Upgrading transformers 2 to 3
Unknown2.11.0 → 3.5.1
Fetching 10 releases from GitHub…0 / 10
UnknownUpgrade verdict
DepDelta is still reading the notes for this range, so breaking changes can't be ruled out yet.
In plain terms
- DepDelta is still reading the release notes in this range, so it can't rule out breaking changes yet.
- This page updates as the notes come in.
Must act Still reading release notes
DepDelta is still reading the release notes in this range. This updates as they come in.
Adds protobuf.
Installs 17 packages in all (was 16).
Moves sentencepiece 0.2.2 → 0.1.91, tokenizers 0.7.0 → 0.9.3. Your build may pick other versions.
API changes
Generated by DepDelta from published Python source. Not mixed with prose notes.
- removed:
export DefaultDataCollator - removed:
export TransfoXLTokenizerFast - added:
export Adafactor - added:
export AddedToken - added:
export AlbertForMultipleChoice - added:
export AlbertTokenizerFast - added:
export AutoModelForCausalLM - added:
export AutoModelForMaskedLM - added:
export AutoModelForNextSentencePrediction - added:
export AutoModelForSeq2SeqLM - added:
export BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export BartForQuestionAnswering - added:
export BartTokenizerFast - added:
export BatchEncoding - added:
export BeamScorer - added:
export BeamSearchScorer - added:
export BertGenerationConfig - added:
export BertGenerationDecoder - added:
export BertGenerationEncoder - added:
export BertGenerationTokenizer - added:
export BertLMHeadModel - added:
export BertweetTokenizer - added:
export BlenderbotConfig - added:
export BlenderbotForConditionalGeneration - added:
export BlenderbotSmallTokenizer - added:
export BlenderbotTokenizer - added:
export CamembertForCausalLM - added:
export CamembertTokenizerFast - added:
export CharSpan - added:
export Conversation - added:
export ConversationalPipeline - added:
export DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export DPRConfig - added:
export DPRContextEncoder - added:
export DPRContextEncoderTokenizer - added:
export DPRContextEncoderTokenizerFast - added:
export DPRPretrainedContextEncoder - added:
export DPRPretrainedQuestionEncoder - added:
export DPRPretrainedReader - added:
export DPRQuestionEncoder - added:
export DPRQuestionEncoderTokenizer - added:
export DPRQuestionEncoderTokenizerFast - added:
export DPRReader - added:
export DPRReaderOutput - added:
export DPRReaderTokenizer - added:
export DPRReaderTokenizerFast - added:
export DPR_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export DataCollatorForPermutationLanguageModeling - added:
export DataCollatorForSOP - added:
export DataCollatorForTokenClassification - added:
export DataCollatorForWholeWordMask - added:
export DataCollatorWithPadding - added:
export DebertaConfig - added:
export DebertaForSequenceClassification - added:
export DebertaModel - added:
export DebertaPreTrainedModel - added:
export DebertaTokenizer - added:
export DefaultFlowCallback - added:
export DistilBertForMultipleChoice - added:
export ElectraForMultipleChoice - added:
export ElectraForQuestionAnswering - added:
export EvaluationStrategy - added:
export FSMTConfig - added:
export FSMTForConditionalGeneration - added:
export FSMTModel - added:
export FSMTTokenizer - added:
export FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export FlaubertForMultipleChoice - added:
export FlaubertForTokenClassification - added:
export FlaxBertModel - added:
export FlaxRobertaModel - added:
export FunnelBaseModel - added:
export FunnelConfig - added:
export FunnelForMaskedLM - added:
export FunnelForMultipleChoice - added:
export FunnelForPreTraining - added:
export FunnelForQuestionAnswering - added:
export FunnelForSequenceClassification - added:
export FunnelForTokenClassification - added:
export FunnelModel - added:
export FunnelTokenizer - added:
export FunnelTokenizerFast - added:
export GPT2ForSequenceClassification - added:
export HerbertTokenizer - added:
export HerbertTokenizerFast - added:
export LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export LayoutLMConfig - added:
export LayoutLMForMaskedLM - added:
export LayoutLMForTokenClassification - added:
export LayoutLMModel - added:
export LayoutLMTokenizer - added:
export LayoutLMTokenizerFast - added:
export LineByLineWithRefDataset - added:
export LineByLineWithSOPTextDataset - added:
export LogitsProcessor - added:
export LogitsProcessorList - added:
export LogitsWarper - added:
export LongformerSelfAttention - added:
export LxmertConfig - added:
export LxmertEncoder - added:
export LxmertForPreTraining - added:
export LxmertForQuestionAnswering - added:
export LxmertModel - added:
export LxmertPreTrainedModel - added:
export LxmertTokenizer - added:
export LxmertTokenizerFast - added:
export LxmertVisualFeatureEncoder - added:
export LxmertXLayer - added:
export MBartConfig - added:
export MBartForConditionalGeneration - added:
export MBartTokenizerFast - added:
export MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export MODEL_FOR_CAUSAL_LM_MAPPING - added:
export MODEL_FOR_MASKED_LM_MAPPING - added:
export MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
export MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
export MinLengthLogitsProcessor - added:
export MobileBertConfig - added:
export MobileBertForMaskedLM - added:
export MobileBertForMultipleChoice - added:
export MobileBertForNextSentencePrediction - added:
export MobileBertForPreTraining - added:
export MobileBertForQuestionAnswering - added:
export MobileBertForSequenceClassification - added:
export MobileBertForTokenClassification - added:
export MobileBertLayer - added:
export MobileBertModel - added:
export MobileBertPreTrainedModel - added:
export MobileBertTokenizer - added:
export MobileBertTokenizerFast - added:
export NoBadWordsLogitsProcessor - added:
export NoRepeatNGramLogitsProcessor - added:
export OpenAIGPTForSequenceClassification - added:
export PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export PegasusConfig - added:
export PegasusForConditionalGeneration - added:
export PegasusTokenizer - added:
export PegasusTokenizerFast - added:
export PhobertTokenizer - added:
export PreTrainedTokenizerBase - added:
export PreTrainedTokenizerFast - added:
export PretrainedBartModel - added:
export PretrainedFSMTModel - added:
export PrinterCallback - added:
export ProgressCallback - added:
export ProphetNetConfig - added:
export ProphetNetDecoder - added:
export ProphetNetEncoder - added:
export ProphetNetForCausalLM - added:
export ProphetNetForConditionalGeneration - added:
export ProphetNetModel - added:
export ProphetNetPreTrainedModel - added:
export ProphetNetTokenizer - added:
export RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export RagConfig - added:
export RagModel - added:
export RagRetriever - added:
export RagSequenceForGeneration - added:
export RagTokenForGeneration - added:
export RagTokenizer - added:
export ReformerForMaskedLM - added:
export ReformerForQuestionAnswering - added:
export ReformerForSequenceClassification - added:
export ReformerTokenizerFast - added:
export RepetitionPenaltyLogitsProcessor - added:
export RetriBertConfig - added:
export RetriBertModel - added:
export RetriBertPreTrainedModel - added:
export RetriBertTokenizer - added:
export RetriBertTokenizerFast - added:
export RobertaForCausalLM - added:
export SLOW_TO_FAST_CONVERTERS - added:
export SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export SpecialTokensMixin - added:
export SquadDataTrainingArguments - added:
export SquadDataset - added:
export SqueezeBertConfig - added:
export SqueezeBertForMaskedLM - added:
export SqueezeBertForMultipleChoice - added:
export SqueezeBertForQuestionAnswering - added:
export SqueezeBertForSequenceClassification - added:
export SqueezeBertForTokenClassification - added:
