(ArgumentHandler) - added:
class ZeroShotClassificationPipeline(ChunkPipeline) - added:
constructor ZeroShotClassificationPipeline(args_parser=…, **kwargs) - added:
method ZeroShotClassificationPipeline.postprocess(model_outputs, multi_label=…) - added:
method ZeroShotClassificationPipeline.preprocess(inputs, candidate_labels=…, hypothesis_template=…) - added:
property ZeroShotClassificationPipeline.entailment_id - added:
variable logger
- added:
class ZeroShotImageClassificationPipeline(Pipeline) - added:
constructor ZeroShotImageClassificationPipeline(**kwargs) - added:
method ZeroShotImageClassificationPipeline.postprocess(model_outputs) - added:
method ZeroShotImageClassificationPipeline.preprocess(image, candidate_labels=…, hypothesis_template=…, timeout=…, tokenizer_kwargs=…) - added:
variable logger
- added:
class ZeroShotObjectDetectionPipeline(ChunkPipeline) - added:
constructor ZeroShotObjectDetectionPipeline(**kwargs) - added:
method ZeroShotObjectDetectionPipeline.postprocess(model_outputs, threshold=…, top_k=…) - added:
method ZeroShotObjectDetectionPipeline.preprocess(inputs, timeout=…) - added:
variable logger
- added:
class AllKwargsForChatTemplate(TypedDict) - added:
class AudioKwargs(TypedDict) - added:
class ChatTemplateLoadKwargs(TypedDict) - added:
class CommonKwargs(TypedDict) - added:
class ImagesKwargs(TypedDict) - added:
class MultiModalData - added:
class ProcessingKwargs(TypedDict) - added:
class ProcessorChatTemplateKwargs(ChatTemplateLoadKwargs, TokenizerChatTemplateKwargs) - added:
class ProcessorMixin(PushToHubMixin) - added:
class TextKwargs(TypedDict) - added:
class TokenizerChatTemplateKwargs(TypedDict) - added:
class VideosKwargs(TypedDict) - added:
constructor ProcessorMixin(*args, **kwargs) - added:
field AllKwargsForChatTemplate.mm_load_kwargs: ChatTemplateLoadKwargs - added:
field AllKwargsForChatTemplate.processor_kwargs: ProcessingKwargs - added:
field AllKwargsForChatTemplate.template_kwargs: ProcessorChatTemplateKwargs - added:
field AudioKwargs.max_length: Optional[int] - added:
field AudioKwargs.pad_to_multiple_of: Optional[int] - added:
field AudioKwargs.padding: Optional[Union[bool, str, PaddingStrategy]] - added:
field AudioKwargs.raw_speech: Optional[Union[np.ndarray, list[float], list[np.ndarray], list[list[float]]]] - added:
field AudioKwargs.return_attention_mask: Optional[bool] - added:
field AudioKwargs.sampling_rate: Optional[int] - added:
field AudioKwargs.truncation: Optional[bool] - added:
field ChatTemplateLoadKwargs.load_audio_from_video: Optional[bool] - added:
field ChatTemplateLoadKwargs.sampling_rate: Optional[int] - added:
field CommonKwargs.return_tensors: Optional[Union[str, TensorType]] - added:
field ImagesKwargs.crop_size: Optional[dict[str, int]] - added:
field ImagesKwargs.data_format: Optional[ChannelDimension] - added:
field ImagesKwargs.device: Optional[str] - added:
field ImagesKwargs.do_center_crop: Optional[bool] - added:
field ImagesKwargs.do_normalize: Optional[bool] - added:
field ImagesKwargs.do_pad: Optional[bool] - added:
field ImagesKwargs.do_rescale: Optional[bool] - added:
field ImagesKwargs.do_resize: Optional[bool] - added:
field ImagesKwargs.image_mean: Optional[Union[float, list[float]]] - added:
field ImagesKwargs.image_std: Optional[Union[float, list[float]]] - added:
field ImagesKwargs.input_data_format: Optional[Union[str, ChannelDimension]] - added:
field ImagesKwargs.pad_size: Optional[dict[str, int]] - added:
field ImagesKwargs.resample: Optional[Union[PILImageResampling, int]] - added:
field ImagesKwargs.rescale_factor: Optional[float] - added:
field ImagesKwargs.size: Optional[dict[str, int]] - added:
field MultiModalData.num_audio_tokens: Optional[list[int]] - added:
field MultiModalData.num_image_patches: Optional[list[int]] - added:
field MultiModalData.num_image_tokens: Optional[list[int]] - added:
field MultiModalData.num_video_tokens: Optional[list[int]] - added:
field ProcessingKwargs.audio_kwargs: AudioKwargs - added:
field ProcessingKwargs.common_kwargs: CommonKwargs - added:
field ProcessingKwargs.images_kwargs: ImagesKwargs - added:
field ProcessingKwargs.text_kwargs: TextKwargs - added:
field ProcessingKwargs.videos_kwargs: VideosKwargs - added:
field ProcessorChatTemplateKwargs.return_dict: Optional[bool] - added:
field ProcessorChatTemplateKwargs.tokenize: Optional[bool] - added:
field ProcessorMixin.attributes - added:
field ProcessorMixin.feature_extractor_class - added:
field ProcessorMixin.optional_attributes - added:
field ProcessorMixin.optional_call_args: list[str] - added:
field ProcessorMixin.tokenizer_class - added:
field ProcessorMixin.valid_processor_kwargs - added:
field TextKwargs.add_special_tokens: Optional[bool] - added:
field TextKwargs.is_split_into_words: Optional[bool] - added:
field TextKwargs.max_length: Optional[int] - added:
field TextKwargs.pad_to_multiple_of: Optional[int] - added:
field TextKwargs.padding: Union[bool, str, PaddingStrategy] - added:
field TextKwargs.padding_side: Optional[str] - added:
field TextKwargs.return_attention_mask: Optional[bool] - added:
field TextKwargs.return_length: Optional[bool] - added:
field TextKwargs.return_mm_token_type_ids: Optional[bool] - added:
field TextKwargs.return_offsets_mapping: Optional[bool] - added:
field TextKwargs.return_overflowing_tokens: Optional[bool] - added:
field TextKwargs.return_special_tokens_mask: Optional[bool] - added:
field TextKwargs.return_token_type_ids: Optional[bool] - added:
field TextKwargs.stride: Optional[int] - added:
field TextKwargs.text_pair: Optional[Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]]] - added:
field TextKwargs.text_pair_target: Optional[Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]]] - added:
field TextKwargs.text_target: Union[TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]] - added:
field TextKwargs.truncation: Union[bool, str, TruncationStrategy] - added:
field TextKwargs.verbose: Optional[bool] - added:
field TokenizerChatTemplateKwargs.add_generation_prompt: Optional[bool] - added:
field TokenizerChatTemplateKwargs.continue_final_message: Optional[bool] - added:
field TokenizerChatTemplateKwargs.documents: Optional[list[dict[str, str]]] - added:
field TokenizerChatTemplateKwargs.return_assistant_tokens_mask: Optional[bool] - added:
field TokenizerChatTemplateKwargs.tools: Optional[list[dict]] - added:
field VideosKwargs.crop_size: Optional[dict[str, int]] - added:
field VideosKwargs.data_format: Optional[ChannelDimension] - added:
field VideosKwargs.default_to_square: Optional[bool] - added:
field VideosKwargs.device: Optional[str] - added:
field VideosKwargs.do_center_crop: Optional[bool] - added:
field VideosKwargs.do_convert_rgb: Optional[bool] - added:
field VideosKwargs.do_normalize: Optional[bool] - added:
field VideosKwargs.do_rescale: Optional[bool] - added:
field VideosKwargs.do_resize: Optional[bool] - added:
field VideosKwargs.do_sample_frames: Optional[bool] - added:
field VideosKwargs.fps: Optional[Union[int, float]] - added:
field VideosKwargs.image_mean: Optional[Union[float, list[float]]] - added:
field VideosKwargs.image_std: Optional[Union[float, list[float]]] - added:
field VideosKwargs.input_data_format: Optional[Union[str, ChannelDimension]] - added:
field VideosKwargs.num_frames: Optional[int] - added:
field VideosKwargs.resample: Optional[PILImageResampling] - added:
field VideosKwargs.rescale_factor: Optional[float] - added:
field VideosKwargs.return_metadata: Optional[bool] - added:
field VideosKwargs.size: Optional[dict[str, int]] - added:
field VideosKwargs.video_metadata: Optional[Union[VideoMetadata, dict]] - added:
method ProcessorMixin.apply_chat_template(conversation: Union[list[dict[str, str]], list[list[dict[str, str]]]], chat_template: Optional[str]=…, **kwargs: Unpack[AllKwargsForChatTemplate]) -> str - added:
method ProcessorMixin.batch_decode(*args, **kwargs) - added:
method ProcessorMixin.check_argument_for_proper_class(argument_name, argument) - added:
method ProcessorMixin.decode(*args, **kwargs) - added:
method ProcessorMixin.post_process_image_text_to_text(generated_outputs, skip_special_tokens=…, **kwargs) - added:
method ProcessorMixin.save_pretrained(save_directory, push_to_hub: bool=…, legacy_serialization: bool=…, **kwargs) - added:
method ProcessorMixin.to_dict(legacy_serialization=…) -> dict[str, Any] - added:
method ProcessorMixin.to_json_file(json_file_path: Union[str, os.PathLike], legacy_serialization=…) - added:
method ProcessorMixin.to_json_string(legacy_serialization=…) -> str - added:
property ProcessorMixin.model_input_names - added:
static method ProcessorMixin.from_args_and_dict(args, processor_dict: dict[str, Any], **kwargs) - added:
static method ProcessorMixin.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) -> SpecificProcessorType - added:
static method ProcessorMixin.get_possibly_dynamic_module(module_name) - added:
static method ProcessorMixin.get_processor_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]] - added:
static method ProcessorMixin.register_for_auto_class(auto_class=…) - added:
static method ProcessorMixin.validate_init_kwargs(processor_config, valid_kwargs) - added:
variable AUTO_TO_BASE_CLASS_MAPPING - added:
variable SpecificProcessorType - added:
variable Unpack - added:
variable logger - added:
variable transformers_module
- added:
class Conv1D(nn.Module) - added:
constructor Conv1D(nf, nx) - added:
function apply_chunking_to_forward(forward_fn: Callable[..., torch.Tensor], chunk_size: int, chunk_dim: int, *input_tensors) -> torch.Tensor - added:
function compile_compatible_method_lru_cache(*lru_args, **lru_kwargs) - added:
function find_pruneable_heads_and_indices(heads: list[int], n_heads: int, head_size: int, already_pruned_heads: set[int]) -> tuple[set[int], torch.LongTensor] - added:
function id_tensor_storage(tensor: torch.Tensor) -> tuple[torch.device, int, int] - added:
function infer_device() - added:
function isin_mps_friendly(elements: torch.Tensor, test_elements: torch.Tensor | int) -> torch.Tensor - added:
function meshgrid(*tensors: torch.Tensor | list[torch.Tensor], indexing: str | None=…) -> tuple[torch.Tensor, ...] - added:
function prune_conv1d_layer(layer: Conv1D, index: torch.LongTensor, dim: int=…) -> Conv1D - added:
function prune_layer(layer: nn.Linear | Conv1D, index: torch.LongTensor, dim: int | None=…) -> nn.Linear | Conv1D - added:
function prune_linear_layer(layer: nn.Linear, index: torch.LongTensor, dim: int=…) -> nn.Linear - added:
function softmax_backward_data(parent, grad_output, output) - added:
method Conv1D.forward(x) - added:
variable ALL_LAYERNORM_LAYERS - added:
variable is_torch_greater_or_equal_than_1_12 - added:
variable is_torch_greater_or_equal_than_1_13 - added:
variable is_torch_greater_or_equal_than_2_0 - added:
variable is_torch_greater_or_equal_than_2_1 - added:
variable is_torch_greater_or_equal_than_2_2 - added:
variable is_torch_greater_or_equal_than_2_3 - added:
variable is_torch_greater_or_equal_than_2_4 - added:
variable is_torch_greater_or_equal_than_2_6 - added:
variable is_torch_greater_or_equal_than_2_8 - added:
variable logger
- added:
export AutoHfQuantizer - added:
export AutoQuantizationConfig - added:
export HfQuantizer - added:
export get_module_from_name - added:
export register_quantization_config - added:
export register_quantizer
- added:
class AutoHfQuantizer - added:
class AutoQuantizationConfig - added:
function get_hf_quantizer(config, quantization_config, dtype, from_tf, from_flax, device_map, weights_only, user_agent) - added:
function register_quantization_config(method: str) - added:
function register_quantizer(name: str) - added:
static method AutoHfQuantizer.from_config(quantization_config: Union[QuantizationConfigMixin, dict], **kwargs) - added:
static method AutoHfQuantizer.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
static method AutoHfQuantizer.merge_quantization_configs(quantization_config: Union[dict, QuantizationConfigMixin], quantization_config_from_args: Optional[QuantizationConfigMixin]) - added:
static method AutoHfQuantizer.supports_quant_method(quantization_config_dict) - added:
static method AutoQuantizationConfig.from_dict(quantization_config_dict: dict) - added:
static method AutoQuantizationConfig.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
variable AUTO_QUANTIZATION_CONFIG_MAPPING - added:
variable AUTO_QUANTIZER_MAPPING - added:
variable logger
- added:
class HfQuantizer(ABC) - added:
class SequentialLlama4TextExperts(ModuleList) - added:
constructor HfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
constructor SequentialLlama4TextExperts(config) - added:
field HfQuantizer.required_packages - added:
field HfQuantizer.requires_calibration - added:
field HfQuantizer.requires_parameters_quantization - added:
method HfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method HfQuantizer.adjust_target_dtype(dtype: torch.dtype) -> torch.dtype - added:
method HfQuantizer.check_quantized_param(*args, **kwargs) -> bool - added:
method HfQuantizer.create_quantized_param(*args, **kwargs) - added:
method HfQuantizer.dequantize(model) - added:
method HfQuantizer.get_accelerator_warm_up_factor() - added:
method HfQuantizer.get_param_name(param_name: str) -> str - added:
method HfQuantizer.get_special_dtypes_update(model, dtype: torch.dtype) -> dict[str, torch.dtype] - added:
method HfQuantizer.get_state_dict_and_metadata(model, safe_serialization=…) - added:
method HfQuantizer.is_serializable(safe_serialization=…) - added:
method HfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method HfQuantizer.postprocess_model(model: PreTrainedModel, **kwargs) - added:
method HfQuantizer.preprocess_model(model: PreTrainedModel, **kwargs) - added:
method HfQuantizer.remove_quantization_config(model) - added:
method HfQuantizer.update_device_map(device_map: Optional[dict[str, Any]]) -> Optional[dict[str, Any]] - added:
method HfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method HfQuantizer.update_ep_plan(config) - added:
method HfQuantizer.update_expected_keys(model, expected_keys: list[str], loaded_keys: list[str]) -> list[str] - added:
method HfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method HfQuantizer.update_state_dict_with_metadata(state_dict, metadata) - added:
method HfQuantizer.update_torch_dtype(dtype: torch.dtype) -> torch.dtype - added:
method HfQuantizer.update_tp_plan(config) - added:
method HfQuantizer.update_unexpected_keys(model, unexpected_keys: list[str]) -> list[str] - added:
method HfQuantizer.validate_environment(*args, **kwargs) - added:
method SequentialLlama4TextExperts.forward(hidden_states: torch.Tensor) -> torch.Tensor - added:
property HfQuantizer.is_compileable -> bool - added:
property HfQuantizer.is_qat_trainable -> bool - added:
property HfQuantizer.is_trainable - added:
static method HfQuantizer.get_modules_to_not_convert(model: PreTrainedModel, skip_modules: Optional[list[str]]=…, keep_in_fp32_modules: Optional[list[str]]=…, add_default_skips: bool=…) - added:
variable MODULES_TO_PATCH_FOR_QUANTIZATION - added:
variable ModuleList - added:
variable logger
- added:
class AqlmHfQuantizer(HfQuantizer) - added:
constructor AqlmHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field AqlmHfQuantizer.optimum_quantizer - added:
field AqlmHfQuantizer.required_packages - added:
field AqlmHfQuantizer.requires_calibration - added:
method AqlmHfQuantizer.is_serializable(safe_serialization=…) - added:
method AqlmHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method AqlmHfQuantizer.validate_environment(*args, **kwargs) - added:
property AqlmHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class AutoRoundQuantizer(HfQuantizer) - added:
constructor AutoRoundQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field AutoRoundQuantizer.required_packages - added:
field AutoRoundQuantizer.requires_calibration - added:
method AutoRoundQuantizer.is_serializable(safe_serialization=…) - added:
method AutoRoundQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method AutoRoundQuantizer.validate_environment(*args, **kwargs) - added:
property AutoRoundQuantizer.is_trainable -> bool - added:
variable logger
- added:
class AwqQuantizer(HfQuantizer) - added:
constructor AwqQuantizer(quantization_config, **kwargs) - added:
field AwqQuantizer.required_packages - added:
field AwqQuantizer.requires_calibration - added:
method AwqQuantizer.is_serializable(safe_serialization=…) - added:
method AwqQuantizer.update_dtype(dtype) - added:
method AwqQuantizer.validate_environment(device_map, **kwargs) - added:
property AwqQuantizer.is_trainable - added:
variable logger
- added:
class BitNetHfQuantizer(HfQuantizer) - added:
constructor BitNetHfQuantizer(quantization_config, **kwargs) - added:
field BitNetHfQuantizer.required_packages - added:
field BitNetHfQuantizer.requires_calibration - added:
field BitNetHfQuantizer.requires_parameters_quantization - added:
method BitNetHfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method BitNetHfQuantizer.adjust_target_dtype(target_dtype: torch.dtype) -> torch.dtype - added:
method BitNetHfQuantizer.is_serializable(safe_serialization=…) - added:
method BitNetHfQuantizer.validate_environment(*args, **kwargs) - added:
property BitNetHfQuantizer.is_qat_trainable -> bool - added:
property BitNetHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class Bnb4BitHfQuantizer(HfQuantizer) - added:
constructor Bnb4BitHfQuantizer(quantization_config, **kwargs) - added:
field Bnb4BitHfQuantizer.required_packages - added:
field Bnb4BitHfQuantizer.requires_calibration - added:
field Bnb4BitHfQuantizer.requires_parameters_quantization - added:
field Bnb4BitHfQuantizer.use_keep_in_fp32_modules - added:
method Bnb4BitHfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method Bnb4BitHfQuantizer.adjust_target_dtype(target_dtype: torch.dtype) -> torch.dtype - added:
method Bnb4BitHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method Bnb4BitHfQuantizer.get_param_name(param_name: str) -> str - added:
method Bnb4BitHfQuantizer.is_serializable(safe_serialization=…) - added:
method Bnb4BitHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method Bnb4BitHfQuantizer.update_device_map(device_map) - added:
method Bnb4BitHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method Bnb4BitHfQuantizer.update_unexpected_keys(model, unexpected_keys: list[str]) -> list[str] - added:
method Bnb4BitHfQuantizer.validate_environment(*args, **kwargs) - added:
property Bnb4BitHfQuantizer.is_bnb_supports_quant_storage_module -> bool - added:
property Bnb4BitHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class Bnb8BitHfQuantizer(HfQuantizer) - added:
constructor Bnb8BitHfQuantizer(quantization_config, **kwargs) - added:
field Bnb8BitHfQuantizer.required_packages - added:
field Bnb8BitHfQuantizer.requires_calibration - added:
field Bnb8BitHfQuantizer.requires_parameters_quantization - added:
field Bnb8BitHfQuantizer.use_keep_in_fp32_modules - added:
method Bnb8BitHfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method Bnb8BitHfQuantizer.adjust_target_dtype(target_dtype: torch.dtype) -> torch.dtype - added:
method Bnb8BitHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method Bnb8BitHfQuantizer.is_serializable(safe_serialization=…) - added:
method Bnb8BitHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method Bnb8BitHfQuantizer.update_device_map(device_map) - added:
method Bnb8BitHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method Bnb8BitHfQuantizer.update_unexpected_keys(model, unexpected_keys: list[str]) -> list[str] - added:
method Bnb8BitHfQuantizer.validate_environment(*args, **kwargs) - added:
property Bnb8BitHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class CompressedTensorsHfQuantizer(HfQuantizer) - added:
constructor CompressedTensorsHfQuantizer(quantization_config: CompressedTensorsConfig, **kwargs) - added:
field CompressedTensorsHfQuantizer.required_packages - added:
field CompressedTensorsHfQuantizer.requires_calibration - added:
method CompressedTensorsHfQuantizer.is_qat_trainable() -> bool - added:
method CompressedTensorsHfQuantizer.is_serializable(safe_serialization=…) -> bool - added:
method CompressedTensorsHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method CompressedTensorsHfQuantizer.update_tp_plan(config) - added:
method CompressedTensorsHfQuantizer.validate_environment(*args, **kwargs) - added:
property CompressedTensorsHfQuantizer.is_trainable - added:
variable logger
- added:
class EetqHfQuantizer(HfQuantizer) - added:
constructor EetqHfQuantizer(quantization_config, **kwargs) - added:
field EetqHfQuantizer.required_packages - added:
field EetqHfQuantizer.requires_calibration - added:
field EetqHfQuantizer.requires_parameters_quantization - added:
method EetqHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method EetqHfQuantizer.is_serializable(safe_serialization=…) - added:
method EetqHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method EetqHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method EetqHfQuantizer.validate_environment(*args, **kwargs) - added:
property EetqHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class FbgemmFp8HfQuantizer(HfQuantizer) - added:
constructor FbgemmFp8HfQuantizer(quantization_config, **kwargs) - added:
field FbgemmFp8HfQuantizer.required_packages - added:
field FbgemmFp8HfQuantizer.requires_calibration - added:
field FbgemmFp8HfQuantizer.requires_parameters_quantization - added:
method FbgemmFp8HfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method FbgemmFp8HfQuantizer.is_serializable(safe_serialization=…) - added:
method FbgemmFp8HfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method FbgemmFp8HfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method FbgemmFp8HfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method FbgemmFp8HfQuantizer.update_tp_plan(config) - added:
method FbgemmFp8HfQuantizer.validate_environment(*args, **kwargs) - added:
property FbgemmFp8HfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class FineGrainedFP8HfQuantizer(HfQuantizer) - added:
constructor FineGrainedFP8HfQuantizer(quantization_config, **kwargs) - added:
field FineGrainedFP8HfQuantizer.required_packages - added:
field FineGrainedFP8HfQuantizer.requires_calibration - added:
field FineGrainedFP8HfQuantizer.requires_parameters_quantization - added:
method FineGrainedFP8HfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method FineGrainedFP8HfQuantizer.get_accelerator_warm_up_factor() - added:
method FineGrainedFP8HfQuantizer.is_serializable(safe_serialization=…) - added:
method FineGrainedFP8HfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method FineGrainedFP8HfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method FineGrainedFP8HfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method FineGrainedFP8HfQuantizer.update_tp_plan(config) - added:
method FineGrainedFP8HfQuantizer.validate_environment(*args, **kwargs) - added:
property FineGrainedFP8HfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class FPQuantHfQuantizer(HfQuantizer) - added:
constructor FPQuantHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field FPQuantHfQuantizer.is_qat_trainable - added:
field FPQuantHfQuantizer.required_packages - added:
field FPQuantHfQuantizer.requires_calibration - added:
field FPQuantHfQuantizer.requires_parameters_quantization - added:
method FPQuantHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method FPQuantHfQuantizer.is_serializable(safe_serialization=…) - added:
method FPQuantHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method FPQuantHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method FPQuantHfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method FPQuantHfQuantizer.validate_environment(device_map, **kwargs) - added:
property FPQuantHfQuantizer.is_trainable - added:
variable logger
- added:
class GptqHfQuantizer(HfQuantizer) - added:
constructor GptqHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field GptqHfQuantizer.optimum_quantizer - added:
field GptqHfQuantizer.required_packages - added:
field GptqHfQuantizer.requires_calibration - added:
method GptqHfQuantizer.is_serializable(safe_serialization=…) - added:
method GptqHfQuantizer.update_device_map(device_map) - added:
method GptqHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method GptqHfQuantizer.validate_environment(*args, **kwargs) - added:
property GptqHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class HiggsHfQuantizer(HfQuantizer) - added:
constructor HiggsHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field HiggsHfQuantizer.required_packages - added:
field HiggsHfQuantizer.requires_calibration - added:
field HiggsHfQuantizer.requires_parameters_quantization - added:
method HiggsHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method HiggsHfQuantizer.is_serializable(safe_serialization=…) - added:
method HiggsHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method HiggsHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method HiggsHfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method HiggsHfQuantizer.validate_environment(device_map, **kwargs) - added:
property HiggsHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class HqqHfQuantizer(HfQuantizer) - added:
constructor HqqHfQuantizer(quantization_config, **kwargs) - added:
field HqqHfQuantizer.required_packages - added:
field HqqHfQuantizer.requires_calibration - added:
field HqqHfQuantizer.requires_parameters_quantization - added:
field HqqHfQuantizer.use_keep_in_fp32_modules - added:
function weight(self) - added:
method HqqHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method HqqHfQuantizer.is_serializable(safe_serialization=…) - added:
method HqqHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method HqqHfQuantizer.update_expected_keys(model: PreTrainedModel, expected_keys: list[str], loaded_keys: list[str]) -> list[str] - added:
method HqqHfQuantizer.update_missing_keys(model: PreTrainedModel, missing_keys: list[str], prefix: str, **kwargs) -> list[str] - added:
property HqqHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class Mxfp4HfQuantizer(HfQuantizer) - added:
constructor Mxfp4HfQuantizer(quantization_config, **kwargs) - added:
field Mxfp4HfQuantizer.required_packages - added:
field Mxfp4HfQuantizer.requires_calibration - added:
field Mxfp4HfQuantizer.requires_parameters_quantization - added:
method Mxfp4HfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method Mxfp4HfQuantizer.get_param_name(param_name: str) -> str - added:
method Mxfp4HfQuantizer.get_state_dict_and_metadata(model, safe_serialization: bool=…) - added:
method Mxfp4HfQuantizer.is_serializable(safe_serialization=…) - added:
method Mxfp4HfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method Mxfp4HfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method Mxfp4HfQuantizer.update_ep_plan(config) - added:
method Mxfp4HfQuantizer.update_expected_keys(model: PreTrainedModel, expected_keys: list[str], checkpoint_keys: list[str]) - added:
method Mxfp4HfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method Mxfp4HfQuantizer.update_tp_plan(config) - added:
method Mxfp4HfQuantizer.validate_environment(*args, **kwargs) - added:
property Mxfp4HfQuantizer.is_trainable -> bool - added:
variable logger - added:
variable triton_kernels_hub
- added:
class QuantoHfQuantizer(HfQuantizer) - added:
constructor QuantoHfQuantizer(quantization_config: QuantoConfig, **kwargs) - added:
field QuantoHfQuantizer.required_packages - added:
field QuantoHfQuantizer.requires_calibration - added:
field QuantoHfQuantizer.requires_parameters_quantization - added:
method QuantoHfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method QuantoHfQuantizer.adjust_target_dtype(target_dtype: torch.dtype) -> torch.dtype - added:
method QuantoHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method QuantoHfQuantizer.is_serializable(safe_serialization=…) - added:
method QuantoHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method QuantoHfQuantizer.post_init() - added:
method QuantoHfQuantizer.update_device_map(device_map) - added:
method QuantoHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method QuantoHfQuantizer.update_missing_keys(model, missing_keys: list[str], prefix: str) -> list[str] - added:
method QuantoHfQuantizer.validate_environment(*args, **kwargs) - added:
property QuantoHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
class QuarkHfQuantizer(HfQuantizer) - added:
constructor QuarkHfQuantizer(quantization_config, **kwargs) - added:
field QuarkHfQuantizer.required_packages - added:
field QuarkHfQuantizer.requires_calibration - added:
field QuarkHfQuantizer.requires_parameters_quantization - added:
method QuarkHfQuantizer.create_quantized_param(model, param, param_name, param_device, **kwargs) - added:
method QuarkHfQuantizer.is_serializable(safe_serialization=…) - added:
method QuarkHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method QuarkHfQuantizer.validate_environment(*args, **kwargs) - added:
property QuarkHfQuantizer.is_trainable - added:
variable CHECKPOINT_KEYS - added:
variable logger
- added:
class SpQRHfQuantizer(HfQuantizer) - added:
constructor SpQRHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field SpQRHfQuantizer.requires_calibration - added:
method SpQRHfQuantizer.is_serializable(safe_serialization=…) - added:
method SpQRHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method SpQRHfQuantizer.validate_environment(*args, **kwargs) - added:
property SpQRHfQuantizer.is_trainable - added:
variable logger
- added:
class TorchAoHfQuantizer(HfQuantizer) - added:
constructor TorchAoHfQuantizer(quantization_config, **kwargs) - added:
field TorchAoHfQuantizer.required_packages - added:
field TorchAoHfQuantizer.requires_calibration - added:
field TorchAoHfQuantizer.requires_parameters_quantization - added:
function fuzzy_match_size(config_name: str) -> Optional[str] - added:
method TorchAoHfQuantizer.adjust_max_memory(max_memory: dict[str, Union[int, str]]) -> dict[str, Union[int, str]] - added:
method TorchAoHfQuantizer.adjust_target_dtype(dtype: torch.dtype) -> torch.dtype - added:
method TorchAoHfQuantizer.create_quantized_param(model: PreTrainedModel, param_value: torch.Tensor, param_name: str, target_device: torch.device, **kwargs) - added:
method TorchAoHfQuantizer.get_accelerator_warm_up_factor() - added:
method TorchAoHfQuantizer.get_state_dict_and_metadata(model, safe_serialization: Optional[bool]=…) - added:
method TorchAoHfQuantizer.is_serializable(safe_serialization=…) -> bool - added:
method TorchAoHfQuantizer.param_needs_quantization(model: PreTrainedModel, param_name: str, **kwargs) -> bool - added:
method TorchAoHfQuantizer.set_metadata(checkpoint_files: list[str]) - added:
method TorchAoHfQuantizer.update_dtype(dtype) - added:
method TorchAoHfQuantizer.update_unexpected_keys(model, unexpected_keys: list[str]) -> list[str] - added:
method TorchAoHfQuantizer.validate_environment(*args, **kwargs) - added:
property TorchAoHfQuantizer.is_compileable -> bool - added:
property TorchAoHfQuantizer.is_trainable -> bool - added:
variable SUPPORTED_SAFE_SERIALIZATION_CONFIGS - added:
variable TORCHAO_VERSION - added:
variable logger
- added:
class VptqHfQuantizer(HfQuantizer) - added:
constructor VptqHfQuantizer(quantization_config: QuantizationConfigMixin, **kwargs) - added:
field VptqHfQuantizer.required_packages - added:
field VptqHfQuantizer.requires_calibration - added:
method VptqHfQuantizer.is_serializable(safe_serialization=…) - added:
method VptqHfQuantizer.update_dtype(dtype: torch.dtype) -> torch.dtype - added:
method VptqHfQuantizer.validate_environment(*args, **kwargs) - added:
property VptqHfQuantizer.is_trainable -> bool - added:
variable logger
- added:
function get_module_from_name(module, tensor_name: str) -> tuple[Any, str]
- removed:
class CanonicalHFIndex(HFIndexBase) - removed:
class CustomHFIndex(HFIndexBase) - removed:
class HFIndexBase(Index) - removed:
class Index - removed:
class LegacyIndex(Index) - removed:
class RagRetriever - removed:
constructor CanonicalHFIndex(vector_size: int, dataset_name: str=…, dataset_split: str=…, index_name: Optional[str]=…, index_path: Optional[str]=…, use_dummy_dataset=…) - removed:
constructor CustomHFIndex(vector_size: int, dataset, index_path=…) - removed:
constructor HFIndexBase(vector_size, dataset, index_initialized=…) - removed:
constructor LegacyIndex(vector_size, index_path) - removed:
constructor RagRetriever(config, question_encoder_tokenizer, generator_tokenizer, index=…) - removed:
field LegacyIndex.INDEX_FILENAME - removed:
field LegacyIndex.PASSAGE_FILENAME - removed:
method CanonicalHFIndex.init_index() - removed:
method CustomHFIndex.init_index() - removed:
method HFIndexBase.get_doc_dicts(doc_ids: np.ndarray) -> List[dict] - removed:
method HFIndexBase.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - removed:
method HFIndexBase.init_index() - removed:
method HFIndexBase.is_initialized() - removed:
method Index.get_doc_dicts(doc_ids: np.ndarray) -> List[dict] - removed:
method Index.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - removed:
method Index.init_index() - removed:
method Index.is_initialized() - removed:
method LegacyIndex.get_doc_dicts(doc_ids: np.array) - removed:
method LegacyIndex.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - removed:
method LegacyIndex.init_index() - removed:
method LegacyIndex.is_initialized() - removed:
method RagRetriever.init_retrieval() - removed:
method RagRetriever.postprocess_docs(docs, input_strings, prefix, n_docs, return_tensors=…) - removed:
method RagRetriever.retrieve(question_hidden_states: np.ndarray, n_docs: int) -> Tuple[np.ndarray, List[dict]] - removed:
method RagRetriever.save_pretrained(save_directory) - removed:
static method CustomHFIndex.load_from_disk(vector_size, dataset_path, index_path) - removed:
static method RagRetriever.from_pretrained(retriever_name_or_path, indexed_dataset=…, **kwargs) - removed:
variable LEGACY_INDEX_PATH - removed:
variable logger
- added:
function auto_conversion(pretrained_model_name_or_path: str, ignore_errors_during_conversion=…, **cached_file_kwargs) - added:
function get_conversion_pr_reference(api: HfApi, model_id: str, **kwargs) - added:
function previous_pr(api: HfApi, model_id: str, pr_title: str, token: str) -> Optional[Discussion] - added:
function spawn_conversion(token: str, private: bool, model_id: str) - added:
variable logger
- added:
export SageMakerTrainer - added:
export SageMakerTrainingArguments - added:
export is_sagemaker_dp_enabled
- added:
class SageMakerTrainer(Trainer) - added:
constructor SageMakerTrainer(args=…, **kwargs) - added:
variable logger
- added:
class SageMakerTrainingArguments(TrainingArguments) - added:
field SageMakerTrainingArguments.mp_parameters: str - added:
function is_sagemaker_model_parallel_available() - added:
property SageMakerTrainingArguments.place_model_on_device - added:
property SageMakerTrainingArguments.world_size - added:
variable logger
- removed:
function is_pt_tf_cross_test(test_case) - removed:
function require_datasets(test_case) - removed:
function require_flax(test_case) - removed:
function require_tf(test_case) - removed:
function require_torch_multigpu(test_case) - removed:
function require_torch_non_multigpu(test_case) - removed:
function require_torch_tpu(test_case) - removed:
variable DUMMY_UNKWOWN_IDENTIFIER - removed:
variable require_torch_non_multigpu_but_fix_me - changed:
constructor CaptureStd(out=…, err=…) → constructor CaptureStd(out=…, err=…, replay=…) - changed:
constructor CaptureStderr() → constructor CaptureStderr(replay=…) - changed:
constructor CaptureStdout() → constructor CaptureStdout(replay=…) - changed:
method TestCasePlus.get_auto_remove_tmp_dir(tmp_dir=…, after=…, before=…) → method TestCasePlus.get_auto_remove_tmp_dir(tmp_dir=…, before=…, after=…) - added:
class Expectations(UserDict[PackedDeviceProperties, Any]) - added:
class HfDocTestParser(doctest.DocTestParser) - added:
class HfDoctestModule(Module) - added:
class RequestCounter - added:
class SubprocessCallException(Exception) - added:
class TemporaryHubRepo - added:
constructor TemporaryHubRepo(namespace: Optional[str]=…, token: Optional[str]=…) -> None - added:
field HfDocTestParser.skip_cuda_tests: bool - added:
function ExtendSysPath(path: Union[str, os.PathLike]) -> Iterator[None] - added:
function LoggingLevel(level) - added:
function apply_skip_if_not_implemented(cls) - added:
function backend_device_count(device: str) - added:
function backend_empty_cache(device: str) - added:
function backend_manual_seed(device: str, seed: int) - added:
function backend_max_memory_allocated(device: str) - added:
function backend_memory_allocated(device: str) - added:
function backend_reset_max_memory_allocated(device: str) - added:
function backend_reset_peak_memory_stats(device: str) - added:
function backend_synchronize(device: str) - added:
function backend_torch_accelerator_module(device: str) - added:
function check_json_file_has_correct_format(file_path) - added:
function cleanup(device: str, gc_collect=…) - added:
function cmd_exists(cmd) - added:
function compare_pipeline_output_to_hub_spec(output, hub_spec) - added:
function evaluate_side_effect_factory(side_effect_values: list[dict[str, float]]) -> Generator[dict[str, float], None, None] - added:
function get_device_properties() -> DeviceProperties - added:
function get_steps_per_epoch(trainer: Trainer) -> int - added:
function get_torch_dist_unique_port() - added:
function hub_retry(max_attempts: int=…, wait_before_retry: Optional[float]=…) - added:
function is_agent_test(test_case) - added:
function is_flaky(max_attempts: int=…, wait_before_retry: Optional[float]=…, description: Optional[str]=…) - added:
function is_staging_test(test_case) - added:
function mockenv_context(*remove, **update) - added:
function nested_simplify(obj, decimals=…) - added:
function patch_testing_methods_to_collect_info() - added:
function patch_torch_compile_force_graph() - added:
function preprocess_string(string, skip_cuda_tests) - added:
function pytest_xdist_worker_id() - added:
function require_accelerate(test_case, min_version: str=…) - added:
function require_apex(test_case) - added:
function require_apollo_torch(test_case) - added:
function require_aqlm(test_case) - added:
function require_auto_awq(test_case) - added:
function require_auto_round(test_case) - added:
function require_av(test_case) - added:
function require_bitsandbytes(test_case) - added:
function require_bs4(test_case) - added:
function require_clearml(test_case) - added:
function require_compressed_tensors(test_case) - added:
function require_cv2(test_case) - added:
function require_cython(test_case) - added:
function require_decord(test_case) - added:
function require_deepspeed(test_case) - added:
function require_detectron2(test_case) - added:
function require_deterministic_for_xpu(test_case) - added:
function require_eetq(test_case) - added:
function require_essentia(test_case) - added:
function require_fbgemm_gpu(test_case) - added:
function require_flash_attn(test_case) - added:
function require_flash_attn_3(test_case) - added:
function require_flute_hadamard(test_case) - added:
function require_fp8(test_case) - added:
function require_fp_quant(test_case) - added:
function require_fsdp(test_case, min_version: str=…) - added:
function require_ftfy(test_case) - added:
function require_g2p_en(test_case) - added:
function require_galore_torch(test_case) - added:
function require_gguf(test_case, min_version: str=…) - added:
function require_gptq(test_case) - added:
function require_grokadamw(test_case) - added:
function require_hqq(test_case) - added:
function require_huggingface_hub_greater_or_equal(version: str) - added:
function require_intel_extension_for_pytorch(test_case) - added:
function require_jinja(test_case) - added:
function require_jumanpp(test_case) - added:
function require_keras_nlp(test_case) - added:
function require_kernels(test_case) - added:
function require_large_cpu_ram(test_case, memory: float=…) - added:
function require_levenshtein(test_case) - added:
function require_librosa(test_case) - added:
function require_liger_kernel(test_case) - added:
function require_lomo(test_case) - added:
function require_mistral_common(test_case) - added:
function require_natten(test_case) - added:
function require_nltk(test_case) - added:
function require_non_hpu(test_case) - added:
function require_non_xpu(test_case) - added:
function require_onnx(test_case) - added:
function require_openai(test_case) - added:
function require_optimum(test_case) - added:
function require_optimum_quanto(test_case) - added:
function require_pandas(test_case) - added:
function require_peft(test_case) - added:
function require_phonemizer(test_case) - added:
function require_pretty_midi(test_case) - added:
function require_pyctcdecode(test_case) - added:
function require_pytesseract(test_case) - added:
function require_pytorch_quantization(test_case) - added:
function require_quark(test_case) - added:
function require_qutlass(test_case) - added:
function require_read_token(test_case) - added:
function require_rjieba(test_case) - added:
function require_sacremoses(test_case) - added:
function require_safetensors(test_case) - added:
function require_schedulefree(test_case) - added:
function require_scipy(test_case) - added:
function require_seqio(test_case) - added:
function require_sigopt(test_case) - added:
function require_spacy(test_case) - added:
function require_speech(test_case) - added:
function require_spqr(test_case) - added:
function require_sudachi(test_case) - added:
function require_sudachi_projection(test_case) - added:
function require_swanlab(test_case) - added:
function require_tensorboard(test_case) - added:
function require_tiktoken(test_case) - added:
function require_timm(test_case) - added:
function require_torch_accelerator(test_case) - added:
function require_torch_bf16(test_case) - added:
function require_torch_bf16_gpu(test_case) - added:
function require_torch_fp16(test_case) - added:
function require_torch_gpu_if_bnb_not_multi_backend_enabled(test_case) - added:
function require_torch_greater_or_equal(version: str) - added:
function require_torch_large_accelerator(test_case, memory: float=…) - added:
function require_torch_large_gpu(test_case, memory: float=…) - added:
function require_torch_mps(test_case) - added:
function require_torch_multi_accelerator(test_case) - added:
function require_torch_multi_gpu(test_case) - added:
function require_torch_multi_hpu(test_case) - added:
function require_torch_multi_npu(test_case) - added:
function require_torch_multi_xpu(test_case) - added:
function require_torch_neuroncore(test_case) - added:
function require_torch_non_multi_accelerator(test_case) - added:
function require_torch_non_multi_gpu(test_case) - added:
function require_torch_npu(test_case) - added:
function require_torch_optimi(test_case) - added:
function require_torch_or_tf(test_case) - added:
function require_torch_tensorrt_fx(test_case) - added:
function require_torch_tf32(test_case) - added:
function require_torch_up_to_2_accelerators(test_case) - added:
function require_torch_up_to_2_gpus(test_case) - added:
function require_torch_xla(test_case) - added:
function require_torch_xpu(test_case) - added:
function require_torchao(test_case) - added:
function require_torchao_version_greater_or_equal(torchao_version) - added:
function require_torchaudio(test_case) - added:
function require_torchcodec(test_case) - added:
function require_torchdynamo(test_case) - added:
function require_torchvision(test_case) - added:
function require_trackio(test_case) - added:
function require_triton(min_version: str=…) - added:
function require_usr_bin_time(test_case) - added:
function require_vision(test_case) - added:
function require_vptq(test_case) - added:
function require_wandb(test_case) - added:
function run_command(command: list[str], return_stdout=…) - added:
function run_first(test_case) - added:
function run_test_in_subprocess(test_case, target_func, inputs=…, timeout=…) - added:
function run_test_using_subprocess(func) - added:
function set_config_for_less_flaky_test(config) - added:
function set_model_for_less_flaky_test(model) - added:
function skip_if_not_implemented(test_func) - added:
function to_2tuple(x) - added:
function tooslow(test_case) - added:
function torchrun(script: str, nproc_per_node: int, is_torchrun: bool=…, env: Optional[dict]=…) - added:
function unpack_device_properties(properties: Optional[PackedDeviceProperties]=…) -> DeviceProperties - added:
function update_mapping_from_spec(device_fn_dict: dict[str, Callable], attribute_name: str) - added:
method Expectations.find_expectation(properties: DeviceProperties=…) -> Any - added:
method Expectations.get_expectation() -> Any - added:
method Expectations.unpacked() -> list[tuple[DeviceProperties, Any]] - added:
method HfDocTestParser.parse(string, name=…) - added:
method HfDoctestModule.collect() -> Iterable[DoctestItem] - added:
method TestCasePlus.python_one_liner_max_rss(one_liner_str) - added:
property RequestCounter.total_calls -> int - added:
static method Expectations.is_default(expectation_key: PackedDeviceProperties) -> bool - added:
static method Expectations.score(properties: DeviceProperties, other: DeviceProperties) -> float - added:
variable BACKEND_DEVICE_COUNT - added:
variable BACKEND_EMPTY_CACHE - added:
variable BACKEND_MANUAL_SEED - added:
variable BACKEND_MAX_MEMORY_ALLOCATED - added:
variable BACKEND_MEMORY_ALLOCATED - added:
variable BACKEND_RESET_MAX_MEMORY_ALLOCATED - added:
variable BACKEND_RESET_PEAK_MEMORY_STATS - added:
variable BACKEND_SYNCHRONIZE - added:
variable BACKEND_TORCH_ACCELERATOR_MODULE - added:
variable DUMMY_UNKNOWN_IDENTIFIER - added:
variable DeviceProperties - added:
variable DoctestItem - added:
variable ENDPOINT_STAGING - added:
variable IS_CUDA_SYSTEM - added:
variable IS_ROCM_SYSTEM - added:
variable IS_XPU_SYSTEM - added:
variable Module - added:
variable PackedDeviceProperties - added:
variable TOKEN - added:
variable USER - added:
variable backend - added:
variable device_name - added:
variable device_spec_dir - added:
variable device_spec_module - added:
variable device_spec_path - added:
variable import_name - added:
variable jax_device - added:
variable logger - added:
variable msg
- added:
function check_embeddings_within_bounds(tensor: tf.Tensor, embed_dim: int, tensor_name: str=…) -> None - added:
function convert_batch_encoding(*args, **kwargs) - added:
function expand_1d(data) - added:
function flatten(input, start_dim=…, end_dim=…) - added:
function functional_layernorm(inputs, weight, bias, epsilon=…, axis=…) - added:
function invert_attention_mask(encoder_attention_mask: tf.Tensor) -> tf.Tensor - added:
function load_attributes_from_hdf5_group(group, name) - added:
function save_attributes_to_hdf5_group(group, name, data) - added:
function scaled_dot_product_attention(query, key, value, attn_mask=…, dropout_p=…, is_causal=…, scale: Optional[float]=…) - added:
function shape_list(tensor: Union[tf.Tensor, np.ndarray]) -> list[int] - added:
function stable_softmax(logits: tf.Tensor, axis: Optional[int]=…, name: Optional[str]=…) -> tf.Tensor - added:
variable logger
- added:
class AffineTransformed(TransformedDistribution) - added:
class DistributionOutput - added:
class LambdaLayer(nn.Module) - added:
class NegativeBinomialOutput(DistributionOutput) - added:
class NormalOutput(DistributionOutput) - added:
class ParameterProjection(nn.Module) - added:
class StudentTOutput(DistributionOutput) - added:
constructor AffineTransformed(base_distribution: Distribution, loc=…, scale=…, event_dim=…) - added:
constructor DistributionOutput(dim: int=…) -> None - added:
constructor LambdaLayer(function) - added:
constructor ParameterProjection(in_features: int, args_dim: dict[str, int], domain_map: Callable[..., tuple[torch.Tensor]], **kwargs) -> None - added:
field DistributionOutput.args_dim: dict[str, int] - added:
field DistributionOutput.distribution_class: type - added:
field DistributionOutput.in_features: int - added:
field NegativeBinomialOutput.args_dim: dict[str, int] - added:
field NegativeBinomialOutput.distribution_class: type - added:
field NormalOutput.args_dim: dict[str, int] - added:
field NormalOutput.distribution_class: type - added:
field StudentTOutput.args_dim: dict[str, int] - added:
field StudentTOutput.distribution_class: type - added:
method DistributionOutput.distribution(distr_args, loc: Optional[torch.Tensor]=…, scale: Optional[torch.Tensor]=…) -> Distribution - added:
method DistributionOutput.domain_map(*args: torch.Tensor) - added:
method DistributionOutput.get_parameter_projection(in_features: int) -> nn.Module - added:
method LambdaLayer.forward(x, *args) - added:
method NegativeBinomialOutput.distribution(distr_args, loc: Optional[torch.Tensor]=…, scale: Optional[torch.Tensor]=…) -> Distribution - added:
method ParameterProjection.forward(x: torch.Tensor) -> tuple[torch.Tensor] - added:
property AffineTransformed.mean - added:
property AffineTransformed.stddev - added:
property AffineTransformed.variance - added:
property DistributionOutput.event_dim -> int - added:
property DistributionOutput.event_shape -> tuple - added:
property DistributionOutput.value_in_support -> float - added:
static method DistributionOutput.squareplus(x: torch.Tensor) -> torch.Tensor - added:
static method NegativeBinomialOutput.domain_map(total_count: torch.Tensor, logits: torch.Tensor) - added:
static method NormalOutput.domain_map(loc: torch.Tensor, scale: torch.Tensor) - added:
static method StudentTOutput.domain_map(df: torch.Tensor, loc: torch.Tensor, scale: torch.Tensor)
- removed:
class AlbertTokenizer(PreTrainedTokenizer) - removed:
constructor AlbertTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - removed:
field AlbertTokenizer.max_model_input_sizes - removed:
field AlbertTokenizer.pretrained_vocab_files_map - removed:
field AlbertTokenizer.vocab_files_names - removed:
method AlbertTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method AlbertTokenizer.convert_tokens_to_string(tokens) - removed:
method AlbertTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method AlbertTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method AlbertTokenizer.get_vocab() - removed:
method AlbertTokenizer.preprocess_text(inputs) - removed:
method AlbertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property AlbertTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class AlbertTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor AlbertTokenizerFast(vocab_file, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - removed:
field AlbertTokenizerFast.max_model_input_sizes - removed:
field AlbertTokenizerFast.pretrained_vocab_files_map - removed:
field AlbertTokenizerFast.slow_tokenizer_class - removed:
field AlbertTokenizerFast.vocab_files_names - removed:
method AlbertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method AlbertTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method AlbertTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method AlbertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable AlbertTokenizer - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class AutoTokenizer - removed:
constructor AutoTokenizer() - removed:
static method AutoTokenizer.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs) - removed:
variable AlbertTokenizer - removed:
variable AlbertTokenizerFast - removed:
variable BartTokenizerFast - removed:
variable BertGenerationTokenizer - removed:
variable BertTokenizerFast - removed:
variable CamembertTokenizer - removed:
variable CamembertTokenizerFast - removed:
variable DPRQuestionEncoderTokenizerFast - removed:
variable DistilBertTokenizerFast - removed:
variable ElectraTokenizerFast - removed:
variable FunnelTokenizerFast - removed:
variable GPT2TokenizerFast - removed:
variable HerbertTokenizerFast - removed:
variable LayoutLMTokenizerFast - removed:
variable LongformerTokenizerFast - removed:
variable LxmertTokenizerFast - removed:
variable MBartTokenizer - removed:
variable MBartTokenizerFast - removed:
variable MarianTokenizer - removed:
variable MobileBertTokenizerFast - removed:
variable OpenAIGPTTokenizerFast - removed:
variable PegasusTokenizer - removed:
variable PegasusTokenizerFast - removed:
variable ReformerTokenizer - removed:
variable ReformerTokenizerFast - removed:
variable RetriBertTokenizerFast - removed:
variable RobertaTokenizerFast - removed:
variable SLOW_TOKENIZER_MAPPING - removed:
variable SqueezeBertTokenizerFast - removed:
variable T5Tokenizer - removed:
variable T5TokenizerFast - removed:
variable TOKENIZER_MAPPING - removed:
variable XLMProphetNetTokenizer - removed:
variable XLMRobertaTokenizer - removed:
variable XLMRobertaTokenizerFast - removed:
variable XLNetTokenizer - removed:
variable XLNetTokenizerFast - removed:
variable logger
- removed:
class BartTokenizer(RobertaTokenizer) - removed:
field BartTokenizer.max_model_input_sizes - removed:
field BartTokenizer.pretrained_vocab_files_map - removed:
method BartTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation=…, **kwargs) -> BatchEncoding - removed:
variable logger - removed:
variable merges_url - removed:
variable vocab_url
- removed:
class BartTokenizerFast(RobertaTokenizerFast) - removed:
field BartTokenizerFast.max_model_input_sizes - removed:
field BartTokenizerFast.pretrained_vocab_files_map - removed:
field BartTokenizerFast.slow_tokenizer_class - removed:
method BartTokenizerFast.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation=…, **kwargs) -> BatchEncoding - removed:
variable logger - removed:
variable merges_url - removed:
variable tokenizer_url - removed:
variable vocab_url
- removed:
class BasicTokenizer(object) - removed:
class BertTokenizer(PreTrainedTokenizer) - removed:
class WordpieceTokenizer(object) - removed:
constructor BasicTokenizer(do_lower_case=…, never_split=…, tokenize_chinese_chars=…, strip_accents=…) - removed:
constructor BertTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - removed:
constructor WordpieceTokenizer(vocab, unk_token, max_input_chars_per_word=…) - removed:
field BertTokenizer.max_model_input_sizes - removed:
field BertTokenizer.pretrained_init_configuration - removed:
field BertTokenizer.pretrained_vocab_files_map - removed:
field BertTokenizer.vocab_files_names - removed:
function load_vocab(vocab_file) - removed:
function whitespace_tokenize(text) - removed:
method BasicTokenizer.tokenize(text, never_split=…) - removed:
method BertTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method BertTokenizer.convert_tokens_to_string(tokens) - removed:
method BertTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method BertTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method BertTokenizer.get_vocab() - removed:
method BertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
method WordpieceTokenizer.tokenize(text) - removed:
property BertTokenizer.do_lower_case - removed:
property BertTokenizer.vocab_size - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class BertTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor BertTokenizerFast(vocab_file, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - removed:
field BertTokenizerFast.max_model_input_sizes - removed:
field BertTokenizerFast.pretrained_init_configuration - removed:
field BertTokenizerFast.pretrained_vocab_files_map - removed:
field BertTokenizerFast.slow_tokenizer_class - removed:
field BertTokenizerFast.vocab_files_names - removed:
method BertTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - removed:
method BertTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method BertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class BertGenerationTokenizer(PreTrainedTokenizer) - removed:
constructor BertGenerationTokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, sep_token=…, **kwargs) - removed:
field BertGenerationTokenizer.max_model_input_sizes - removed:
field BertGenerationTokenizer.prefix_tokens: List[int] - removed:
field BertGenerationTokenizer.pretrained_vocab_files_map - removed:
field BertGenerationTokenizer.vocab_files_names - removed:
method BertGenerationTokenizer.convert_tokens_to_string(tokens) - removed:
method BertGenerationTokenizer.get_vocab() - removed:
method BertGenerationTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property BertGenerationTokenizer.vocab_size - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger - removed:
variable tokenizer_url
- removed:
class BertJapaneseTokenizer(BertTokenizer) - removed:
class CharacterTokenizer - removed:
class MecabTokenizer - removed:
constructor BertJapaneseTokenizer(vocab_file, do_lower_case=…, do_word_tokenize=…, do_subword_tokenize=…, word_tokenizer_type=…, subword_tokenizer_type=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, mecab_kwargs=…, **kwargs) - removed:
constructor CharacterTokenizer(vocab, unk_token, normalize_text=…) - removed:
constructor MecabTokenizer(do_lower_case=…, never_split=…, normalize_text=…, mecab_dic: Optional[str]=…, mecab_option: Optional[str]=…) - removed:
field BertJapaneseTokenizer.max_model_input_sizes - removed:
field BertJapaneseTokenizer.pretrained_init_configuration - removed:
field BertJapaneseTokenizer.pretrained_vocab_files_map - removed:
field BertJapaneseTokenizer.vocab_files_names - removed:
method CharacterTokenizer.tokenize(text) - removed:
method MecabTokenizer.tokenize(text, never_split=…, **kwargs) - removed:
property BertJapaneseTokenizer.do_lower_case - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class BertweetTokenizer(PreTrainedTokenizer) - removed:
class TweetTokenizer - removed:
constructor BertweetTokenizer(vocab_file, merges_file, normalization=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - removed:
constructor TweetTokenizer(preserve_case=…, reduce_len=…, strip_handles=…) - removed:
field BertweetTokenizer.max_model_input_sizes - removed:
field BertweetTokenizer.pretrained_vocab_files_map - removed:
field BertweetTokenizer.vocab_files_names - removed:
function casual_tokenize(text, preserve_case=…, reduce_len=…, strip_handles=…) - removed:
function get_pairs(word) - removed:
function reduce_lengthening(text) - removed:
function remove_handles(text) - removed:
method BertweetTokenizer.add_from_file(f) - removed:
method BertweetTokenizer.bpe(token) - removed:
method BertweetTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method BertweetTokenizer.convert_tokens_to_string(tokens) - removed:
method BertweetTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method BertweetTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method BertweetTokenizer.get_vocab() - removed:
method BertweetTokenizer.normalizeToken(token) - removed:
method BertweetTokenizer.normalizeTweet(tweet) - removed:
method BertweetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
method TweetTokenizer.tokenize(text) - removed:
property BertweetTokenizer.vocab_size - removed:
variable EMOTICONS - removed:
variable EMOTICON_RE - removed:
variable ENT_RE - removed:
variable HANG_RE - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable REGEXPS - removed:
variable URLS - removed:
variable VOCAB_FILES_NAMES - removed:
variable WORD_RE - removed:
variable logger
- removed:
class BlenderbotSmallTokenizer(PreTrainedTokenizer) - removed:
class BlenderbotTokenizer(RobertaTokenizer) - removed:
constructor BlenderbotSmallTokenizer(vocab_file, merges_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, **kwargs) - removed:
field BlenderbotSmallTokenizer.max_model_input_sizes - removed:
field BlenderbotSmallTokenizer.pretrained_vocab_files_map - removed:
field BlenderbotSmallTokenizer.vocab_files_names - removed:
field BlenderbotTokenizer.max_model_input_sizes - removed:
field BlenderbotTokenizer.pretrained_vocab_files_map - removed:
field BlenderbotTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
method BlenderbotSmallTokenizer.bpe(token: str) -> str - removed:
method BlenderbotSmallTokenizer.convert_tokens_to_string(tokens: List[str]) -> str - removed:
method BlenderbotSmallTokenizer.get_vocab() -> Dict - removed:
method BlenderbotSmallTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
method BlenderbotTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: List[int]=…) - removed:
property BlenderbotSmallTokenizer.vocab_size -> int - removed:
variable CKPT_3B - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class CamembertTokenizer(PreTrainedTokenizer) - removed:
constructor CamembertTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - removed:
field CamembertTokenizer.max_model_input_sizes - removed:
field CamembertTokenizer.model_input_names - removed:
field CamembertTokenizer.pretrained_vocab_files_map - removed:
field CamembertTokenizer.vocab_files_names - removed:
method CamembertTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method CamembertTokenizer.convert_tokens_to_string(tokens) - removed:
method CamembertTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method CamembertTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method CamembertTokenizer.get_vocab() - removed:
method CamembertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property CamembertTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SHARED_MODEL_IDENTIFIERS - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class CamembertTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor CamembertTokenizerFast(vocab_file, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - removed:
field CamembertTokenizerFast.max_model_input_sizes - removed:
field CamembertTokenizerFast.model_input_names - removed:
field CamembertTokenizerFast.pretrained_vocab_files_map - removed:
field CamembertTokenizerFast.slow_tokenizer_class - removed:
field CamembertTokenizerFast.vocab_files_names - removed:
method CamembertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method CamembertTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method CamembertTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method CamembertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable CamembertTokenizer - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SHARED_MODEL_IDENTIFIERS - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class CTRLTokenizer(PreTrainedTokenizer) - removed:
constructor CTRLTokenizer(vocab_file, merges_file, unk_token=…, **kwargs) - removed:
field CTRLTokenizer.control_codes - removed:
field CTRLTokenizer.max_model_input_sizes - removed:
field CTRLTokenizer.pretrained_vocab_files_map - removed:
field CTRLTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
method CTRLTokenizer.bpe(token) - removed:
method CTRLTokenizer.convert_tokens_to_string(tokens) - removed:
method CTRLTokenizer.get_vocab() - removed:
method CTRLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property CTRLTokenizer.vocab_size - removed:
variable CONTROL_CODES - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class DebertaTokenizer(PreTrainedTokenizer) - removed:
constructor DebertaTokenizer(vocab_file, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - removed:
field DebertaTokenizer.max_model_input_sizes - removed:
field DebertaTokenizer.pretrained_init_configuration - removed:
field DebertaTokenizer.pretrained_vocab_files_map - removed:
field DebertaTokenizer.vocab_files_names - removed:
method DebertaTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - removed:
method DebertaTokenizer.convert_tokens_to_string(tokens) - removed:
method DebertaTokenizer.create_token_type_ids_from_sequences(token_ids_0, token_ids_1=…) - removed:
method DebertaTokenizer.get_special_tokens_mask(token_ids_0, token_ids_1=…, already_has_special_tokens=…) - removed:
method DebertaTokenizer.get_vocab() - removed:
method DebertaTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - removed:
method DebertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property DebertaTokenizer.vocab - removed:
property DebertaTokenizer.vocab_size
- removed:
class DistilBertTokenizer(BertTokenizer) - removed:
field DistilBertTokenizer.max_model_input_sizes - removed:
field DistilBertTokenizer.model_input_names - removed:
field DistilBertTokenizer.pretrained_init_configuration - removed:
field DistilBertTokenizer.pretrained_vocab_files_map - removed:
field DistilBertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class DistilBertTokenizerFast(BertTokenizerFast) - removed:
field DistilBertTokenizerFast.max_model_input_sizes - removed:
field DistilBertTokenizerFast.model_input_names - removed:
field DistilBertTokenizerFast.pretrained_init_configuration - removed:
field DistilBertTokenizerFast.pretrained_vocab_files_map - removed:
field DistilBertTokenizerFast.slow_tokenizer_class - removed:
field DistilBertTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class CustomDPRReaderTokenizerMixin - removed:
class DPRContextEncoderTokenizer(BertTokenizer) - removed:
class DPRQuestionEncoderTokenizer(BertTokenizer) - removed:
class DPRReaderTokenizer(CustomDPRReaderTokenizerMixin, BertTokenizer) - removed:
field DPRContextEncoderTokenizer.max_model_input_sizes - removed:
field DPRContextEncoderTokenizer.pretrained_init_configuration - removed:
field DPRContextEncoderTokenizer.pretrained_vocab_files_map - removed:
field DPRContextEncoderTokenizer.vocab_files_names - removed:
field DPRQuestionEncoderTokenizer.max_model_input_sizes - removed:
field DPRQuestionEncoderTokenizer.pretrained_init_configuration - removed:
field DPRQuestionEncoderTokenizer.pretrained_vocab_files_map - removed:
field DPRQuestionEncoderTokenizer.vocab_files_names - removed:
field DPRReaderTokenizer.max_model_input_sizes - removed:
field DPRReaderTokenizer.model_input_names - removed:
field DPRReaderTokenizer.pretrained_init_configuration - removed:
field DPRReaderTokenizer.pretrained_vocab_files_map - removed:
field DPRReaderTokenizer.vocab_files_names - removed:
method CustomDPRReaderTokenizerMixin.decode_best_spans(reader_input: BatchEncoding, reader_output: DPRReaderOutput, num_spans: int=…, max_answer_length: int=…, num_spans_per_passage: int=…) -> List[DPRSpanPrediction] - removed:
variable CONTEXT_ENCODER_PRETRAINED_INIT_CONFIGURATION - removed:
variable CONTEXT_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable CONTEXT_ENCODER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable CUSTOM_DPR_READER_DOCSTRING - removed:
variable DPRReaderOutput - removed:
variable DPRSpanPrediction - removed:
variable QUESTION_ENCODER_PRETRAINED_INIT_CONFIGURATION - removed:
variable QUESTION_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable QUESTION_ENCODER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable READER_PRETRAINED_INIT_CONFIGURATION - removed:
variable READER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable READER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class CustomDPRReaderTokenizerMixin - removed:
class DPRContextEncoderTokenizerFast(BertTokenizerFast) - removed:
class DPRQuestionEncoderTokenizerFast(BertTokenizerFast) - removed:
class DPRReaderTokenizerFast(CustomDPRReaderTokenizerMixin, BertTokenizerFast) - removed:
field DPRContextEncoderTokenizerFast.max_model_input_sizes - removed:
field DPRContextEncoderTokenizerFast.pretrained_init_configuration - removed:
field DPRContextEncoderTokenizerFast.pretrained_vocab_files_map - removed:
field DPRContextEncoderTokenizerFast.slow_tokenizer_class - removed:
field DPRContextEncoderTokenizerFast.vocab_files_names - removed:
field DPRQuestionEncoderTokenizerFast.max_model_input_sizes - removed:
field DPRQuestionEncoderTokenizerFast.pretrained_init_configuration - removed:
field DPRQuestionEncoderTokenizerFast.pretrained_vocab_files_map - removed:
field DPRQuestionEncoderTokenizerFast.slow_tokenizer_class - removed:
field DPRQuestionEncoderTokenizerFast.vocab_files_names - removed:
field DPRReaderTokenizerFast.max_model_input_sizes - removed:
field DPRReaderTokenizerFast.model_input_names - removed:
field DPRReaderTokenizerFast.pretrained_init_configuration - removed:
field DPRReaderTokenizerFast.pretrained_vocab_files_map - removed:
field DPRReaderTokenizerFast.slow_tokenizer_class - removed:
field DPRReaderTokenizerFast.vocab_files_names - removed:
method CustomDPRReaderTokenizerMixin.decode_best_spans(reader_input: BatchEncoding, reader_output: DPRReaderOutput, num_spans: int=…, max_answer_length: int=…, num_spans_per_passage: int=…) -> List[DPRSpanPrediction] - removed:
variable CONTEXT_ENCODER_PRETRAINED_INIT_CONFIGURATION - removed:
variable CONTEXT_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable CONTEXT_ENCODER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable CUSTOM_DPR_READER_DOCSTRING - removed:
variable DPRReaderOutput - removed:
variable DPRSpanPrediction - removed:
variable QUESTION_ENCODER_PRETRAINED_INIT_CONFIGURATION - removed:
variable QUESTION_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable QUESTION_ENCODER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable READER_PRETRAINED_INIT_CONFIGURATION - removed:
variable READER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable READER_PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class ElectraTokenizer(BertTokenizer) - removed:
field ElectraTokenizer.max_model_input_sizes - removed:
field ElectraTokenizer.pretrained_init_configuration - removed:
field ElectraTokenizer.pretrained_vocab_files_map - removed:
field ElectraTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES
- removed:
class ElectraTokenizerFast(BertTokenizerFast) - removed:
field ElectraTokenizerFast.max_model_input_sizes - removed:
field ElectraTokenizerFast.pretrained_init_configuration - removed:
field ElectraTokenizerFast.pretrained_vocab_files_map - removed:
field ElectraTokenizerFast.slow_tokenizer_class - removed:
field ElectraTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES
- removed:
class FlaubertTokenizer(XLMTokenizer) - removed:
constructor FlaubertTokenizer(do_lowercase=…, **kwargs) - removed:
field FlaubertTokenizer.max_model_input_sizes - removed:
field FlaubertTokenizer.pretrained_init_configuration - removed:
field FlaubertTokenizer.pretrained_vocab_files_map - removed:
field FlaubertTokenizer.vocab_files_names - removed:
function convert_to_unicode(text) - removed:
method FlaubertTokenizer.preprocess_text(text) - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class FSMTTokenizer(PreTrainedTokenizer) - removed:
constructor FSMTTokenizer(langs=…, src_vocab_file=…, tgt_vocab_file=…, merges_file=…, do_lower_case=…, unk_token=…, bos_token=…, sep_token=…, pad_token=…, **kwargs) - removed:
field FSMTTokenizer.max_model_input_sizes - removed:
field FSMTTokenizer.pretrained_init_configuration - removed:
field FSMTTokenizer.pretrained_vocab_files_map - removed:
field FSMTTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
function remove_non_printing_char(text) - removed:
function replace_unicode_punct(text) - removed:
method FSMTTokenizer.bpe(token) - removed:
method FSMTTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method FSMTTokenizer.convert_tokens_to_string(tokens) - removed:
method FSMTTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method FSMTTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method FSMTTokenizer.get_src_vocab() - removed:
method FSMTTokenizer.get_tgt_vocab() - removed:
method FSMTTokenizer.get_vocab() -> Dict[str, int] - removed:
method FSMTTokenizer.moses_detokenize(tokens, lang) - removed:
method FSMTTokenizer.moses_pipeline(text, lang) - removed:
method FSMTTokenizer.moses_punct_norm(text, lang) - removed:
method FSMTTokenizer.moses_tokenize(text, lang) - removed:
method FSMTTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, return_tensors: str=…, truncation=…, padding=…, **unused) -> BatchEncoding - removed:
method FSMTTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property FSMTTokenizer.src_vocab_size - removed:
property FSMTTokenizer.tgt_vocab_size - removed:
property FSMTTokenizer.vocab_size -> int - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class FunnelTokenizer(BertTokenizer) - removed:
constructor FunnelTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, bos_token=…, eos_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - removed:
field FunnelTokenizer.cls_token_type_id: int - removed:
field FunnelTokenizer.max_model_input_sizes - removed:
field FunnelTokenizer.pretrained_init_configuration - removed:
field FunnelTokenizer.pretrained_vocab_files_map - removed:
field FunnelTokenizer.vocab_files_names - removed:
method FunnelTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class FunnelTokenizerFast(BertTokenizerFast) - removed:
constructor FunnelTokenizerFast(vocab_file, tokenizer_file=…, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, bos_token=…, eos_token=…, clean_text=…, tokenize_chinese_chars=…, strip_accents=…, wordpieces_prefix=…, **kwargs) - removed:
field FunnelTokenizerFast.cls_token_type_id: int - removed:
field FunnelTokenizerFast.max_model_input_sizes - removed:
field FunnelTokenizerFast.pretrained_init_configuration - removed:
field FunnelTokenizerFast.pretrained_vocab_files_map - removed:
field FunnelTokenizerFast.slow_tokenizer_class - removed:
field FunnelTokenizerFast.vocab_files_names - removed:
method FunnelTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class GPT2Tokenizer(PreTrainedTokenizer) - removed:
constructor GPT2Tokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, **kwargs) - removed:
field GPT2Tokenizer.max_model_input_sizes - removed:
field GPT2Tokenizer.model_input_names - removed:
field GPT2Tokenizer.pretrained_vocab_files_map - removed:
field GPT2Tokenizer.vocab_files_names - removed:
function bytes_to_unicode() - removed:
function get_pairs(word) - removed:
method GPT2Tokenizer.bpe(token) - removed:
method GPT2Tokenizer.convert_tokens_to_string(tokens) - removed:
method GPT2Tokenizer.get_vocab() - removed:
method GPT2Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - removed:
method GPT2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property GPT2Tokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class GPT2TokenizerFast(PreTrainedTokenizerFast) - removed:
constructor GPT2TokenizerFast(vocab_file, merges_file, tokenizer_file=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, **kwargs) - removed:
field GPT2TokenizerFast.max_model_input_sizes - removed:
field GPT2TokenizerFast.model_input_names - removed:
field GPT2TokenizerFast.pretrained_vocab_files_map - removed:
field GPT2TokenizerFast.slow_tokenizer_class - removed:
field GPT2TokenizerFast.vocab_files_names - removed:
method GPT2TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class HerbertTokenizer(XLMTokenizer) - removed:
constructor HerbertTokenizer(**kwargs) - removed:
field HerbertTokenizer.max_model_input_sizes - removed:
field HerbertTokenizer.pretrained_init_configuration - removed:
field HerbertTokenizer.pretrained_vocab_files_map - removed:
field HerbertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class HerbertTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor HerbertTokenizerFast(vocab_file, merges_file, tokenizer_file=…, **kwargs) - removed:
field HerbertTokenizerFast.max_model_input_sizes - removed:
field HerbertTokenizerFast.pretrained_init_configuration - removed:
field HerbertTokenizerFast.pretrained_vocab_files_map - removed:
field HerbertTokenizerFast.slow_tokenizer_class - removed:
field HerbertTokenizerFast.vocab_files_names - removed:
method HerbertTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method HerbertTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method HerbertTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method HerbertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class LayoutLMTokenizer(BertTokenizer) - removed:
field LayoutLMTokenizer.max_model_input_sizes - removed:
field LayoutLMTokenizer.pretrained_init_configuration - removed:
field LayoutLMTokenizer.pretrained_vocab_files_map - removed:
field LayoutLMTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class LayoutLMTokenizerFast(BertTokenizerFast) - removed:
field LayoutLMTokenizerFast.max_model_input_sizes - removed:
field LayoutLMTokenizerFast.pretrained_init_configuration - removed:
field LayoutLMTokenizerFast.pretrained_vocab_files_map - removed:
field LayoutLMTokenizerFast.slow_tokenizer_class - removed:
field LayoutLMTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class LongformerTokenizer(RobertaTokenizer) - removed:
field LongformerTokenizer.max_model_input_sizes - removed:
field LongformerTokenizer.pretrained_vocab_files_map - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable logger - removed:
variable merges_url - removed:
variable vocab_url
- removed:
class LongformerTokenizerFast(RobertaTokenizerFast) - removed:
field LongformerTokenizerFast.max_model_input_sizes - removed:
field LongformerTokenizerFast.pretrained_vocab_files_map - removed:
field LongformerTokenizerFast.slow_tokenizer_class - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable logger - removed:
variable merges_url - removed:
variable tokenizer_url - removed:
variable vocab_url
- removed:
class LxmertTokenizer(BertTokenizer) - removed:
field LxmertTokenizer.max_model_input_sizes - removed:
field LxmertTokenizer.pretrained_init_configuration - removed:
field LxmertTokenizer.pretrained_vocab_files_map - removed:
field LxmertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES
- removed:
class LxmertTokenizerFast(BertTokenizerFast) - removed:
field LxmertTokenizerFast.max_model_input_sizes - removed:
field LxmertTokenizerFast.pretrained_init_configuration - removed:
field LxmertTokenizerFast.pretrained_vocab_files_map - removed:
field LxmertTokenizerFast.slow_tokenizer_class - removed:
field LxmertTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES
- removed:
class MarianTokenizer(PreTrainedTokenizer) - removed:
constructor MarianTokenizer(vocab, source_spm, target_spm, source_lang=…, target_lang=…, unk_token=…, eos_token=…, pad_token=…, model_max_length=…, **kwargs) - removed:
field MarianTokenizer.language_code_re - removed:
field MarianTokenizer.max_model_input_sizes - removed:
field MarianTokenizer.model_input_names - removed:
field MarianTokenizer.pretrained_init_configuration - removed:
field MarianTokenizer.pretrained_vocab_files_map - removed:
field MarianTokenizer.vocab_files_names - removed:
function load_json(path: str) -> Union[Dict, List] - removed:
function load_spm(path: str) -> sentencepiece.SentencePieceProcessor - removed:
function save_json(data, path: str) -> None - removed:
method MarianTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> List[int] - removed:
method MarianTokenizer.convert_tokens_to_string(tokens: List[str]) -> str - removed:
method MarianTokenizer.get_special_tokens_mask(token_ids_0: List, token_ids_1: Optional[List]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method MarianTokenizer.get_vocab() -> Dict - removed:
method MarianTokenizer.normalize(x: str) -> str - removed:
method MarianTokenizer.num_special_tokens_to_add(**unused) - removed:
method MarianTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, return_tensors: str=…, truncation=…, padding=…, **unused) -> BatchEncoding - removed:
method MarianTokenizer.remove_language_code(text: str) - removed:
method MarianTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property MarianTokenizer.vocab_size -> int - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable vocab_files_names
- removed:
class MBartTokenizer(XLMRobertaTokenizer) - removed:
constructor MBartTokenizer(*args, tokenizer_file=…, **kwargs) - removed:
field MBartTokenizer.max_model_input_sizes - removed:
field MBartTokenizer.prefix_tokens: List[int] - removed:
field MBartTokenizer.pretrained_vocab_files_map - removed:
field MBartTokenizer.suffix_tokens: List[int] - removed:
field MBartTokenizer.vocab_files_names - removed:
method MBartTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method MBartTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method MBartTokenizer.prepare_seq2seq_batch(src_texts: List[str], src_lang: str=…, tgt_texts: Optional[List[str]]=…, tgt_lang: str=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, truncation: bool=…, padding: str=…, return_tensors: str=…, add_prefix_space: bool=…, **kwargs) -> BatchEncoding - removed:
method MBartTokenizer.set_src_lang_special_tokens(src_lang) -> None - removed:
method MBartTokenizer.set_tgt_lang_special_tokens(lang: str) -> None - removed:
property MBartTokenizer.vocab_size - removed:
variable FAIRSEQ_LANGUAGE_CODES - removed:
variable SPM_URL - removed:
variable logger
- removed:
class MBartTokenizerFast(XLMRobertaTokenizerFast) - removed:
constructor MBartTokenizerFast(*args, tokenizer_file=…, **kwargs) - removed:
field MBartTokenizerFast.max_model_input_sizes - removed:
field MBartTokenizerFast.prefix_tokens: List[int] - removed:
field MBartTokenizerFast.pretrained_vocab_files_map - removed:
field MBartTokenizerFast.slow_tokenizer_class - removed:
field MBartTokenizerFast.suffix_tokens: List[int] - removed:
field MBartTokenizerFast.vocab_files_names - removed:
method MBartTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method MBartTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method MBartTokenizerFast.prepare_seq2seq_batch(src_texts: List[str], src_lang: str=…, tgt_texts: Optional[List[str]]=…, tgt_lang: str=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, truncation: bool=…, padding: str=…, return_tensors: str=…, **kwargs) -> BatchEncoding - removed:
method MBartTokenizerFast.set_src_lang_special_tokens(src_lang) -> None - removed:
method MBartTokenizerFast.set_tgt_lang_special_tokens(lang: str) -> None - removed:
variable FAIRSEQ_LANGUAGE_CODES - removed:
variable MBartTokenizer - removed:
variable SPM_URL - removed:
variable logger - removed:
variable tokenizer_URL
- added:
class MistralCommonTokenizer(PushToHubMixin) - added:
class MistralTokenizerType(str, Enum) - added:
constructor MistralCommonTokenizer(tokenizer_path: Union[str, os.PathLike, Path], mode: ValidationMode=…, model_max_length: int=…, padding_side: str=…, truncation_side: str=…, model_input_names: Optional[list[str]]=…, clean_up_tokenization_spaces: bool=…, **kwargs) - added:
field MistralCommonTokenizer.model_input_names: list[str] - added:
field MistralCommonTokenizer.padding_side: str - added:
field MistralCommonTokenizer.truncation_side: str - added:
field MistralTokenizerType.spm - added:
field MistralTokenizerType.tekken - added:
method MistralCommonTokenizer.apply_chat_template(conversation: Union[list[dict[str, str]], list[list[dict[str, str]]]], tools: Optional[list[Union[dict, Callable]]]=…, continue_final_message: bool=…, tokenize: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: bool=…, max_length: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_dict: bool=…, **kwargs) ->… - added:
method MistralCommonTokenizer.batch_decode(sequences: Union[list[int], list[list[int]], np.ndarray, torch.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, **kwargs) -> list[str] - added:
method MistralCommonTokenizer.convert_ids_to_tokens(ids: int, skip_special_tokens: bool=…) -> str | (ids: list[int], skip_special_tokens: bool=…) -> list[str] - added:
method MistralCommonTokenizer.convert_tokens_to_ids(tokens: Union[str, list[str]]) -> Union[int, list[int]] - added:
method MistralCommonTokenizer.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, **kwargs) -> str - added:
method MistralCommonTokenizer.encode(text: Union[TextInput, EncodedInput], text_pair: None=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_tensors: Optional[Union[str, TensorType]]=…, verbose:… - added:
method MistralCommonTokenizer.get_special_tokens_mask(token_ids_0: list, token_ids_1: None=…, already_has_special_tokens: bool=…) -> list[int] - added:
method MistralCommonTokenizer.get_vocab() -> dict[str, int] - added:
method MistralCommonTokenizer.pad(encoded_inputs: Union[BatchEncoding, list[BatchEncoding], dict[str, EncodedInput], dict[str, list[EncodedInput]], list[dict[str, EncodedInput]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_attention_mask: Optional[bool]=…, return_tensors: Optional[Union[str,… - added:
method MistralCommonTokenizer.prepare_for_model(ids: list[int], pair_ids: None=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_attention_mask… - added:
method MistralCommonTokenizer.save_pretrained(save_directory: Union[str, os.PathLike, Path], push_to_hub: bool=…, token: Optional[Union[str, bool]]=…, commit_message: Optional[str]=…, repo_id: Optional[str]=…, private: Optional[bool]=…, repo_url: Optional[str]=…, organization: Optional[str]=…, **kwargs) -> tuple[str, ...] - added:
method MistralCommonTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
method MistralCommonTokenizer.truncate_sequences(ids: list[int], pair_ids: None=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…, **kwargs) -> tuple[list[int], None, list[int]] - added:
property MistralCommonTokenizer.bos_token -> str - added:
property MistralCommonTokenizer.bos_token_id -> int - added:
property MistralCommonTokenizer.eos_token -> str - added:
property MistralCommonTokenizer.eos_token_id -> int - added:
property MistralCommonTokenizer.pad_token -> str - added:
property MistralCommonTokenizer.pad_token_id -> int - added:
property MistralCommonTokenizer.unk_token -> str - added:
property MistralCommonTokenizer.unk_token_id -> int - added:
property MistralCommonTokenizer.vocab_size -> int - added:
static method MistralCommonTokenizer.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], *init_inputs, mode: ValidationMode=…, cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, model_max_length: int=…, padding_side: str=…, truncation_side: str=…, model_input_names: Optional[li… - added:
variable ENCODE_KWARGS_DOCSTRING - added:
variable ENCODE_PLUS_ADDITIONAL_KWARGS_DOCSTRING - added:
variable logger
- removed:
class MobileBertTokenizer(BertTokenizer) - removed:
field MobileBertTokenizer.max_model_input_sizes - removed:
field MobileBertTokenizer.pretrained_init_configuration - removed:
field MobileBertTokenizer.pretrained_vocab_files_map - removed:
field MobileBertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class MobileBertTokenizerFast(BertTokenizerFast) - removed:
field MobileBertTokenizerFast.max_model_input_sizes - removed:
field MobileBertTokenizerFast.pretrained_init_configuration - removed:
field MobileBertTokenizerFast.pretrained_vocab_files_map - removed:
field MobileBertTokenizerFast.slow_tokenizer_class - removed:
field MobileBertTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class OpenAIGPTTokenizer(PreTrainedTokenizer) - removed:
constructor OpenAIGPTTokenizer(vocab_file, merges_file, unk_token=…, **kwargs) - removed:
field OpenAIGPTTokenizer.max_model_input_sizes - removed:
field OpenAIGPTTokenizer.model_input_names - removed:
field OpenAIGPTTokenizer.pretrained_vocab_files_map - removed:
field OpenAIGPTTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
function text_standardize(text) - removed:
method OpenAIGPTTokenizer.bpe(token) - removed:
method OpenAIGPTTokenizer.convert_tokens_to_string(tokens) - removed:
method OpenAIGPTTokenizer.get_vocab() - removed:
method OpenAIGPTTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property OpenAIGPTTokenizer.do_lower_case - removed:
property OpenAIGPTTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class OpenAIGPTTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor OpenAIGPTTokenizerFast(vocab_file, merges_file, tokenizer_file=…, unk_token=…, **kwargs) - removed:
field OpenAIGPTTokenizerFast.max_model_input_sizes - removed:
field OpenAIGPTTokenizerFast.model_input_names - removed:
field OpenAIGPTTokenizerFast.pretrained_vocab_files_map - removed:
field OpenAIGPTTokenizerFast.slow_tokenizer_class - removed:
field OpenAIGPTTokenizerFast.vocab_files_names - removed:
method OpenAIGPTTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property OpenAIGPTTokenizerFast.do_lower_case - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class PegasusTokenizer(ReformerTokenizer) - removed:
constructor PegasusTokenizer(*args, pad_token=…, **kwargs) - removed:
field PegasusTokenizer.max_model_input_sizes - removed:
field PegasusTokenizer.offset - removed:
field PegasusTokenizer.pretrained_vocab_files_map - removed:
field PegasusTokenizer.vocab_files_names - removed:
method PegasusTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> List[int] - removed:
method PegasusTokenizer.get_special_tokens_mask(token_ids_0: List, token_ids_1: Optional[List]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method PegasusTokenizer.num_special_tokens_to_add(pair=…) - removed:
method PegasusTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, return_tensors: str=…, truncation=…, padding=…, **unused) -> BatchEncoding - removed:
property PegasusTokenizer.vocab_size -> int - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES
- removed:
class PegasusTokenizerFast(ReformerTokenizerFast) - removed:
field PegasusTokenizerFast.max_model_input_sizes - removed:
field PegasusTokenizerFast.offset - removed:
field PegasusTokenizerFast.pretrained_vocab_files_map - removed:
field PegasusTokenizerFast.slow_tokenizer_class - removed:
field PegasusTokenizerFast.vocab_files_names - removed:
method PegasusTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) -> List[int] - removed:
method PegasusTokenizerFast.get_special_tokens_mask(token_ids_0: List, token_ids_1: Optional[List]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method PegasusTokenizerFast.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, return_tensors: str=…, truncation=…, padding=…, **unused) -> BatchEncoding - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable PegasusTokenizer - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES
- removed:
class PhobertTokenizer(PreTrainedTokenizer) - removed:
constructor PhobertTokenizer(vocab_file, merges_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - removed:
field PhobertTokenizer.max_model_input_sizes - removed:
field PhobertTokenizer.pretrained_vocab_files_map - removed:
field PhobertTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
method PhobertTokenizer.add_from_file(f) - removed:
method PhobertTokenizer.bpe(token) - removed:
method PhobertTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method PhobertTokenizer.convert_tokens_to_string(tokens) - removed:
method PhobertTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method PhobertTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method PhobertTokenizer.get_vocab() - removed:
method PhobertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property PhobertTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class ProphetNetTokenizer(PreTrainedTokenizer) - removed:
constructor ProphetNetTokenizer(vocab_file, do_lower_case=…, do_basic_tokenize=…, never_split=…, unk_token=…, sep_token=…, x_sep_token=…, pad_token=…, mask_token=…, tokenize_chinese_chars=…, strip_accents=…, **kwargs) - removed:
field ProphetNetTokenizer.max_model_input_sizes - removed:
field ProphetNetTokenizer.pretrained_init_configuration - removed:
field ProphetNetTokenizer.pretrained_vocab_files_map - removed:
field ProphetNetTokenizer.vocab_files_names - removed:
function load_vocab(vocab_file) - removed:
method ProphetNetTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method ProphetNetTokenizer.convert_tokens_to_string(tokens) - removed:
method ProphetNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method ProphetNetTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method ProphetNetTokenizer.get_vocab() - removed:
method ProphetNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property ProphetNetTokenizer.vocab_size - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class RagTokenizer - removed:
constructor RagTokenizer(question_encoder, generator) - removed:
method RagTokenizer.batch_decode(*args, **kwargs) - removed:
method RagTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation=…, **kwargs) -> BatchEncoding - removed:
method RagTokenizer.save_pretrained(save_directory) - removed:
static method RagTokenizer.from_pretrained(pretrained_model_name_or_path, **kwargs) - removed:
variable logger
- removed:
class ReformerTokenizer(PreTrainedTokenizer) - removed:
constructor ReformerTokenizer(vocab_file, eos_token=…, unk_token=…, additional_special_tokens=…, **kwargs) - removed:
field ReformerTokenizer.max_model_input_sizes - removed:
field ReformerTokenizer.model_input_names - removed:
field ReformerTokenizer.pretrained_vocab_files_map - removed:
field ReformerTokenizer.vocab_files_names - removed:
method ReformerTokenizer.convert_tokens_to_string(tokens) - removed:
method ReformerTokenizer.get_vocab() -> Dict[str, int] - removed:
method ReformerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property ReformerTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class ReformerTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor ReformerTokenizerFast(vocab_file, tokenizer_file=…, eos_token=…, unk_token=…, additional_special_tokens=…, **kwargs) - removed:
field ReformerTokenizerFast.max_model_input_sizes - removed:
field ReformerTokenizerFast.model_input_names - removed:
field ReformerTokenizerFast.pretrained_vocab_files_map - removed:
field ReformerTokenizerFast.slow_tokenizer_class - removed:
field ReformerTokenizerFast.vocab_files_names - removed:
method ReformerTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable ReformerTokenizer - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class RetriBertTokenizer(BertTokenizer) - removed:
field RetriBertTokenizer.max_model_input_sizes - removed:
field RetriBertTokenizer.model_input_names - removed:
field RetriBertTokenizer.pretrained_init_configuration - removed:
field RetriBertTokenizer.pretrained_vocab_files_map - removed:
field RetriBertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class RetriBertTokenizerFast(BertTokenizerFast) - removed:
field RetriBertTokenizerFast.max_model_input_sizes - removed:
field RetriBertTokenizerFast.model_input_names - removed:
field RetriBertTokenizerFast.pretrained_init_configuration - removed:
field RetriBertTokenizerFast.pretrained_vocab_files_map - removed:
field RetriBertTokenizerFast.slow_tokenizer_class - removed:
field RetriBertTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class RobertaTokenizer(GPT2Tokenizer) - removed:
constructor RobertaTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - removed:
field RobertaTokenizer.max_model_input_sizes - removed:
field RobertaTokenizer.model_input_names - removed:
field RobertaTokenizer.pretrained_vocab_files_map - removed:
field RobertaTokenizer.vocab_files_names - removed:
method RobertaTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method RobertaTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method RobertaTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method RobertaTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class RobertaTokenizerFast(GPT2TokenizerFast) - removed:
constructor RobertaTokenizerFast(vocab_file, merges_file, tokenizer_file=…, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, **kwargs) - removed:
field RobertaTokenizerFast.max_model_input_sizes - removed:
field RobertaTokenizerFast.model_input_names - removed:
field RobertaTokenizerFast.pretrained_vocab_files_map - removed:
field RobertaTokenizerFast.slow_tokenizer_class - removed:
field RobertaTokenizerFast.vocab_files_names - removed:
method RobertaTokenizerFast.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - removed:
method RobertaTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class SqueezeBertTokenizer(BertTokenizer) - removed:
field SqueezeBertTokenizer.max_model_input_sizes - removed:
field SqueezeBertTokenizer.pretrained_init_configuration - removed:
field SqueezeBertTokenizer.pretrained_vocab_files_map - removed:
field SqueezeBertTokenizer.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class SqueezeBertTokenizerFast(BertTokenizerFast) - removed:
field SqueezeBertTokenizerFast.max_model_input_sizes - removed:
field SqueezeBertTokenizerFast.pretrained_init_configuration - removed:
field SqueezeBertTokenizerFast.pretrained_vocab_files_map - removed:
field SqueezeBertTokenizerFast.slow_tokenizer_class - removed:
field SqueezeBertTokenizerFast.vocab_files_names - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class T5Tokenizer(PreTrainedTokenizer) - removed:
constructor T5Tokenizer(vocab_file, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, **kwargs) - removed:
field T5Tokenizer.max_model_input_sizes - removed:
field T5Tokenizer.model_input_names - removed:
field T5Tokenizer.pretrained_vocab_files_map - removed:
field T5Tokenizer.vocab_files_names - removed:
method T5Tokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method T5Tokenizer.convert_tokens_to_string(tokens) - removed:
method T5Tokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method T5Tokenizer.get_vocab() - removed:
method T5Tokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation: bool=…, **kwargs) -> BatchEncoding - removed:
method T5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property T5Tokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class T5TokenizerFast(PreTrainedTokenizerFast) - removed:
constructor T5TokenizerFast(vocab_file, tokenizer_file=…, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, **kwargs) - removed:
field T5TokenizerFast.max_model_input_sizes - removed:
field T5TokenizerFast.model_input_names - removed:
field T5TokenizerFast.prefix_tokens: List[int] - removed:
field T5TokenizerFast.pretrained_vocab_files_map - removed:
field T5TokenizerFast.slow_tokenizer_class - removed:
field T5TokenizerFast.vocab_files_names - removed:
method T5TokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method T5TokenizerFast.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation: bool=…, **kwargs) -> BatchEncoding - removed:
method T5TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable T5Tokenizer - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class LMMultiFileIterator(LMShuffledIterator) - removed:
class LMOrderedIterator(object) - removed:
class LMShuffledIterator(object) - removed:
class TransfoXLCorpus(object) - removed:
class TransfoXLTokenizer(PreTrainedTokenizer) - removed:
constructor LMMultiFileIterator(paths, vocab, bsz, bptt, device=…, ext_len=…, shuffle=…) - removed:
constructor LMOrderedIterator(data, bsz, bptt, device=…, ext_len=…) - removed:
constructor LMShuffledIterator(data, bsz, bptt, device=…, ext_len=…, shuffle=…) - removed:
constructor TransfoXLCorpus(*args, **kwargs) - removed:
constructor TransfoXLTokenizer(special=…, min_freq=…, max_size=…, lower_case=…, delimiter=…, vocab_file=…, pretrained_vocab_file: str=…, never_split=…, unk_token=…, eos_token=…, additional_special_tokens=…, language=…, **kwargs) - removed:
field TransfoXLTokenizer.max_model_input_sizes - removed:
field TransfoXLTokenizer.model_input_names - removed:
field TransfoXLTokenizer.pretrained_vocab_files_map - removed:
field TransfoXLTokenizer.vocab_files_names - removed:
function detokenize_numbers(text: str) -> str - removed:
function get_lm_corpus(datadir, dataset) - removed:
function tokenize_numbers(text_array: List[str]) -> List[str] - removed:
method LMMultiFileIterator.get_sent_stream(path) - removed:
method LMOrderedIterator.get_batch(i, bptt=…) - removed:
method LMOrderedIterator.get_fixlen_iter(start=…) - removed:
method LMOrderedIterator.get_varlen_iter(start=…, std=…, min_len=…, max_deviation=…) - removed:
method LMShuffledIterator.get_sent_stream() - removed:
method LMShuffledIterator.stream_iterator(sent_stream) - removed:
method TransfoXLCorpus.build_corpus(path, dataset) - removed:
method TransfoXLCorpus.get_iterator(split, *args, **kwargs) - removed:
method TransfoXLTokenizer.add_special(sym) - removed:
method TransfoXLTokenizer.add_symbol(sym) - removed:
method TransfoXLTokenizer.build_vocab() - removed:
method TransfoXLTokenizer.convert_to_tensor(symbols) - removed:
method TransfoXLTokenizer.convert_tokens_to_string(tokens) - removed:
method TransfoXLTokenizer.count_file(path, verbose=…, add_eos=…) - removed:
method TransfoXLTokenizer.count_sents(sents, verbose=…) - removed:
method TransfoXLTokenizer.encode_file(path, ordered=…, verbose=…, add_eos=…, add_double_eos=…) - removed:
method TransfoXLTokenizer.encode_sents(sents, ordered=…, verbose=…) - removed:
method TransfoXLTokenizer.get_vocab() - removed:
method TransfoXLTokenizer.moses_pipeline(text: str) -> List[str] - removed:
method TransfoXLTokenizer.moses_punct_norm(text) - removed:
method TransfoXLTokenizer.moses_tokenize(text) - removed:
method TransfoXLTokenizer.move_added_token(token: str, target_idx: int) - removed:
method TransfoXLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property TransfoXLTokenizer.do_lower_case - removed:
property TransfoXLTokenizer.vocab_size - removed:
static method TransfoXLCorpus.from_pretrained(pretrained_model_name_or_path, cache_dir=…, *inputs, **kwargs) - removed:
variable CORPUS_NAME - removed:
variable DETOKENIZE_NUMBERS - removed:
variable MATCH_NUMBERS - removed:
variable PRETRAINED_CORPUS_ARCHIVE_MAP - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
method PreTrainedTokenizer.prepare_seq2seq_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: str=…, truncation=…, **kwargs) -> BatchEncoding - changed:
method PreTrainedTokenizer.convert_ids_to_tokens(ids: int, skip_special_tokens: bool=…) -> str | (ids: List[int], skip_special_tokens: bool=…) -> List[str] → method PreTrainedTokenizer.convert_ids_to_tokens(ids: int, skip_special_tokens: bool=…) -> str | (ids: list[int], skip_special_tokens: bool=…) -> list[str] - changed:
method PreTrainedTokenizer.convert_tokens_to_ids(tokens: Union[str, List[str]]) -> Union[int, List[int]] → method PreTrainedTokenizer.convert_tokens_to_ids(tokens: Union[str, list[str]]) -> Union[int, list[int]] - changed:
method PreTrainedTokenizer.convert_tokens_to_string(tokens: List[str]) -> str → method PreTrainedTokenizer.convert_tokens_to_string(tokens: list[str]) -> str - changed:
method PreTrainedTokenizer.get_added_vocab() -> Dict[str, int] → method PreTrainedTokenizer.get_added_vocab() -> dict[str, int] - changed:
method PreTrainedTokenizer.get_special_tokens_mask(token_ids_0: List, token_ids_1: Optional[List]=…, already_has_special_tokens: bool=…) -> List[int] → method PreTrainedTokenizer.get_special_tokens_mask(token_ids_0: list, token_ids_1: Optional[list]=…, already_has_special_tokens: bool=…) -> list[int] - changed:
method PreTrainedTokenizer.prepare_for_tokenization(text: str, is_split_into_words: bool=…, **kwargs) -> Tuple[str, Dict[str, Any]] → method PreTrainedTokenizer.prepare_for_tokenization(text: str, is_split_into_words: bool=…, **kwargs) -> tuple[str, dict[str, Any]] - changed:
method PreTrainedTokenizer.tokenize(text: TextInput, **kwargs) -> List[str] → method PreTrainedTokenizer.tokenize(text: TextInput, **kwargs) -> list[str] - added:
class ExtensionsTrie(Trie) - added:
class Trie - added:
constructor ExtensionsTrie(*args) - added:
constructor Trie(*args) - added:
method ExtensionsTrie.extensions(prefix: str) - added:
method Trie.add(word: str) - added:
method Trie.cut_text(text, offsets) - added:
method Trie.split(text: str) -> list[str] - added:
method Trie.update(*args) - added:
property PreTrainedTokenizer.added_tokens_decoder -> dict[int, AddedToken] - added:
property PreTrainedTokenizer.added_tokens_encoder -> dict[str, int]
- removed:
class EncodingFast - removed:
class ExplicitEnum(Enum) - removed:
class PaddingStrategy(ExplicitEnum) - removed:
class TensorType(ExplicitEnum) - removed:
field AddedToken.content: str - removed:
field AddedToken.lstrip: bool - removed:
field AddedToken.normalized: bool - removed:
field AddedToken.rstrip: bool - removed:
field AddedToken.single_word: bool - removed:
field PaddingStrategy.DO_NOT_PAD - removed:
field PaddingStrategy.LONGEST - removed:
field PaddingStrategy.MAX_LENGTH - removed:
field PreTrainedTokenizerBase.max_model_input_sizes: Dict[str, Optional[int]] - removed:
field PreTrainedTokenizerBase.pretrained_init_configuration: Dict[str, Dict[str, Any]] - removed:
field TensorType.JAX - removed:
field TensorType.NUMPY - removed:
field TensorType.PYTORCH - removed:
field TensorType.TENSORFLOW - removed:
function to_py_obj(obj) - removed:
method BatchEncoding.items() - removed:
method BatchEncoding.keys() - removed:
method BatchEncoding.values() - removed:
property BatchEncoding.encodings -> Optional[List[EncodingFast]] - removed:
property PreTrainedTokenizerBase.max_len -> int - removed:
property SpecialTokensMixin.additional_special_tokens -> List[str] - removed:
property SpecialTokensMixin.additional_special_tokens_ids -> List[int] - removed:
property SpecialTokensMixin.all_special_ids -> List[int] - removed:
property SpecialTokensMixin.all_special_tokens -> List[str] - removed:
property SpecialTokensMixin.all_special_tokens_extended -> List[Union[str, AddedToken]] - removed:
property SpecialTokensMixin.bos_token -> str - removed:
property SpecialTokensMixin.bos_token_id -> Optional[int] - removed:
property SpecialTokensMixin.cls_token -> str - removed:
property SpecialTokensMixin.cls_token_id -> Optional[int] - removed:
property SpecialTokensMixin.eos_token -> str - removed:
property SpecialTokensMixin.eos_token_id -> Optional[int] - removed:
property SpecialTokensMixin.mask_token -> str - removed:
property SpecialTokensMixin.mask_token_id -> Optional[int] - removed:
property SpecialTokensMixin.pad_token -> str - removed:
property SpecialTokensMixin.pad_token_id -> Optional[int] - removed:
property SpecialTokensMixin.sep_token -> str - removed:
property SpecialTokensMixin.sep_token_id -> Optional[int] - removed:
property SpecialTokensMixin.special_tokens_map -> Dict[str, Union[str, List[str]]] - removed:
property SpecialTokensMixin.special_tokens_map_extended -> Dict[str, Union[str, AddedToken, List[Union[str, AddedToken]]]] - removed:
property SpecialTokensMixin.unk_token -> str - removed:
property SpecialTokensMixin.unk_token_id -> Optional[int] - removed:
variable PREPARE_SEQ2SEQ_BATCH_DOCSTRING - changed:
class PreTrainedTokenizerBase(SpecialTokensMixin) → class PreTrainedTokenizerBase(SpecialTokensMixin, PushToHubMixin) - changed:
constructor BatchEncoding(data: Optional[Dict[str, Any]]=…, encoding: Optional[Union[EncodingFast, Sequence[EncodingFast]]]=…, tensor_type: Union[None, str, TensorType]=…, prepend_batch_axis: bool=…) → constructor BatchEncoding(data: Optional[dict[str, Any]]=…, encoding: Optional[Union[EncodingFast, Sequence[EncodingFast]]]=…, tensor_type: Union[None, str, TensorType]=…, prepend_batch_axis: bool=…, n_sequences: Optional[int]=…) - changed:
field PreTrainedTokenizerBase.model_input_names: List[str] → field PreTrainedTokenizerBase.model_input_names: list[str] - changed:
field PreTrainedTokenizerBase.pretrained_vocab_files_map: Dict[str, Dict[str, str]] → field PreTrainedTokenizerBase.pretrained_vocab_files_map: dict[str, dict[str, str]] - changed:
field PreTrainedTokenizerBase.vocab_files_names: Dict[str, str] → field PreTrainedTokenizerBase.vocab_files_names: dict[str, str] - changed:
method BatchEncoding.char_to_token(batch_or_char_index: int, char_index: Optional[int]=…) -> int → method BatchEncoding.char_to_token(batch_or_char_index: int, char_index: Optional[int]=…, sequence_index: int=…) -> int - changed:
method BatchEncoding.char_to_word(batch_or_char_index: int, char_index: Optional[int]=…) -> int → method BatchEncoding.char_to_word(batch_or_char_index: int, char_index: Optional[int]=…, sequence_index: int=…) -> int - changed:
method BatchEncoding.to(device: Union[str, torch.device]) -> BatchEncoding → method BatchEncoding.to(device: Union[str, torch.device], *, non_blocking: bool=…) -> BatchEncoding - changed:
method BatchEncoding.token_to_chars(batch_or_token_index: int, token_index: Optional[int]=…) -> CharSpan → method BatchEncoding.token_to_chars(batch_or_token_index: int, token_index: Optional[int]=…) -> Optional[CharSpan] - changed:
method BatchEncoding.tokens(batch_index: int=…) -> List[str] → method BatchEncoding.tokens(batch_index: int=…) -> list[str] - changed:
method BatchEncoding.word_to_chars(batch_or_word_index: int, word_index: Optional[int]=…) -> CharSpan → method BatchEncoding.word_to_chars(batch_or_word_index: int, word_index: Optional[int]=…, sequence_index: int=…) -> CharSpan - changed:
method BatchEncoding.word_to_tokens(batch_or_word_index: int, word_index: Optional[int]=…) -> Optional[TokenSpan] → method BatchEncoding.word_to_tokens(batch_or_word_index: int, word_index: Optional[int]=…, sequence_index: int=…) -> Optional[TokenSpan] - changed:
method BatchEncoding.words(batch_index: int=…) -> List[Optional[int]] → method BatchEncoding.words(batch_index: int=…) -> list[Optional[int]] - changed:
method PreTrainedTokenizerBase.batch_decode(sequences: Union[List[int], List[List[int]], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: bool=…, **kwargs) -> List[str] → method PreTrainedTokenizerBase.batch_decode(sequences: Union[list[int], list[list[int]], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, **kwargs) -> list[str] - changed:
method PreTrainedTokenizerBase.batch_encode_plus(batch_text_or_text_pairs: Union[List[TextInput], List[TextInputPair], List[PreTokenizedInput], List[PreTokenizedInputPair], List[EncodedInput], List[EncodedInputPair]], add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, is_split_in… → method PreTrainedTokenizerBase.batch_encode_plus(batch_text_or_text_pairs: Union[list[TextInput], list[TextInputPair], list[PreTokenizedInput], list[PreTokenizedInputPair], list[EncodedInput], list[EncodedInputPair]], add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, is_sp… - changed:
method PreTrainedTokenizerBase.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] → method PreTrainedTokenizerBase.build_inputs_with_special_tokens(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - changed:
method PreTrainedTokenizerBase.convert_tokens_to_string(tokens: List[str]) -> str → method PreTrainedTokenizerBase.convert_tokens_to_string(tokens: list[str]) -> str - changed:
method PreTrainedTokenizerBase.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] → method PreTrainedTokenizerBase.create_token_type_ids_from_sequences(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…) -> list[int] - changed:
method PreTrainedTokenizerBase.decode(token_ids: Union[int, List[int], np.ndarray, torch.Tensor, tf.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: bool=…, **kwargs) -> str → method PreTrainedTokenizerBase.decode(token_ids: Union[int, list[int], np.ndarray, torch.Tensor], skip_special_tokens: bool=…, clean_up_tokenization_spaces: Optional[bool]=…, **kwargs) -> str - changed:
method PreTrainedTokenizerBase.encode(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, return_tensors: Optional[Union[str, TensorType]]=…, **kwargs… → method PreTrainedTokenizerBase.encode(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, padding_side: Optional[str]=…, return_tensors: Optiona… - changed:
method PreTrainedTokenizerBase.encode_plus(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, is_split_into_words: bool=…, pad_to_multiple_of: Option… → method PreTrainedTokenizerBase.encode_plus(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, is_split_into_words: bool=…, pad_to_multiple_of:… - changed:
method PreTrainedTokenizerBase.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] → method PreTrainedTokenizerBase.get_special_tokens_mask(token_ids_0: list[int], token_ids_1: Optional[list[int]]=…, already_has_special_tokens: bool=…) -> list[int] - changed:
method PreTrainedTokenizerBase.get_vocab() -> Dict[str, int] → method PreTrainedTokenizerBase.get_vocab() -> dict[str, int] - changed:
method PreTrainedTokenizerBase.pad(encoded_inputs: Union[BatchEncoding, List[BatchEncoding], Dict[str, EncodedInput], Dict[str, List[EncodedInput]], List[Dict[str, EncodedInput]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, return_attention_mask: Optional[bool]=…, return_tensors: Optional[Union[str, TensorType]]=…, verbose: bool… → method PreTrainedTokenizerBase.pad(encoded_inputs: Union[BatchEncoding, list[BatchEncoding], dict[str, EncodedInput], dict[str, list[EncodedInput]], list[dict[str, EncodedInput]]], padding: Union[bool, str, PaddingStrategy]=…, max_length: Optional[int]=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_attention_mask: Optional[bool]=…, return_tensors: Optional[Union[str… - changed:
method PreTrainedTokenizerBase.prepare_for_model(ids: List[int], pair_ids: Optional[List[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, return_tensors: Optional[Union[str, TensorType]]=…, return_token_type_ids: Optional[bool]=…, r… → method PreTrainedTokenizerBase.prepare_for_model(ids: list[int], pair_ids: Optional[list[int]]=…, add_special_tokens: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: Union[bool, str, TruncationStrategy, None]=…, max_length: Optional[int]=…, stride: int=…, pad_to_multiple_of: Optional[int]=…, padding_side: Optional[str]=…, return_tensors: Optional[Union[str, TensorType]]=…, retur… - changed:
method PreTrainedTokenizerBase.save_pretrained(save_directory: str, legacy_format: bool=…, filename_prefix: Optional[str]=…) -> Tuple[str] → method PreTrainedTokenizerBase.save_pretrained(save_directory: Union[str, os.PathLike], legacy_format: Optional[bool]=…, filename_prefix: Optional[str]=…, push_to_hub: bool=…, **kwargs) -> tuple[str, ...] - changed:
method PreTrainedTokenizerBase.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] → method PreTrainedTokenizerBase.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> tuple[str, ...] - changed:
method PreTrainedTokenizerBase.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> List[str] → method PreTrainedTokenizerBase.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str] - changed:
method PreTrainedTokenizerBase.truncate_sequences(ids: List[int], pair_ids: Optional[List[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> Tuple[List[int], List[int], List[int]] → method PreTrainedTokenizerBase.truncate_sequences(ids: list[int], pair_ids: Optional[list[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: Union[str, TruncationStrategy]=…, stride: int=…) -> tuple[list[int], list[int], list[int]] - changed:
method SpecialTokensMixin.add_special_tokens(special_tokens_dict: Dict[str, Union[str, AddedToken]]) -> int → method SpecialTokensMixin.add_special_tokens(special_tokens_dict: dict[str, Union[str, AddedToken, Sequence[Union[str, AddedToken]]]], replace_additional_special_tokens=…) -> int - changed:
method SpecialTokensMixin.add_tokens(new_tokens: Union[str, AddedToken, List[Union[str, AddedToken]]], special_tokens: bool=…) -> int → method SpecialTokensMixin.add_tokens(new_tokens: Union[str, AddedToken, Sequence[Union[str, AddedToken]]], special_tokens: bool=…) -> int - changed:
static method PreTrainedTokenizerBase.from_pretrained(pretrained_model_name_or_path, *init_inputs, **kwargs) → static method PreTrainedTokenizerBase.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], *init_inputs, cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, trust_remote_code=…, **kwargs) - added:
constructor AddedToken(content: str, single_word=…, lstrip=…, rstrip=…, special=…, normalized=…) - added:
field PreTrainedTokenizerBase.truncation_side: str - added:
function flatten(arr: list) - added:
function get_fast_tokenizer_file(tokenization_files: list[str]) -> str - added:
function import_protobuf_decode_error(error_message=…) - added:
method BatchEncoding.sequence_ids(batch_index: int=…) -> list[Optional[int]] - added:
method BatchEncoding.token_to_sequence(batch_or_token_index: int, token_index: Optional[int]=…) -> int - added:
method BatchEncoding.word_ids(batch_index: int=…) -> list[Optional[int]] - added:
method PreTrainedTokenizerBase.apply_chat_template(conversation: Union[list[dict[str, str]], list[list[dict[str, str]]]], tools: Optional[list[Union[dict, Callable]]]=…, documents: Optional[list[dict[str, str]]]=…, chat_template: Optional[str]=…, add_generation_prompt: bool=…, continue_final_message: bool=…, tokenize: bool=…, padding: Union[bool, str, PaddingStrategy]=…, truncation: bool=…, max_l… - added:
method PreTrainedTokenizerBase.as_target_tokenizer() - added:
method PreTrainedTokenizerBase.encode_message_with_chat_template(message: dict[str, str], conversation_history: Optional[list[dict[str, str]]]=…, **kwargs) -> list[int] - added:
method PreTrainedTokenizerBase.get_chat_template(chat_template: Optional[str]=…, tools: Optional[list[dict]]=…) -> str - added:
method PreTrainedTokenizerBase.prepare_seq2seq_batch(src_texts: list[str], tgt_texts: Optional[list[str]]=…, max_length: Optional[int]=…, max_target_length: Optional[int]=…, padding: str=…, return_tensors: Optional[str]=…, truncation: bool=…, **kwargs) -> BatchEncoding - added:
method PreTrainedTokenizerBase.save_chat_templates(save_directory: Union[str, os.PathLike], tokenizer_config: dict, filename_prefix: Optional[str], save_jinja_files: bool) - added:
property BatchEncoding.encodings -> Optional[list[EncodingFast]] - added:
property BatchEncoding.n_sequences -> Optional[int] - added:
property PreTrainedTokenizerBase.added_tokens_decoder -> dict[int, AddedToken] - added:
property SpecialTokensMixin.all_special_ids -> list[int] - added:
property SpecialTokensMixin.all_special_tokens -> list[str] - added:
property SpecialTokensMixin.all_special_tokens_extended -> list[Union[str, AddedToken]] - added:
property SpecialTokensMixin.special_tokens_map -> dict[str, Union[str, list[str]]] - added:
property SpecialTokensMixin.special_tokens_map_extended -> dict[str, Union[str, AddedToken, list[Union[str, AddedToken]]]] - added:
static method PreTrainedTokenizerBase.convert_added_tokens(obj: Union[AddedToken, Any], save=…, add_type_field=…) - added:
static method PreTrainedTokenizerBase.register_for_auto_class(auto_class=…) - added:
variable AudioInput
- removed:
property PreTrainedTokenizerFast.vocab -> Dict[str, int] - changed:
field PreTrainedTokenizerFast.slow_tokenizer_class: PreTrainedTokenizer → field PreTrainedTokenizerFast.slow_tokenizer_class: Optional[type[PreTrainedTokenizer]] - changed:
method PreTrainedTokenizerFast.convert_ids_to_tokens(ids: Union[int, List[int]], skip_special_tokens: bool=…) -> Union[str, List[str]] → method PreTrainedTokenizerFast.convert_ids_to_tokens(ids: Union[int, list[int]], skip_special_tokens: bool=…) -> Union[str, list[str]] - changed:
method PreTrainedTokenizerFast.convert_tokens_to_ids(tokens: Union[str, List[str]]) -> Union[int, List[int]] → method PreTrainedTokenizerFast.convert_tokens_to_ids(tokens: Union[str, Iterable[str]]) -> Union[int, list[int]] - changed:
method PreTrainedTokenizerFast.convert_tokens_to_string(tokens: List[str]) -> str → method PreTrainedTokenizerFast.convert_tokens_to_string(tokens: list[str]) -> str - changed:
method PreTrainedTokenizerFast.get_added_vocab() -> Dict[str, int] → method PreTrainedTokenizerFast.get_added_vocab() -> dict[str, int] - changed:
method PreTrainedTokenizerFast.get_vocab() -> Dict[str, int] → method PreTrainedTokenizerFast.get_vocab() -> dict[str, int] - changed:
method PreTrainedTokenizerFast.set_truncation_and_padding(padding_strategy: PaddingStrategy, truncation_strategy: TruncationStrategy, max_length: int, stride: int, pad_to_multiple_of: Optional[int]) → method PreTrainedTokenizerFast.set_truncation_and_padding(padding_strategy: PaddingStrategy, truncation_strategy: TruncationStrategy, max_length: int, stride: int, pad_to_multiple_of: Optional[int], padding_side: Optional[str]) - changed:
method PreTrainedTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> List[str] → method PreTrainedTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> list[str] - added:
field PreTrainedTokenizerFast.vocab_files_names - added:
method PreTrainedTokenizerFast.train_new_from_iterator(text_iterator, vocab_size, length=…, new_special_tokens=…, special_tokens_map=…, **kwargs) - added:
property PreTrainedTokenizerFast.added_tokens_decoder -> dict[int, AddedToken] - added:
property PreTrainedTokenizerFast.added_tokens_encoder -> dict[str, int] - added:
property PreTrainedTokenizerFast.can_save_slow_tokenizer -> bool - added:
property PreTrainedTokenizerFast.vocab -> dict[str, int] - added:
variable MODEL_TO_TRAINER_MAPPING - added:
variable TIKTOKEN_VOCAB_FILE - added:
variable VOCAB_FILES_NAMES
- removed:
class XLMTokenizer(PreTrainedTokenizer) - removed:
constructor XLMTokenizer(vocab_file, merges_file, unk_token=…, bos_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, lang2id=…, id2lang=…, do_lowercase_and_remove_accent=…, **kwargs) - removed:
field XLMTokenizer.max_model_input_sizes - removed:
field XLMTokenizer.pretrained_init_configuration - removed:
field XLMTokenizer.pretrained_vocab_files_map - removed:
field XLMTokenizer.vocab_files_names - removed:
function get_pairs(word) - removed:
function lowercase_and_remove_accent(text) - removed:
function remove_non_printing_char(text) - removed:
function replace_unicode_punct(text) - removed:
function romanian_preprocessing(text) - removed:
method XLMTokenizer.bpe(token) - removed:
method XLMTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMTokenizer.convert_tokens_to_string(tokens) - removed:
method XLMTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLMTokenizer.get_vocab() - removed:
method XLMTokenizer.ja_tokenize(text) - removed:
method XLMTokenizer.moses_pipeline(text, lang) - removed:
method XLMTokenizer.moses_punct_norm(text, lang) - removed:
method XLMTokenizer.moses_tokenize(text, lang) - removed:
method XLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property XLMTokenizer.do_lower_case - removed:
property XLMTokenizer.vocab_size - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class XLMProphetNetTokenizer(PreTrainedTokenizer) - removed:
constructor XLMProphetNetTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, unk_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - removed:
field XLMProphetNetTokenizer.max_model_input_sizes - removed:
field XLMProphetNetTokenizer.model_input_names - removed:
field XLMProphetNetTokenizer.pretrained_vocab_files_map - removed:
field XLMProphetNetTokenizer.vocab_files_names - removed:
function load_vocab(vocab_file) - removed:
method XLMProphetNetTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMProphetNetTokenizer.convert_tokens_to_string(tokens) - removed:
method XLMProphetNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMProphetNetTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLMProphetNetTokenizer.get_vocab() - removed:
method XLMProphetNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property XLMProphetNetTokenizer.vocab_size - removed:
variable PRETRAINED_INIT_CONFIGURATION - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class XLMRobertaTokenizer(PreTrainedTokenizer) - removed:
constructor XLMRobertaTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - removed:
field XLMRobertaTokenizer.max_model_input_sizes - removed:
field XLMRobertaTokenizer.model_input_names - removed:
field XLMRobertaTokenizer.pretrained_vocab_files_map - removed:
field XLMRobertaTokenizer.vocab_files_names - removed:
method XLMRobertaTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMRobertaTokenizer.convert_tokens_to_string(tokens) - removed:
method XLMRobertaTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMRobertaTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLMRobertaTokenizer.get_vocab() - removed:
method XLMRobertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property XLMRobertaTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class XLMRobertaTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor XLMRobertaTokenizerFast(vocab_file, tokenizer_file=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - removed:
field XLMRobertaTokenizerFast.max_model_input_sizes - removed:
field XLMRobertaTokenizerFast.model_input_names - removed:
field XLMRobertaTokenizerFast.pretrained_vocab_files_map - removed:
field XLMRobertaTokenizerFast.slow_tokenizer_class - removed:
field XLMRobertaTokenizerFast.vocab_files_names - removed:
method XLMRobertaTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMRobertaTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLMRobertaTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLMRobertaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable VOCAB_FILES_NAMES - removed:
variable XLMRobertaTokenizer - removed:
variable logger
- removed:
class XLNetTokenizer(PreTrainedTokenizer) - removed:
constructor XLNetTokenizer(vocab_file, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - removed:
field XLNetTokenizer.max_model_input_sizes - removed:
field XLNetTokenizer.padding_side - removed:
field XLNetTokenizer.pretrained_vocab_files_map - removed:
field XLNetTokenizer.vocab_files_names - removed:
method XLNetTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLNetTokenizer.convert_tokens_to_string(tokens) - removed:
method XLNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLNetTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLNetTokenizer.get_vocab() - removed:
method XLNetTokenizer.preprocess_text(inputs) - removed:
method XLNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
property XLNetTokenizer.vocab_size - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SEG_ID_A - removed:
variable SEG_ID_B - removed:
variable SEG_ID_CLS - removed:
variable SEG_ID_PAD - removed:
variable SEG_ID_SEP - removed:
variable VOCAB_FILES_NAMES - removed:
variable logger
- removed:
class XLNetTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor XLNetTokenizerFast(vocab_file, tokenizer_file=…, do_lower_case=…, remove_space=…, keep_accents=…, bos_token=…, eos_token=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, additional_special_tokens=…, **kwargs) - removed:
field XLNetTokenizerFast.max_model_input_sizes - removed:
field XLNetTokenizerFast.padding_side - removed:
field XLNetTokenizerFast.pretrained_vocab_files_map - removed:
field XLNetTokenizerFast.slow_tokenizer_class - removed:
field XLNetTokenizerFast.vocab_files_names - removed:
method XLNetTokenizerFast.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLNetTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - removed:
method XLNetTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - removed:
method XLNetTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - removed:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - removed:
variable PRETRAINED_VOCAB_FILES_MAP - removed:
variable SEG_ID_A - removed:
variable SEG_ID_B - removed:
variable SEG_ID_CLS - removed:
variable SEG_ID_PAD - removed:
variable SEG_ID_SEP - removed:
variable SPIECE_UNDERLINE - removed:
variable VOCAB_FILES_NAMES - removed:
variable XLNetTokenizer - removed:
variable logger
- removed:
method Trainer.is_local_master() -> bool - removed:
method Trainer.is_world_master() -> bool - changed:
constructor Trainer(model: Union[PreTrainedModel, torch.nn.Module]=…, args: TrainingArguments=…, data_collator: Optional[DataCollator]=…, train_dataset: Optional[Dataset]=…, eval_dataset: Optional[Dataset]=…, tokenizer: Optional[PreTrainedTokenizerBase]=…, model_init: Callable[[], PreTrainedModel]=…, compute_metrics: Optional[Callable[[EvalPrediction], Dict]]=…, callbacks: Optional[List[TrainerCa… → constructor Trainer(model: Union[PreTrainedModel, nn.Module, None]=…, args: Optional[TrainingArguments]=…, data_collator: Optional[DataCollator]=…, train_dataset: Optional[Union[Dataset, IterableDataset, datasets.Dataset]]=…, eval_dataset: Optional[Union[Dataset, dict[str, Dataset], datasets.Dataset]]=…, processing_class: Optional[Union[PreTrainedTokenizerBase, BaseImageProcessor, FeatureExtracti… - changed:
method Trainer.compute_loss(model, inputs) → method Trainer.compute_loss(model: nn.Module, inputs: dict[str, Union[torch.Tensor, Any]], return_outputs: bool=…, num_items_in_batch: Optional[torch.Tensor]=…) - changed:
method Trainer.evaluate(eval_dataset: Optional[Dataset]=…) -> Dict[str, float] → method Trainer.evaluate(eval_dataset: Optional[Union[Dataset, dict[str, Dataset]]]=…, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…) -> dict[str, float] - changed:
method Trainer.floating_point_ops(inputs: Dict[str, Union[torch.Tensor, Any]]) → method Trainer.floating_point_ops(inputs: dict[str, Union[torch.Tensor, Any]]) - changed:
method Trainer.get_eval_dataloader(eval_dataset: Optional[Dataset]=…) -> DataLoader → method Trainer.get_eval_dataloader(eval_dataset: Optional[Union[str, Dataset]]=…) -> DataLoader - changed:
method Trainer.hyperparameter_search(hp_space: Optional[Callable[[optuna.Trial], Dict[str, float]]]=…, compute_objective: Optional[Callable[[Dict[str, float]], float]]=…, n_trials: int=…, direction: str=…, backend: Optional[Union[str, HPSearchBackend]]=…, hp_name: Optional[Callable[[optuna.Trial], str]]=…, **kwargs) -> BestRun → method Trainer.hyperparameter_search(hp_space: Optional[Callable[[optuna.Trial], dict[str, float]]]=…, compute_objective: Optional[Callable[[dict[str, float]], float]]=…, n_trials: int=…, direction: Union[str, list[str]]=…, backend: Optional[Union[str, HPSearchBackend]]=…, hp_name: Optional[Callable[[optuna.Trial], str]]=…, **kwargs) -> Union[BestRun, list[BestRun]] - changed:
method Trainer.log(logs: Dict[str, float]) -> None → method Trainer.log(logs: dict[str, float], start_time: Optional[float]=…) -> None - changed:
method Trainer.predict(test_dataset: Dataset) -> PredictionOutput → method Trainer.predict(test_dataset: Dataset, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…) -> PredictionOutput - changed:
method Trainer.prediction_loop(dataloader: DataLoader, description: str, prediction_loss_only: Optional[bool]=…) -> PredictionOutput → method Trainer.prediction_loop(dataloader: DataLoader, description: str, prediction_loss_only: Optional[bool]=…, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…) -> EvalLoopOutput - changed:
method Trainer.prediction_step(model: nn.Module, inputs: Dict[str, Union[torch.Tensor, Any]], prediction_loss_only: bool) -> Tuple[Optional[float], Optional[torch.Tensor], Optional[torch.Tensor]] → method Trainer.prediction_step(model: nn.Module, inputs: dict[str, Union[torch.Tensor, Any]], prediction_loss_only: bool, ignore_keys: Optional[list[str]]=…) -> tuple[Optional[torch.Tensor], Optional[torch.Tensor], Optional[torch.Tensor]] - changed:
method Trainer.save_model(output_dir: Optional[str]=…) → method Trainer.save_model(output_dir: Optional[str]=…, _internal_call: bool=…) - changed:
method Trainer.train(model_path: Optional[str]=…, trial: Union[optuna.Trial, Dict[str, Any]]=…) → method Trainer.train(resume_from_checkpoint: Optional[Union[str, bool]]=…, trial: Union[optuna.Trial, dict[str, Any], None]=…, ignore_keys_for_eval: Optional[list[str]]=…, **kwargs: Any) - changed:
method Trainer.training_step(model: nn.Module, inputs: Dict[str, Union[torch.Tensor, Any]]) -> torch.Tensor → method Trainer.training_step(model: nn.Module, inputs: dict[str, Union[torch.Tensor, Any]], num_items_in_batch: Optional[torch.Tensor]=…) -> torch.Tensor - added:
function safe_globals() - added:
method Trainer.autocast_smart_context_manager(cache_enabled: Optional[bool]=…) - added:
method Trainer.compare_trainer_and_checkpoint_args(training_args, trainer_state) - added:
method Trainer.compute_loss_context_manager() - added:
method Trainer.create_accelerator_and_postprocess() - added:
method Trainer.create_model_card(language: Optional[str]=…, license: Optional[str]=…, tags: Union[str, list[str], None]=…, model_name: Optional[str]=…, finetuned_from: Optional[str]=…, tasks: Union[str, list[str], None]=…, dataset_tags: Union[str, list[str], None]=…, dataset: Union[str, list[str], None]=…, dataset_args: Union[str, list[str], None]=…) - added:
method Trainer.create_optimizer() - added:
method Trainer.create_scheduler(num_training_steps: int, optimizer: torch.optim.Optimizer=…) - added:
method Trainer.evaluation_loop(dataloader: DataLoader, description: str, prediction_loss_only: Optional[bool]=…, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…) -> EvalLoopOutput - added:
method Trainer.get_batch_samples(epoch_iterator: Iterator, num_batches: int, device: torch.device) -> tuple[list, Optional[Union[torch.Tensor, int]]] - added:
method Trainer.get_decay_parameter_names(model) -> list[str] - added:
method Trainer.get_learning_rates() - added:
method Trainer.get_num_trainable_parameters() - added:
method Trainer.get_optimizer_group(param: Optional[Union[str, torch.nn.parameter.Parameter]]=…) - added:
method Trainer.get_total_train_batch_size(args) -> int - added:
method Trainer.get_tp_size() -> int - added:
method Trainer.init_hf_repo(token: Optional[str]=…) - added:
method Trainer.propagate_args_to_deepspeed(auto_find_batch_size=…) - added:
method Trainer.push_to_hub(commit_message: Optional[str]=…, blocking: bool=…, token: Optional[str]=…, revision: Optional[str]=…, **kwargs) -> str - added:
method Trainer.set_initial_training_values(args: TrainingArguments, dataloader: DataLoader, total_train_batch_size: int) - added:
method Trainer.torch_jit_model_eval(model, dataloader, training=…) - added:
property Trainer.tokenizer -> Optional[PreTrainedTokenizerBase] - added:
static method Trainer.get_optimizer_cls_and_kwargs(args: TrainingArguments, model: Optional[PreTrainedModel]=…) -> tuple[Any, Any] - added:
static method Trainer.num_tokens(train_dl: DataLoader, max_steps: Optional[int]=…) -> int - added:
variable DATA_SAMPLERS - added:
variable FSDP_MODEL_NAME - added:
variable IS_SAGEMAKER_MP_POST_1_10 - added:
variable IS_XLA_FSDPV2_POST_2_2 - added:
variable OPTIMIZER_NAME - added:
variable OPTIMIZER_NAME_BIN - added:
variable SCALER_NAME - added:
variable SCHEDULER_NAME - added:
variable TRAINER_STATE_NAME - added:
variable TRAINING_ARGS_NAME
- changed:
class TrainerControl → class TrainerControl(ExportableState) - changed:
constructor CallbackHandler(callbacks, model, optimizer, lr_scheduler) → constructor CallbackHandler(callbacks, model, processing_class, optimizer, lr_scheduler) - changed:
constructor ProgressCallback() → constructor ProgressCallback(max_str_len: int=…) - changed:
field TrainerState.log_history: List[Dict[str, float]] → field TrainerState.log_history: list[dict[str, float]] - changed:
field TrainerState.total_flos: int → field TrainerState.total_flos: float - changed:
field TrainerState.trial_name: str → field TrainerState.trial_name: Optional[str] - changed:
field TrainerState.trial_params: Dict[str, Union[str, float, int, bool]] → field TrainerState.trial_params: Optional[dict[str, Union[str, float, int, bool]]] - added:
class EarlyStoppingCallback(TrainerCallback, ExportableState) - added:
class ExportableState - added:
constructor EarlyStoppingCallback(early_stopping_patience: int=…, early_stopping_threshold: Optional[float]=…) - added:
field TrainerState.best_global_step: Optional[int] - added:
field TrainerState.eval_steps: int - added:
field TrainerState.logging_steps: int - added:
field TrainerState.num_input_tokens_seen: int - added:
field TrainerState.save_steps: int - added:
field TrainerState.stateful_callbacks: Optional[list[TrainerCallback]] - added:
field TrainerState.train_batch_size: Optional[int] - added:
method CallbackHandler.on_optimizer_step(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_pre_optimizer_step(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_predict(args: TrainingArguments, state: TrainerState, control: TrainerControl, metrics) - added:
method CallbackHandler.on_substep_end(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method EarlyStoppingCallback.check_metric_value(args, state, control, metric_value) - added:
method EarlyStoppingCallback.on_evaluate(args, state, control, metrics, **kwargs) - added:
method EarlyStoppingCallback.on_train_begin(args, state, control, **kwargs) - added:
method EarlyStoppingCallback.state() -> dict - added:
method ExportableState.state() -> dict - added:
method ProgressCallback.on_predict(args, state, control, **kwargs) - added:
method TrainerCallback.on_optimizer_step(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_pre_optimizer_step(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_predict(args: TrainingArguments, state: TrainerState, control: TrainerControl, metrics, **kwargs) - added:
method TrainerCallback.on_substep_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerControl.state() -> dict - added:
method TrainerState.compute_steps(args, max_steps) - added:
method TrainerState.init_training_references(trainer, max_steps, num_train_epochs, trial) - added:
static method ExportableState.from_state(state)
- removed:
function nested_expand_like(arrays, new_seq_length, padding_index=…) - removed:
variable SAVE_STATE_WARNING - changed:
constructor SequentialDistributedSampler(dataset, num_replicas=…, rank=…) → constructor SequentialDistributedSampler(dataset, num_replicas=…, rank=…, batch_size=…) - changed:
function distributed_broadcast_scalars(scalars: List[Union[int, float]], num_total_examples: Optional[int]=…) -> torch.Tensor → function distributed_broadcast_scalars(scalars: list[Union[int, float]], num_total_examples: Optional[int]=…, device: Optional[torch.device]=…) -> torch.Tensor - changed:
function distributed_concat(tensor: torch.Tensor, num_total_examples: Optional[int]=…) -> torch.Tensor → function distributed_concat(tensor: Any, num_total_examples: Optional[int]=…) -> Any - changed:
function get_tpu_sampler(dataset: torch.utils.data.dataset.Dataset) → function get_tpu_sampler(dataset: torch.utils.data.Dataset, batch_size: int) - added:
class AcceleratorConfig - added:
class DistributedLengthGroupedSampler(DistributedSampler) - added:
class DistributedSamplerWithLoop(DistributedSampler) - added:
class EvalLoopContainer - added:
class IterableDatasetShard(IterableDataset) - added:
class LabelSmoother - added:
class LayerWiseDummyOptimizer(torch.optim.Optimizer) - added:
class LayerWiseDummyScheduler(LRScheduler) - added:
class LengthGroupedSampler(Sampler) - added:
class ShardSampler(Sampler) - added:
constructor DistributedLengthGroupedSampler(batch_size: int, dataset: Optional[Dataset]=…, num_replicas: Optional[int]=…, rank: Optional[int]=…, seed: int=…, drop_last: bool=…, lengths: Optional[list[int]]=…, model_input_name: Optional[str]=…) - added:
constructor DistributedSamplerWithLoop(dataset, batch_size, **kwargs) - added:
constructor EvalLoopContainer(do_nested_concat: bool=…, padding_index: int=…) - added:
constructor IterableDatasetShard(dataset: IterableDataset, batch_size: int=…, drop_last: bool=…, num_processes: int=…, process_index: int=…, seed: int=…) - added:
constructor LayerWiseDummyOptimizer(optimizer_dict=…, **kwargs) - added:
constructor LayerWiseDummyScheduler(*args, **kwargs) - added:
constructor LengthGroupedSampler(batch_size: int, dataset: Optional[Dataset]=…, lengths: Optional[list[int]]=…, model_input_name: Optional[str]=…, generator=…) - added:
constructor ShardSampler(dataset: Dataset, batch_size: int=…, drop_last: bool=…, num_processes: int=…, process_index: int=…) - added:
field AcceleratorConfig.dispatch_batches: Optional[bool] - added:
field AcceleratorConfig.even_batches: bool - added:
field AcceleratorConfig.gradient_accumulation_kwargs: Optional[dict] - added:
field AcceleratorConfig.non_blocking: bool - added:
field AcceleratorConfig.split_batches: bool - added:
field AcceleratorConfig.use_configured_state: bool - added:
field AcceleratorConfig.use_seedable_sampler: bool - added:
field LabelSmoother.epsilon: float - added:
field LabelSmoother.ignore_index: int - added:
function atleast_1d(tensor_or_array: Union[torch.Tensor, np.ndarray]) - added:
function expand_like(arrays, new_seq_length, padding_index=…) - added:
function find_batch_size(tensors) - added:
function get_dataloader_sampler(dataloader) - added:
function get_length_grouped_indices(lengths, batch_size, mega_batch_mult=…, generator=…) - added:
function get_model_param_count(model, trainable_only=…) - added:
function get_module_class_from_name(module, name) - added:
function get_parameter_names(model, forbidden_layer_types, forbidden_layer_names=…) - added:
function log_metrics(self, split, metrics) - added:
function metrics_format(metrics: dict[str, float]) -> dict[str, float] - added:
function remove_dummy_checkpoint(is_main_process, output_dir, filenames) - added:
function save_metrics(self, split, metrics, combined=…) - added:
function save_state(self) - added:
function set_rng_state_for_device(device_name, device_module, checkpoint_rng_state, is_distributed) - added:
function smp_forward_backward(model, inputs, gradient_accumulation_steps=…) - added:
function smp_forward_only(model, inputs) - added:
function smp_gather(tensor) - added:
function smp_nested_concat(tensor) - added:
method AcceleratorConfig.pop(key, default=…) - added:
method AcceleratorConfig.to_dict() - added:
method EvalLoopContainer.add(tensors) -> None - added:
method EvalLoopContainer.get_arrays() - added:
method EvalLoopContainer.to_cpu_and_numpy() -> None - added:
method IterableDatasetShard.set_epoch(epoch) - added:
method LayerWiseDummyOptimizer.step(closure=…) -> Optional[float] - added:
method LayerWiseDummyOptimizer.zero_grad(set_to_none: bool=…) -> None - added:
method LayerWiseDummyScheduler.get_lr() - added:
static method AcceleratorConfig.from_json_file(json_file)
- added:
class Seq2SeqTrainer(Trainer) - added:
constructor Seq2SeqTrainer(model: Optional[Union[PreTrainedModel, nn.Module]]=…, args: Optional[TrainingArguments]=…, data_collator: Optional[DataCollator]=…, train_dataset: Optional[Union[Dataset, IterableDataset, datasets.Dataset]]=…, eval_dataset: Optional[Union[Dataset, dict[str, Dataset]]]=…, processing_class: Optional[Union[PreTrainedTokenizerBase, BaseImageProcessor, FeatureExtractionMixin… - added:
method Seq2SeqTrainer.evaluate(eval_dataset: Optional[Dataset]=…, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…, **gen_kwargs) -> dict[str, float] - added:
method Seq2SeqTrainer.predict(test_dataset: Dataset, ignore_keys: Optional[list[str]]=…, metric_key_prefix: str=…, **gen_kwargs) -> PredictionOutput - added:
method Seq2SeqTrainer.prediction_step(model: nn.Module, inputs: dict[str, Union[torch.Tensor, Any]], prediction_loss_only: bool, ignore_keys: Optional[list[str]]=…, **gen_kwargs) -> tuple[Optional[float], Optional[torch.Tensor], Optional[torch.Tensor]] - added:
static method Seq2SeqTrainer.load_generation_config(gen_config_arg: Union[str, GenerationConfig]) -> GenerationConfig - added:
variable logger
- removed:
class TFTrainer - removed:
constructor TFTrainer(model: TFPreTrainedModel, args: TFTrainingArguments, train_dataset: Optional[tf.data.Dataset]=…, eval_dataset: Optional[tf.data.Dataset]=…, compute_metrics: Optional[Callable[[EvalPrediction], Dict]]=…, tb_writer: Optional[tf.summary.SummaryWriter]=…, optimizers: Tuple[tf.keras.optimizers.Optimizer, tf.keras.optimizers.schedules.LearningRateSchedule]=…, **kwargs) - removed:
method TFTrainer.apply_gradients(features, labels, nb_instances_in_global_batch) - removed:
method TFTrainer.create_optimizer_and_scheduler(num_training_steps: int) - removed:
method TFTrainer.distributed_prediction_steps(batch) - removed:
method TFTrainer.distributed_training_steps(batch) - removed:
method TFTrainer.evaluate(eval_dataset: Optional[tf.data.Dataset]=…) -> Dict[str, float] - removed:
method TFTrainer.get_eval_tfdataset(eval_dataset: Optional[tf.data.Dataset]=…) -> tf.data.Dataset - removed:
method TFTrainer.get_test_tfdataset(test_dataset: tf.data.Dataset) -> tf.data.Dataset - removed:
method TFTrainer.get_train_tfdataset() -> tf.data.Dataset - removed:
method TFTrainer.log(logs: Dict[str, float]) -> None - removed:
method TFTrainer.predict(test_dataset: tf.data.Dataset) -> PredictionOutput - removed:
method TFTrainer.prediction_loop(dataset: tf.data.Dataset, steps: int, num_examples: int, description: str, prediction_loss_only: Optional[bool]=…) -> PredictionOutput - removed:
method TFTrainer.prediction_step(features: tf.Tensor, labels: tf.Tensor, nb_instances_in_global_batch: tf.Tensor) -> tf.Tensor - removed:
method TFTrainer.run_model(features, labels, training) - removed:
method TFTrainer.save_model(output_dir: Optional[str]=…) - removed:
method TFTrainer.setup_comet() - removed:
method TFTrainer.setup_wandb() - removed:
method TFTrainer.train() -> None - removed:
method TFTrainer.training_step(features, labels, nb_instances_in_global_batch) - removed:
variable logger
- removed:
field EvalPrediction.label_ids: np.ndarray - removed:
field EvalPrediction.predictions: Union[np.ndarray, Tuple[np.ndarray]] - removed:
variable default_hp_space - changed:
class EvalPrediction(NamedTuple) → class EvalPrediction - changed:
field BestRun.hyperparameters: Dict[str, Any] → field BestRun.hyperparameters: dict[str, Any] - changed:
field BestRun.objective: float → field BestRun.objective: Union[float, list[float]] - changed:
field PredictionOutput.label_ids: Optional[np.ndarray] → field PredictionOutput.label_ids: Optional[Union[np.ndarray, tuple[np.ndarray]]] - changed:
field PredictionOutput.metrics: Optional[Dict[str, float]] → field PredictionOutput.metrics: Optional[dict[str, float]] - changed:
field PredictionOutput.predictions: Union[np.ndarray, Tuple[np.ndarray]] → field PredictionOutput.predictions: Union[np.ndarray, tuple[np.ndarray]] - changed:
function default_compute_objective(metrics: Dict[str, float]) -> float → function default_compute_objective(metrics: dict[str, float]) -> float - changed:
function default_hp_space_optuna(trial) -> Dict[str, float] → function default_hp_space_optuna(trial) -> dict[str, float] - changed:
function default_hp_space_ray(trial) -> Dict[str, float] → function default_hp_space_ray(trial) -> dict[str, Any] - changed:
function set_seed(seed: int) → function set_seed(seed: int, deterministic: bool=…) - added:
class EvalLoopOutput(NamedTuple) - added:
class FSDPOption(ExplicitEnum) - added:
class HubStrategy(ExplicitEnum) - added:
class IntervalStrategy(ExplicitEnum) - added:
class RemoveColumnsCollator - added:
class SaveStrategy(ExplicitEnum) - added:
class SchedulerType(ExplicitEnum) - added:
class TrainerMemoryTracker - added:
constructor EvalPrediction(predictions: Union[np.ndarray, tuple[np.ndarray]], label_ids: Union[np.ndarray, tuple[np.ndarray]], inputs: Optional[Union[np.ndarray, tuple[np.ndarray]]]=…, losses: Optional[Union[np.ndarray, tuple[np.ndarray]]]=…) - added:
constructor RemoveColumnsCollator(data_collator, signature_columns, logger=…, model_name: Optional[str]=…, description: Optional[str]=…) - added:
constructor TrainerMemoryTracker(skip_memory_metrics=…) - added:
field BestRun.run_summary: Optional[Any] - added:
field EvalLoopOutput.label_ids: Optional[Union[np.ndarray, tuple[np.ndarray]]] - added:
field EvalLoopOutput.metrics: Optional[dict[str, float]] - added:
field EvalLoopOutput.num_samples: Optional[int] - added:
field EvalLoopOutput.predictions: Union[np.ndarray, tuple[np.ndarray]] - added:
field FSDPOption.AUTO_WRAP - added:
field FSDPOption.FULL_SHARD - added:
field FSDPOption.HYBRID_SHARD - added:
field FSDPOption.HYBRID_SHARD_ZERO2 - added:
field FSDPOption.NO_SHARD - added:
field FSDPOption.OFFLOAD - added:
field FSDPOption.SHARD_GRAD_OP - added:
field HPSearchBackend.SIGOPT - added:
field HPSearchBackend.WANDB - added:
field HubStrategy.ALL_CHECKPOINTS - added:
field HubStrategy.CHECKPOINT - added:
field HubStrategy.END - added:
field HubStrategy.EVERY_SAVE - added:
field IntervalStrategy.EPOCH - added:
field IntervalStrategy.NO - added:
field IntervalStrategy.STEPS - added:
field SaveStrategy.BEST - added:
field SaveStrategy.EPOCH - added:
field SaveStrategy.NO - added:
field SaveStrategy.STEPS - added:
field SchedulerType.CONSTANT - added:
field SchedulerType.CONSTANT_WITH_WARMUP - added:
field SchedulerType.COSINE - added:
field SchedulerType.COSINE_WARMUP_WITH_MIN_LR - added:
field SchedulerType.COSINE_WITH_MIN_LR - added:
field SchedulerType.COSINE_WITH_RESTARTS - added:
field SchedulerType.INVERSE_SQRT - added:
field SchedulerType.LINEAR - added:
field SchedulerType.POLYNOMIAL - added:
field SchedulerType.REDUCE_ON_PLATEAU - added:
field SchedulerType.WARMUP_STABLE_DECAY - added:
field TrainOutput.metrics: dict[str, float] - added:
field TrainerMemoryTracker.stages - added:
function check_target_module_exists(optim_target_modules, key: str, return_is_regex: bool=…) - added:
function default_hp_space_sigopt(trial) - added:
function default_hp_space_wandb(trial) -> dict[str, Any] - added:
function denumpify_detensorize(metrics) - added:
function enable_full_determinism(seed: int, warn_only: bool=…) - added:
function find_executable_batch_size(function: Optional[Callable]=…, starting_batch_size: int=…, auto_find_batch_size: bool=…) - added:
function get_last_checkpoint(folder) - added:
function has_length(dataset) - added:
function neftune_post_forward_hook(module, input, output) - added:
function number_of_arguments(func) - added:
function seed_worker(worker_id: int, num_workers: int, rank: int) - added:
function speed_metrics(split, start_time, num_samples=…, num_steps=…, num_tokens=…) - added:
method TrainerMemoryTracker.cpu_mem_used() - added:
method TrainerMemoryTracker.derive_stage() - added:
method TrainerMemoryTracker.peak_monitor_func() - added:
method TrainerMemoryTracker.start() - added:
method TrainerMemoryTracker.stop(stage) - added:
method TrainerMemoryTracker.stop_and_update_metrics(metrics=…) - added:
method TrainerMemoryTracker.update_metrics(stage, metrics)
- removed:
field TrainingArguments.evaluate_during_training: bool - removed:
field TrainingArguments.evaluation_strategy: EvaluationStrategy - changed:
field TrainingArguments.debug: bool → field TrainingArguments.debug: Union[str, list[DebugOption]] - changed:
field TrainingArguments.eval_steps: int → field TrainingArguments.eval_steps: Optional[float] - changed:
field TrainingArguments.label_names: Optional[List[str]] → field TrainingArguments.label_names: Optional[list[str]] - changed:
field TrainingArguments.load_best_model_at_end: Optional[bool] → field TrainingArguments.load_best_model_at_end: bool - changed:
field TrainingArguments.logging_steps: int → field TrainingArguments.logging_steps: float - changed:
field TrainingArguments.output_dir: str → field TrainingArguments.output_dir: Optional[str] - changed:
field TrainingArguments.remove_unused_columns: Optional[bool] → field TrainingArguments.remove_unused_columns: bool - changed:
field TrainingArguments.save_steps: int → field TrainingArguments.save_steps: float - changed:
method TrainingArguments.to_sanitized_dict() -> Dict[str, Any] → method TrainingArguments.to_sanitized_dict() -> dict[str, Any] - added:
class OptimizerNames(ExplicitEnum) - added:
class ParallelMode(Enum) - added:
field OptimizerNames.ADAFACTOR - added:
field OptimizerNames.ADAGRAD - added:
field OptimizerNames.ADALOMO - added:
field OptimizerNames.ADAMW_8BIT - added:
field OptimizerNames.ADAMW_ANYPRECISION - added:
field OptimizerNames.ADAMW_APEX_FUSED - added:
field OptimizerNames.ADAMW_BNB - added:
field OptimizerNames.ADAMW_TORCH - added:
field OptimizerNames.ADAMW_TORCH_4BIT - added:
field OptimizerNames.ADAMW_TORCH_8BIT - added:
field OptimizerNames.ADAMW_TORCH_FUSED - added:
field OptimizerNames.ADAMW_TORCH_NPU_FUSED - added:
field OptimizerNames.ADAMW_TORCH_XLA - added:
field OptimizerNames.ADEMAMIX - added:
field OptimizerNames.ADEMAMIX_8BIT - added:
field OptimizerNames.APOLLO_ADAMW - added:
field OptimizerNames.APOLLO_ADAMW_LAYERWISE - added:
field OptimizerNames.GALORE_ADAFACTOR - added:
field OptimizerNames.GALORE_ADAFACTOR_LAYERWISE - added:
field OptimizerNames.GALORE_ADAMW - added:
field OptimizerNames.GALORE_ADAMW_8BIT - added:
field OptimizerNames.GALORE_ADAMW_8BIT_LAYERWISE - added:
field OptimizerNames.GALORE_ADAMW_LAYERWISE - added:
field OptimizerNames.GROKADAMW - added:
field OptimizerNames.LION - added:
field OptimizerNames.LION_8BIT - added:
field OptimizerNames.LOMO - added:
field OptimizerNames.PAGED_ADAMW - added:
field OptimizerNames.PAGED_ADAMW_8BIT - added:
field OptimizerNames.PAGED_ADEMAMIX - added:
field OptimizerNames.PAGED_ADEMAMIX_8BIT - added:
field OptimizerNames.PAGED_LION - added:
field OptimizerNames.PAGED_LION_8BIT - added:
field OptimizerNames.RMSPROP - added:
field OptimizerNames.RMSPROP_32BIT - added:
field OptimizerNames.RMSPROP_8BIT - added:
field OptimizerNames.RMSPROP_BNB - added:
field OptimizerNames.SCHEDULE_FREE_ADAMW - added:
field OptimizerNames.SCHEDULE_FREE_RADAM - added:
field OptimizerNames.SCHEDULE_FREE_SGD - added:
field OptimizerNames.SGD - added:
field OptimizerNames.STABLE_ADAMW - added:
field ParallelMode.DISTRIBUTED - added:
field ParallelMode.NOT_DISTRIBUTED - added:
field ParallelMode.NOT_PARALLEL - added:
field ParallelMode.SAGEMAKER_DATA_PARALLEL - added:
field ParallelMode.SAGEMAKER_MODEL_PARALLEL - added:
field ParallelMode.TPU - added:
field TrainingArguments.accelerator_config: Optional[Union[dict, str]] - added:
field TrainingArguments.adafactor: bool - added:
field TrainingArguments.auto_find_batch_size: bool - added:
field TrainingArguments.average_tokens_across_devices: bool - added:
field TrainingArguments.batch_eval_metrics: bool - added:
field TrainingArguments.bf16: bool - added:
field TrainingArguments.bf16_full_eval: bool - added:
field TrainingArguments.data_seed: Optional[int] - added:
field TrainingArguments.dataloader_persistent_workers: bool - added:
field TrainingArguments.dataloader_pin_memory: bool - added:
field TrainingArguments.dataloader_prefetch_factor: Optional[int] - added:
field TrainingArguments.ddp_backend: Optional[str] - added:
field TrainingArguments.ddp_broadcast_buffers: Optional[bool] - added:
field TrainingArguments.ddp_bucket_cap_mb: Optional[int] - added:
field TrainingArguments.ddp_find_unused_parameters: Optional[bool] - added:
field TrainingArguments.ddp_timeout: int - added:
field TrainingArguments.deepspeed: Optional[Union[dict, str]] - added:
field TrainingArguments.default_optim - added:
field TrainingArguments.eval_delay: float - added:
field TrainingArguments.eval_do_concat_batches: bool - added:
field TrainingArguments.eval_on_start: bool - added:
field TrainingArguments.eval_strategy: Union[IntervalStrategy, str] - added:
field TrainingArguments.eval_use_gather_object: bool - added:
field TrainingArguments.fp16_backend: str - added:
field TrainingArguments.fp16_full_eval: bool - added:
field TrainingArguments.framework - added:
field TrainingArguments.fsdp: Optional[Union[list[FSDPOption], str]] - added:
field TrainingArguments.fsdp_config: Optional[Union[dict[str, Any], str]] - added:
field TrainingArguments.fsdp_min_num_params: int - added:
field TrainingArguments.fsdp_transformer_layer_cls_to_wrap: Optional[str] - added:
field TrainingArguments.full_determinism: bool - added:
field TrainingArguments.gradient_checkpointing: bool - added:
field TrainingArguments.gradient_checkpointing_kwargs: Optional[Union[dict[str, Any], str]] - added:
field TrainingArguments.group_by_length: bool - added:
field TrainingArguments.half_precision_backend: str - added:
field TrainingArguments.hub_always_push: bool - added:
field TrainingArguments.hub_model_id: Optional[str] - added:
field TrainingArguments.hub_private_repo: Optional[bool] - added:
field TrainingArguments.hub_revision: Optional[str] - added:
field TrainingArguments.hub_strategy: Union[HubStrategy, str] - added:
field TrainingArguments.hub_token: Optional[str] - added:
field TrainingArguments.ignore_data_skip: bool - added:
field TrainingArguments.include_for_metrics: list[str] - added:
field TrainingArguments.include_inputs_for_metrics: bool - added:
field TrainingArguments.include_num_input_tokens_seen: Union[str, bool] - added:
field TrainingArguments.include_tokens_per_second: bool - added:
field TrainingArguments.jit_mode_eval: bool - added:
field TrainingArguments.label_smoothing_factor: float - added:
field TrainingArguments.length_column_name: str - added:
field TrainingArguments.liger_kernel_config: Optional[dict[str, bool]] - added:
field TrainingArguments.log_level: str - added:
field TrainingArguments.log_level_replica: str - added:
field TrainingArguments.log_on_each_node: bool - added:
field TrainingArguments.logging_nan_inf_filter: bool - added:
field TrainingArguments.logging_strategy: Union[IntervalStrategy, str] - added:
field TrainingArguments.lr_scheduler_kwargs: Optional[Union[dict, str]] - added:
field TrainingArguments.lr_scheduler_type: Union[SchedulerType, str] - added:
field TrainingArguments.mp_parameters: str - added:
field TrainingArguments.neftune_noise_alpha: Optional[float] - added:
field TrainingArguments.optim: Union[OptimizerNames, str] - added:
field TrainingArguments.optim_args: Optional[str] - added:
field TrainingArguments.optim_target_modules: Union[None, str, list[str]] - added:
field TrainingArguments.parallelism_config: Optional[ParallelismConfig] - added:
field TrainingArguments.project: str - added:
field TrainingArguments.push_to_hub: bool - added:
field TrainingArguments.push_to_hub_model_id: Optional[str] - added:
field TrainingArguments.push_to_hub_organization: Optional[str] - added:
field TrainingArguments.push_to_hub_token: Optional[str] - added:
field TrainingArguments.ray_scope: Optional[str] - added:
field TrainingArguments.report_to: Union[None, str, list[str]] - added:
field TrainingArguments.restore_callback_states_from_checkpoint: bool - added:
field TrainingArguments.resume_from_checkpoint: Optional[str] - added:
field TrainingArguments.save_on_each_node: bool - added:
field TrainingArguments.save_only_model: bool - added:
field TrainingArguments.save_safetensors: bool - added:
field TrainingArguments.save_strategy: Union[SaveStrategy, str] - added:
field TrainingArguments.skip_memory_metrics: bool - added:
field TrainingArguments.tf32: Optional[bool] - added:
field TrainingArguments.torch_compile: bool - added:
field TrainingArguments.torch_compile_backend: Optional[str] - added:
field TrainingArguments.torch_compile_mode: Optional[str] - added:
field TrainingArguments.torch_empty_cache_steps: Optional[int] - added:
field TrainingArguments.torchdynamo: Optional[str] - added:
field TrainingArguments.trackio_space_id: Optional[str] - added:
field TrainingArguments.use_cpu: bool - added:
field TrainingArguments.use_legacy_prediction_loop: bool - added:
field TrainingArguments.use_liger_kernel: bool - added:
field TrainingArguments.use_mps_device: bool - added:
field TrainingArguments.warmup_ratio: float - added:
function get_int_from_env(env_keys, default) - added:
function get_xla_device_type(device: torch.device) -> Optional[str] - added:
method TrainingArguments.get_process_log_level() - added:
method TrainingArguments.get_warmup_steps(num_training_steps: int) - added:
method TrainingArguments.main_process_first(local=…, desc=…) - added:
method TrainingArguments.set_dataloader(train_batch_size: int=…, eval_batch_size: int=…, drop_last: bool=…, num_workers: int=…, pin_memory: bool=…, persistent_workers: bool=…, prefetch_factor: Optional[int]=…, auto_find_batch_size: bool=…, ignore_data_skip: bool=…, sampler_seed: Optional[int]=…) - added:
method TrainingArguments.set_evaluate(strategy: Union[str, IntervalStrategy]=…, steps: int=…, batch_size: int=…, accumulation_steps: Optional[int]=…, delay: Optional[float]=…, loss_only: bool=…, jit_mode: bool=…) - added:
method TrainingArguments.set_logging(strategy: Union[str, IntervalStrategy]=…, steps: int=…, report_to: Union[str, list[str]]=…, level: str=…, first_step: bool=…, nan_inf_filter: bool=…, on_each_node: bool=…, replica_level: str=…) - added:
method TrainingArguments.set_lr_scheduler(name: Union[str, SchedulerType]=…, num_epochs: float=…, max_steps: int=…, warmup_ratio: float=…, warmup_steps: int=…) - added:
method TrainingArguments.set_optimizer(name: Union[str, OptimizerNames]=…, learning_rate: float=…, weight_decay: float=…, beta1: float=…, beta2: float=…, epsilon: float=…, args: Optional[str]=…) - added:
method TrainingArguments.set_push_to_hub(model_id: str, strategy: Union[str, HubStrategy]=…, token: Optional[str]=…, private_repo: Optional[bool]=…, always_push: bool=…, revision: Optional[str]=…) - added:
method TrainingArguments.set_save(strategy: Union[str, IntervalStrategy]=…, steps: int=…, total_limit: Optional[int]=…, on_each_node: bool=…) - added:
method TrainingArguments.set_testing(batch_size: int=…, loss_only: bool=…, jit_mode: bool=…) - added:
method TrainingArguments.set_training(learning_rate: float=…, batch_size: int=…, weight_decay: float=…, num_epochs: float=…, max_steps: int=…, gradient_accumulation_steps: int=…, seed: int=…, gradient_checkpointing: bool=…) - added:
property TrainingArguments.ddp_timeout_delta -> timedelta - added:
property TrainingArguments.local_process_index - added:
property TrainingArguments.parallel_mode - added:
property TrainingArguments.place_model_on_device - added:
property TrainingArguments.process_index - added:
property TrainingArguments.should_log - added:
property TrainingArguments.should_save - added:
property TrainingArguments.world_size - added:
variable ParallelismConfig - added:
variable log_levels - added:
variable trainer_log_levels
- added:
class Seq2SeqTrainingArguments(TrainingArguments) - added:
field Seq2SeqTrainingArguments.generation_config: Optional[Union[str, Path, GenerationConfig]] - added:
field Seq2SeqTrainingArguments.generation_max_length: Optional[int] - added:
field Seq2SeqTrainingArguments.generation_num_beams: Optional[int] - added:
field Seq2SeqTrainingArguments.predict_with_generate: bool - added:
field Seq2SeqTrainingArguments.sortish_sampler: bool - added:
method Seq2SeqTrainingArguments.to_dict() - added:
variable logger
- changed:
field TFTrainingArguments.tpu_name: str → field TFTrainingArguments.tpu_name: Optional[str] - added:
field TFTrainingArguments.framework - added:
field TFTrainingArguments.gcp_project: Optional[str] - added:
field TFTrainingArguments.tpu_zone: Optional[str] - added:
property TFTrainingArguments.should_log
- added:
export ACCELERATE_MIN_VERSION - added:
export ADAPTER_CONFIG_NAME - added:
export ADAPTER_SAFE_WEIGHTS_NAME - added:
export ADAPTER_WEIGHTS_NAME - added:
export BackboneConfigMixin - added:
export BackboneMixin - added:
export CHAT_TEMPLATE_DIR - added:
export CHAT_TEMPLATE_FILE - added:
export CLOUDFRONT_DISTRIB_PREFIX - added:
export ClassAttrs - added:
export ClassDocstring - added:
export ContextManagers - added:
export DocstringParsingException - added:
export DummyObject - added:
export ENV_VARS_TRUE_AND_AUTO_VALUES - added:
export ENV_VARS_TRUE_VALUES - added:
export EntryNotFoundError - added:
export ExplicitEnum - added:
export GGUF_MIN_VERSION - added:
export HF_MODULES_CACHE - added:
export HUGGINGFACE_CO_PREFIX - added:
export HUGGINGFACE_CO_RESOLVE_ENDPOINT - added:
export IMAGENET_DEFAULT_MEAN - added:
export IMAGENET_DEFAULT_STD - added:
export IMAGENET_STANDARD_MEAN - added:
export IMAGENET_STANDARD_STD - added:
export ImageProcessorArgs - added:
export LEGACY_PROCESSOR_CHAT_TEMPLATE_FILE - added:
export ModelArgs - added:
export ModelOutput - added:
export ModelOutputArgs - added:
export OptionalDependencyNotAvailable - added:
export PYTORCH_PRETRAINED_BERT_CACHE - added:
export PYTORCH_TRANSFORMERS_CACHE - added:
export PaddingStrategy - added:
export PushInProgress - added:
export PushToHubMixin - added:
export RepositoryNotFoundError - added:
export RevisionNotFoundError - added:
export S3_BUCKET_PREFIX - added:
export TORCH_FX_REQUIRED_VERSION - added:
export TRANSFORMERS_CACHE - added:
export TRANSFORMERS_DYNAMIC_MODULE_NAME - added:
export TRITON_MIN_VERSION - added:
export TensorType - added:
export TransformersKwargs - added:
export TypeHintParsingException - added:
export USE_JAX - added:
export USE_TF - added:
export USE_TORCH - added:
export XLA_FSDPV2_MIN_VERSION - added:
export add_code_sample_docstrings - added:
export add_end_docstrings - added:
export add_start_docstrings - added:
export add_start_docstrings_to_model_forward - added:
export auto_class_docstring - added:
export auto_docstring - added:
export cached_file - added:
export can_return_loss - added:
export can_return_tuple - added:
export ccl_version - added:
export check_peft_version - added:
export check_torch_load_is_safe - added:
export copy_func - added:
export default_cache_path - added:
export define_sagemaker_information - added:
export direct_transformers_import - added:
export download_url - added:
export expand_dims - added:
export extract_commit_hash - added:
export filter_out_non_signature_kwargs - added:
export find_adapter_config_file - added:
export find_labels - added:
export flatten_dict - added:
export get_args_doc_from_source - added:
export get_json_schema - added:
export get_torch_version - added:
export has_file - added:
export http_user_agent - added:
export infer_framework - added:
export is_accelerate_available - added:
export is_apex_available - added:
export is_apollo_torch_available - added:
export is_aqlm_available - added:
export is_auto_awq_available - added:
export is_auto_gptq_available - added:
export is_auto_round_available - added:
export is_av_available - added:
export is_bitsandbytes_available - added:
export is_bitsandbytes_multi_backend_available - added:
export is_bs4_available - added:
export is_ccl_available - added:
export is_coloredlogs_available - added:
export is_compressed_tensors_available - added:
export is_cuda_platform - added:
export is_cv2_available - added:
export is_cython_available - added:
export is_datasets_available - added:
export is_decord_available - added:
export is_detectron2_available - added:
export is_eetq_available - added:
export is_essentia_available - added:
export is_faiss_available - added:
export is_fbgemm_gpu_available - added:
export is_flash_attn_2_available - added:
export is_flash_attn_3_available - added:
export is_flash_attn_greater_or_equal - added:
export is_flash_attn_greater_or_equal_2_10 - added:
export is_flax_available - added:
export is_flute_available - added:
export is_fp_quant_available - added:
export is_fsdp_available - added:
export is_ftfy_available - added:
export is_g2p_en_available - added:
export is_galore_torch_available - added:
export is_gguf_available - added:
export is_gptqmodel_available - added:
export is_grokadamw_available - added:
export is_habana_gaudi1 - added:
export is_hadamard_available - added:
export is_hqq_available - added:
export is_huggingface_hub_greater_or_equal - added:
export is_in_notebook - added:
export is_ipex_available - added:
export is_jax_tensor - added:
export is_jinja_available - added:
export is_jumanpp_available - added:
export is_kenlm_available - added:
export is_keras_nlp_available - added:
export is_kernels_available - added:
export is_levenshtein_available - added:
export is_libcst_available - added:
export is_librosa_available - added:
export is_liger_kernel_available - added:
export is_lomo_available - added:
export is_matplotlib_available - added:
export is_mistral_common_available - added:
export is_mlx_available - added:
export is_natten_available - added:
export is_ninja_available - added:
export is_nltk_available - added:
export is_num2words_available - added:
export is_numpy_array - added:
export is_offline_mode - added:
export is_onnx_available - added:
export is_openai_available - added:
export is_optimum_available - added:
export is_optimum_quanto_available - added:
export is_pandas_available - added:
export is_peft_available - added:
export is_phonemizer_available - added:
export is_pretty_midi_available - added:
export is_protobuf_available - added:
export is_psutil_available - added:
export is_py3nvml_available - added:
export is_pyctcdecode_available - added:
export is_pytesseract_available - added:
export is_pytest_available - added:
export is_pytorch_quantization_available - added:
export is_quanto_greater - added:
export is_quark_available - added:
export is_qutlass_available - added:
export is_remote_url - added:
export is_rich_available - added:
export is_rjieba_available - added:
export is_rocm_platform - added:
export is_sacremoses_available - added:
export is_safetensors_available - added:
export is_sagemaker_dp_enabled - added:
export is_sagemaker_mp_enabled - added:
export is_schedulefree_available - added:
export is_scipy_available - added:
export is_sentencepiece_available - added:
export is_seqio_available - added:
export is_sklearn_available - added:
export is_soundfile_available - added:
export is_spacy_available - added:
export is_speech_available - added:
export is_spqr_available - added:
export is_sudachi_available - added:
export is_sudachi_projection_available - added:
export is_tensor - added:
export is_tensorflow_probability_available - added:
export is_tensorflow_text_available - added:
export is_tf2onnx_available - added:
export is_tf_available - added:
export is_tf_symbolic_tensor - added:
export is_tf_tensor - added:
export is_tiktoken_available - added:
export is_timm_available - added:
export is_timm_config_dict - added:
export is_timm_local_checkpoint - added:
export is_tokenizers_available - added:
export is_torch_accelerator_available - added:
export is_torch_available - added:
export is_torch_bf16_available - added:
export is_torch_bf16_available_on_device - added:
export is_torch_bf16_cpu_available - added:
export is_torch_bf16_gpu_available - added:
export is_torch_compile_available - added:
export is_torch_cuda_available - added:
export is_torch_deterministic - added:
export is_torch_device - added:
export is_torch_dtype - added:
export is_torch_flex_attn_available - added:
export is_torch_fp16_available_on_device - added:
export is_torch_fx_available - added:
export is_torch_fx_proxy - added:
export is_torch_greater_or_equal - added:
export is_torch_hpu_available - added:
export is_torch_mlu_available - added:
export is_torch_mps_available - added:
export is_torch_musa_available - added:
export is_torch_neuroncore_available - added:
export is_torch_npu_available - added:
export is_torch_optimi_available - added:
export is_torch_sdpa_available - added:
export is_torch_tensor - added:
export is_torch_tensorrt_fx_available - added:
export is_torch_tf32_available - added:
export is_torch_xla_available - added:
export is_torch_xpu_available - added:
export is_torchao_available - added:
export is_torchaudio_available - added:
export is_torchcodec_available - added:
export is_torchdistx_available - added:
export is_torchdynamo_available - added:
export is_torchdynamo_compiling - added:
export is_torchdynamo_exporting - added:
export is_torchvision_available - added:
export is_torchvision_v2_available - added:
export is_training_run_on_sagemaker - added:
export is_triton_available - added:
export is_uroman_available - added:
export is_vision_available - added:
export is_vptq_available - added:
export is_xlstm_available - added:
export is_yt_dlp_available - added:
export list_repo_templates - added:
export parse_docstring - added:
export replace_return_docstrings - added:
export requires_backends - added:
export reshape - added:
export set_min_indent - added:
export squeeze - added:
export strtobool - added:
export tensor_size - added:
export to_numpy - added:
export to_py_obj - added:
export torch_float - added:
export torch_int - added:
export torch_only_method - added:
export transpose - added:
export try_to_load_from_cache - added:
export working_or_temp_dir - added:
function check_min_version(min_version) - added:
function get_available_devices() -> frozenset[str] - added:
variable AUDIO_TOKENIZER_NAME - added:
variable CONFIG_NAME - added:
variable DUMMY_INPUTS - added:
variable DUMMY_MASK - added:
variable FEATURE_EXTRACTOR_NAME - added:
variable FLAX_WEIGHTS_INDEX_NAME - added:
variable FLAX_WEIGHTS_NAME - added:
variable GENERATION_CONFIG_NAME - added:
variable IMAGE_PROCESSOR_NAME - added:
variable MODEL_CARD_NAME - added:
variable MULTIPLE_CHOICE_DUMMY_INPUTS - added:
variable PROCESSOR_NAME - added:
variable SAFE_WEIGHTS_INDEX_NAME - added:
variable SAFE_WEIGHTS_NAME - added:
variable SENTENCEPIECE_UNDERLINE - added:
variable SPIECE_UNDERLINE - added:
variable TF2_WEIGHTS_INDEX_NAME - added:
variable TF2_WEIGHTS_NAME - added:
variable TF_WEIGHTS_NAME - added:
variable VIDEO_PROCESSOR_NAME - added:
variable WEIGHTS_INDEX_NAME - added:
variable WEIGHTS_NAME
- added:
class AttentionMaskVisualizer - added:
constructor AttentionMaskVisualizer(model_name: str) - added:
function generate_attention_matrix_from_mask(words, mask, img_token=…, sliding_window=…, token_type_ids=…, image_seq_length=…) - added:
method AttentionMaskVisualizer.visualize_attention_mask(input_sentence: str, suffix=…) - added:
variable BLACK_SQUARE - added:
variable GREEN - added:
variable RESET - added:
variable WHITE_SQUARE - added:
variable YELLOW
- added:
class ClassAttrs - added:
class ClassDocstring - added:
class ImageProcessorArgs - added:
class ModelArgs - added:
class ModelOutputArgs - added:
field ClassAttrs.base_model_prefix - added:
field ClassAttrs.supports_gradient_checkpointing - added:
field ClassDocstring.Backbone - added:
field ClassDocstring.Decoder - added:
field ClassDocstring.ForAudioClassification - added:
field ClassDocstring.ForAudioFrameClassification - added:
field ClassDocstring.ForCausalLM - added:
field ClassDocstring.ForConditionalGeneration - added:
field ClassDocstring.ForImageClassification - added:
field ClassDocstring.ForMaskedLM - added:
field ClassDocstring.ForMultipleChoice - added:
field ClassDocstring.ForPreTraining - added:
field ClassDocstring.ForPrediction - added:
field ClassDocstring.ForQuestionAnswering - added:
field ClassDocstring.ForSemanticSegmentation - added:
field ClassDocstring.ForSequenceClassification - added:
field ClassDocstring.ForTokenClassification - added:
field ClassDocstring.ImageProcessorFast - added:
field ClassDocstring.Model - added:
field ClassDocstring.PreTrainedModel - added:
field ClassDocstring.TextModel - added:
field ClassDocstring.WithProjection - added:
field ImageProcessorArgs.crop_size - added:
field ImageProcessorArgs.data_format - added:
field ImageProcessorArgs.default_to_square - added:
field ImageProcessorArgs.device - added:
field ImageProcessorArgs.disable_grouping - added:
field ImageProcessorArgs.do_center_crop - added:
field ImageProcessorArgs.do_convert_rgb - added:
field ImageProcessorArgs.do_normalize - added:
field ImageProcessorArgs.do_pad - added:
field ImageProcessorArgs.do_rescale - added:
field ImageProcessorArgs.do_resize - added:
field ImageProcessorArgs.image_mean - added:
field ImageProcessorArgs.image_std - added:
field ImageProcessorArgs.images - added:
field ImageProcessorArgs.input_data_format - added:
field ImageProcessorArgs.pad_size - added:
field ImageProcessorArgs.resample - added:
field ImageProcessorArgs.rescale_factor - added:
field ImageProcessorArgs.return_tensors - added:
field ImageProcessorArgs.size - added:
field ImageProcessorArgs.videos - added:
field ModelArgs.attention_mask - added:
field ModelArgs.cache_position - added:
field ModelArgs.config - added:
field ModelArgs.cross_attn_head_mask - added:
field ModelArgs.decoder_attention_mask - added:
field ModelArgs.decoder_head_mask - added:
field ModelArgs.decoder_input_ids - added:
field ModelArgs.decoder_inputs_embeds - added:
field ModelArgs.encoder_attention_mask - added:
field ModelArgs.encoder_hidden_states - added:
field ModelArgs.encoder_outputs - added:
field ModelArgs.end_positions - added:
field ModelArgs.head_mask - added:
field ModelArgs.hidden_states - added:
field ModelArgs.image_sizes - added:
field ModelArgs.input_features - added:
field ModelArgs.input_ids - added:
field ModelArgs.input_values - added:
field ModelArgs.inputs_embeds - added:
field ModelArgs.interpolate_pos_encoding - added:
field ModelArgs.labels - added:
field ModelArgs.logits_to_keep - added:
field ModelArgs.num_logits_to_keep - added:
field ModelArgs.output_attentions - added:
field ModelArgs.output_hidden_states - added:
field ModelArgs.output_router_logits - added:
field ModelArgs.past_key_values - added:
field ModelArgs.pixel_mask - added:
field ModelArgs.pixel_values - added:
field ModelArgs.pixel_values_videos - added:
field ModelArgs.position_embeddings - added:
field ModelArgs.position_ids - added:
field ModelArgs.return_dict - added:
field ModelArgs.start_positions - added:
field ModelArgs.token_type_ids - added:
field ModelArgs.use_cache - added:
field ModelArgs.vision_feature_layer - added:
field ModelArgs.vision_feature_select_strategy - added:
field ModelOutputArgs.attentions - added:
field ModelOutputArgs.aux_loss - added:
field ModelOutputArgs.cross_attentions - added:
field ModelOutputArgs.decoder_attentions - added:
field ModelOutputArgs.decoder_hidden_states - added:
field ModelOutputArgs.embeddings - added:
field ModelOutputArgs.encoder_attentions - added:
field ModelOutputArgs.encoder_hidden_states - added:
field ModelOutputArgs.encoder_last_hidden_state - added:
field ModelOutputArgs.end_logits - added:
field ModelOutputArgs.extract_features - added:
field ModelOutputArgs.feature_maps - added:
field ModelOutputArgs.hidden_states - added:
field ModelOutputArgs.image_hidden_states - added:
field ModelOutputArgs.last_hidden_state - added:
field ModelOutputArgs.loc - added:
field ModelOutputArgs.params - added:
field ModelOutputArgs.past_key_values - added:
field ModelOutputArgs.pooler_output - added:
field ModelOutputArgs.predicted_depth - added:
field ModelOutputArgs.projection_state - added:
field ModelOutputArgs.reconstruction - added:
field ModelOutputArgs.router_logits - added:
field ModelOutputArgs.router_probs - added:
field ModelOutputArgs.scale - added:
field ModelOutputArgs.sequences - added:
field ModelOutputArgs.spectrogram - added:
field ModelOutputArgs.start_logits - added:
field ModelOutputArgs.static_features - added:
field ModelOutputArgs.video_hidden_states - added:
field ModelOutputArgs.z_loss - added:
function add_intro_docstring(func, class_name, indent_level=…) - added:
function auto_class_docstring(cls, custom_intro=…, custom_args=…, checkpoint=…) - added:
function auto_docstring(obj=…, *, custom_intro=…, custom_args=…, checkpoint=…) - added:
function auto_method_docstring(func, parent_class=…, custom_intro=…, custom_args=…, checkpoint=…, source_args_dict=…) - added:
function contains_type(type_hint, target_type) -> tuple[bool, Optional[object]] - added:
function equalize_indent(docstring, indent_level) - added:
function find_sig_line(lines, line_end) - added:
function format_args_docstring(docstring, model_name) - added:
function get_args_doc_from_source(args_classes: Union[object, list[object]]) -> dict - added:
function get_checkpoint_from_config_class(config_class) - added:
function get_indent_level(func) - added:
function get_model_name(obj) - added:
function get_placeholders_dict(placeholders: list, model_name: str) -> dict - added:
function parse_default(docstring) - added:
function parse_docstring(docstring, max_indent_level=…, return_intro=…) - added:
function parse_shape(docstring) - added:
function set_min_indent(docstring, indent_level) - added:
variable ARGS_TO_IGNORE - added:
variable AUTODOC_FILES - added:
variable HARDCODED_CONFIG_FOR_MODELS - added:
variable PATH_TO_TRANSFORMERS - added:
variable PLACEHOLDER_TO_AUTO_MODULE - added:
variable UNROLL_KWARGS_CLASSES - added:
variable UNROLL_KWARGS_METHODS
- added:
class BackboneConfigMixin - added:
class BackboneMixin - added:
class BackboneType(enum.Enum) - added:
field BackboneMixin.backbone_type: Optional[BackboneType] - added:
field BackboneMixin.has_attentions: bool - added:
field BackboneType.TIMM - added:
field BackboneType.TRANSFORMERS - added:
function get_aligned_output_features_output_indices(out_features: Optional[list[str]], out_indices: Optional[Union[list[int], tuple[int]]], stage_names: list[str]) -> tuple[list[str], list[int]] - added:
function load_backbone(config) - added:
function verify_backbone_config_arguments(use_timm_backbone: bool, use_pretrained_backbone: bool, backbone: Optional[str], backbone_config: Optional[Union[dict, PretrainedConfig]], backbone_kwargs: Optional[dict]) - added:
function verify_out_features_out_indices(out_features: Optional[Iterable[str]], out_indices: Optional[Iterable[int]], stage_names: Optional[Iterable[str]]) - added:
method BackboneConfigMixin.to_dict() - added:
method BackboneMixin.forward(pixel_values, output_hidden_states: Optional[bool]=…, output_attentions: Optional[bool]=…, return_dict: Optional[bool]=…) - added:
method BackboneMixin.forward_with_filtered_kwargs(*args, **kwargs) - added:
method BackboneMixin.to_dict() - added:
property BackboneConfigMixin.out_features - added:
property BackboneConfigMixin.out_indices - added:
property BackboneMixin.channels - added:
property BackboneMixin.out_feature_channels - added:
property BackboneMixin.out_features - added:
property BackboneMixin.out_indices
- added:
class DocstringParsingException(Exception) - added:
class TypeHintParsingException(Exception) - added:
function get_json_schema(func: Callable) -> dict - added:
function parse_google_format_docstring(docstring: str) -> tuple[Optional[str], Optional[dict], Optional[str]] - added:
function render_jinja_template(conversations: list[list[dict[str, str]]], tools: Optional[list[Union[dict, Callable]]]=…, documents: Optional[list[dict[str, str]]]=…, chat_template: Optional[str]=…, return_assistant_tokens_mask: bool=…, continue_final_message: bool=…, add_generation_prompt: bool=…, **kwargs) -> str - added:
variable BASIC_TYPES - added:
variable args_re - added:
variable args_split_re - added:
variable description_re - added:
variable jinja2 - added:
variable logger - added:
variable returns_re
- added:
variable IMAGENET_DEFAULT_MEAN - added:
variable IMAGENET_DEFAULT_STD - added:
variable IMAGENET_STANDARD_MEAN - added:
variable IMAGENET_STANDARD_STD - added:
variable OPENAI_CLIP_MEAN - added:
variable OPENAI_CLIP_STD
- added:
class Action(ExplicitEnum) - added:
field Action.NONE - added:
field Action.NOTIFY - added:
field Action.NOTIFY_ALWAYS - added:
field Action.RAISE - added:
function deprecate_kwarg(old_name: str, version: str, new_name: Optional[str]=…, warn_if_greater_or_equal_version: bool=…, raise_if_greater_or_equal_version: bool=…, raise_if_both_names: bool=…, additional_message: Optional[str]=…)
- added:
class LayoutLMv2Model - added:
constructor LayoutLMv2Model(*args, **kwargs) - added:
static method LayoutLMv2Model.from_pretrained(*args, **kwargs)
- added:
class Pop2PianoFeatureExtractor - added:
class Pop2PianoProcessor - added:
class Pop2PianoTokenizer - added:
constructor Pop2PianoFeatureExtractor(*args, **kwargs) - added:
constructor Pop2PianoProcessor(*args, **kwargs) - added:
constructor Pop2PianoTokenizer(*args, **kwargs)
- removed:
class FlaxBertModel - removed:
class FlaxRobertaModel - removed:
constructor FlaxBertModel(*args, **kwargs) - removed:
constructor FlaxRobertaModel(*args, **kwargs) - removed:
static method FlaxBertModel.from_pretrained(*args, **kwargs) - removed:
static method FlaxRobertaModel.from_pretrained(*args, **kwargs) - added:
class FlaxForceTokensLogitsProcessor - added:
class FlaxForcedBOSTokenLogitsProcessor - added:
class FlaxForcedEOSTokenLogitsProcessor - added:
class FlaxGenerationMixin - added:
class FlaxLogitsProcessor - added:
class FlaxLogitsProcessorList - added:
class FlaxLogitsWarper - added:
class FlaxMinLengthLogitsProcessor - added:
class FlaxPreTrainedModel - added:
class FlaxSuppressTokensAtBeginLogitsProcessor - added:
class FlaxSuppressTokensLogitsProcessor - added:
class FlaxTemperatureLogitsWarper - added:
class FlaxTopKLogitsWarper - added:
class FlaxTopPLogitsWarper - added:
class FlaxWhisperTimeStampLogitsProcessor - added:
constructor FlaxForceTokensLogitsProcessor(*args, **kwargs) - added:
constructor FlaxForcedBOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor FlaxForcedEOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor FlaxGenerationMixin(*args, **kwargs) - added:
constructor FlaxLogitsProcessor(*args, **kwargs) - added:
constructor FlaxLogitsProcessorList(*args, **kwargs) - added:
constructor FlaxLogitsWarper(*args, **kwargs) - added:
constructor FlaxMinLengthLogitsProcessor(*args, **kwargs) - added:
constructor FlaxPreTrainedModel(*args, **kwargs) - added:
constructor FlaxSuppressTokensAtBeginLogitsProcessor(*args, **kwargs) - added:
constructor FlaxSuppressTokensLogitsProcessor(*args, **kwargs) - added:
constructor FlaxTemperatureLogitsWarper(*args, **kwargs) - added:
constructor FlaxTopKLogitsWarper(*args, **kwargs) - added:
constructor FlaxTopPLogitsWarper(*args, **kwargs) - added:
constructor FlaxWhisperTimeStampLogitsProcessor(*args, **kwargs)
- added:
class MistralCommonTokenizer - added:
constructor MistralCommonTokenizer(*args, **kwargs)
- added:
class Pop2PianoFeatureExtractor - added:
class Pop2PianoTokenizer - added:
constructor Pop2PianoFeatureExtractor(*args, **kwargs) - added:
constructor Pop2PianoTokenizer(*args, **kwargs)
- removed:
class AdamW - removed:
class AdaptiveEmbedding - removed:
class AlbertForMaskedLM - removed:
class AlbertForMultipleChoice - removed:
class AlbertForPreTraining - removed:
class AlbertForQuestionAnswering - removed:
class AlbertForSequenceClassification - removed:
class AlbertForTokenClassification - removed:
class AlbertModel - removed:
class AlbertPreTrainedModel - removed:
class AutoModel - removed:
class AutoModelForCausalLM - removed:
class AutoModelForMaskedLM - removed:
class AutoModelForMultipleChoice - removed:
class AutoModelForNextSentencePrediction - removed:
class AutoModelForPreTraining - removed:
class AutoModelForQuestionAnswering - removed:
class AutoModelForSeq2SeqLM - removed:
class AutoModelForSequenceClassification - removed:
class AutoModelForTokenClassification - removed:
class AutoModelWithLMHead - removed:
class BartForConditionalGeneration - removed:
class BartForQuestionAnswering - removed:
class BartForSequenceClassification - removed:
class BartModel - removed:
class BeamSearchScorer - removed:
class BertForMaskedLM - removed:
class BertForMultipleChoice - removed:
class BertForNextSentencePrediction - removed:
class BertForPreTraining - removed:
class BertForQuestionAnswering - removed:
class BertForSequenceClassification - removed:
class BertForTokenClassification - removed:
class BertGenerationDecoder - removed:
class BertGenerationEncoder - removed:
class BertLMHeadModel - removed:
class BertLayer - removed:
class BertModel - removed:
class BertPreTrainedModel - removed:
class BlenderbotForConditionalGeneration - removed:
class CTRLLMHeadModel - removed:
class CTRLModel - removed:
class CTRLPreTrainedModel - removed:
class CamembertForCausalLM - removed:
class CamembertForMaskedLM - removed:
class CamembertForMultipleChoice - removed:
class CamembertForQuestionAnswering - removed:
class CamembertForSequenceClassification - removed:
class CamembertForTokenClassification - removed:
class CamembertModel - removed:
class DPRContextEncoder - removed:
class DPRPretrainedContextEncoder - removed:
class DPRPretrainedQuestionEncoder - removed:
class DPRPretrainedReader - removed:
class DPRQuestionEncoder - removed:
class DPRReader - removed:
class DataCollator - removed:
class DataCollatorForLanguageModeling - removed:
class DataCollatorForPermutationLanguageModeling - removed:
class DataCollatorForSOP - removed:
class DataCollatorForTokenClassification - removed:
class DataCollatorForWholeWordMask - removed:
class DataCollatorWithPadding - removed:
class DebertaForSequenceClassification - removed:
class DebertaModel - removed:
class DebertaPreTrainedModel - removed:
class DistilBertForMaskedLM - removed:
class DistilBertForMultipleChoice - removed:
class DistilBertForQuestionAnswering - removed:
class DistilBertForSequenceClassification - removed:
class DistilBertForTokenClassification - removed:
class DistilBertModel - removed:
class DistilBertPreTrainedModel - removed:
class ElectraForMaskedLM - removed:
class ElectraForMultipleChoice - removed:
class ElectraForPreTraining - removed:
class ElectraForQuestionAnswering - removed:
class ElectraForSequenceClassification - removed:
class ElectraForTokenClassification - removed:
class ElectraModel - removed:
class ElectraPreTrainedModel - removed:
class EncoderDecoderModel - removed:
class FSMTForConditionalGeneration - removed:
class FSMTModel - removed:
class FlaubertForMultipleChoice - removed:
class FlaubertForQuestionAnswering - removed:
class FlaubertForQuestionAnsweringSimple - removed:
class FlaubertForSequenceClassification - removed:
class FlaubertForTokenClassification - removed:
class FlaubertModel - removed:
class FlaubertWithLMHeadModel - removed:
class FunnelBaseModel - removed:
class FunnelForMaskedLM - removed:
class FunnelForMultipleChoice - removed:
class FunnelForPreTraining - removed:
class FunnelForQuestionAnswering - removed:
class FunnelForSequenceClassification - removed:
class FunnelForTokenClassification - removed:
class FunnelModel - removed:
class GPT2DoubleHeadsModel - removed:
class GPT2ForSequenceClassification - removed:
class GPT2LMHeadModel - removed:
class GPT2Model - removed:
class GPT2PreTrainedModel - removed:
class LayoutLMForMaskedLM - removed:
class LayoutLMForTokenClassification - removed:
class LayoutLMModel - removed:
class LogitsWarper - removed:
class LongformerForMaskedLM - removed:
class LongformerForMultipleChoice - removed:
class LongformerForQuestionAnswering - removed:
class LongformerForSequenceClassification - removed:
class LongformerForTokenClassification - removed:
class LongformerModel - removed:
class LongformerSelfAttention - removed:
class LxmertEncoder - removed:
class LxmertForPreTraining - removed:
class LxmertForQuestionAnswering - removed:
class LxmertModel - removed:
class LxmertPreTrainedModel - removed:
class LxmertVisualFeatureEncoder - removed:
class LxmertXLayer - removed:
class MBartForConditionalGeneration - removed:
class MMBTForClassification - removed:
class MMBTModel - removed:
class MarianMTModel - removed:
class MobileBertForMaskedLM - removed:
class MobileBertForMultipleChoice - removed:
class MobileBertForNextSentencePrediction - removed:
class MobileBertForPreTraining - removed:
class MobileBertForQuestionAnswering - removed:
class MobileBertForSequenceClassification - removed:
class MobileBertForTokenClassification - removed:
class MobileBertLayer - removed:
class MobileBertModel - removed:
class MobileBertPreTrainedModel - removed:
class ModalEmbeddings - removed:
class OpenAIGPTDoubleHeadsModel - removed:
class OpenAIGPTForSequenceClassification - removed:
class OpenAIGPTLMHeadModel - removed:
class OpenAIGPTModel - removed:
class OpenAIGPTPreTrainedModel - removed:
class PegasusForConditionalGeneration - removed:
class PretrainedBartModel - removed:
class PretrainedFSMTModel - removed:
class ProphetNetDecoder - removed:
class ProphetNetEncoder - removed:
class ProphetNetForCausalLM - removed:
class ProphetNetForConditionalGeneration - removed:
class ProphetNetModel - removed:
class ProphetNetPreTrainedModel - removed:
class PyTorchBenchmark - removed:
class PyTorchBenchmarkArguments - removed:
class RagModel - removed:
class RagSequenceForGeneration - removed:
class RagTokenForGeneration - removed:
class ReformerAttention - removed:
class ReformerForMaskedLM - removed:
class ReformerForQuestionAnswering - removed:
class ReformerForSequenceClassification - removed:
class ReformerLayer - removed:
class ReformerModel - removed:
class ReformerModelWithLMHead - removed:
class RetriBertModel - removed:
class RetriBertPreTrainedModel - removed:
class RobertaForCausalLM - removed:
class RobertaForMaskedLM - removed:
class RobertaForMultipleChoice - removed:
class RobertaForQuestionAnswering - removed:
class RobertaForSequenceClassification - removed:
class RobertaForTokenClassification - removed:
class RobertaModel - removed:
class SqueezeBertForMaskedLM - removed:
class SqueezeBertForMultipleChoice - removed:
class SqueezeBertForQuestionAnswering - removed:
class SqueezeBertForSequenceClassification - removed:
class SqueezeBertForTokenClassification - removed:
class SqueezeBertModel - removed:
class SqueezeBertModule - removed:
class SqueezeBertPreTrainedModel - removed:
class T5ForConditionalGeneration - removed:
class T5Model - removed:
class T5PreTrainedModel - removed:
class TransfoXLLMHeadModel - removed:
class TransfoXLModel - removed:
class TransfoXLPreTrainedModel - removed:
class XLMForMultipleChoice - removed:
class XLMForQuestionAnswering - removed:
class XLMForQuestionAnsweringSimple - removed:
class XLMForSequenceClassification - removed:
class XLMForTokenClassification - removed:
class XLMModel - removed:
class XLMPreTrainedModel - removed:
class XLMProphetNetDecoder - removed:
class XLMProphetNetEncoder - removed:
class XLMProphetNetForCausalLM - removed:
class XLMProphetNetForConditionalGeneration - removed:
class XLMProphetNetModel - removed:
class XLMRobertaForCausalLM - removed:
class XLMRobertaForMaskedLM - removed:
class XLMRobertaForMultipleChoice - removed:
class XLMRobertaForQuestionAnswering - removed:
class XLMRobertaForSequenceClassification - removed:
class XLMRobertaForTokenClassification - removed:
class XLMRobertaModel - removed:
class XLMWithLMHeadModel - removed:
class XLNetForMultipleChoice - removed:
class XLNetForQuestionAnswering - removed:
class XLNetForQuestionAnsweringSimple - removed:
class XLNetForSequenceClassification - removed:
class XLNetForTokenClassification - removed:
class XLNetLMHeadModel - removed:
class XLNetModel - removed:
class XLNetPreTrainedModel - removed:
constructor AdamW(*args, **kwargs) - removed:
constructor AdaptiveEmbedding(*args, **kwargs) - removed:
constructor AlbertForMaskedLM(*args, **kwargs) - removed:
constructor AlbertForMultipleChoice(*args, **kwargs) - removed:
constructor AlbertForPreTraining(*args, **kwargs) - removed:
constructor AlbertForQuestionAnswering(*args, **kwargs) - removed:
constructor AlbertForSequenceClassification(*args, **kwargs) - removed:
constructor AlbertForTokenClassification(*args, **kwargs) - removed:
constructor AlbertModel(*args, **kwargs) - removed:
constructor AlbertPreTrainedModel(*args, **kwargs) - removed:
constructor AutoModel(*args, **kwargs) - removed:
constructor AutoModelForCausalLM(*args, **kwargs) - removed:
constructor AutoModelForMaskedLM(*args, **kwargs) - removed:
constructor AutoModelForMultipleChoice(*args, **kwargs) - removed:
constructor AutoModelForNextSentencePrediction(*args, **kwargs) - removed:
constructor AutoModelForPreTraining(*args, **kwargs) - removed:
constructor AutoModelForQuestionAnswering(*args, **kwargs) - removed:
constructor AutoModelForSeq2SeqLM(*args, **kwargs) - removed:
constructor AutoModelForSequenceClassification(*args, **kwargs) - removed:
constructor AutoModelForTokenClassification(*args, **kwargs) - removed:
constructor AutoModelWithLMHead(*args, **kwargs) - removed:
constructor BartForConditionalGeneration(*args, **kwargs) - removed:
constructor BartForQuestionAnswering(*args, **kwargs) - removed:
constructor BartForSequenceClassification(*args, **kwargs) - removed:
constructor BartModel(*args, **kwargs) - removed:
constructor BeamSearchScorer(*args, **kwargs) - removed:
constructor BertForMaskedLM(*args, **kwargs) - removed:
constructor BertForMultipleChoice(*args, **kwargs) - removed:
constructor BertForNextSentencePrediction(*args, **kwargs) - removed:
constructor BertForPreTraining(*args, **kwargs) - removed:
constructor BertForQuestionAnswering(*args, **kwargs) - removed:
constructor BertForSequenceClassification(*args, **kwargs) - removed:
constructor BertForTokenClassification(*args, **kwargs) - removed:
constructor BertGenerationDecoder(*args, **kwargs) - removed:
constructor BertGenerationEncoder(*args, **kwargs) - removed:
constructor BertLMHeadModel(*args, **kwargs) - removed:
constructor BertLayer(*args, **kwargs) - removed:
constructor BertModel(*args, **kwargs) - removed:
constructor BertPreTrainedModel(*args, **kwargs) - removed:
constructor BlenderbotForConditionalGeneration(*args, **kwargs) - removed:
constructor CTRLLMHeadModel(*args, **kwargs) - removed:
constructor CTRLModel(*args, **kwargs) - removed:
constructor CTRLPreTrainedModel(*args, **kwargs) - removed:
constructor CamembertForCausalLM(*args, **kwargs) - removed:
constructor CamembertForMaskedLM(*args, **kwargs) - removed:
constructor CamembertForMultipleChoice(*args, **kwargs) - removed:
constructor CamembertForQuestionAnswering(*args, **kwargs) - removed:
constructor CamembertForSequenceClassification(*args, **kwargs) - removed:
constructor CamembertForTokenClassification(*args, **kwargs) - removed:
constructor CamembertModel(*args, **kwargs) - removed:
constructor DPRContextEncoder(*args, **kwargs) - removed:
constructor DPRPretrainedContextEncoder(*args, **kwargs) - removed:
constructor DPRPretrainedQuestionEncoder(*args, **kwargs) - removed:
constructor DPRPretrainedReader(*args, **kwargs) - removed:
constructor DPRQuestionEncoder(*args, **kwargs) - removed:
constructor DPRReader(*args, **kwargs) - removed:
constructor DataCollator(*args, **kwargs) - removed:
constructor DataCollatorForLanguageModeling(*args, **kwargs) - removed:
constructor DataCollatorForPermutationLanguageModeling(*args, **kwargs) - removed:
constructor DataCollatorForSOP(*args, **kwargs) - removed:
constructor DataCollatorForTokenClassification(*args, **kwargs) - removed:
constructor DataCollatorForWholeWordMask(*args, **kwargs) - removed:
constructor DataCollatorWithPadding(*args, **kwargs) - removed:
constructor DebertaForSequenceClassification(*args, **kwargs) - removed:
constructor DebertaModel(*args, **kwargs) - removed:
constructor DebertaPreTrainedModel(*args, **kwargs) - removed:
constructor DistilBertForMaskedLM(*args, **kwargs) - removed:
constructor DistilBertForMultipleChoice(*args, **kwargs) - removed:
constructor DistilBertForQuestionAnswering(*args, **kwargs) - removed:
constructor DistilBertForSequenceClassification(*args, **kwargs) - removed:
constructor DistilBertForTokenClassification(*args, **kwargs) - removed:
constructor DistilBertModel(*args, **kwargs) - removed:
constructor DistilBertPreTrainedModel(*args, **kwargs) - removed:
constructor ElectraForMaskedLM(*args, **kwargs) - removed:
constructor ElectraForMultipleChoice(*args, **kwargs) - removed:
constructor ElectraForPreTraining(*args, **kwargs) - removed:
constructor ElectraForQuestionAnswering(*args, **kwargs) - removed:
constructor ElectraForSequenceClassification(*args, **kwargs) - removed:
constructor ElectraForTokenClassification(*args, **kwargs) - removed:
constructor ElectraModel(*args, **kwargs) - removed:
constructor ElectraPreTrainedModel(*args, **kwargs) - removed:
constructor EncoderDecoderModel(*args, **kwargs) - removed:
constructor FSMTForConditionalGeneration(*args, **kwargs) - removed:
constructor FSMTModel(*args, **kwargs) - removed:
constructor FlaubertForMultipleChoice(*args, **kwargs) - removed:
constructor FlaubertForQuestionAnswering(*args, **kwargs) - removed:
constructor FlaubertForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor FlaubertForSequenceClassification(*args, **kwargs) - removed:
constructor FlaubertForTokenClassification(*args, **kwargs) - removed:
constructor FlaubertModel(*args, **kwargs) - removed:
constructor FlaubertWithLMHeadModel(*args, **kwargs) - removed:
constructor FunnelBaseModel(*args, **kwargs) - removed:
constructor FunnelForMaskedLM(*args, **kwargs) - removed:
constructor FunnelForMultipleChoice(*args, **kwargs) - removed:
constructor FunnelForPreTraining(*args, **kwargs) - removed:
constructor FunnelForQuestionAnswering(*args, **kwargs) - removed:
constructor FunnelForSequenceClassification(*args, **kwargs) - removed:
constructor FunnelForTokenClassification(*args, **kwargs) - removed:
constructor FunnelModel(*args, **kwargs) - removed:
constructor GPT2DoubleHeadsModel(*args, **kwargs) - removed:
constructor GPT2ForSequenceClassification(*args, **kwargs) - removed:
constructor GPT2LMHeadModel(*args, **kwargs) - removed:
constructor GPT2Model(*args, **kwargs) - removed:
constructor GPT2PreTrainedModel(*args, **kwargs) - removed:
constructor LayoutLMForMaskedLM(*args, **kwargs) - removed:
constructor LayoutLMForTokenClassification(*args, **kwargs) - removed:
constructor LayoutLMModel(*args, **kwargs) - removed:
constructor LogitsWarper(*args, **kwargs) - removed:
constructor LongformerForMaskedLM(*args, **kwargs) - removed:
constructor LongformerForMultipleChoice(*args, **kwargs) - removed:
constructor LongformerForQuestionAnswering(*args, **kwargs) - removed:
constructor LongformerForSequenceClassification(*args, **kwargs) - removed:
constructor LongformerForTokenClassification(*args, **kwargs) - removed:
constructor LongformerModel(*args, **kwargs) - removed:
constructor LongformerSelfAttention(*args, **kwargs) - removed:
constructor LxmertEncoder(*args, **kwargs) - removed:
constructor LxmertForPreTraining(*args, **kwargs) - removed:
constructor LxmertForQuestionAnswering(*args, **kwargs) - removed:
constructor LxmertModel(*args, **kwargs) - removed:
constructor LxmertPreTrainedModel(*args, **kwargs) - removed:
constructor LxmertVisualFeatureEncoder(*args, **kwargs) - removed:
constructor LxmertXLayer(*args, **kwargs) - removed:
constructor MBartForConditionalGeneration(*args, **kwargs) - removed:
constructor MMBTForClassification(*args, **kwargs) - removed:
constructor MMBTModel(*args, **kwargs) - removed:
constructor MarianMTModel(*args, **kwargs) - removed:
constructor MobileBertForMaskedLM(*args, **kwargs) - removed:
constructor MobileBertForMultipleChoice(*args, **kwargs) - removed:
constructor MobileBertForNextSentencePrediction(*args, **kwargs) - removed:
constructor MobileBertForPreTraining(*args, **kwargs) - removed:
constructor MobileBertForQuestionAnswering(*args, **kwargs) - removed:
constructor MobileBertForSequenceClassification(*args, **kwargs) - removed:
constructor MobileBertForTokenClassification(*args, **kwargs) - removed:
constructor MobileBertLayer(*args, **kwargs) - removed:
constructor MobileBertModel(*args, **kwargs) - removed:
constructor MobileBertPreTrainedModel(*args, **kwargs) - removed:
constructor ModalEmbeddings(*args, **kwargs) - removed:
constructor OpenAIGPTDoubleHeadsModel(*args, **kwargs) - removed:
constructor OpenAIGPTForSequenceClassification(*args, **kwargs) - removed:
constructor OpenAIGPTLMHeadModel(*args, **kwargs) - removed:
constructor OpenAIGPTModel(*args, **kwargs) - removed:
constructor OpenAIGPTPreTrainedModel(*args, **kwargs) - removed:
constructor PegasusForConditionalGeneration(*args, **kwargs) - removed:
constructor PretrainedBartModel(*args, **kwargs) - removed:
constructor PretrainedFSMTModel(*args, **kwargs) - removed:
constructor ProphetNetDecoder(*args, **kwargs) - removed:
constructor ProphetNetEncoder(*args, **kwargs) - removed:
constructor ProphetNetForCausalLM(*args, **kwargs) - removed:
constructor ProphetNetForConditionalGeneration(*args, **kwargs) - removed:
constructor ProphetNetModel(*args, **kwargs) - removed:
constructor ProphetNetPreTrainedModel(*args, **kwargs) - removed:
constructor PyTorchBenchmark(*args, **kwargs) - removed:
constructor PyTorchBenchmarkArguments(*args, **kwargs) - removed:
constructor RagModel(*args, **kwargs) - removed:
constructor RagSequenceForGeneration(*args, **kwargs) - removed:
constructor RagTokenForGeneration(*args, **kwargs) - removed:
constructor ReformerAttention(*args, **kwargs) - removed:
constructor ReformerForMaskedLM(*args, **kwargs) - removed:
constructor ReformerForQuestionAnswering(*args, **kwargs) - removed:
constructor ReformerForSequenceClassification(*args, **kwargs) - removed:
constructor ReformerLayer(*args, **kwargs) - removed:
constructor ReformerModel(*args, **kwargs) - removed:
constructor ReformerModelWithLMHead(*args, **kwargs) - removed:
constructor RetriBertModel(*args, **kwargs) - removed:
constructor RetriBertPreTrainedModel(*args, **kwargs) - removed:
constructor RobertaForCausalLM(*args, **kwargs) - removed:
constructor RobertaForMaskedLM(*args, **kwargs) - removed:
constructor RobertaForMultipleChoice(*args, **kwargs) - removed:
constructor RobertaForQuestionAnswering(*args, **kwargs) - removed:
constructor RobertaForSequenceClassification(*args, **kwargs) - removed:
constructor RobertaForTokenClassification(*args, **kwargs) - removed:
constructor RobertaModel(*args, **kwargs) - removed:
constructor SqueezeBertForMaskedLM(*args, **kwargs) - removed:
constructor SqueezeBertForMultipleChoice(*args, **kwargs) - removed:
constructor SqueezeBertForQuestionAnswering(*args, **kwargs) - removed:
constructor SqueezeBertForSequenceClassification(*args, **kwargs) - removed:
constructor SqueezeBertForTokenClassification(*args, **kwargs) - removed:
constructor SqueezeBertModel(*args, **kwargs) - removed:
constructor SqueezeBertModule(*args, **kwargs) - removed:
constructor SqueezeBertPreTrainedModel(*args, **kwargs) - removed:
constructor T5ForConditionalGeneration(*args, **kwargs) - removed:
constructor T5Model(*args, **kwargs) - removed:
constructor T5PreTrainedModel(*args, **kwargs) - removed:
constructor TransfoXLLMHeadModel(*args, **kwargs) - removed:
constructor TransfoXLModel(*args, **kwargs) - removed:
constructor TransfoXLPreTrainedModel(*args, **kwargs) - removed:
constructor XLMForMultipleChoice(*args, **kwargs) - removed:
constructor XLMForQuestionAnswering(*args, **kwargs) - removed:
constructor XLMForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor XLMForSequenceClassification(*args, **kwargs) - removed:
constructor XLMForTokenClassification(*args, **kwargs) - removed:
constructor XLMModel(*args, **kwargs) - removed:
constructor XLMPreTrainedModel(*args, **kwargs) - removed:
constructor XLMProphetNetDecoder(*args, **kwargs) - removed:
constructor XLMProphetNetEncoder(*args, **kwargs) - removed:
constructor XLMProphetNetForCausalLM(*args, **kwargs) - removed:
constructor XLMProphetNetForConditionalGeneration(*args, **kwargs) - removed:
constructor XLMProphetNetModel(*args, **kwargs) - removed:
constructor XLMRobertaForCausalLM(*args, **kwargs) - removed:
constructor XLMRobertaForMaskedLM(*args, **kwargs) - removed:
constructor XLMRobertaForMultipleChoice(*args, **kwargs) - removed:
constructor XLMRobertaForQuestionAnswering(*args, **kwargs) - removed:
constructor XLMRobertaForSequenceClassification(*args, **kwargs) - removed:
constructor XLMRobertaForTokenClassification(*args, **kwargs) - removed:
constructor XLMRobertaModel(*args, **kwargs) - removed:
constructor XLMWithLMHeadModel(*args, **kwargs) - removed:
constructor XLNetForMultipleChoice(*args, **kwargs) - removed:
constructor XLNetForQuestionAnswering(*args, **kwargs) - removed:
constructor XLNetForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor XLNetForSequenceClassification(*args, **kwargs) - removed:
constructor XLNetForTokenClassification(*args, **kwargs) - removed:
constructor XLNetLMHeadModel(*args, **kwargs) - removed:
constructor XLNetModel(*args, **kwargs) - removed:
constructor XLNetPreTrainedModel(*args, **kwargs) - removed:
function default_data_collator(*args, **kwargs) - removed:
function load_tf_weights_in_albert(*args, **kwargs) - removed:
function load_tf_weights_in_bert(*args, **kwargs) - removed:
function load_tf_weights_in_bert_generation(*args, **kwargs) - removed:
function load_tf_weights_in_electra(*args, **kwargs) - removed:
function load_tf_weights_in_funnel(*args, **kwargs) - removed:
function load_tf_weights_in_gpt2(*args, **kwargs) - removed:
function load_tf_weights_in_mobilebert(*args, **kwargs) - removed:
function load_tf_weights_in_openai_gpt(*args, **kwargs) - removed:
function load_tf_weights_in_t5(*args, **kwargs) - removed:
function load_tf_weights_in_transfo_xl(*args, **kwargs) - removed:
function load_tf_weights_in_xlnet(*args, **kwargs) - removed:
function top_k_top_p_filtering(*args, **kwargs) - removed:
static method AlbertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method AlbertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method AlbertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method AlbertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method AlbertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method AlbertModel.from_pretrained(*args, **kwargs) - removed:
static method AlbertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method AutoModel.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForCausalLM.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForNextSentencePrediction.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForPreTraining.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForSeq2SeqLM.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method AutoModelForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method AutoModelWithLMHead.from_pretrained(*args, **kwargs) - removed:
static method BartForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method BartForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method BartForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method BartModel.from_pretrained(*args, **kwargs) - removed:
static method BertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method BertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method BertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method BertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method BertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method BertLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method BertModel.from_pretrained(*args, **kwargs) - removed:
static method BertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method BlenderbotForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method CTRLLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method CTRLModel.from_pretrained(*args, **kwargs) - removed:
static method CTRLPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method CamembertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method CamembertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method CamembertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method CamembertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method CamembertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method CamembertModel.from_pretrained(*args, **kwargs) - removed:
static method DataCollatorForLanguageModeling.from_pretrained(*args, **kwargs) - removed:
static method DataCollatorForPermutationLanguageModeling.from_pretrained(*args, **kwargs) - removed:
static method DataCollatorForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method DebertaForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method DebertaModel.from_pretrained(*args, **kwargs) - removed:
static method DebertaPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method DistilBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method DistilBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method DistilBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method DistilBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method DistilBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method DistilBertModel.from_pretrained(*args, **kwargs) - removed:
static method DistilBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method ElectraForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method ElectraForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method ElectraForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method ElectraForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method ElectraForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method ElectraModel.from_pretrained(*args, **kwargs) - removed:
static method ElectraPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method EncoderDecoderModel.from_pretrained(*args, **kwargs) - removed:
static method FSMTForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method FSMTModel.from_pretrained(*args, **kwargs) - removed:
static method FlaubertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method FlaubertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method FlaubertForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method FlaubertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method FlaubertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method FlaubertModel.from_pretrained(*args, **kwargs) - removed:
static method FlaubertWithLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method FunnelBaseModel.from_pretrained(*args, **kwargs) - removed:
static method FunnelForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method FunnelForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method FunnelForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method FunnelForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method FunnelForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method FunnelModel.from_pretrained(*args, **kwargs) - removed:
static method GPT2DoubleHeadsModel.from_pretrained(*args, **kwargs) - removed:
static method GPT2ForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method GPT2LMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method GPT2Model.from_pretrained(*args, **kwargs) - removed:
static method GPT2PreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method LayoutLMForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method LayoutLMForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method LayoutLMModel.from_pretrained(*args, **kwargs) - removed:
static method LongformerForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method LongformerForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method LongformerForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method LongformerForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method LongformerForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method LongformerModel.from_pretrained(*args, **kwargs) - removed:
static method LxmertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method LxmertModel.from_pretrained(*args, **kwargs) - removed:
static method LxmertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method MBartForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method MMBTModel.from_pretrained(*args, **kwargs) - removed:
static method MarianMTModel.from_pretrained(*args, **kwargs) - removed:
static method MobileBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method MobileBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method MobileBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method MobileBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method MobileBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method MobileBertModel.from_pretrained(*args, **kwargs) - removed:
static method MobileBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTDoubleHeadsModel.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTModel.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method PegasusForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method PreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method PretrainedBartModel.from_pretrained(*args, **kwargs) - removed:
static method PretrainedFSMTModel.from_pretrained(*args, **kwargs) - removed:
static method ProphetNetForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method ProphetNetModel.from_pretrained(*args, **kwargs) - removed:
static method ProphetNetPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method RagModel.from_pretrained(*args, **kwargs) - removed:
static method ReformerForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method ReformerForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method ReformerForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method ReformerModel.from_pretrained(*args, **kwargs) - removed:
static method ReformerModelWithLMHead.from_pretrained(*args, **kwargs) - removed:
static method RetriBertModel.from_pretrained(*args, **kwargs) - removed:
static method RetriBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method RobertaForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method RobertaForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method RobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method RobertaForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method RobertaForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method RobertaModel.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertModel.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method T5ForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method T5Model.from_pretrained(*args, **kwargs) - removed:
static method T5PreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TransfoXLLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TransfoXLModel.from_pretrained(*args, **kwargs) - removed:
static method TransfoXLPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method XLMForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method XLMForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method XLMForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method XLMForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method XLMForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method XLMModel.from_pretrained(*args, **kwargs) - removed:
static method XLMPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method XLMProphetNetForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method XLMProphetNetModel.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaModel.from_pretrained(*args, **kwargs) - removed:
static method XLMWithLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method XLNetForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method XLNetForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method XLNetForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method XLNetForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method XLNetForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method XLNetLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method XLNetModel.from_pretrained(*args, **kwargs) - removed:
static method XLNetPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
variable ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BART_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable MODEL_FOR_CAUSAL_LM_MAPPING - removed:
variable MODEL_FOR_MASKED_LM_MAPPING - removed:
variable MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
variable MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - removed:
variable MODEL_FOR_PRETRAINING_MAPPING - removed:
variable MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
variable MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
variable MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
variable MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
variable MODEL_MAPPING - removed:
variable MODEL_WITH_LM_HEAD_MAPPING - removed:
variable OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
class AlternatingCodebooksLogitsProcessor - added:
class AttentionInterface - added:
class AttentionMaskInterface - added:
class BayesianDetectorConfig - added:
class BayesianDetectorModel - added:
class Cache - added:
class ClassifierFreeGuidanceLogitsProcessor - added:
class ConstrainedBeamSearchScorer - added:
class Constraint - added:
class ConstraintListState - added:
class DisjunctiveConstraint - added:
class DynamicCache - added:
class EncoderDecoderCache - added:
class EncoderNoRepeatNGramLogitsProcessor - added:
class EncoderRepetitionPenaltyLogitsProcessor - added:
class EosTokenCriteria - added:
class EpsilonLogitsWarper - added:
class EtaLogitsWarper - added:
class ExponentialDecayLengthPenalty - added:
class ForcedBOSTokenLogitsProcessor - added:
class ForcedEOSTokenLogitsProcessor - added:
class GenerationMixin - added:
class GradientCheckpointingLayer - added:
class HQQQuantizedCache - added:
class HybridCache - added:
class InfNanRemoveLogitsProcessor - added:
class LogitNormalization - added:
class MaxLengthCriteria - added:
class MaxTimeCriteria - added:
class MinNewTokensLengthLogitsProcessor - added:
class MinPLogitsWarper - added:
class OffloadedCache - added:
class OffloadedStaticCache - added:
class PhrasalConstraint - added:
class PrefixConstrainedLogitsProcessor - added:
class QuantizedCache - added:
class QuantoQuantizedCache - added:
class Seq2SeqTrainer - added:
class SequenceBiasLogitsProcessor - added:
class SinkCache - added:
class SlidingWindowCache - added:
class StaticCache - added:
class StopStringCriteria - added:
class StoppingCriteria - added:
class StoppingCriteriaList - added:
class SuppressTokensAtBeginLogitsProcessor - added:
class SuppressTokensLogitsProcessor - added:
class SynthIDTextWatermarkDetector - added:
class SynthIDTextWatermarkLogitsProcessor - added:
class SynthIDTextWatermarkingConfig - added:
class TorchExportableModuleWithStaticCache - added:
class TypicalLogitsWarper - added:
class UnbatchedClassifierFreeGuidanceLogitsProcessor - added:
class WatermarkDetector - added:
class WatermarkLogitsProcessor - added:
class WhisperTimeStampLogitsProcessor - added:
constructor AlternatingCodebooksLogitsProcessor(*args, **kwargs) - added:
constructor AttentionInterface(*args, **kwargs) - added:
constructor AttentionMaskInterface(*args, **kwargs) - added:
constructor BayesianDetectorConfig(*args, **kwargs) - added:
constructor BayesianDetectorModel(*args, **kwargs) - added:
constructor Cache(*args, **kwargs) - added:
constructor ClassifierFreeGuidanceLogitsProcessor(*args, **kwargs) - added:
constructor ConstrainedBeamSearchScorer(*args, **kwargs) - added:
constructor Constraint(*args, **kwargs) - added:
constructor ConstraintListState(*args, **kwargs) - added:
constructor DisjunctiveConstraint(*args, **kwargs) - added:
constructor DynamicCache(*args, **kwargs) - added:
constructor EncoderDecoderCache(*args, **kwargs) - added:
constructor EncoderNoRepeatNGramLogitsProcessor(*args, **kwargs) - added:
constructor EncoderRepetitionPenaltyLogitsProcessor(*args, **kwargs) - added:
constructor EosTokenCriteria(*args, **kwargs) - added:
constructor EpsilonLogitsWarper(*args, **kwargs) - added:
constructor EtaLogitsWarper(*args, **kwargs) - added:
constructor ExponentialDecayLengthPenalty(*args, **kwargs) - added:
constructor ForcedBOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor ForcedEOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor GenerationMixin(*args, **kwargs) - added:
constructor GradientCheckpointingLayer(*args, **kwargs) - added:
constructor HQQQuantizedCache(*args, **kwargs) - added:
constructor HybridCache(*args, **kwargs) - added:
constructor InfNanRemoveLogitsProcessor(*args, **kwargs) - added:
constructor LogitNormalization(*args, **kwargs) - added:
constructor MaxLengthCriteria(*args, **kwargs) - added:
constructor MaxTimeCriteria(*args, **kwargs) - added:
constructor MinNewTokensLengthLogitsProcessor(*args, **kwargs) - added:
constructor MinPLogitsWarper(*args, **kwargs) - added:
constructor OffloadedCache(*args, **kwargs) - added:
constructor OffloadedStaticCache(*args, **kwargs) - added:
constructor PhrasalConstraint(*args, **kwargs) - added:
constructor PrefixConstrainedLogitsProcessor(*args, **kwargs) - added:
constructor QuantizedCache(*args, **kwargs) - added:
constructor QuantoQuantizedCache(*args, **kwargs) - added:
constructor Seq2SeqTrainer(*args, **kwargs) - added:
constructor SequenceBiasLogitsProcessor(*args, **kwargs) - added:
constructor SinkCache(*args, **kwargs) - added:
constructor SlidingWindowCache(*args, **kwargs) - added:
constructor StaticCache(*args, **kwargs) - added:
constructor StopStringCriteria(*args, **kwargs) - added:
constructor StoppingCriteria(*args, **kwargs) - added:
constructor StoppingCriteriaList(*args, **kwargs) - added:
constructor SuppressTokensAtBeginLogitsProcessor(*args, **kwargs) - added:
constructor SuppressTokensLogitsProcessor(*args, **kwargs) - added:
constructor SynthIDTextWatermarkDetector(*args, **kwargs) - added:
constructor SynthIDTextWatermarkLogitsProcessor(*args, **kwargs) - added:
constructor SynthIDTextWatermarkingConfig(*args, **kwargs) - added:
constructor TorchExportableModuleWithStaticCache(*args, **kwargs) - added:
constructor TypicalLogitsWarper(*args, **kwargs) - added:
constructor UnbatchedClassifierFreeGuidanceLogitsProcessor(*args, **kwargs) - added:
constructor WatermarkDetector(*args, **kwargs) - added:
constructor WatermarkLogitsProcessor(*args, **kwargs) - added:
constructor WhisperTimeStampLogitsProcessor(*args, **kwargs) - added:
function convert_and_export_with_cache(*args, **kwargs) - added:
function dynamic_rope_update(*args, **kwargs) - added:
function get_inverse_sqrt_schedule(*args, **kwargs) - added:
function get_scheduler(*args, **kwargs) - added:
function get_wsd_schedule(*args, **kwargs) - added:
function model_addition_debugger_context(*args, **kwargs) - added:
variable ROPE_INIT_FUNCTIONS
- added:
function convert_slow_tokenizer(*args, **kwargs) - added:
variable SLOW_TO_FAST_CONVERTERS
- removed:
static method AlbertTokenizer.from_pretrained(*args, **kwargs) - removed:
static method BertGenerationTokenizer.from_pretrained(*args, **kwargs) - removed:
static method CamembertTokenizer.from_pretrained(*args, **kwargs) - removed:
static method MBartTokenizer.from_pretrained(*args, **kwargs) - removed:
static method MarianTokenizer.from_pretrained(*args, **kwargs) - removed:
static method PegasusTokenizer.from_pretrained(*args, **kwargs) - removed:
static method ReformerTokenizer.from_pretrained(*args, **kwargs) - removed:
static method T5Tokenizer.from_pretrained(*args, **kwargs) - removed:
static method XLMProphetNetTokenizer.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaTokenizer.from_pretrained(*args, **kwargs) - removed:
static method XLNetTokenizer.from_pretrained(*args, **kwargs) - added:
class BarthezTokenizer - added:
class BartphoTokenizer - added:
class BigBirdTokenizer - added:
class CodeLlamaTokenizer - added:
class CpmTokenizer - added:
class DebertaV2Tokenizer - added:
class ErnieMTokenizer - added:
class FNetTokenizer - added:
class GPTSw3Tokenizer - added:
class GemmaTokenizer - added:
class LayoutXLMTokenizer - added:
class LlamaTokenizer - added:
class M2M100Tokenizer - added:
class MBart50Tokenizer - added:
class MLukeTokenizer - added:
class MT5Tokenizer - added:
class NllbTokenizer - added:
class PLBartTokenizer - added:
class RemBertTokenizer - added:
class SeamlessM4TTokenizer - added:
class SiglipTokenizer - added:
class Speech2TextTokenizer - added:
class SpeechT5Tokenizer - added:
class UdopTokenizer - added:
class XGLMTokenizer - added:
constructor BarthezTokenizer(*args, **kwargs) - added:
constructor BartphoTokenizer(*args, **kwargs) - added:
constructor BigBirdTokenizer(*args, **kwargs) - added:
constructor CodeLlamaTokenizer(*args, **kwargs) - added:
constructor CpmTokenizer(*args, **kwargs) - added:
constructor DebertaV2Tokenizer(*args, **kwargs) - added:
constructor ErnieMTokenizer(*args, **kwargs) - added:
constructor FNetTokenizer(*args, **kwargs) - added:
constructor GPTSw3Tokenizer(*args, **kwargs) - added:
constructor GemmaTokenizer(*args, **kwargs) - added:
constructor LayoutXLMTokenizer(*args, **kwargs) - added:
constructor LlamaTokenizer(*args, **kwargs) - added:
constructor M2M100Tokenizer(*args, **kwargs) - added:
constructor MBart50Tokenizer(*args, **kwargs) - added:
constructor MLukeTokenizer(*args, **kwargs) - added:
constructor MT5Tokenizer(*args, **kwargs) - added:
constructor NllbTokenizer(*args, **kwargs) - added:
constructor PLBartTokenizer(*args, **kwargs) - added:
constructor RemBertTokenizer(*args, **kwargs) - added:
constructor SeamlessM4TTokenizer(*args, **kwargs) - added:
constructor SiglipTokenizer(*args, **kwargs) - added:
constructor Speech2TextTokenizer(*args, **kwargs) - added:
constructor SpeechT5Tokenizer(*args, **kwargs) - added:
constructor UdopTokenizer(*args, **kwargs) - added:
constructor XGLMTokenizer(*args, **kwargs)
- added:
class ASTFeatureExtractor - added:
class Speech2TextFeatureExtractor - added:
constructor ASTFeatureExtractor(*args, **kwargs) - added:
constructor Speech2TextFeatureExtractor(*args, **kwargs)
- added:
class TFBertTokenizer - added:
constructor TFBertTokenizer(*args, **kwargs)
- removed:
class TFAdaptiveEmbedding - removed:
class TFAlbertForMaskedLM - removed:
class TFAlbertForMultipleChoice - removed:
class TFAlbertForPreTraining - removed:
class TFAlbertForQuestionAnswering - removed:
class TFAlbertForSequenceClassification - removed:
class TFAlbertForTokenClassification - removed:
class TFAlbertMainLayer - removed:
class TFAlbertModel - removed:
class TFAlbertPreTrainedModel - removed:
class TFAutoModel - removed:
class TFAutoModelForCausalLM - removed:
class TFAutoModelForMaskedLM - removed:
class TFAutoModelForMultipleChoice - removed:
class TFAutoModelForPreTraining - removed:
class TFAutoModelForQuestionAnswering - removed:
class TFAutoModelForSeq2SeqLM - removed:
class TFAutoModelForSequenceClassification - removed:
class TFAutoModelForTokenClassification - removed:
class TFAutoModelWithLMHead - removed:
class TFBartForConditionalGeneration - removed:
class TFBartModel - removed:
class TFBertEmbeddings - removed:
class TFBertForMaskedLM - removed:
class TFBertForMultipleChoice - removed:
class TFBertForNextSentencePrediction - removed:
class TFBertForPreTraining - removed:
class TFBertForQuestionAnswering - removed:
class TFBertForSequenceClassification - removed:
class TFBertForTokenClassification - removed:
class TFBertLMHeadModel - removed:
class TFBertMainLayer - removed:
class TFBertModel - removed:
class TFBertPreTrainedModel - removed:
class TFBlenderbotForConditionalGeneration - removed:
class TFCTRLLMHeadModel - removed:
class TFCTRLModel - removed:
class TFCTRLPreTrainedModel - removed:
class TFCamembertForMaskedLM - removed:
class TFCamembertForMultipleChoice - removed:
class TFCamembertForQuestionAnswering - removed:
class TFCamembertForSequenceClassification - removed:
class TFCamembertForTokenClassification - removed:
class TFCamembertModel - removed:
class TFDistilBertForMaskedLM - removed:
class TFDistilBertForMultipleChoice - removed:
class TFDistilBertForQuestionAnswering - removed:
class TFDistilBertForSequenceClassification - removed:
class TFDistilBertForTokenClassification - removed:
class TFDistilBertMainLayer - removed:
class TFDistilBertModel - removed:
class TFDistilBertPreTrainedModel - removed:
class TFElectraForMaskedLM - removed:
class TFElectraForMultipleChoice - removed:
class TFElectraForPreTraining - removed:
class TFElectraForQuestionAnswering - removed:
class TFElectraForSequenceClassification - removed:
class TFElectraForTokenClassification - removed:
class TFElectraModel - removed:
class TFElectraPreTrainedModel - removed:
class TFFlaubertForMultipleChoice - removed:
class TFFlaubertForQuestionAnsweringSimple - removed:
class TFFlaubertForSequenceClassification - removed:
class TFFlaubertForTokenClassification - removed:
class TFFlaubertModel - removed:
class TFFlaubertWithLMHeadModel - removed:
class TFFunnelBaseModel - removed:
class TFFunnelForMaskedLM - removed:
class TFFunnelForMultipleChoice - removed:
class TFFunnelForPreTraining - removed:
class TFFunnelForQuestionAnswering - removed:
class TFFunnelForSequenceClassification - removed:
class TFFunnelForTokenClassification - removed:
class TFFunnelModel - removed:
class TFGPT2DoubleHeadsModel - removed:
class TFGPT2LMHeadModel - removed:
class TFGPT2MainLayer - removed:
class TFGPT2Model - removed:
class TFGPT2PreTrainedModel - removed:
class TFLongformerForMaskedLM - removed:
class TFLongformerForQuestionAnswering - removed:
class TFLongformerModel - removed:
class TFLongformerSelfAttention - removed:
class TFLxmertForPreTraining - removed:
class TFLxmertMainLayer - removed:
class TFLxmertModel - removed:
class TFLxmertPreTrainedModel - removed:
class TFLxmertVisualFeatureEncoder - removed:
class TFMBartForConditionalGeneration - removed:
class TFMarianMTModel - removed:
class TFMobileBertForMaskedLM - removed:
class TFMobileBertForMultipleChoice - removed:
class TFMobileBertForNextSentencePrediction - removed:
class TFMobileBertForPreTraining - removed:
class TFMobileBertForQuestionAnswering - removed:
class TFMobileBertForSequenceClassification - removed:
class TFMobileBertForTokenClassification - removed:
class TFMobileBertMainLayer - removed:
class TFMobileBertModel - removed:
class TFMobileBertPreTrainedModel - removed:
class TFOpenAIGPTDoubleHeadsModel - removed:
class TFOpenAIGPTLMHeadModel - removed:
class TFOpenAIGPTMainLayer - removed:
class TFOpenAIGPTModel - removed:
class TFOpenAIGPTPreTrainedModel - removed:
class TFPegasusForConditionalGeneration - removed:
class TFRobertaForMaskedLM - removed:
class TFRobertaForMultipleChoice - removed:
class TFRobertaForQuestionAnswering - removed:
class TFRobertaForSequenceClassification - removed:
class TFRobertaForTokenClassification - removed:
class TFRobertaMainLayer - removed:
class TFRobertaModel - removed:
class TFRobertaPreTrainedModel - removed:
class TFT5ForConditionalGeneration - removed:
class TFT5Model - removed:
class TFT5PreTrainedModel - removed:
class TFTrainer - removed:
class TFTransfoXLLMHeadModel - removed:
class TFTransfoXLMainLayer - removed:
class TFTransfoXLModel - removed:
class TFTransfoXLPreTrainedModel - removed:
class TFXLMForMultipleChoice - removed:
class TFXLMForQuestionAnsweringSimple - removed:
class TFXLMForSequenceClassification - removed:
class TFXLMForTokenClassification - removed:
class TFXLMMainLayer - removed:
class TFXLMModel - removed:
class TFXLMPreTrainedModel - removed:
class TFXLMRobertaForMaskedLM - removed:
class TFXLMRobertaForMultipleChoice - removed:
class TFXLMRobertaForQuestionAnswering - removed:
class TFXLMRobertaForSequenceClassification - removed:
class TFXLMRobertaForTokenClassification - removed:
class TFXLMRobertaModel - removed:
class TFXLMWithLMHeadModel - removed:
class TFXLNetForMultipleChoice - removed:
class TFXLNetForQuestionAnsweringSimple - removed:
class TFXLNetForSequenceClassification - removed:
class TFXLNetForTokenClassification - removed:
class TFXLNetLMHeadModel - removed:
class TFXLNetMainLayer - removed:
class TFXLNetModel - removed:
class TFXLNetPreTrainedModel - removed:
class TensorFlowBenchmark - removed:
class TensorFlowBenchmarkArguments - removed:
constructor TFAdaptiveEmbedding(*args, **kwargs) - removed:
constructor TFAlbertForMaskedLM(*args, **kwargs) - removed:
constructor TFAlbertForMultipleChoice(*args, **kwargs) - removed:
constructor TFAlbertForPreTraining(*args, **kwargs) - removed:
constructor TFAlbertForQuestionAnswering(*args, **kwargs) - removed:
constructor TFAlbertForSequenceClassification(*args, **kwargs) - removed:
constructor TFAlbertForTokenClassification(*args, **kwargs) - removed:
constructor TFAlbertMainLayer(*args, **kwargs) - removed:
constructor TFAlbertModel(*args, **kwargs) - removed:
constructor TFAlbertPreTrainedModel(*args, **kwargs) - removed:
constructor TFAutoModel(*args, **kwargs) - removed:
constructor TFAutoModelForCausalLM(*args, **kwargs) - removed:
constructor TFAutoModelForMaskedLM(*args, **kwargs) - removed:
constructor TFAutoModelForMultipleChoice(*args, **kwargs) - removed:
constructor TFAutoModelForPreTraining(*args, **kwargs) - removed:
constructor TFAutoModelForQuestionAnswering(*args, **kwargs) - removed:
constructor TFAutoModelForSeq2SeqLM(*args, **kwargs) - removed:
constructor TFAutoModelForSequenceClassification(*args, **kwargs) - removed:
constructor TFAutoModelForTokenClassification(*args, **kwargs) - removed:
constructor TFAutoModelWithLMHead(*args, **kwargs) - removed:
constructor TFBartForConditionalGeneration(*args, **kwargs) - removed:
constructor TFBartModel(*args, **kwargs) - removed:
constructor TFBertEmbeddings(*args, **kwargs) - removed:
constructor TFBertForMaskedLM(*args, **kwargs) - removed:
constructor TFBertForMultipleChoice(*args, **kwargs) - removed:
constructor TFBertForNextSentencePrediction(*args, **kwargs) - removed:
constructor TFBertForPreTraining(*args, **kwargs) - removed:
constructor TFBertForQuestionAnswering(*args, **kwargs) - removed:
constructor TFBertForSequenceClassification(*args, **kwargs) - removed:
constructor TFBertForTokenClassification(*args, **kwargs) - removed:
constructor TFBertLMHeadModel(*args, **kwargs) - removed:
constructor TFBertMainLayer(*args, **kwargs) - removed:
constructor TFBertModel(*args, **kwargs) - removed:
constructor TFBertPreTrainedModel(*args, **kwargs) - removed:
constructor TFBlenderbotForConditionalGeneration(*args, **kwargs) - removed:
constructor TFCTRLLMHeadModel(*args, **kwargs) - removed:
constructor TFCTRLModel(*args, **kwargs) - removed:
constructor TFCTRLPreTrainedModel(*args, **kwargs) - removed:
constructor TFCamembertForMaskedLM(*args, **kwargs) - removed:
constructor TFCamembertForMultipleChoice(*args, **kwargs) - removed:
constructor TFCamembertForQuestionAnswering(*args, **kwargs) - removed:
constructor TFCamembertForSequenceClassification(*args, **kwargs) - removed:
constructor TFCamembertForTokenClassification(*args, **kwargs) - removed:
constructor TFCamembertModel(*args, **kwargs) - removed:
constructor TFDistilBertForMaskedLM(*args, **kwargs) - removed:
constructor TFDistilBertForMultipleChoice(*args, **kwargs) - removed:
constructor TFDistilBertForQuestionAnswering(*args, **kwargs) - removed:
constructor TFDistilBertForSequenceClassification(*args, **kwargs) - removed:
constructor TFDistilBertForTokenClassification(*args, **kwargs) - removed:
constructor TFDistilBertMainLayer(*args, **kwargs) - removed:
constructor TFDistilBertModel(*args, **kwargs) - removed:
constructor TFDistilBertPreTrainedModel(*args, **kwargs) - removed:
constructor TFElectraForMaskedLM(*args, **kwargs) - removed:
constructor TFElectraForMultipleChoice(*args, **kwargs) - removed:
constructor TFElectraForPreTraining(*args, **kwargs) - removed:
constructor TFElectraForQuestionAnswering(*args, **kwargs) - removed:
constructor TFElectraForSequenceClassification(*args, **kwargs) - removed:
constructor TFElectraForTokenClassification(*args, **kwargs) - removed:
constructor TFElectraModel(*args, **kwargs) - removed:
constructor TFElectraPreTrainedModel(*args, **kwargs) - removed:
constructor TFFlaubertForMultipleChoice(*args, **kwargs) - removed:
constructor TFFlaubertForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor TFFlaubertForSequenceClassification(*args, **kwargs) - removed:
constructor TFFlaubertForTokenClassification(*args, **kwargs) - removed:
constructor TFFlaubertModel(*args, **kwargs) - removed:
constructor TFFlaubertWithLMHeadModel(*args, **kwargs) - removed:
constructor TFFunnelBaseModel(*args, **kwargs) - removed:
constructor TFFunnelForMaskedLM(*args, **kwargs) - removed:
constructor TFFunnelForMultipleChoice(*args, **kwargs) - removed:
constructor TFFunnelForPreTraining(*args, **kwargs) - removed:
constructor TFFunnelForQuestionAnswering(*args, **kwargs) - removed:
constructor TFFunnelForSequenceClassification(*args, **kwargs) - removed:
constructor TFFunnelForTokenClassification(*args, **kwargs) - removed:
constructor TFFunnelModel(*args, **kwargs) - removed:
constructor TFGPT2DoubleHeadsModel(*args, **kwargs) - removed:
constructor TFGPT2LMHeadModel(*args, **kwargs) - removed:
constructor TFGPT2MainLayer(*args, **kwargs) - removed:
constructor TFGPT2Model(*args, **kwargs) - removed:
constructor TFGPT2PreTrainedModel(*args, **kwargs) - removed:
constructor TFLongformerForMaskedLM(*args, **kwargs) - removed:
constructor TFLongformerForQuestionAnswering(*args, **kwargs) - removed:
constructor TFLongformerModel(*args, **kwargs) - removed:
constructor TFLongformerSelfAttention(*args, **kwargs) - removed:
constructor TFLxmertForPreTraining(*args, **kwargs) - removed:
constructor TFLxmertMainLayer(*args, **kwargs) - removed:
constructor TFLxmertModel(*args, **kwargs) - removed:
constructor TFLxmertPreTrainedModel(*args, **kwargs) - removed:
constructor TFLxmertVisualFeatureEncoder(*args, **kwargs) - removed:
constructor TFMBartForConditionalGeneration(*args, **kwargs) - removed:
constructor TFMarianMTModel(*args, **kwargs) - removed:
constructor TFMobileBertForMaskedLM(*args, **kwargs) - removed:
constructor TFMobileBertForMultipleChoice(*args, **kwargs) - removed:
constructor TFMobileBertForNextSentencePrediction(*args, **kwargs) - removed:
constructor TFMobileBertForPreTraining(*args, **kwargs) - removed:
constructor TFMobileBertForQuestionAnswering(*args, **kwargs) - removed:
constructor TFMobileBertForSequenceClassification(*args, **kwargs) - removed:
constructor TFMobileBertForTokenClassification(*args, **kwargs) - removed:
constructor TFMobileBertMainLayer(*args, **kwargs) - removed:
constructor TFMobileBertModel(*args, **kwargs) - removed:
constructor TFMobileBertPreTrainedModel(*args, **kwargs) - removed:
constructor TFOpenAIGPTDoubleHeadsModel(*args, **kwargs) - removed:
constructor TFOpenAIGPTLMHeadModel(*args, **kwargs) - removed:
constructor TFOpenAIGPTMainLayer(*args, **kwargs) - removed:
constructor TFOpenAIGPTModel(*args, **kwargs) - removed:
constructor TFOpenAIGPTPreTrainedModel(*args, **kwargs) - removed:
constructor TFPegasusForConditionalGeneration(*args, **kwargs) - removed:
constructor TFRobertaForMaskedLM(*args, **kwargs) - removed:
constructor TFRobertaForMultipleChoice(*args, **kwargs) - removed:
constructor TFRobertaForQuestionAnswering(*args, **kwargs) - removed:
constructor TFRobertaForSequenceClassification(*args, **kwargs) - removed:
constructor TFRobertaForTokenClassification(*args, **kwargs) - removed:
constructor TFRobertaMainLayer(*args, **kwargs) - removed:
constructor TFRobertaModel(*args, **kwargs) - removed:
constructor TFRobertaPreTrainedModel(*args, **kwargs) - removed:
constructor TFT5ForConditionalGeneration(*args, **kwargs) - removed:
constructor TFT5Model(*args, **kwargs) - removed:
constructor TFT5PreTrainedModel(*args, **kwargs) - removed:
constructor TFTrainer(*args, **kwargs) - removed:
constructor TFTransfoXLLMHeadModel(*args, **kwargs) - removed:
constructor TFTransfoXLMainLayer(*args, **kwargs) - removed:
constructor TFTransfoXLModel(*args, **kwargs) - removed:
constructor TFTransfoXLPreTrainedModel(*args, **kwargs) - removed:
constructor TFXLMForMultipleChoice(*args, **kwargs) - removed:
constructor TFXLMForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor TFXLMForSequenceClassification(*args, **kwargs) - removed:
constructor TFXLMForTokenClassification(*args, **kwargs) - removed:
constructor TFXLMMainLayer(*args, **kwargs) - removed:
constructor TFXLMModel(*args, **kwargs) - removed:
constructor TFXLMPreTrainedModel(*args, **kwargs) - removed:
constructor TFXLMRobertaForMaskedLM(*args, **kwargs) - removed:
constructor TFXLMRobertaForMultipleChoice(*args, **kwargs) - removed:
constructor TFXLMRobertaForQuestionAnswering(*args, **kwargs) - removed:
constructor TFXLMRobertaForSequenceClassification(*args, **kwargs) - removed:
constructor TFXLMRobertaForTokenClassification(*args, **kwargs) - removed:
constructor TFXLMRobertaModel(*args, **kwargs) - removed:
constructor TFXLMWithLMHeadModel(*args, **kwargs) - removed:
constructor TFXLNetForMultipleChoice(*args, **kwargs) - removed:
constructor TFXLNetForQuestionAnsweringSimple(*args, **kwargs) - removed:
constructor TFXLNetForSequenceClassification(*args, **kwargs) - removed:
constructor TFXLNetForTokenClassification(*args, **kwargs) - removed:
constructor TFXLNetLMHeadModel(*args, **kwargs) - removed:
constructor TFXLNetMainLayer(*args, **kwargs) - removed:
constructor TFXLNetModel(*args, **kwargs) - removed:
constructor TFXLNetPreTrainedModel(*args, **kwargs) - removed:
constructor TensorFlowBenchmark(*args, **kwargs) - removed:
constructor TensorFlowBenchmarkArguments(*args, **kwargs) - removed:
function tf_top_k_top_p_filtering(*args, **kwargs) - removed:
static method TFAlbertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertModel.from_pretrained(*args, **kwargs) - removed:
static method TFAlbertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModel.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForCausalLM.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForPreTraining.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForSeq2SeqLM.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFAutoModelWithLMHead.from_pretrained(*args, **kwargs) - removed:
static method TFBartForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method TFBartModel.from_pretrained(*args, **kwargs) - removed:
static method TFBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFBertLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFBertModel.from_pretrained(*args, **kwargs) - removed:
static method TFBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFBlenderbotForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method TFCTRLLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFCTRLModel.from_pretrained(*args, **kwargs) - removed:
static method TFCTRLPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFCamembertModel.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertModel.from_pretrained(*args, **kwargs) - removed:
static method TFDistilBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFElectraForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFElectraForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFElectraForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFElectraForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFElectraForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFElectraModel.from_pretrained(*args, **kwargs) - removed:
static method TFElectraPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertModel.from_pretrained(*args, **kwargs) - removed:
static method TFFlaubertWithLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelBaseModel.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFFunnelModel.from_pretrained(*args, **kwargs) - removed:
static method TFGPT2DoubleHeadsModel.from_pretrained(*args, **kwargs) - removed:
static method TFGPT2LMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFGPT2Model.from_pretrained(*args, **kwargs) - removed:
static method TFGPT2PreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFLongformerForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFLongformerForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFLongformerModel.from_pretrained(*args, **kwargs) - removed:
static method TFLxmertModel.from_pretrained(*args, **kwargs) - removed:
static method TFLxmertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFMBartForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method TFMarianMTModel.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertModel.from_pretrained(*args, **kwargs) - removed:
static method TFMobileBertPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFOpenAIGPTDoubleHeadsModel.from_pretrained(*args, **kwargs) - removed:
static method TFOpenAIGPTLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFOpenAIGPTModel.from_pretrained(*args, **kwargs) - removed:
static method TFOpenAIGPTPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFPegasusForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method TFPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaModel.from_pretrained(*args, **kwargs) - removed:
static method TFRobertaPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFT5ForConditionalGeneration.from_pretrained(*args, **kwargs) - removed:
static method TFT5Model.from_pretrained(*args, **kwargs) - removed:
static method TFT5PreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFTransfoXLLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFTransfoXLModel.from_pretrained(*args, **kwargs) - removed:
static method TFTransfoXLPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLMForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFXLMForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method TFXLMForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLMForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLMModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLMPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaForMaskedLM.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLMRobertaModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLMWithLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetForMultipleChoice.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetForSequenceClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetForTokenClassification.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetLMHeadModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetModel.from_pretrained(*args, **kwargs) - removed:
static method TFXLNetPreTrainedModel.from_pretrained(*args, **kwargs) - removed:
variable TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_MODEL_FOR_CAUSAL_LM_MAPPING - removed:
variable TF_MODEL_FOR_MASKED_LM_MAPPING - removed:
variable TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - removed:
variable TF_MODEL_FOR_PRETRAINING_MAPPING - removed:
variable TF_MODEL_FOR_QUESTION_ANSWERING_MAPPING - removed:
variable TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - removed:
variable TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - removed:
variable TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - removed:
variable TF_MODEL_MAPPING - removed:
variable TF_MODEL_WITH_LM_HEAD_MAPPING - removed:
variable TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - removed:
variable TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
class KerasMetricCallback - added:
class PushToHubCallback - added:
class TFForceTokensLogitsProcessor - added:
class TFForcedBOSTokenLogitsProcessor - added:
class TFForcedEOSTokenLogitsProcessor - added:
class TFGenerationMixin - added:
class TFLogitsProcessor - added:
class TFLogitsProcessorList - added:
class TFLogitsWarper - added:
class TFMinLengthLogitsProcessor - added:
class TFNoBadWordsLogitsProcessor - added:
class TFNoRepeatNGramLogitsProcessor - added:
class TFRepetitionPenaltyLogitsProcessor - added:
class TFSuppressTokensAtBeginLogitsProcessor - added:
class TFSuppressTokensLogitsProcessor - added:
class TFTemperatureLogitsWarper - added:
class TFTopKLogitsWarper - added:
class TFTopPLogitsWarper - added:
constructor KerasMetricCallback(*args, **kwargs) - added:
constructor PushToHubCallback(*args, **kwargs) - added:
constructor TFForceTokensLogitsProcessor(*args, **kwargs) - added:
constructor TFForcedBOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor TFForcedEOSTokenLogitsProcessor(*args, **kwargs) - added:
constructor TFGenerationMixin(*args, **kwargs) - added:
constructor TFLogitsProcessor(*args, **kwargs) - added:
constructor TFLogitsProcessorList(*args, **kwargs) - added:
constructor TFLogitsWarper(*args, **kwargs) - added:
constructor TFMinLengthLogitsProcessor(*args, **kwargs) - added:
constructor TFNoBadWordsLogitsProcessor(*args, **kwargs) - added:
constructor TFNoRepeatNGramLogitsProcessor(*args, **kwargs) - added:
constructor TFRepetitionPenaltyLogitsProcessor(*args, **kwargs) - added:
constructor TFSuppressTokensAtBeginLogitsProcessor(*args, **kwargs) - added:
constructor TFSuppressTokensLogitsProcessor(*args, **kwargs) - added:
constructor TFTemperatureLogitsWarper(*args, **kwargs) - added:
constructor TFTopKLogitsWarper(*args, **kwargs) - added:
constructor TFTopPLogitsWarper(*args, **kwargs)
- added:
class TimmWrapperImageProcessor - added:
constructor TimmWrapperImageProcessor(*args, **kwargs)
- removed:
class AlbertTokenizerFast - removed:
class BartTokenizerFast - removed:
class BertTokenizerFast - removed:
class CamembertTokenizerFast - removed:
class DPRContextEncoderTokenizerFast - removed:
class DPRQuestionEncoderTokenizerFast - removed:
class DPRReaderTokenizerFast - removed:
class DistilBertTokenizerFast - removed:
class ElectraTokenizerFast - removed:
class FunnelTokenizerFast - removed:
class GPT2TokenizerFast - removed:
class HerbertTokenizerFast - removed:
class LayoutLMTokenizerFast - removed:
class LongformerTokenizerFast - removed:
class LxmertTokenizerFast - removed:
class MBartTokenizerFast - removed:
class MobileBertTokenizerFast - removed:
class OpenAIGPTTokenizerFast - removed:
class PegasusTokenizerFast - removed:
class ReformerTokenizerFast - removed:
class RetriBertTokenizerFast - removed:
class RobertaTokenizerFast - removed:
class SqueezeBertTokenizerFast - removed:
class T5TokenizerFast - removed:
class XLMRobertaTokenizerFast - removed:
class XLNetTokenizerFast - removed:
constructor AlbertTokenizerFast(*args, **kwargs) - removed:
constructor BartTokenizerFast(*args, **kwargs) - removed:
constructor BertTokenizerFast(*args, **kwargs) - removed:
constructor CamembertTokenizerFast(*args, **kwargs) - removed:
constructor DPRContextEncoderTokenizerFast(*args, **kwargs) - removed:
constructor DPRQuestionEncoderTokenizerFast(*args, **kwargs) - removed:
constructor DPRReaderTokenizerFast(*args, **kwargs) - removed:
constructor DistilBertTokenizerFast(*args, **kwargs) - removed:
constructor ElectraTokenizerFast(*args, **kwargs) - removed:
constructor FunnelTokenizerFast(*args, **kwargs) - removed:
constructor GPT2TokenizerFast(*args, **kwargs) - removed:
constructor HerbertTokenizerFast(*args, **kwargs) - removed:
constructor LayoutLMTokenizerFast(*args, **kwargs) - removed:
constructor LongformerTokenizerFast(*args, **kwargs) - removed:
constructor LxmertTokenizerFast(*args, **kwargs) - removed:
constructor MBartTokenizerFast(*args, **kwargs) - removed:
constructor MobileBertTokenizerFast(*args, **kwargs) - removed:
constructor OpenAIGPTTokenizerFast(*args, **kwargs) - removed:
constructor PegasusTokenizerFast(*args, **kwargs) - removed:
constructor ReformerTokenizerFast(*args, **kwargs) - removed:
constructor RetriBertTokenizerFast(*args, **kwargs) - removed:
constructor RobertaTokenizerFast(*args, **kwargs) - removed:
constructor SqueezeBertTokenizerFast(*args, **kwargs) - removed:
constructor T5TokenizerFast(*args, **kwargs) - removed:
constructor XLMRobertaTokenizerFast(*args, **kwargs) - removed:
constructor XLNetTokenizerFast(*args, **kwargs) - removed:
function convert_slow_tokenizer(*args, **kwargs) - removed:
static method AlbertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method BartTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method BertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method CamembertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method DPRContextEncoderTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method DPRQuestionEncoderTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method DPRReaderTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method DistilBertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method ElectraTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method FunnelTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method GPT2TokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method HerbertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method LayoutLMTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method LongformerTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method LxmertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method MBartTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method MobileBertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method OpenAIGPTTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method PegasusTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method PreTrainedTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method ReformerTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method RetriBertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method RobertaTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method SqueezeBertTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method T5TokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method XLMRobertaTokenizerFast.from_pretrained(*args, **kwargs) - removed:
static method XLNetTokenizerFast.from_pretrained(*args, **kwargs) - removed:
variable SLOW_TO_FAST_CONVERTERS
- added:
class GraniteSpeechFeatureExtractor - added:
class GraniteSpeechProcessor - added:
class MusicgenMelodyFeatureExtractor - added:
class MusicgenMelodyProcessor - added:
constructor GraniteSpeechFeatureExtractor(*args, **kwargs) - added:
constructor GraniteSpeechProcessor(*args, **kwargs) - added:
constructor MusicgenMelodyFeatureExtractor(*args, **kwargs) - added:
constructor MusicgenMelodyProcessor(*args, **kwargs)
- added:
class BaseImageProcessorFast - added:
class BaseVideoProcessor - added:
constructor BaseImageProcessorFast(*args, **kwargs) - added:
constructor BaseVideoProcessor(*args, **kwargs)
- added:
class BaseImageProcessor - added:
class ImageFeatureExtractionMixin - added:
class ImageProcessingMixin - added:
constructor BaseImageProcessor(*args, **kwargs) - added:
constructor ImageFeatureExtractionMixin(*args, **kwargs) - added:
constructor ImageProcessingMixin(*args, **kwargs)
- added:
class HFAttribute(HFProxy) - added:
class HFCacheProxy(HFProxy) - added:
class HFProxy(Proxy) - added:
class HFProxyableClassMeta(type) - added:
class HFTracer(Tracer) - added:
class MetaDeviceAttribute(HFAttribute) - added:
constructor HFAttribute(root, attr: str) - added:
constructor HFTracer(autowrap_modules=…, autowrap_functions=…) - added:
field HFTracer.allow_insert_stateless_mods: bool - added:
field HFTracer.proxy_buffer_attributes: bool - added:
field HFTracer.supported_archs - added:
function check_if_model_is_supported(model: PreTrainedModel) - added:
function create_cache_proxy_factory_fn(orig_cache_cls: type[Cache]) -> Callable[[Node], HFCacheProxy] - added:
function create_wrapper(function: Callable, op_type: Union[Literal['call_function'], Literal['call_method'], Literal['get_attr']], proxy_factory_fn: Optional[Callable[[Node], Proxy]]=…) -> Callable - added:
function gen_constructor_wrapper(target: Callable) -> tuple[Callable, Callable] - added:
function get_concrete_args(model: nn.Module, input_names: list[str]) - added:
function is_model_supported(model: PreTrainedModel) - added:
function operator_getitem(a, b) - added:
function symbolic_trace(model: PreTrainedModel, input_names: Optional[list[str]]=…, disable_check: bool=…, tracer_cls: type[HFTracer]=…) -> GraphModule - added:
function torch_abs(input, *, out=…) - added:
function torch_add(input, other, *, alpha=…, out=…) - added:
function torch_arange(*args, **kwargs) - added:
function torch_baddbmm(input, batch1, batch2, *, beta=…, alpha=…, out=…) - added:
function torch_bmm(input, mat2, *, out=…) - added:
function torch_cat(tensors, dim=…, axis=…, *, out=…) - added:
function torch_einsum(equation, *operands) - added:
function torch_flip(input, dims) - added:
function torch_full(*args, **kwargs) - added:
function torch_gather(input, dim, index, *, sparse_grad=…, out=…) - added:
function torch_index_select(input, dim, index, *, out=…) - added:
function torch_matmul(input, other, *, out=…) - added:
function torch_mul(input, other, *, out=…) - added:
function torch_nn_bcewithlogitsloss(self, input, target) - added:
function torch_nn_conv1d(self, input) - added:
function torch_nn_conv2d(self, input) - added:
function torch_nn_crossentropyloss(self, input, target) - added:
function torch_nn_embedding(self, input) - added:
function torch_nn_functional_embedding(input, weight, padding_idx=…, max_norm=…, norm_type=…, scale_grad_by_freq=…, sparse=…) - added:
function torch_nn_functional_one_hot(tensor, num_classes=…) - added:
function torch_nn_functional_relu(x, inplace=…) - added:
function torch_nn_functional_scaled_dot_product_attention(query, key, value, attn_mask=…, dropout_p=…, is_causal=…, scale=…) - added:
function torch_nn_groupnorm(self, input) - added:
function torch_nn_layernorm(self, input) - added:
function torch_nn_linear(self, input) - added:
function torch_nn_mseloss(self, input, target) - added:
function torch_nn_relu(self, x) - added:
function torch_relu(x) - added:
function torch_repeat_interleave(*args, dim=…, output_size=…) - added:
function torch_roll(input, shifts, dims=…) - added:
function torch_squeeze(input, dim=…) - added:
function torch_stack(tensors, dim=…, axis=…, *, out=…) - added:
function torch_tensor_baddbmm(self, batch1, batch2, *, beta=…, alpha=…, out=…) - added:
function torch_tensor_flip(self, dims) - added:
function torch_tensor_gather(self, dim, index) - added:
function torch_tensor_index_select(self, dim, index) - added:
function torch_tensor_mul(self, other) - added:
function torch_tensor_repeat(self, *sizes) - added:
function torch_tensor_squeeze(self, dim=…) - added:
function torch_tensor_unsqueeze(self, dim) - added:
function torch_unique_consecutive(input, **kwargs) - added:
function torch_unsqueeze(input, dim) - added:
function torch_where(condition, x, y) - added:
method HFCacheProxy.install_orig_cache_cls(orig_cache_cls: type[Cache]) - added:
method HFProxy.install_metadata(metadata) - added:
method HFTracer.call_module(m, forward, args, kwargs) - added:
method HFTracer.create_proxy(kind, target, args, kwargs, name=…, type_expr=…, proxy_factory_fn=…) - added:
method HFTracer.getattr(attr: str, attr_val: Any, parameter_proxy_cache: dict[str, Any]) - added:
method HFTracer.is_leaf_module(m: torch.nn.Module, module_qualified_name: str) -> bool - added:
method HFTracer.keys(obj: Proxy) -> Any - added:
method HFTracer.patch_for_tracing(root: Union[torch.nn.Module, Callable[..., Any]]) - added:
method HFTracer.path_of_module(mod: nn.Module) -> str - added:
method HFTracer.proxy(node) - added:
method HFTracer.trace(root: Union[torch.nn.Module, Callable[..., Any]], concrete_args: Optional[dict[str, Any]]=…, dummy_inputs: Optional[dict[str, Any]]=…, complete_concrete_args_with_inputs_not_in_dummy_inputs: bool=…) -> Graph - added:
property HFAttribute.node - added:
property HFProxy.device - added:
property HFProxy.shape - added:
variable ProxyableCache - added:
variable ProxyableDynamicCache - added:
variable ProxyableStaticCache - added:
variable logger
- added:
class ContextManagers - added:
class ExplicitEnum(str, Enum) - added:
class GeneralInterface(MutableMapping) - added:
class ModelOutput(OrderedDict) - added:
class OutputRecorder - added:
class PaddingStrategy(ExplicitEnum) - added:
class TensorType(ExplicitEnum) - added:
class TransformersKwargs(TypedDict) - added:
constructor ContextManagers(context_managers: list[AbstractContextManager]) - added:
constructor GeneralInterface() - added:
constructor ModelOutput(*args, **kwargs) - added:
field OutputRecorder.class_name: Optional[str] - added:
field OutputRecorder.index: int - added:
field OutputRecorder.layer_name: Optional[str] - added:
field OutputRecorder.target_class: type[torch.nn.Module] - added:
field PaddingStrategy.DO_NOT_PAD - added:
field PaddingStrategy.LONGEST - added:
field PaddingStrategy.MAX_LENGTH - added:
field TensorType.JAX - added:
field TensorType.MLX - added:
field TensorType.NUMPY - added:
field TensorType.PYTORCH - added:
field TensorType.TENSORFLOW - added:
field TransformersKwargs.cu_seq_lens_k: Optional[torch.LongTensor] - added:
field TransformersKwargs.cu_seq_lens_q: Optional[torch.LongTensor] - added:
field TransformersKwargs.max_length_k: Optional[int] - added:
field TransformersKwargs.max_length_q: Optional[int] - added:
field TransformersKwargs.num_items_in_batch: Optional[torch.Tensor] - added:
field TransformersKwargs.output_attentions: Optional[bool] - added:
field TransformersKwargs.output_hidden_states: Optional[bool] - added:
field TransformersKwargs.output_router_logits: Optional[bool] - added:
function can_return_loss(model_class) - added:
function can_return_tuple(func) - added:
function check_model_inputs(func=…, *, tie_last_hidden_states=…) - added:
function del_attribute_from_modules(module: torch.nn.Module, key: str) - added:
function expand_dims(array, axis) - added:
function filter_out_non_signature_kwargs(extra: Optional[list]=…) - added:
function find_labels(model_class) - added:
function flatten_dict(d: MutableMapping, parent_key: str=…, delimiter: str=…) - added:
function infer_framework(model_class) - added:
function infer_framework_from_repr(x) - added:
function is_jax_tensor(x) - added:
function is_mlx_array(x) - added:
function is_numpy_array(x) - added:
function is_tensor(x) - added:
function is_tf_symbolic_tensor(x) - added:
function is_tf_tensor(x) - added:
function is_timm_config_dict(config_dict: dict[str, Any]) -> bool - added:
function is_timm_local_checkpoint(pretrained_model_path: str) -> bool - added:
function is_torch_device(x) - added:
function is_torch_dtype(x) - added:
function is_torch_tensor(x) - added:
function reshape(array, newshape) - added:
function set_attribute_for_modules(module: torch.nn.Module, key: str, value: Any) - added:
function squeeze(array, axis=…) - added:
function strtobool(val) - added:
function tensor_size(array) - added:
function to_numpy(obj) - added:
function to_py_obj(obj) - added:
function torch_float(x) - added:
function torch_int(x) - added:
function transpose(array, axes=…) - added:
function working_or_temp_dir(working_dir, use_temp_dir: bool=…) - added:
method GeneralInterface.valid_keys() -> list[str] - added:
method ModelOutput.pop(*args, **kwargs) - added:
method ModelOutput.setdefault(*args, **kwargs) - added:
method ModelOutput.to_tuple() -> tuple - added:
method ModelOutput.update(*args, **kwargs) - added:
static method GeneralInterface.register(key: str, value: Callable) - added:
variable logger
- added:
class PushInProgress - added:
class PushToHubMixin - added:
constructor PushInProgress(jobs: Optional[futures.Future]=…) -> None - added:
function cached_file(path_or_repo_id: Union[str, os.PathLike], filename: str, **kwargs) -> Optional[str] - added:
function cached_files(path_or_repo_id: Union[str, os.PathLike], filenames: list[str], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, resume_download: Optional[bool]=…, proxies: Optional[dict[str, str]]=…, token: Optional[Union[bool, str]]=…, revision: Optional[str]=…, local_files_only: bool=…, subfolder: str=…, repo_type: Optional[str]=…, user_agent: Optional[Union[str, d… - added:
function convert_file_size_to_int(size: Union[int, str]) - added:
function create_and_tag_model_card(repo_id: str, tags: Optional[list[str]]=…, token: Optional[str]=…, ignore_metadata_errors: bool=…) - added:
function define_sagemaker_information() - added:
function download_url(url, proxies=…) - added:
function extract_commit_hash(resolved_file: Optional[str], commit_hash: Optional[str]) -> Optional[str] - added:
function get_checkpoint_shard_files(pretrained_model_name_or_path, index_filename, cache_dir=…, force_download=…, proxies=…, resume_download=…, local_files_only=…, token=…, user_agent=…, revision=…, subfolder=…, _commit_hash=…, **deprecated_kwargs) - added:
function has_file(path_or_repo: Union[str, os.PathLike], filename: str, revision: Optional[str]=…, proxies: Optional[dict[str, str]]=…, token: Optional[Union[bool, str]]=…, *, local_files_only: bool=…, cache_dir: Union[str, Path, None]=…, repo_type: Optional[str]=…, **deprecated_kwargs) - added:
function http_user_agent(user_agent: Union[dict, str, None]=…) -> str - added:
function is_offline_mode() - added:
function is_remote_url(url_or_filename) - added:
function list_repo_templates(repo_id: str, *, local_files_only: bool, revision: Optional[str]=…, cache_dir: Optional[str]=…, token: Union[bool, str, None]=…) -> list[str] - added:
method PushInProgress.cancel() -> None - added:
method PushInProgress.is_done() - added:
method PushInProgress.wait_until_done() - added:
method PushToHubMixin.push_to_hub(repo_id: str, use_temp_dir: Optional[bool]=…, commit_message: Optional[str]=…, private: Optional[bool]=…, token: Optional[Union[bool, str]]=…, max_shard_size: Optional[Union[int, str]]=…, create_pr: bool=…, safe_serialization: bool=…, revision: Optional[str]=…, commit_description: Optional[str]=…, tags: Optional[list[str]]=…, **deprecated_kwargs) -> str - added:
variable CHAT_TEMPLATE_DIR - added:
variable CHAT_TEMPLATE_FILE - added:
variable CLOUDFRONT_DISTRIB_PREFIX - added:
variable HF_MODULES_CACHE - added:
variable HUGGINGFACE_CO_EXAMPLES_TELEMETRY - added:
variable HUGGINGFACE_CO_PREFIX - added:
variable HUGGINGFACE_CO_RESOLVE_ENDPOINT - added:
variable LEGACY_PROCESSOR_CHAT_TEMPLATE_FILE - added:
variable PYTORCH_PRETRAINED_BERT_CACHE - added:
variable PYTORCH_TRANSFORMERS_CACHE - added:
variable S3_BUCKET_PREFIX - added:
variable SESSION_ID - added:
variable TRANSFORMERS_CACHE - added:
variable TRANSFORMERS_DYNAMIC_MODULE_NAME - added:
variable default_cache_path - added:
variable logger - added:
variable torch_cache_home
- added:
class Backend - added:
class DummyObject(type) - added:
class OptionalDependencyNotAvailable(BaseException) - added:
class VersionComparison(Enum) - added:
constructor Backend(backend_requirement: str) - added:
field DummyObject.is_dummy - added:
field VersionComparison.EQUAL - added:
field VersionComparison.GREATER_THAN - added:
field VersionComparison.GREATER_THAN_OR_EQUAL - added:
field VersionComparison.LESS_THAN - added:
field VersionComparison.LESS_THAN_OR_EQUAL - added:
field VersionComparison.NOT_EQUAL - added:
function check_torch_load_is_safe() -> None - added:
function clear_import_cache() -> None - added:
function create_import_structure_from_path(module_path) - added:
function define_import_structure(module_path: str, prefix: Optional[str]=…) -> IMPORT_STRUCTURE_T - added:
function direct_transformers_import(path: str, file=…) -> ModuleType - added:
function fetch__all__(file_content) -> list[str] - added:
function get_sudachi_version() -> bool - added:
function get_torch_major_and_minor_version() -> str - added:
function get_torch_version() -> str - added:
function is_accelerate_available(min_version: str=…) -> bool - added:
function is_apex_available() -> Union[tuple[bool, str], bool] - added:
function is_apollo_torch_available() -> Union[tuple[bool, str], bool] - added:
function is_aqlm_available() -> Union[tuple[bool, str], bool] - added:
function is_auto_awq_available() -> bool - added:
function is_auto_gptq_available() -> Union[tuple[bool, str], bool] - added:
function is_auto_round_available(min_version: str=…) -> bool - added:
function is_av_available() -> bool - added:
function is_bitsandbytes_available(check_library_only: bool=…) -> bool - added:
function is_bitsandbytes_multi_backend_available() -> bool - added:
function is_bs4_available() -> Union[tuple[bool, str], bool] - added:
function is_causal_conv1d_available() -> Union[tuple[bool, str], bool] - added:
function is_ccl_available() -> bool - added:
function is_coloredlogs_available() -> Union[tuple[bool, str], bool] - added:
function is_compressed_tensors_available() -> bool - added:
function is_cuda_platform() -> bool - added:
function is_cv2_available() -> Union[tuple[bool, str], bool] - added:
function is_cython_available() -> bool - added:
function is_datasets_available() -> Union[tuple[bool, str], bool] - added:
function is_decord_available() -> bool - added:
function is_detectron2_available() -> Union[tuple[bool, str], bool] - added:
function is_eetq_available() -> Union[tuple[bool, str], bool] - added:
function is_essentia_available() -> Union[tuple[bool, str], bool] - added:
function is_faiss_available() -> bool - added:
function is_fastapi_available() -> Union[tuple[bool, str], bool] - added:
function is_fbgemm_gpu_available() -> Union[tuple[bool, str], bool] - added:
function is_flash_attn_2_available() -> bool - added:
function is_flash_attn_3_available() -> bool - added:
function is_flash_attn_greater_or_equal(library_version: str) -> bool - added:
function is_flash_attn_greater_or_equal_2_10() -> bool - added:
function is_flash_linear_attention_available() - added:
function is_flax_available() -> bool - added:
function is_flute_available() -> bool - added:
function is_fp_quant_available() - added:
function is_fsdp_available(min_version: str=…) -> bool - added:
function is_ftfy_available() -> Union[tuple[bool, str], bool] - added:
function is_g2p_en_available() -> Union[tuple[bool, str], bool] - added:
function is_galore_torch_available() -> Union[tuple[bool, str], bool] - added:
function is_gguf_available(min_version: str=…) -> bool - added:
function is_gptqmodel_available() -> Union[tuple[bool, str], bool] - added:
function is_grokadamw_available() -> Union[tuple[bool, str], bool] - added:
function is_habana_gaudi1() -> bool - added:
function is_hadamard_available() -> Union[tuple[bool, str], bool] - added:
function is_hqq_available(min_version: str=…) -> bool - added:
function is_huggingface_hub_greater_or_equal(library_version: str, accept_dev: bool=…) -> bool - added:
function is_in_notebook() -> bool - added:
function is_ipex_available(min_version: str=…) -> bool - added:
function is_jinja_available() -> Union[tuple[bool, str], bool] - added:
function is_jumanpp_available() -> bool - added:
function is_kenlm_available() -> Union[tuple[bool, str], bool] - added:
function is_keras_nlp_available() -> Union[tuple[bool, str], bool] - added:
function is_kernels_available() -> Union[tuple[bool, str], bool] - added:
function is_levenshtein_available() -> Union[tuple[bool, str], bool] - added:
function is_libcst_available() -> Union[tuple[bool, str], bool] - added:
function is_librosa_available() -> Union[tuple[bool, str], bool] - added:
function is_liger_kernel_available() -> bool - added:
function is_lomo_available() -> Union[tuple[bool, str], bool] - added:
function is_mamba_2_ssm_available() -> bool - added:
function is_mamba_ssm_available() -> Union[tuple[bool, str], bool] - added:
function is_mambapy_available() -> Union[tuple[bool, str], bool] - added:
function is_matplotlib_available() -> Union[tuple[bool, str], bool] - added:
function is_mistral_common_available() -> Union[tuple[bool, str], bool] - added:
function is_mlx_available() -> Union[tuple[bool, str], bool] - added:
function is_natten_available() -> Union[tuple[bool, str], bool] - added:
function is_ninja_available() -> bool - added:
function is_nltk_available() -> Union[tuple[bool, str], bool] - added:
function is_num2words_available() -> Union[tuple[bool, str], bool] - added:
function is_onnx_available() -> Union[tuple[bool, str], bool] - added:
function is_openai_available() -> Union[tuple[bool, str], bool] - added:
function is_optimum_available() -> Union[tuple[bool, str], bool] - added:
function is_optimum_neuron_available() -> Union[tuple[bool, str], bool] - added:
function is_optimum_quanto_available() - added:
function is_pandas_available() -> Union[tuple[bool, str], bool] - added:
function is_peft_available() -> Union[tuple[bool, str], bool] - added:
function is_phonemizer_available() -> Union[tuple[bool, str], bool] - added:
function is_pretty_midi_available() -> Union[tuple[bool, str], bool] - added:
function is_protobuf_available() -> bool - added:
function is_psutil_available() -> Union[tuple[bool, str], bool] - added:
function is_py3nvml_available() -> Union[tuple[bool, str], bool] - added:
function is_pyctcdecode_available() -> Union[tuple[bool, str], bool] - added:
function is_pydantic_available() -> Union[tuple[bool, str], bool] - added:
function is_pygments_available() -> Union[tuple[bool, str], bool] - added:
function is_pytesseract_available() -> Union[tuple[bool, str], bool] - added:
function is_pytest_available() -> Union[tuple[bool, str], bool] - added:
function is_pytorch_quantization_available() -> Union[tuple[bool, str], bool] - added:
function is_quanto_greater(library_version: str, accept_dev: bool=…) -> bool - added:
function is_quark_available() -> Union[tuple[bool, str], bool] - added:
function is_qutlass_available() - added:
function is_rich_available() -> Union[tuple[bool, str], bool] - added:
function is_rjieba_available() -> Union[tuple[bool, str], bool] - added:
function is_rocm_platform() -> bool - added:
function is_sacremoses_available() -> Union[tuple[bool, str], bool] - added:
function is_safetensors_available() -> Union[tuple[bool, str], bool] - added:
function is_sagemaker_dp_enabled() -> bool - added:
function is_sagemaker_mp_enabled() -> bool - added:
function is_schedulefree_available(min_version: str=…) -> bool - added:
function is_scipy_available() -> Union[tuple[bool, str], bool] - added:
function is_sentencepiece_available() -> Union[tuple[bool, str], bool] - added:
function is_seqio_available() -> Union[tuple[bool, str], bool] - added:
function is_sklearn_available() -> Union[tuple[bool, str], bool] - added:
function is_soundfile_available() -> Union[tuple[bool, str], bool] - added:
function is_spacy_available() -> Union[tuple[bool, str], bool] - added:
function is_speech_available() -> Union[tuple[bool, str], bool] - added:
function is_spqr_available() -> Union[tuple[bool, str], bool] - added:
function is_sudachi_available() -> bool - added:
function is_sudachi_projection_available() -> bool - added:
function is_tensorflow_probability_available() -> Union[tuple[bool, str], bool] - added:
function is_tensorflow_text_available() -> Union[tuple[bool, str], bool] - added:
function is_tf2onnx_available() -> Union[tuple[bool, str], bool] - added:
function is_tf_available() -> bool - added:
function is_tiktoken_available() -> Union[tuple[bool, str], bool] - added:
function is_timm_available() -> Union[tuple[bool, str], bool] - added:
function is_tokenizers_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_accelerator_available() -> bool - added:
function is_torch_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_bf16_available() -> bool - added:
function is_torch_bf16_available_on_device(device: str) -> bool - added:
function is_torch_bf16_cpu_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_bf16_gpu_available() -> bool - added:
function is_torch_compile_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_cuda_available() -> bool - added:
function is_torch_deterministic() -> bool - added:
function is_torch_flex_attn_available() -> bool - added:
function is_torch_fp16_available_on_device(device: str) -> bool - added:
function is_torch_fx_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_fx_proxy(x) - added:
function is_torch_greater_or_equal(library_version: str, accept_dev: bool=…) -> bool - added:
function is_torch_hpu_available() -> bool - added:
function is_torch_less_or_equal(library_version: str, accept_dev: bool=…) -> bool - added:
function is_torch_mlu_available() -> bool - added:
function is_torch_mps_available(min_version: Optional[str]=…) -> bool - added:
function is_torch_musa_available(check_device=…) -> bool - added:
function is_torch_neuroncore_available(check_device=…) -> bool - added:
function is_torch_npu_available(check_device=…) -> bool - added:
function is_torch_optimi_available() -> Union[tuple[bool, str], bool] - added:
function is_torch_sdpa_available() - added:
function is_torch_tensorrt_fx_available() -> bool - added:
function is_torch_tf32_available() -> bool - added:
function is_torch_xla_available(check_is_tpu=…, check_is_gpu=…) -> bool - added:
function is_torch_xpu_available(check_device: bool=…) -> bool - added:
function is_torchao_available(min_version: str=…) -> bool - added:
function is_torchaudio_available() -> Union[tuple[bool, str], bool] - added:
function is_torchcodec_available() -> bool - added:
function is_torchdistx_available() - added:
function is_torchdynamo_available() -> Union[tuple[bool, str], bool] - added:
function is_torchdynamo_compiling() -> Union[tuple[bool, str], bool] - added:
function is_torchdynamo_exporting() -> bool - added:
function is_torchvision_available() -> bool - added:
function is_torchvision_v2_available() -> bool - added:
function is_training_run_on_sagemaker() -> bool - added:
function is_triton_available(min_version: str=…) -> bool - added:
function is_uroman_available() -> Union[tuple[bool, str], bool] - added:
function is_uvicorn_available() -> Union[tuple[bool, str], bool] - added:
function is_vision_available() -> bool - added:
function is_vptq_available(min_version: str=…) -> bool - added:
function is_xlstm_available() -> Union[tuple[bool, str], bool] - added:
function is_yt_dlp_available() -> Union[tuple[bool, str], bool] - added:
function requires(*, backends=…) - added:
function requires_backends(obj, backends) - added:
function split_package_version(package_version_str) -> tuple[str, str, str] - added:
function spread_import_structure(nested_import_structure) - added:
function torch_only_method(fn: Callable) -> Callable - added:
method Backend.is_satisfied() -> bool - added:
property Backend.error_message - added:
static method VersionComparison.from_string(version_string: str) -> VersionComparison - added:
variable ACCELERATE_IMPORT_ERROR - added:
variable ACCELERATE_MIN_VERSION - added:
variable AUTOROUND_MIN_VERSION - added:
variable AV_IMPORT_ERROR - added:
variable BACKENDS_MAPPING - added:
variable BACKENDS_T - added:
variable BASE_FILE_REQUIREMENTS - added:
variable BS4_IMPORT_ERROR - added:
variable CCL_IMPORT_ERROR - added:
variable CV2_IMPORT_ERROR - added:
variable CYTHON_IMPORT_ERROR - added:
variable DATASETS_IMPORT_ERROR - added:
variable DECORD_IMPORT_ERROR - added:
variable DETECTRON2_IMPORT_ERROR - added:
variable ENV_VARS_TRUE_AND_AUTO_VALUES - added:
variable ENV_VARS_TRUE_VALUES - added:
variable ESSENTIA_IMPORT_ERROR - added:
variable FAISS_IMPORT_ERROR - added:
variable FASTAPI_IMPORT_ERROR - added:
variable FLAX_IMPORT_ERROR - added:
variable FORCE_TF_AVAILABLE - added:
variable FSDP_MIN_VERSION - added:
variable FTFY_IMPORT_ERROR - added:
variable G2P_EN_IMPORT_ERROR - added:
variable GGUF_MIN_VERSION - added:
variable HQQ_MIN_VERSION - added:
variable IMPORT_STRUCTURE_T - added:
variable JINJA_IMPORT_ERROR - added:
variable KERAS_NLP_IMPORT_ERROR - added:
variable LEVENSHTEIN_IMPORT_ERROR - added:
variable LIBROSA_IMPORT_ERROR - added:
variable MISTRAL_COMMON_IMPORT_ERROR - added:
variable NATTEN_IMPORT_ERROR - added:
variable NLTK_IMPORT_ERROR - added:
variable NUMEXPR_IMPORT_ERROR - added:
variable OPENAI_IMPORT_ERROR - added:
variable PANDAS_IMPORT_ERROR - added:
variable PEFT_IMPORT_ERROR - added:
variable PHONEMIZER_IMPORT_ERROR - added:
variable PRETTY_MIDI_IMPORT_ERROR - added:
variable PROTOBUF_IMPORT_ERROR - added:
variable PYCTCDECODE_IMPORT_ERROR - added:
variable PYDANTIC_IMPORT_ERROR - added:
variable PYTESSERACT_IMPORT_ERROR - added:
variable PYTORCH_IMPORT_ERROR - added:
variable PYTORCH_IMPORT_ERROR_WITH_TF - added:
variable PYTORCH_QUANTIZATION_IMPORT_ERROR - added:
variable RICH_IMPORT_ERROR - added:
variable RJIEBA_IMPORT_ERROR - added:
variable SACREMOSES_IMPORT_ERROR - added:
variable SCHEDULEFREE_MIN_VERSION - added:
variable SCIPY_IMPORT_ERROR - added:
variable SENTENCEPIECE_IMPORT_ERROR - added:
variable SKLEARN_IMPORT_ERROR - added:
variable SPEECH_IMPORT_ERROR - added:
variable TENSORFLOW_IMPORT_ERROR - added:
variable TENSORFLOW_PROBABILITY_IMPORT_ERROR - added:
variable TENSORFLOW_TEXT_IMPORT_ERROR - added:
variable TF_IMPORT_ERROR_WITH_PYTORCH - added:
variable TIMM_IMPORT_ERROR - added:
variable TOKENIZERS_IMPORT_ERROR - added:
variable TORCHAO_MIN_VERSION - added:
variable TORCHAUDIO_IMPORT_ERROR - added:
variable TORCHCODEC_IMPORT_ERROR - added:
variable TORCHVISION_IMPORT_ERROR - added:
variable TORCH_FX_REQUIRED_VERSION - added:
variable TRITON_MIN_VERSION - added:
variable UROMAN_IMPORT_ERROR - added:
variable USE_JAX - added:
variable USE_TF - added:
variable USE_TORCH - added:
variable USE_TORCH_XLA - added:
variable UVICORN_IMPORT_ERROR - added:
variable VISION_IMPORT_ERROR - added:
variable VPTQ_MIN_VERSION - added:
variable XLA_FSDPV2_MIN_VERSION - added:
variable YT_DLP_IMPORT_ERROR - added:
variable candidates - added:
variable ccl_version - added:
variable logger
- added:
class EmptyTqdm - added:
constructor EmptyTqdm(*args, **kwargs) - added:
function add_handler(handler: logging.Handler) -> None - added:
function captureWarnings(capture) - added:
function disable_progress_bar() - added:
function enable_progress_bar() - added:
function get_log_levels_dict() - added:
function info_once(self, *args, **kwargs) - added:
function is_progress_bar_enabled() -> bool - added:
function remove_handler(handler: logging.Handler) -> None - added:
function warning_advice(self, *args, **kwargs) - added:
function warning_once(self, *args, **kwargs) - added:
variable tqdm
- added:
class ContinuousBatchProcessorMetrics - added:
class RequestStatus(Enum) - added:
constructor ContinuousBatchProcessorMetrics(max_batch_tokens: int) - added:
field RequestStatus.DECODING - added:
field RequestStatus.FAILED - added:
field RequestStatus.FINISHED - added:
field RequestStatus.PENDING - added:
field RequestStatus.PREFILLING - added:
field RequestStatus.PREFILLING_SPLIT - added:
field RequestStatus.SPLIT_PENDING_REMAINDER - added:
function attach_tracer(tracer_name_template=…) - added:
function traced(func=…, *, span_name=…, standalone=…, additional_attributes: Optional[list[tuple[str, str, Union[Any, Callable[[Any], Any]]]]]=…) - added:
method ContinuousBatchProcessorMetrics.record_batch_metrics(requests_in_batch: list) -> None - added:
method ContinuousBatchProcessorMetrics.record_kv_cache_memory_metrics(cache) -> None - added:
method ContinuousBatchProcessorMetrics.record_queue_metrics(active_requests: int, waiting_requests: int) -> None - added:
method ContinuousBatchProcessorMetrics.record_request_completion(created_time: float, request_id: str) -> None - added:
method ContinuousBatchProcessorMetrics.record_ttft_metric(created_time: float, request_id: str) -> None - added:
variable logger
- added:
function assert_device_map(device_map, num_blocks) - added:
function get_device_map(n_layers, devices)
- changed:
method NotebookProgressBar.update(value: int, force_update: bool=…, comment: str=…) → method NotebookProgressBar.update(value: int, force_update: bool=…, comment: Optional[str]=…) - added:
method NotebookProgressCallback.on_predict(args, state, control, **kwargs)
- added:
function check_peft_version(min_version: str) -> None - added:
function find_adapter_config_file(model_id: str, cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, resume_download: Optional[bool]=…, proxies: Optional[dict[str, str]]=…, token: Optional[Union[bool, str]]=…, revision: Optional[str]=…, local_files_only: bool=…, subfolder: str=…, _commit_hash: Optional[str]=…) -> Optional[str] - added:
variable ADAPTER_CONFIG_NAME - added:
variable ADAPTER_SAFE_WEIGHTS_NAME - added:
variable ADAPTER_WEIGHTS_NAME
- added:
class AWQLinearVersion(str, Enum) - added:
class AqlmConfig(QuantizationConfigMixin) - added:
class AutoRoundConfig(QuantizationConfigMixin) - added:
class AwqBackendPackingMethod(str, Enum) - added:
class AwqConfig(QuantizationConfigMixin) - added:
class BitNetQuantConfig(QuantizationConfigMixin) - added:
class BitsAndBytesConfig(QuantizationConfigMixin) - added:
class CompressedTensorsConfig(QuantizationConfigMixin) - added:
class EetqConfig(QuantizationConfigMixin) - added:
class ExllamaVersion(int, Enum) - added:
class FPQuantConfig(QuantizationConfigMixin) - added:
class FbgemmFp8Config(QuantizationConfigMixin) - added:
class FineGrainedFP8Config(QuantizationConfigMixin) - added:
class GPTQConfig(QuantizationConfigMixin) - added:
class HiggsConfig(QuantizationConfigMixin) - added:
class HqqConfig(QuantizationConfigMixin) - added:
class Mxfp4Config(QuantizationConfigMixin) - added:
class QuantizationConfigMixin - added:
class QuantizationMethod(str, Enum) - added:
class QuantoConfig(QuantizationConfigMixin) - added:
class QuarkConfig(QuantizationConfigMixin) - added:
class SpQRConfig(QuantizationConfigMixin) - added:
class TorchAoConfig(QuantizationConfigMixin) - added:
class VptqConfig(QuantizationConfigMixin) - added:
class VptqLayerConfig(QuantizationConfigMixin) - added:
constructor AqlmConfig(in_group_size: int=…, out_group_size: int=…, num_codebooks: int=…, nbits_per_codebook: int=…, linear_weights_not_to_quantize: Optional[list[str]]=…, **kwargs) - added:
constructor AutoRoundConfig(bits: int=…, group_size: int=…, sym: bool=…, backend: str=…, **kwargs) - added:
constructor AwqConfig(bits: int=…, group_size: int=…, zero_point: bool=…, version: AWQLinearVersion=…, backend: AwqBackendPackingMethod=…, do_fuse: Optional[bool]=…, fuse_max_seq_len: Optional[int]=…, modules_to_fuse: Optional[dict]=…, modules_to_not_convert: Optional[list]=…, exllama_config: Optional[dict[str, int]]=…, **kwargs) - added:
constructor BitNetQuantConfig(modules_to_not_convert: Optional[list]=…, linear_class: str=…, quantization_mode: str=…, use_rms_norm: bool=…, rms_norm_eps: Optional[float]=…, **kwargs) - added:
constructor BitsAndBytesConfig(load_in_8bit=…, load_in_4bit=…, llm_int8_threshold=…, llm_int8_skip_modules=…, llm_int8_enable_fp32_cpu_offload=…, llm_int8_has_fp16_weight=…, bnb_4bit_compute_dtype=…, bnb_4bit_quant_type=…, bnb_4bit_use_double_quant=…, bnb_4bit_quant_storage=…, **kwargs) - added:
constructor CompressedTensorsConfig(config_groups: Optional[dict[str, Union[QuantizationScheme, list[str]]]]=…, format: str=…, quantization_status: QuantizationStatus=…, kv_cache_scheme: Optional[QuantizationArgs]=…, global_compression_ratio: Optional[float]=…, ignore: Optional[list[str]]=…, sparsity_config: Optional[dict[str, Any]]=…, quant_method: str=…, run_compressed: bool=…, **kwargs) - added:
constructor EetqConfig(weights: str=…, modules_to_not_convert: Optional[list]=…, **kwargs) - added:
constructor FPQuantConfig(forward_dtype: str=…, forward_method: str=…, backward_dtype: str=…, store_master_weights: bool=…, hadamard_group_size: Optional[int]=…, pseudoquantization: bool=…, transform_init: str=…, modules_to_not_convert: Optional[list[str]]=…, **kwargs) - added:
constructor FbgemmFp8Config(activation_scale_ub: float=…, modules_to_not_convert: Optional[list]=…, **kwargs) - added:
constructor FineGrainedFP8Config(activation_scheme: str=…, weight_block_size: tuple[int, int]=…, modules_to_not_convert: Optional[list]=…, **kwargs) - added:
constructor GPTQConfig(bits: int, tokenizer: Any=…, dataset: Optional[Union[list[str], str]]=…, group_size: int=…, damp_percent: float=…, desc_act: bool=…, sym: bool=…, true_sequential: bool=…, checkpoint_format: str=…, meta: Optional[dict[str, Any]]=…, backend: Optional[str]=…, use_cuda_fp16: bool=…, model_seqlen: Optional[int]=…, block_name_to_quantize: Optional[str]=…, module_name_preceding_fi… - added:
constructor HiggsConfig(bits: int=…, p: int=…, modules_to_not_convert: Optional[list[str]]=…, hadamard_size: int=…, group_size: int=…, tune_metadata: Optional[dict[str, Any]]=…, **kwargs) - added:
constructor HqqConfig(nbits: int=…, group_size: int=…, view_as_float: bool=…, axis: Optional[int]=…, dynamic_config: Optional[dict]=…, skip_modules: list[str]=…, **kwargs) - added:
constructor Mxfp4Config(modules_to_not_convert: Optional[list]=…, dequantize: bool=…, **kwargs) - added:
constructor QuantoConfig(weights=…, activations=…, modules_to_not_convert: Optional[list]=…, **kwargs) - added:
constructor QuarkConfig(**kwargs) - added:
constructor SpQRConfig(bits: int=…, beta1: int=…, beta2: int=…, shapes: Optional[dict[str, int]]=…, modules_to_not_convert: Optional[list[str]]=…, **kwargs) - added:
constructor TorchAoConfig(quant_type: Union[str, AOBaseConfig], modules_to_not_convert: Optional[list]=…, include_input_output_embeddings: bool=…, untie_embedding_weights: bool=…, **kwargs) - added:
constructor VptqConfig(enable_proxy_error: bool=…, config_for_layers: dict[str, Any]=…, shared_layer_config: dict[str, Any]=…, modules_to_not_convert: Optional[list]=…, **kwargs) - added:
constructor VptqLayerConfig(enable_norm: bool=…, enable_perm: bool=…, group_num: int=…, group_size: int=…, in_features: int=…, indices_as_float: bool=…, is_indice_packed: bool=…, num_centroids: tuple=…, num_res_centroids: tuple=…, out_features: int=…, outlier_size: int=…, vector_lens: tuple=…, **kwargs) - added:
field AWQLinearVersion.EXLLAMA - added:
field AWQLinearVersion.GEMM - added:
field AWQLinearVersion.GEMV - added:
field AWQLinearVersion.IPEX - added:
field AwqBackendPackingMethod.AUTOAWQ - added:
field AwqBackendPackingMethod.LLMAWQ - added:
field ExllamaVersion.ONE - added:
field ExllamaVersion.TWO - added:
field QuantizationConfigMixin.quant_method: QuantizationMethod - added:
field QuantizationMethod.AQLM - added:
field QuantizationMethod.AUTOROUND - added:
field QuantizationMethod.AWQ - added:
field QuantizationMethod.BITNET - added:
field QuantizationMethod.BITS_AND_BYTES - added:
field QuantizationMethod.COMPRESSED_TENSORS - added:
field QuantizationMethod.EETQ - added:
field QuantizationMethod.FBGEMM_FP8 - added:
field QuantizationMethod.FP8 - added:
field QuantizationMethod.FPQUANT - added:
field QuantizationMethod.GPTQ - added:
field QuantizationMethod.HIGGS - added:
field QuantizationMethod.HQQ - added:
field QuantizationMethod.MXFP4 - added:
field QuantizationMethod.QUANTO - added:
field QuantizationMethod.QUARK - added:
field QuantizationMethod.SPQR - added:
field QuantizationMethod.TORCHAO - added:
field QuantizationMethod.VPTQ - added:
field TorchAoConfig.include_input_output_embeddings: bool - added:
field TorchAoConfig.modules_to_not_convert: Optional[list] - added:
field TorchAoConfig.quant_method: QuantizationMethod - added:
field TorchAoConfig.quant_type: Union[str, AOBaseConfig] - added:
field TorchAoConfig.quant_type_kwargs: dict[str, Any] - added:
field TorchAoConfig.untie_embedding_weights: bool - added:
method AqlmConfig.post_init() - added:
method AutoRoundConfig.get_loading_attributes() - added:
method AutoRoundConfig.post_init() - added:
method AutoRoundConfig.to_dict() - added:
method AwqConfig.get_loading_attributes() - added:
method AwqConfig.post_init() - added:
method BitNetQuantConfig.post_init() - added:
method BitsAndBytesConfig.is_quantizable() - added:
method BitsAndBytesConfig.post_init() - added:
method BitsAndBytesConfig.quantization_method() - added:
method BitsAndBytesConfig.to_dict() -> dict[str, Any] - added:
method BitsAndBytesConfig.to_diff_dict() -> dict[str, Any] - added:
method CompressedTensorsConfig.get_loading_attributes() - added:
method CompressedTensorsConfig.post_init() - added:
method CompressedTensorsConfig.to_dict() -> dict[str, Any] - added:
method CompressedTensorsConfig.to_diff_dict() -> dict[str, Any] - added:
method EetqConfig.post_init() - added:
method FPQuantConfig.post_init() - added:
method FbgemmFp8Config.get_loading_attributes() - added:
method FineGrainedFP8Config.post_init() - added:
method GPTQConfig.get_loading_attributes() - added:
method GPTQConfig.post_init() - added:
method GPTQConfig.to_dict() -> dict[str, Any] - added:
method GPTQConfig.to_dict_optimum() - added:
method HiggsConfig.post_init() - added:
method HqqConfig.post_init() - added:
method HqqConfig.to_dict() -> dict[str, Any] - added:
method HqqConfig.to_diff_dict() -> dict[str, Any] - added:
method Mxfp4Config.get_loading_attributes() - added:
method Mxfp4Config.to_dict() -> dict[str, Any] - added:
method QuantizationConfigMixin.to_dict() -> dict[str, Any] - added:
method QuantizationConfigMixin.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method QuantizationConfigMixin.to_json_string(use_diff: bool=…) -> str - added:
method QuantizationConfigMixin.update(**kwargs) - added:
method QuantoConfig.post_init() - added:
method SpQRConfig.post_init() - added:
method TorchAoConfig.get_apply_tensor_subclass() - added:
method TorchAoConfig.post_init() - added:
method TorchAoConfig.to_dict() - added:
method VptqConfig.post_init() - added:
method VptqLayerConfig.post_init() - added:
property BitsAndBytesConfig.load_in_4bit - added:
property BitsAndBytesConfig.load_in_8bit - added:
property CompressedTensorsConfig.is_quantization_compressed - added:
property CompressedTensorsConfig.is_quantized - added:
property CompressedTensorsConfig.is_sparsification_compressed - added:
static method AWQLinearVersion.from_str(version: str) - added:
static method AutoRoundConfig.from_dict(config_dict, return_unused_kwargs=…, **kwargs) - added:
static method CompressedTensorsConfig.from_dict(config_dict, return_unused_kwargs=…, **kwargs) - added:
static method GPTQConfig.from_dict_optimum(config_dict) - added:
static method HqqConfig.from_dict(config: dict[str, Any]) - added:
static method QuantizationConfigMixin.from_dict(config_dict, return_unused_kwargs=…, **kwargs) - added:
static method TorchAoConfig.from_dict(config_dict, return_unused_kwargs=…, **kwargs) - added:
variable logger
- added:
variable DESCRIPTOR
- added:
function require_version(requirement: str, hint: Optional[str]=…) -> None - added:
function require_version_core(requirement) - added:
variable ops
- added:
class BaseVideoProcessor(BaseImageProcessorFast) - added:
constructor BaseVideoProcessor(**kwargs: Unpack[VideosKwargs]) -> None - added:
field BaseVideoProcessor.crop_size - added:
field BaseVideoProcessor.default_to_square - added:
field BaseVideoProcessor.do_center_crop - added:
field BaseVideoProcessor.do_convert_rgb - added:
field BaseVideoProcessor.do_normalize - added:
field BaseVideoProcessor.do_rescale - added:
field BaseVideoProcessor.do_resize - added:
field BaseVideoProcessor.do_sample_frames - added:
field BaseVideoProcessor.fps - added:
field BaseVideoProcessor.image_mean - added:
field BaseVideoProcessor.image_std - added:
field BaseVideoProcessor.model_input_names - added:
field BaseVideoProcessor.num_frames - added:
field BaseVideoProcessor.resample - added:
field BaseVideoProcessor.rescale_factor - added:
field BaseVideoProcessor.return_metadata - added:
field BaseVideoProcessor.size - added:
field BaseVideoProcessor.size_divisor - added:
field BaseVideoProcessor.valid_kwargs - added:
field BaseVideoProcessor.video_metadata - added:
method BaseVideoProcessor.convert_to_rgb(video: torch.Tensor) -> VideoInput - added:
method BaseVideoProcessor.fetch_videos(video_url_or_urls: Union[str, list[str], list[list[str]]], sample_indices_fn=…) - added:
method BaseVideoProcessor.preprocess(videos: VideoInput, **kwargs: Unpack[VideosKwargs]) -> BatchFeature - added:
method BaseVideoProcessor.sample_frames(metadata: VideoMetadata, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, **kwargs) - added:
method BaseVideoProcessor.save_pretrained(save_directory: Union[str, os.PathLike], push_to_hub: bool=…, **kwargs) - added:
method BaseVideoProcessor.to_dict() -> dict[str, Any] - added:
method BaseVideoProcessor.to_json_file(json_file_path: Union[str, os.PathLike]) - added:
method BaseVideoProcessor.to_json_string() -> str - added:
static method BaseVideoProcessor.from_dict(video_processor_dict: dict[str, Any], **kwargs) - added:
static method BaseVideoProcessor.from_json_file(json_file: Union[str, os.PathLike]) - added:
static method BaseVideoProcessor.from_pretrained(pretrained_model_name_or_path: Union[str, os.PathLike], cache_dir: Optional[Union[str, os.PathLike]]=…, force_download: bool=…, local_files_only: bool=…, token: Optional[Union[str, bool]]=…, revision: str=…, **kwargs) - added:
static method BaseVideoProcessor.get_video_processor_dict(pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> tuple[dict[str, Any], dict[str, Any]] - added:
static method BaseVideoProcessor.register_for_auto_class(auto_class=…) - added:
variable BASE_VIDEO_PROCESSOR_DOCSTRING - added:
variable logger
- added:
class VideoMetadata(Mapping) - added:
field VideoMetadata.duration: Optional[float] - added:
field VideoMetadata.fps: Optional[float] - added:
field VideoMetadata.frames_indices: Optional[list[int]] - added:
field VideoMetadata.height: Optional[int] - added:
field VideoMetadata.total_num_frames: int - added:
field VideoMetadata.video_backend: Optional[str] - added:
field VideoMetadata.width: Optional[int] - added:
function convert_pil_frames_to_video(videos: list[VideoInput]) -> list[Union[np.ndarray, torch.Tensor]] - added:
function convert_to_rgb(video: np.ndarray, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function default_sample_indices_fn(metadata: VideoMetadata, num_frames=…, fps=…, **kwargs) - added:
function get_uniform_frame_indices(total_num_frames: int, num_frames: Optional[int]=…) - added:
function get_video_size(video: np.ndarray, channel_dim: Optional[ChannelDimension]=…) -> tuple[int, int] - added:
function group_videos_by_shape(videos: list[torch.Tensor]) -> tuple[dict[tuple[int, int], torch.Tensor], dict[int, tuple[tuple[int, int], int]]] - added:
function is_batched_video(videos) - added:
function is_scaled_video(video: np.ndarray) -> bool - added:
function is_valid_video(video) - added:
function is_valid_video_frame(frame) - added:
function load_video(video: VideoInput, num_frames: Optional[int]=…, fps: Optional[Union[int, float]]=…, backend: str=…, sample_indices_fn: Optional[Callable]=…, **kwargs) -> np.ndarray - added:
function make_batched_metadata(videos: VideoInput, video_metadata: Union[VideoMetadata, dict]) - added:
function make_batched_videos(videos) -> list[Union[np.ndarray, torch.Tensor, URL, Path]] - added:
function pad(video: np.ndarray, padding: Union[int, tuple[int, int], Iterable[tuple[int, int]]], mode: PaddingMode=…, constant_values: Union[float, Iterable[float]]=…, data_format: Optional[Union[str, ChannelDimension]]=…, input_data_format: Optional[Union[str, ChannelDimension]]=…) -> np.ndarray - added:
function read_video_decord(video_path: Union[URL, Path], sample_indices_fn: Callable, **kwargs) - added:
function read_video_opencv(video_path: Union[URL, Path], sample_indices_fn: Callable, **kwargs) -> tuple[np.ndarray, VideoMetadata] - added:
function read_video_pyav(video_path: Union[URL, Path], sample_indices_fn: Callable, **kwargs) - added:
function read_video_torchcodec(video_path: Union[URL, Path], sample_indices_fn: Callable, **kwargs) - added:
function read_video_torchvision(video_path: Union[URL, Path], sample_indices_fn: Callable, **kwargs) - added:
function reorder_videos(processed_videos: dict[tuple[int, int], torch.Tensor], grouped_videos_index: dict[int, tuple[tuple[int, int], int]]) -> list[torch.Tensor] - added:
function valid_videos(videos) - added:
method VideoMetadata.update(dictionary) - added:
property VideoMetadata.timestamps -> list[float] - added:
variable Path - added:
variable URL - added:
variable VIDEO_DECODERS - added:
variable VideoInput - added:
variable logger