export SqueezeBertModel - added:
export SqueezeBertModule - added:
export SqueezeBertPreTrainedModel - added:
export SqueezeBertTokenizer - added:
export SqueezeBertTokenizerFast - added:
export T5TokenizerFast - added:
export TFAlbertForTokenClassification - added:
export TFAutoModelForCausalLM - added:
export TFAutoModelForMaskedLM - added:
export TFAutoModelForSeq2SeqLM - added:
export TFBartForConditionalGeneration - added:
export TFBartModel - added:
export TFBertLMHeadModel - added:
export TFBlenderbotForConditionalGeneration - added:
export TFCamembertForMultipleChoice - added:
export TFCamembertForQuestionAnswering - added:
export TFDistilBertForMultipleChoice - added:
export TFElectraForMultipleChoice - added:
export TFElectraForQuestionAnswering - added:
export TFElectraForSequenceClassification - added:
export TFFlaubertForMultipleChoice - added:
export TFFlaubertForQuestionAnsweringSimple - added:
export TFFlaubertForTokenClassification - added:
export TFFunnelBaseModel - added:
export TFFunnelForMaskedLM - added:
export TFFunnelForMultipleChoice - added:
export TFFunnelForPreTraining - added:
export TFFunnelForQuestionAnswering - added:
export TFFunnelForSequenceClassification - added:
export TFFunnelForTokenClassification - added:
export TFFunnelModel - added:
export TFLongformerForMaskedLM - added:
export TFLongformerForQuestionAnswering - added:
export TFLongformerModel - added:
export TFLongformerSelfAttention - added:
export TFLxmertForPreTraining - added:
export TFLxmertMainLayer - added:
export TFLxmertModel - added:
export TFLxmertPreTrainedModel - added:
export TFLxmertVisualFeatureEncoder - added:
export TFMBartForConditionalGeneration - added:
export TFMarianMTModel - added:
export TFMobileBertForMaskedLM - added:
export TFMobileBertForMultipleChoice - added:
export TFMobileBertForNextSentencePrediction - added:
export TFMobileBertForPreTraining - added:
export TFMobileBertForQuestionAnswering - added:
export TFMobileBertForSequenceClassification - added:
export TFMobileBertForTokenClassification - added:
export TFMobileBertMainLayer - added:
export TFMobileBertModel - added:
export TFMobileBertPreTrainedModel - added:
export TFPegasusForConditionalGeneration - added:
export TFRobertaForMultipleChoice - added:
export TFXLMForMultipleChoice - added:
export TFXLMForTokenClassification - added:
export TFXLMRobertaForMultipleChoice - added:
export TFXLMRobertaForQuestionAnswering - added:
export TFXLNetForMultipleChoice - added:
export TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export TF_MODEL_FOR_CAUSAL_LM_MAPPING - added:
export TF_MODEL_FOR_MASKED_LM_MAPPING - added:
export TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
export TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
export TemperatureLogitsWarper - added:
export TensorFlowBenchmark - added:
export TensorFlowBenchmarkArguments - added:
export TensorType - added:
export Text2TextGenerationPipeline - added:
export TextDatasetForNextSentencePrediction - added:
export TokenSpan - added:
export TopKLogitsWarper - added:
export TopPLogitsWarper - added:
export TrainerCallback - added:
export TrainerControl - added:
export TrainerState - added:
export XLMForMultipleChoice - added:
export XLMProphetNetConfig - added:
export XLMProphetNetDecoder - added:
export XLMProphetNetEncoder - added:
export XLMProphetNetForCausalLM - added:
export XLMProphetNetForConditionalGeneration - added:
export XLMProphetNetModel - added:
export XLMProphetNetTokenizer - added:
export XLMRobertaForCausalLM - added:
export XLMRobertaForQuestionAnswering - added:
export XLMRobertaTokenizerFast - added:
export XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
export XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
export XLNetTokenizerFast - added:
export ZeroShotClassificationPipeline - added:
export convert_slow_tokenizer - added:
export default_data_collator - added:
export get_polynomial_decay_schedule_with_warmup - added:
export is_apex_available - added:
export is_comet_available - added:
export is_datasets_available - added:
export is_faiss_available - added:
export is_flax_available - added:
export is_optuna_available - added:
export is_psutil_available - added:
export is_py3nvml_available - added:
export is_ray_available - added:
export is_sentencepiece_available - added:
export is_tensorboard_available - added:
export is_tokenizers_available - added:
export is_torch_tpu_available - added:
export is_wandb_available - added:
export load_tf_weights_in_bert_generation - added:
export load_tf_weights_in_funnel - added:
export load_tf_weights_in_mobilebert - added:
export logging
- removed:
function swish(x) - added:
function linear_act(x) - added:
function mish(x) - added:
variable silu
- added:
function gelu(x) - added:
function gelu_fast(x) - added:
function gelu_new(x) - added:
function get_tf_activation(activation_string) - added:
function mish(x) - added:
variable ACT2FN
- removed:
variable config - removed:
variable inputs - removed:
variable model - removed:
variable outputs - removed:
variable text - removed:
variable tok
- removed:
class Frame(NamedTuple) - removed:
class Memory(NamedTuple) - removed:
class MemoryState(NamedTuple) - removed:
class MemorySummary(NamedTuple) - removed:
class UsedMemoryState(NamedTuple) - removed:
field Frame.event: str - removed:
field Frame.filename: str - removed:
field Frame.line_number: int - removed:
field Frame.line_text: str - removed:
field Frame.module: str - removed:
field Memory.bytes: int - removed:
field MemoryState.cpu: Memory - removed:
field MemoryState.cpu_gpu: Memory - removed:
field MemoryState.frame: Frame - removed:
field MemoryState.gpu: Memory - removed:
field MemorySummary.cumulative: List[MemoryState] - removed:
field MemorySummary.sequential: List[MemoryState] - removed:
field MemorySummary.total: Memory - removed:
field UsedMemoryState.cpu_memory: int - removed:
field UsedMemoryState.frame: Frame - removed:
field UsedMemoryState.gpu_memory: int - removed:
function bytes_to_human_readable(memory_amount: int) -> str - removed:
function is_memory_tracing_enabled() - removed:
function start_memory_tracing(modules_to_trace: Optional[Union[str, Iterable[str]]]=…, modules_not_to_trace: Optional[Union[str, Iterable[str]]]=…, events_to_trace: str=…, gpus_to_trace: Optional[List[int]]=…) -> MemoryTrace - removed:
function stop_memory_tracing(memory_trace: Optional[MemoryTrace]=…, ignore_released_memory: bool=…) -> Optional[MemorySummary] - removed:
variable MemoryTrace - removed:
variable logger
- added:
variable IMPORT_ERROR_MESSAGE
- removed:
method ListObjsCommand.tabulate(rows: List[List[Union[str, int]]], headers: List[str]) -> str - added:
class DeprecatedUploadCommand(BaseUserCommand) - added:
class ListReposObjsCommand(BaseUserCommand) - added:
class RepoCreateCommand(BaseUserCommand) - added:
function tabulate(rows: List[List[Union[str, int]]], headers: List[str]) -> str - added:
method DeprecatedUploadCommand.run() - added:
method ListReposObjsCommand.run() - added:
method RepoCreateCommand.run() - added:
static method ANSI.gray(s)
- removed:
variable logger - added:
function replace_list_option_in_docstrings(config_to_class=…, use_model_types=…) - added:
variable MODEL_NAMES_MAPPING
- changed:
constructor BartConfig(activation_dropout=…, 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_encoder_decoder=…, pad_tok… → 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…
- changed:
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=…, **kwargs) → 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)
- 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=…, gradient_checkpointing=…, **kwargs) - added:
field BertGenerationConfig.model_type
- added:
class BlenderbotConfig(BartConfig) - added:
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… - added:
field BlenderbotConfig.model_type - added:
variable BLENDERBOT_PRETRAINED_CONFIG_ARCHIVE_MAP
- added:
class DebertaConfig(PretrainedConfig) - 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:
variable DEBERTA_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- 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=…, gradient_checkpointing=…, projection_dim: int=…, **kwargs) - added:
field DPRConfig.model_type - added:
variable DPR_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- changed:
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=…, pad_token_id=…, **kwargs) → 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)
- changed:
static method EncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig) -> PretrainedConfig → static method EncoderDecoderConfig.from_encoder_decoder_configs(encoder_config: PretrainedConfig, decoder_config: PretrainedConfig, **kwargs) -> PretrainedConfig - added:
field EncoderDecoderConfig.is_composition
- added:
class DecoderConfig(PretrainedConfig) - added:
class FSMTConfig(PretrainedConfig) - added:
constructor DecoderConfig(vocab_size=…, bos_token_id=…) - 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=…, pad_token_id=…, bos_… - added:
field DecoderConfig.model_type - added:
field FSMTConfig.model_type - added:
method FSMTConfig.to_dict() - added:
property FSMTConfig.hidden_size -> int - added:
property FSMTConfig.num_attention_heads -> int - added:
variable FSMT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- 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=…, 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… - added:
field FunnelConfig.model_type - added:
property FunnelConfig.hidden_size - added:
property FunnelConfig.num_attention_heads - added:
property FunnelConfig.num_blocks - added:
property FunnelConfig.num_hidden_layers - added:
variable FUNNEL_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- changed:
constructor GPT2Config(vocab_size=…, n_positions=…, n_ctx=…, n_embd=…, n_layer=…, n_head=…, 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=…, **kwargs) → 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)
- added:
class LayoutLMConfig(BertConfig) - 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=…, gradient_checkpointing=…, max_2d_position_embeddings=…, **kwargs) - added:
field LayoutLMConfig.model_type - added:
variable LAYOUTLM_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- added:
class LxmertConfig(PretrainedConfig) - added:
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… - added:
field LxmertConfig.model_type - added:
variable LXMERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- added:
class MBartConfig(BartConfig) - added:
field MBartConfig.model_type - added:
variable MBART_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- added:
class MobileBertConfig(PretrainedConfig) - 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:
field MobileBertConfig.pretrained_config_archive_map - added:
variable MOBILEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- added:
class PegasusConfig(BartConfig) - added:
field PegasusConfig.model_type - added:
variable DEFAULTS - added:
variable logger - added:
variable task_specific_params
- added:
class ProphetNetConfig(PretrainedConfig) - added:
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_… - added:
field ProphetNetConfig.model_type - added:
property ProphetNetConfig.num_attention_heads -> int - added:
property ProphetNetConfig.num_hidden_layers -> int - added:
variable PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- 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.is_composition - added:
field RagConfig.model_type - added:
method RagConfig.to_dict() - added:
static method RagConfig.from_question_encoder_generator_configs(question_encoder_config: PretrainedConfig, generator_config: PretrainedConfig, **kwargs) -> PretrainedConfig - added:
variable RAG_CONFIG_DOC
- changed:
constructor ReformerConfig(attention_head_size=…, attn_layers=…, axial_norm_std=…, axial_pos_embds=…, axial_pos_shape=…, axial_pos_embds_dim=…, chunk_size_lm_head=…, chunk_size_feed_forward=…, 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=…… → 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:
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:
variable RETRIBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- added:
class SqueezeBertConfig(PretrainedConfig) - 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:
field SqueezeBertConfig.pretrained_config_archive_map - added:
variable SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- changed:
constructor T5Config(vocab_size=…, n_positions=…, d_model=…, d_kv=…, d_ff=…, num_layers=…, num_heads=…, relative_attention_num_buckets=…, dropout_rate=…, layer_norm_epsilon=…, initializer_factor=…, is_encoder_decoder=…, pad_token_id=…, eos_token_id=…, **kwargs) → 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)
- changed:
constructor TransfoXLConfig(vocab_size=…, cutoffs=…, d_model=…, d_embed=…, n_head=…, d_head=…, d_inner=…, div_val=…, pre_lnorm=…, n_layer=…, tgt_len=…, ext_len=…, mem_len=…, clamp_len=…, same_length=…, proj_share_all_but_first=…, attn_type=…, sample_softmax=…, adaptive=…, tie_weight=…, dropout=…, dropatt=…, untie_r=…, init=…, init_range=…, proj_init_std=…, init_std=…, layer_norm_epsilon=…, eos_to… → 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:
property PretrainedConfig.num_labels - changed:
method PretrainedConfig.save_pretrained(save_directory) → method PretrainedConfig.save_pretrained(save_directory: str) - changed:
method PretrainedConfig.to_dict() → method PretrainedConfig.to_dict() -> Dict[str, Any] - changed:
method PretrainedConfig.to_diff_dict() → method PretrainedConfig.to_diff_dict() -> Dict[str, Any] - changed:
method PretrainedConfig.to_json_file(json_file_path, use_diff=…) → method PretrainedConfig.to_json_file(json_file_path: str, use_diff: bool=…) - changed:
method PretrainedConfig.to_json_string(use_diff=…) → method PretrainedConfig.to_json_string(use_diff: bool=…) -> str - changed:
method PretrainedConfig.update(config_dict: Dict) → method PretrainedConfig.update(config_dict: Dict[str, Any]) - changed:
static method PretrainedConfig.from_dict(config_dict: Dict, **kwargs) -> PretrainedConfig → static method PretrainedConfig.from_dict(config_dict: Dict[str, Any], **kwargs) -> PretrainedConfig - changed:
static method PretrainedConfig.from_pretrained(pretrained_model_name_or_path, **kwargs) -> PretrainedConfig → static method PretrainedConfig.from_pretrained(pretrained_model_name_or_path: str, **kwargs) -> PretrainedConfig - changed:
static method PretrainedConfig.get_config_dict(pretrained_model_name_or_path: str, **kwargs) -> Tuple[Dict, Dict] → static method PretrainedConfig.get_config_dict(pretrained_model_name_or_path: str, **kwargs) -> Tuple[Dict[str, Any], Dict[str, Any]] - added:
field PretrainedConfig.is_composition: bool - added:
property PretrainedConfig.name_or_path -> str - added:
property PretrainedConfig.num_labels -> int - added:
property PretrainedConfig.use_return_dict -> bool
- added:
class XLMProphetNetConfig(ProphetNetConfig) - added:
field XLMProphetNetConfig.model_type - added:
variable XLM_PROPHETNET_PRETRAINED_CONFIG_ARCHIVE_MAP - added:
variable logger
- removed:
function convert_checkpoint_from_disk(checkpoint_path, **config_kwargs)
- added:
function convert_tf2_checkpoint_to_pytorch(tf_checkpoint_path, config_path, pytorch_dump_path) - added:
function load_tf2_weights_in_bert(model, tf_checkpoint_path, config) - added:
variable args - added:
variable logger - added:
variable parser
- added:
function convert_parlai_checkpoint(checkpoint_path, pytorch_dump_folder_path, config_json_path) - added:
function rename_layernorm_keys(sd) - added:
function rename_state_dict_key(k) - added:
variable IGNORE_KEYS - added:
variable PATTERNS - added:
variable args - added:
variable logger - added:
variable parser
- added:
class DPRContextEncoderState(DPRState) - added:
class DPRQuestionEncoderState(DPRState) - added:
class DPRReaderState(DPRState) - added:
class DPRState - added:
constructor DPRState(src_file: Path) - added:
function convert(comp_type: str, src_file: Path, dest_dir: Path) - added:
function load_states_from_checkpoint(model_file: str) -> CheckpointState - added:
method DPRContextEncoderState.load_dpr_model() - added:
method DPRQuestionEncoderState.load_dpr_model() - added:
method DPRReaderState.load_dpr_model() - added:
method DPRState.load_dpr_model() - added:
static method DPRState.from_type(comp_type: str, *args, **kwargs) -> DPRState - added:
variable CheckpointState - added:
variable args - added:
variable dest_dir - added:
variable parser - added:
variable src_file
- added:
function convert_fsmt_checkpoint_to_pytorch(fsmt_checkpoint_path, pytorch_dump_folder_path) - added:
function rewrite_dict_keys(d) - added:
variable args - added:
variable best_score_hparams - added:
variable json_indent - added:
variable org_names - added:
variable parser
- added:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path) - added:
variable args - added:
variable parser
- changed:
function convert(framework: str, model: str, output: str, opset: int, tokenizer: Optional[str]=…, use_external_format: bool=…) → function convert(framework: str, model: str, output: Path, opset: int, tokenizer: Optional[str]=…, use_external_format: bool=…, pipeline_name: str=…) - changed:
function convert_pytorch(nlp: Pipeline, opset: int, output: str, use_external_format: bool) → function convert_pytorch(nlp: Pipeline, opset: int, output: Path, use_external_format: bool) - changed:
function convert_tensorflow(nlp: Pipeline, opset: int, output: str) → function convert_tensorflow(nlp: Pipeline, opset: int, output: Path) - changed:
function load_graph_from_args(framework: str, model: str, tokenizer: Optional[str]=…) -> Pipeline → function load_graph_from_args(pipeline_name: str, framework: str, model: str, tokenizer: Optional[str]=…) -> Pipeline - changed:
function verify(path: str) → function verify(path: Path) - added:
function check_onnxruntime_requirements(minimum_version: Version) - added:
function generate_identified_filename(filename: Path, identifier: str) -> Path - added:
function optimize(onnx_model_path: Path) -> Path - added:
function quantize(onnx_model_path: Path) -> Path - added:
variable ORT_QUANTIZE_MINIMUM_VERSION - added:
variable SUPPORTED_PIPELINES
- added:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path) - added:
variable args - added:
variable parser
- added:
class TatoebaConverter - added:
constructor TatoebaConverter(save_dir=…) - added:
function dedup(lst) - added:
function l2front_matter(langs) - added:
method TatoebaConverter.convert_models(tatoeba_ids, dry_run=…) - added:
method TatoebaConverter.download_metadata() - added:
method TatoebaConverter.expand_group_to_two_letter_codes(grp_name) - added:
method TatoebaConverter.get_tags(code, ref_name) - added:
method TatoebaConverter.get_two_letter_code(three_letter_code) - added:
method TatoebaConverter.resolve_lang_code(r) -> Tuple[List[str], str, str] - added:
method TatoebaConverter.write_model_card(hf_model_id: str, repo_root=…, dry_run=…) -> str - added:
static method TatoebaConverter.make_tatoeba_registry(repo_path=…) - added:
variable DEFAULT_MODEL_DIR - added:
variable DEFAULT_REPO - added:
variable GROUP_MEMBERS - added:
variable ISO_PATH - added:
variable ISO_URL - added:
variable LANG_CODE_PATH - added:
variable LANG_CODE_URL - added:
variable args - added:
variable parser - added:
variable resolver
- removed:
function get_clean_model_id_mapping(multiling_model_ids) - removed:
function save_tokenizer(self, save_directory) - changed:
function convert_all_sentencepiece_models(model_list=…, repo_path=…) → function convert_all_sentencepiece_models(model_list=…, repo_path=…, dest_dir=…) - changed:
function write_model_card(hf_model_name: str, repo_path=…, dry_run=…, model_card_dir=…) -> str → function write_model_card(hf_model_name: str, repo_root=…, save_dir=…, dry_run=…, extra_metadata=…) -> str - added:
function get_system_metadata(repo_root) - added:
function remove_suffix(text: str, suffix: str) - added:
variable DEFAULT_MODEL_DIR - added:
variable DEFAULT_REPO - added:
variable FRONT_MATTER_TEMPLATE
- added:
function convert_fairseq_mbart_checkpoint_from_disk(checkpoint_path, hf_config_path=…) - added:
variable args - added:
variable model - added:
variable parser
- added:
function convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, mobilebert_config_file, pytorch_dump_path) - added:
variable args - added:
variable parser
- added:
function convert_pegasus(tf_weights: dict, cfg_updates: dict) -> PegasusForConditionalGeneration - added:
function convert_pegasus_ckpt_to_pytorch(ckpt_path: str, save_dir: str) - added:
function get_tf_weights_as_numpy(path=…) -> Dict - added:
function rename_state_dict_key(k) - added:
variable PATTERNS - added:
variable args - added:
variable dataset - added:
variable parser
- added:
function convert_prophetnet_checkpoint_to_pytorch(prophetnet_checkpoint_path: str, pytorch_dump_folder_path: str) - added:
variable args - added:
variable logger - added:
variable parser
- added:
class AlbertConverter(SpmConverter) - added:
class BertConverter(Converter) - added:
class BertGenerationConverter(SpmConverter) - added:
class CamembertConverter(SpmConverter) - added:
class Converter - added:
class FunnelConverter(Converter) - added:
class GPT2Converter(Converter) - added:
class HerbertConverter(Converter) - added:
class MBartConverter(SpmConverter) - added:
class OpenAIGPTConverter(Converter) - added:
class PegasusConverter(SpmConverter) - added:
class ReformerConverter(SpmConverter) - added:
class RobertaConverter(Converter) - added:
class SentencePieceExtractor - added:
class SpmConverter(Converter) - added:
class T5Converter(SpmConverter) - added:
class XLMRobertaConverter(SpmConverter) - added:
class XLNetConverter(SpmConverter) - added:
constructor Converter(original_tokenizer) - added:
constructor SentencePieceExtractor(model: str) - added:
constructor SpmConverter(*args) - added:
function check_number_comma(piece: str) -> bool - added:
function convert_slow_tokenizer(transformer_tokenizer) -> Tokenizer - added:
function get_proto(filename: str) - added:
method AlbertConverter.normalizer(proto) - added:
method AlbertConverter.post_processor() - added:
method AlbertConverter.vocab(proto) - added:
method BertConverter.converted() -> Tokenizer - added:
method CamembertConverter.post_processor() - added:
method CamembertConverter.unk_id(proto) - added:
method CamembertConverter.vocab(proto) - added:
method Converter.converted() -> Tokenizer - added:
method FunnelConverter.converted() -> Tokenizer - added:
method GPT2Converter.converted() -> Tokenizer - added:
method HerbertConverter.converted() -> Tokenizer - added:
method MBartConverter.post_processor() - added:
method MBartConverter.unk_id(proto) - added:
method MBartConverter.vocab(proto) - added:
method OpenAIGPTConverter.converted() -> Tokenizer - added:
method PegasusConverter.post_processor() - added:
method PegasusConverter.unk_id(proto) - added:
method PegasusConverter.vocab(proto) - added:
method RobertaConverter.converted() -> Tokenizer - added:
method SentencePieceExtractor.extract() -> Tuple[Dict[str, int], List[Tuple]] - added:
method SpmConverter.converted() -> Tokenizer - added:
method SpmConverter.normalizer(proto) - added:
method SpmConverter.post_processor() - added:
method SpmConverter.tokenizer(proto) - added:
method SpmConverter.unk_id(proto) - added:
method SpmConverter.vocab(proto) - added:
method T5Converter.post_processor() - added:
method T5Converter.vocab(proto) - added:
method XLMRobertaConverter.post_processor() - added:
method XLMRobertaConverter.unk_id(proto) - added:
method XLMRobertaConverter.vocab(proto) - added:
method XLNetConverter.normalizer(proto) - added:
method XLNetConverter.post_processor() - added:
method XLNetConverter.vocab(proto) - added:
variable SLOW_TO_FAST_CONVERTERS
- added:
function convert_slow_checkpoint_to_fast(tokenizer_name, checkpoint_name, dump_path, force_download) - added:
variable TOKENIZER_CLASSES - added:
variable args - added:
variable logger - added:
variable parser
- added:
function convert_tf_checkpoint_to_pytorch(tf_hub_path, pytorch_dump_path, is_encoder_named_decoder, vocab_size, is_encoder) - added:
variable args - added:
variable parser
- removed:
export is_sklearn_available
- removed:
class DataCollator(ABC) - removed:
class DefaultDataCollator(DataCollator) - removed:
method DataCollator.collate_batch() -> Dict[str, torch.Tensor] - removed:
method DataCollatorForLanguageModeling.collate_batch(examples: List[torch.Tensor]) -> Dict[str, torch.Tensor] - removed:
method DefaultDataCollator.collate_batch(features: List[InputDataClass]) -> Dict[str, torch.Tensor] - changed:
class DataCollatorForLanguageModeling(DataCollator) → class DataCollatorForLanguageModeling - changed:
field DataCollatorForLanguageModeling.tokenizer: PreTrainedTokenizer → field DataCollatorForLanguageModeling.tokenizer: PreTrainedTokenizerBase - changed:
method DataCollatorForLanguageModeling.mask_tokens(inputs: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor] → method DataCollatorForLanguageModeling.mask_tokens(inputs: torch.Tensor, special_tokens_mask: Optional[torch.Tensor]=…) -> Tuple[torch.Tensor, torch.Tensor] - added:
class DataCollatorForPermutationLanguageModeling - added:
class DataCollatorForSOP(DataCollatorForLanguageModeling) - added:
class DataCollatorForTokenClassification - added:
class DataCollatorForWholeWordMask(DataCollatorForLanguageModeling) - added:
class DataCollatorWithPadding - added:
constructor DataCollatorForSOP(*args, **kwargs) - added:
field DataCollatorForPermutationLanguageModeling.max_span_length: int - added:
field DataCollatorForPermutationLanguageModeling.plm_probability: float - added:
field DataCollatorForPermutationLanguageModeling.tokenizer: PreTrainedTokenizerBase - added:
field DataCollatorForTokenClassification.label_pad_token_id: int - added:
field DataCollatorForTokenClassification.max_length: Optional[int] - added:
field DataCollatorForTokenClassification.pad_to_multiple_of: Optional[int] - added:
field DataCollatorForTokenClassification.padding: Union[bool, str, PaddingStrategy] - added:
field DataCollatorForTokenClassification.tokenizer: PreTrainedTokenizerBase - added:
field DataCollatorWithPadding.max_length: Optional[int] - added:
field DataCollatorWithPadding.pad_to_multiple_of: Optional[int] - added:
field DataCollatorWithPadding.padding: Union[bool, str, PaddingStrategy] - added:
field DataCollatorWithPadding.tokenizer: PreTrainedTokenizerBase - added:
function default_data_collator(features: List[InputDataClass]) -> Dict[str, torch.Tensor] - added:
function tolist(x: Union[List[Any], torch.Tensor]) - added:
method DataCollatorForPermutationLanguageModeling.mask_tokens(inputs: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor] - added:
method DataCollatorForSOP.mask_tokens(inputs: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor] - added:
method DataCollatorForWholeWordMask.mask_tokens(inputs: torch.Tensor, mask_labels: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor] - added:
variable DataCollator
- added:
export LineByLineWithRefDataset - added:
export LineByLineWithSOPTextDataset - added:
export SquadDataTrainingArguments - added:
export SquadDataset - added:
export TextDatasetForNextSentencePrediction
- changed:
constructor GlueDataset(args: GlueDataTrainingArguments, tokenizer: PreTrainedTokenizer, limit_length: Optional[int]=…, mode: Union[str, Split]=…) → constructor GlueDataset(args: GlueDataTrainingArguments, tokenizer: PreTrainedTokenizerBase, limit_length: Optional[int]=…, mode: Union[str, Split]=…, cache_dir: Optional[str]=…)
- changed:
constructor TextDataset(tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, overwrite_cache=…) → constructor TextDataset(tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, overwrite_cache=…, cache_dir: Optional[str]=…) - added:
class LineByLineWithRefDataset(Dataset) - added:
class LineByLineWithSOPTextDataset(Dataset) - added:
class TextDatasetForNextSentencePrediction(Dataset) - added:
constructor LineByLineWithRefDataset(tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, ref_path: str) - added:
constructor LineByLineWithSOPTextDataset(tokenizer: PreTrainedTokenizer, file_dir: str, block_size: int) - added:
constructor TextDatasetForNextSentencePrediction(tokenizer: PreTrainedTokenizer, file_path: str, block_size: int, overwrite_cache=…, short_seq_probability=…, nsp_probability=…) - added:
method LineByLineWithSOPTextDataset.create_examples_from_document(document, block_size, tokenizer, short_seq_prob=…) - added:
method TextDatasetForNextSentencePrediction.create_examples_from_document(document: List[List[int]], doc_index: int) - added:
variable DEPRECATION_WARNING
- added:
class Split(Enum) - added:
class SquadDataTrainingArguments - added:
class SquadDataset(Dataset) - added:
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]=…) - added:
field Split.dev - added:
field Split.train - added:
field SquadDataTrainingArguments.data_dir: str - added:
field SquadDataTrainingArguments.doc_stride: int - added:
field SquadDataTrainingArguments.lang_id: int - added:
field SquadDataTrainingArguments.max_answer_length: int - added:
field SquadDataTrainingArguments.max_query_length: int - added:
field SquadDataTrainingArguments.max_seq_length: int - added:
field SquadDataTrainingArguments.model_type: str - added:
field SquadDataTrainingArguments.n_best_size: int - added:
field SquadDataTrainingArguments.null_score_diff_threshold: float - added:
field SquadDataTrainingArguments.overwrite_cache: bool - added:
field SquadDataTrainingArguments.threads: int - added:
field SquadDataTrainingArguments.version_2_with_negative: bool - added:
field SquadDataset.args: SquadDataTrainingArguments - added:
field SquadDataset.features: List[SquadFeatures] - added:
field SquadDataset.is_language_sensitive: bool - added:
field SquadDataset.mode: Split - added:
variable MODEL_CONFIG_CLASSES - added:
variable MODEL_TYPES - added:
variable logger
- removed:
function is_sklearn_available() - added:
export is_sklearn_available - added:
export requires_sklearn - added:
variable DEPRECATION_WARNING
- added:
constructor ColaProcessor(*args, **kwargs) - added:
constructor MnliMismatchedProcessor(*args, **kwargs) - added:
constructor MnliProcessor(*args, **kwargs) - added:
constructor MrpcProcessor(*args, **kwargs) - added:
constructor QnliProcessor(*args, **kwargs) - added:
constructor QqpProcessor(*args, **kwargs) - added:
constructor RteProcessor(*args, **kwargs) - added:
constructor Sst2Processor(*args, **kwargs) - added:
constructor StsbProcessor(*args, **kwargs) - added:
constructor WnliProcessor(*args, **kwargs) - added:
variable DEPRECATION_WARNING
- changed:
class SquadExample(object) → class SquadExample - changed:
class SquadFeatures(object) → class SquadFeatures - changed:
class SquadResult(object) → class SquadResult - changed:
function squad_convert_example_to_features(example, max_seq_length, doc_stride, max_query_length, is_training) → function squad_convert_example_to_features(example, max_seq_length, doc_stride, max_query_length, padding_strategy, is_training) - changed:
function squad_convert_example_to_features_init(tokenizer_for_convert) → function squad_convert_example_to_features_init(tokenizer_for_convert: PreTrainedTokenizerBase) - changed:
function squad_convert_examples_to_features(examples, tokenizer, max_seq_length, doc_stride, max_query_length, is_training, return_dataset=…, threads=…, tqdm_enabled=…) → function squad_convert_examples_to_features(examples, tokenizer, max_seq_length, doc_stride, max_query_length, is_training, padding_strategy=…, return_dataset=…, threads=…, tqdm_enabled=…) - added:
variable MULTI_SEP_TOKENS_TOKENIZERS_SET
- added:
class GenerationUtilsTest(unittest.TestCase) - added:
method GenerationUtilsTest.test_postprocess_next_token_scores() - added:
method GenerationUtilsTest.test_postprocess_next_token_scores_large_bad_words_list() - added:
property GenerationUtilsTest.config - added:
property GenerationUtilsTest.model
- removed:
function add_start_docstrings_to_callable(*docstr) - changed:
function cached_path(url_or_filename, cache_dir=…, force_download=…, proxies=…, resume_download=…, user_agent=…, extract_compressed_file=…, force_extract=…, local_files_only=…) -> Optional[str] → 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] - changed:
function get_from_cache(url, cache_dir=…, force_download=…, proxies=…, etag_timeout=…, resume_download=…, user_agent=…, local_files_only=…) -> Optional[str] → 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] - changed:
function hf_bucket_url(model_id: str, filename: str, use_cdn=…) -> str → function hf_bucket_url(model_id: str, filename: str, revision: Optional[str]=…, mirror=…) -> str - changed:
function http_get(url, temp_file, proxies=…, resume_size=…, user_agent=…) → function http_get(url: str, temp_file: BinaryIO, proxies=…, resume_size=…, user_agent: Union[Dict, str, None]=…) - changed:
function url_to_filename(url, etag=…) → function url_to_filename(url: str, etag: Optional[str]=…) -> str - added:
class ModelOutput(OrderedDict) - added:
function add_code_sample_docstrings(*docstr, tokenizer_class=…, checkpoint=…, output_type=…, config_class=…, mask=…) - added:
function add_start_docstrings_to_model_forward(*docstr) - added:
function http_user_agent(user_agent: Union[Dict, str, None]=…) -> str - added:
function is_apex_available() - added:
function is_datasets_available() - added:
function is_faiss_available() - added:
function is_flax_available() - added:
function is_in_notebook() - added:
function is_psutil_available() - added:
function is_py3nvml_available() - added:
function is_sentencepiece_available() - added:
function is_sklearn_available() - added:
function is_tensor(x) - added:
function is_tokenizers_available() - added:
function is_torch_tpu_available() - added:
function replace_return_docstrings(output_type=…, config_class=…) - added:
function requires_datasets(obj) - added:
function requires_faiss(obj) - added:
function requires_flax(obj) - added:
function requires_pytorch(obj) - added:
function requires_sentencepiece(obj) - added:
function requires_sklearn(obj) - added:
function requires_tf(obj) - added:
function requires_tokenizers(obj) - added:
function torch_only_method(fn) - added:
method ModelOutput.pop(*args, **kwargs) - added:
method ModelOutput.setdefault(*args, **kwargs) - added:
method ModelOutput.to_tuple() -> Tuple[Any] - added:
method ModelOutput.update(*args, **kwargs) - added:
variable DATASETS_IMPORT_ERROR - added:
variable ENV_VARS_TRUE_AND_AUTO_VALUES - added:
variable ENV_VARS_TRUE_VALUES - added:
variable FAISS_IMPORT_ERROR - added:
variable FLAX_IMPORT_ERROR - added:
variable HUGGINGFACE_CO_PREFIX - added:
variable PRESET_MIRROR_DICT - added:
variable PT_BASE_MODEL_SAMPLE - added:
variable PT_CAUSAL_LM_SAMPLE - added:
variable PT_MASKED_LM_SAMPLE - added:
variable PT_MULTIPLE_CHOICE_SAMPLE - added:
variable PT_QUESTION_ANSWERING_SAMPLE - added:
variable PT_RETURN_INTRODUCTION - added:
variable PT_SEQUENCE_CLASSIFICATION_SAMPLE - added:
variable PT_TOKEN_CLASSIFICATION_SAMPLE - added:
variable PYTORCH_IMPORT_ERROR - added:
variable SENTENCEPIECE_IMPORT_ERROR - added:
variable SENTENCEPIECE_UNDERLINE - added:
variable SKLEARN_IMPORT_ERROR - added:
variable SPIECE_UNDERLINE - added:
variable TENSORFLOW_IMPORT_ERROR - added:
variable TF_BASE_MODEL_SAMPLE - added:
variable TF_CAUSAL_LM_SAMPLE - added:
variable TF_MASKED_LM_SAMPLE - added:
variable TF_MULTIPLE_CHOICE_SAMPLE - added:
variable TF_QUESTION_ANSWERING_SAMPLE - added:
variable TF_RETURN_INTRODUCTION - added:
variable TF_SEQUENCE_CLASSIFICATION_SAMPLE - added:
variable TF_TOKEN_CLASSIFICATION_SAMPLE - added:
variable TOKENIZERS_IMPORT_ERROR - added:
variable USE_JAX - added:
variable get_ipython
- removed:
variable context - removed:
variable generated - removed:
variable model - removed:
variable past - removed:
variable sequence - removed:
variable tokenizer
- added:
class BeamHypotheses - added:
class BeamScorer(ABC) - added:
class BeamSearchScorer(BeamScorer) - added:
constructor BeamHypotheses(num_beams: int, max_length: int, length_penalty: float, early_stopping: bool) - added:
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]=…) - added:
method BeamHypotheses.add(hyp: torch.LongTensor, sum_logprobs: float) - added:
method BeamHypotheses.is_done(best_sum_logprobs: float, cur_len: int) -> bool - added:
method BeamScorer.finalize(input_ids: torch.LongTensor, next_scores: torch.FloatTensor, next_tokens: torch.LongTensor, next_indices: torch.LongTensor, **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, pad_token_id: Optional[int]=…, eos_token_id: Optional[int]=…) -> torch.LongTensor - added:
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] - added:
property BeamSearchScorer.is_done -> bool - added:
variable FINALIZE_INPUTS_DOCSTRING - added:
variable PROCESS_INPUTS_DOCSTRING
- added:
class LogitsProcessor(ABC) - added:
class LogitsProcessorList(list) - added:
class LogitsWarper(ABC) - added:
class MinLengthLogitsProcessor(LogitsProcessor) - added:
class NoBadWordsLogitsProcessor(LogitsProcessor) - added:
class NoRepeatNGramLogitsProcessor(LogitsProcessor) - added:
class RepetitionPenaltyLogitsProcessor(LogitsProcessor) - added:
class TemperatureLogitsWarper(LogitsWarper) - added:
class TopKLogitsWarper(LogitsWarper) - added:
class TopPLogitsWarper(LogitsWarper) - added:
constructor MinLengthLogitsProcessor(min_length: int, eos_token_id: int) - added:
constructor NoBadWordsLogitsProcessor(bad_words_ids: Iterable[Iterable[int]], eos_token_id: int) - added:
constructor NoRepeatNGramLogitsProcessor(ngram_size: int) - added:
constructor RepetitionPenaltyLogitsProcessor(penalty: float) - 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:
variable LOGITS_PROCESSOR_INPUTS_DOCSTRING
- added:
class BeamHypotheses(object) - added:
class TFGenerationMixin - added:
constructor BeamHypotheses(num_beams, max_length, length_penalty, early_stopping) - added:
function calc_banned_bad_words_ids(prev_input_ids, bad_words_ids) - added:
function calc_banned_ngram_tokens(prev_input_ids, num_hypos, no_repeat_ngram_size, cur_len) - added:
function sample_without_replacement(logits, num_samples) - added:
function scatter_values_on_batch_indices(values, batch_indices) - added:
function set_tensor_by_indices_to_value(tensor, indices, value) - added:
function shape_list(x) - added:
function tf_top_k_top_p_filtering(logits, top_k=…, top_p=…, filter_value=…, min_tokens_to_keep=…) - added:
method BeamHypotheses.add(hyp, sum_logprobs) - added:
method BeamHypotheses.is_done(best_sum_logprobs, cur_len) - added:
method TFGenerationMixin.adjust_logits_during_generation(logits, **kwargs) - added:
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=…) - added:
method TFGenerationMixin.prepare_inputs_for_generation(inputs, **kwargs) - added:
variable logger
- added:
class GenerationMixin - added:
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 - added:
method GenerationMixin.adjust_logits_during_generation(logits: torch.FloatTensor, **kwargs) -> torch.FloatTensor - added:
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) - added:
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) - added:
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:… - added:
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) - added:
method GenerationMixin.prepare_inputs_for_generation(input_ids: torch.LongTensor, **kwargs) -> Dict[str, Any] - added:
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) - added:
variable logger
- removed:
class S3Object - removed:
constructor S3Object(key: str, etag: str, lastModified: str, size: int, rfilename: str, **kwargs) - changed:
constructor ModelInfo(modelId: str, key: str, author: Optional[str]=…, downloads: Optional[int]=…, tags: List[str]=…, siblings: List[Dict]=…, **kwargs) → constructor ModelInfo(modelId: Optional[str]=…, author: Optional[str]=…, downloads: Optional[int]=…, tags: List[str]=…, pipeline_tag: Optional[str]=…, siblings: Optional[List[Dict]]=…, **kwargs) - added:
class ModelSibling - added:
class RepoObj - added:
constructor ModelSibling(rfilename: str, **kwargs) - added:
constructor RepoObj(filename: str, lastModified: str, commit: str, size: int, **kwargs) - added:
method HfApi.create_repo(token: str, name: str, organization: Optional[str]=…) -> str - added:
method HfApi.delete_repo(token: str, name: str, organization: Optional[str]=…) - added:
method HfApi.list_repos_objs(token: str, organization: Optional[str]=…) -> List[S3Obj]
- changed:
method HfArgumentParser.parse_args_into_dataclasses(args=…, return_remaining_strings=…, look_for_args_file=…) -> Tuple[DataClass, ...] → method HfArgumentParser.parse_args_into_dataclasses(args=…, return_remaining_strings=…, look_for_args_file=…, args_filename=…) -> Tuple[DataClass, ...] - added:
method HfArgumentParser.parse_dict(args: dict) -> Tuple[DataClass, ...]
- added:
class AzureMLCallback(TrainerCallback) - added:
class CometCallback(TrainerCallback) - added:
class MLflowCallback(TrainerCallback) - added:
class TensorBoardCallback(TrainerCallback) - added:
class WandbCallback(TrainerCallback) - added:
constructor AzureMLCallback(azureml_run=…) - added:
constructor CometCallback() - added:
constructor MLflowCallback() - added:
constructor TensorBoardCallback(tb_writer=…) - added:
constructor WandbCallback() - added:
field MLflowCallback.MAX_LOG_SIZE - added:
function default_hp_search_backend() - added:
function hp_params(trial) - added:
function is_azureml_available() - added:
function is_comet_available() - added:
function is_mlflow_available() - added:
function is_optuna_available() - added:
function is_ray_available() - added:
function is_tensorboard_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:
method AzureMLCallback.on_init_end(args, state, control, **kwargs) - added:
method AzureMLCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method CometCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method CometCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method CometCallback.setup(args, state, model) - added:
method MLflowCallback.on_log(args, state, control, logs, model=…, **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 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 WandbCallback.on_log(args, state, control, model=…, logs=…, **kwargs) - added:
method WandbCallback.on_train_begin(args, state, control, model=…, **kwargs) - added:
method WandbCallback.setup(args, state, model, reinit, **kwargs) - added:
variable logger
- removed:
method AlbertForMaskedLM.tie_weights() - removed:
method AlbertForPreTraining.tie_weights() - changed:
class AlbertAttention(BertSelfAttention) → class AlbertAttention(nn.Module) - changed:
class AlbertEmbeddings(BertEmbeddings) → class AlbertEmbeddings(nn.Module) - changed:
constructor AlbertModel(config) → constructor AlbertModel(config, add_pooling_layer=…) - changed:
method AlbertAttention.forward(input_ids, attention_mask=…, head_mask=…) → method AlbertAttention.forward(input_ids, attention_mask=…, head_mask=…, output_attentions=…) - changed:
method AlbertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…) → 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) - changed:
method AlbertForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…, sentence_order_label=…) → 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) - changed:
method AlbertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…) → 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=…) - changed:
method AlbertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method AlbertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method AlbertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method AlbertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method AlbertLayer.forward(hidden_states, attention_mask=…, head_mask=…) → method AlbertLayer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…) - changed:
method AlbertLayerGroup.forward(hidden_states, attention_mask=…, head_mask=…) → method AlbertLayerGroup.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…) - changed:
method AlbertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…) → method AlbertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method AlbertTransformer.forward(hidden_states, attention_mask=…, head_mask=…) → method AlbertTransformer.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
class AlbertForMultipleChoice(AlbertPreTrainedModel) - added:
class AlbertForPreTrainingOutput(ModelOutput) - added:
constructor AlbertForMultipleChoice(config) - added:
field AlbertForMaskedLM.authorized_unexpected_keys - added:
field AlbertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field AlbertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field AlbertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - added:
field AlbertForPreTrainingOutput.prediction_logits: torch.FloatTensor - added:
field AlbertForPreTrainingOutput.sop_logits: torch.FloatTensor - added:
field AlbertForQuestionAnswering.authorized_unexpected_keys - added:
field AlbertForTokenClassification.authorized_unexpected_keys - added:
field AlbertPreTrainedModel.authorized_missing_keys - added:
method AlbertAttention.transpose_for_scores(x) - added:
method AlbertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method AlbertForMaskedLM.get_input_embeddings() - added:
method AlbertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method AlbertForPreTraining.get_input_embeddings() - added:
method AlbertLayer.ff_chunk(attention_output)
- added:
class AutoModelForCausalLM - added:
class AutoModelForMaskedLM - added:
class AutoModelForNextSentencePrediction - added:
class AutoModelForSeq2SeqLM - added:
constructor AutoModelForCausalLM() - added:
constructor AutoModelForMaskedLM() - added:
constructor AutoModelForNextSentencePrediction() - added:
constructor AutoModelForSeq2SeqLM() - added:
static method AutoModelForCausalLM.from_config(config) - added:
static method AutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method AutoModelForMaskedLM.from_config(config) - added:
static method AutoModelForMaskedLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method AutoModelForNextSentencePrediction.from_config(config) - added:
static method AutoModelForNextSentencePrediction.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method AutoModelForSeq2SeqLM.from_config(config) - added:
static method AutoModelForSeq2SeqLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable AUTO_MODEL_PRETRAINED_DOCSTRING - added:
variable MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable MODEL_FOR_MASKED_LM_MAPPING - added:
variable MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
variable MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
- removed:
class SelfAttention(nn.Module) - removed:
constructor SelfAttention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…) - removed:
method BartForConditionalGeneration.prepare_logits_for_generation(logits, cur_len, max_length) - removed:
method SelfAttention.forward(query, key: Optional[Tensor], key_padding_mask: Optional[Tensor]=…, layer_state: Optional[Dict[str, Optional[Tensor]]]=…, attn_mask: Optional[Tensor]=…, need_weights=…) -> Tuple[Tensor, Optional[Tensor]] - changed:
constructor LearnedPositionalEmbedding(num_embeddings: int, embedding_dim: int, padding_idx: int) → constructor LearnedPositionalEmbedding(num_embeddings: int, embedding_dim: int, padding_idx: int, offset) - changed:
method BartDecoder.forward(input_ids, encoder_hidden_states, encoder_padding_mask, decoder_padding_mask, decoder_causal_mask, decoder_cached_states=…, use_cache=…, **unused) → 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) - changed:
method BartEncoder.forward(input_ids, attention_mask=…) → method BartEncoder.forward(input_ids, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method BartForConditionalGeneration.forward(input_ids, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, decoder_cached_states=…, lm_labels=…, use_cache=…, **unused) → 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) - changed:
method BartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past, attention_mask, use_cache, **kwargs) → method BartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - changed:
method BartForSequenceClassification.forward(input_ids, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, labels=…) → 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=…) - changed:
method BartModel.forward(input_ids, attention_mask=…, decoder_input_ids=…, encoder_outputs: Optional[Tuple]=…, decoder_attention_mask=…, decoder_cached_states=…, use_cache=…) → 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) - changed:
method DecoderLayer.forward(x, encoder_hidden_states, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…) → method DecoderLayer.forward(x, encoder_hidden_states, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…, output_attentions=…) - changed:
method EncoderLayer.forward(x, encoder_padding_mask) → method EncoderLayer.forward(x, encoder_padding_mask, output_attentions=…) - changed:
method LearnedPositionalEmbedding.forward(input, use_cache=…) → method LearnedPositionalEmbedding.forward(input_ids, use_cache=…) - added:
class Attention(nn.Module) - added:
class BartForQuestionAnswering(PretrainedBartModel) - added:
constructor Attention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…) - added:
constructor BartForQuestionAnswering(config) - added:
field BartForConditionalGeneration.authorized_missing_keys - added:
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]] - added:
method BartForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
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:
class TransformerBeamSearch(nn.Module) - removed:
constructor TransformerBeamSearch(model, tokenizer, batch_size, beam_size, min_length, max_length, alpha=…, block_repeating_trigram=…) - removed:
function tile(x, count, dim=…) - removed:
method TransformerBeamSearch.enforce_max_length() - removed:
method TransformerBeamSearch.enforce_min_length() - removed:
method TransformerBeamSearch.forward(encoder_input_ids, **kwargs) - removed:
method TransformerBeamSearch.length_penalty() - removed:
method TransformerBeamSearch.remove_repeating_trigrams(log_probabilities, _B) - removed:
method TransformerBeamSearch.step(log_probabilities)
- removed:
function mish(x) - removed:
variable ACT2FN - removed:
variable BertLayerNorm - changed:
constructor BertModel(config) → constructor BertModel(config, add_pooling_layer=…) - changed:
method BertAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…) → method BertAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - changed:
method BertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…) → method BertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method BertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…, encoder_hidden_states=…, encoder_attention_mask=…, lm_labels=…) → 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) - changed:
method BertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method BertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method BertForNextSentencePrediction.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, next_sentence_label=…) → 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) - changed:
method BertForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…, next_sentence_label=…) → 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) - changed:
method BertForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…) → 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=…) - changed:
method BertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method BertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method BertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method BertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method BertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…) → method BertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - changed:
method BertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…) → 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=…) - changed:
method BertSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…) → method BertSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
class BertForPreTrainingOutput(ModelOutput) - added:
class BertLMHeadModel(BertPreTrainedModel) - added:
constructor BertLMHeadModel(config) - added:
field BertForMaskedLM.authorized_missing_keys - added:
field BertForMaskedLM.authorized_unexpected_keys - added:
field BertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - added:
field BertForPreTrainingOutput.prediction_logits: torch.FloatTensor - added:
field BertForPreTrainingOutput.seq_relationship_logits: torch.FloatTensor - added:
field BertForQuestionAnswering.authorized_unexpected_keys - added:
field BertForTokenClassification.authorized_unexpected_keys - added:
field BertLMHeadModel.authorized_missing_keys - added:
field BertLMHeadModel.authorized_unexpected_keys - added:
field BertPreTrainedModel.authorized_missing_keys - added:
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=…) - added:
method BertLMHeadModel.get_output_embeddings() - added:
method BertLMHeadModel.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method BertLayer.feed_forward_chunk(attention_output)
- added:
class BertGenerationDecoder(BertGenerationPreTrainedModel) - added:
class BertGenerationEmbeddings(nn.Module) - added:
class BertGenerationEncoder(BertGenerationPreTrainedModel) - added:
class BertGenerationOnlyLMHead(nn.Module) - added:
class BertGenerationPreTrainedModel(PreTrainedModel) - added:
constructor BertGenerationDecoder(config) - added:
constructor BertGenerationEmbeddings(config) - added:
constructor BertGenerationEncoder(config) - added:
constructor BertGenerationOnlyLMHead(config) - added:
field BertGenerationPreTrainedModel.authorized_missing_keys - added:
field BertGenerationPreTrainedModel.base_model_prefix - added:
field BertGenerationPreTrainedModel.config_class - 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=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, encoder_hidden_states=…, encoder_attention_mask=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method BertGenerationDecoder.get_output_embeddings() - added:
method BertGenerationDecoder.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method BertGenerationEmbeddings.forward(input_ids=…, position_ids=…, inputs_embeds=…) - added:
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=…) - added:
method BertGenerationEncoder.get_input_embeddings() - added:
method BertGenerationEncoder.set_input_embeddings(value) - added:
method BertGenerationOnlyLMHead.forward(hidden_states) - added:
variable BERT_GENERATION_INPUTS_DOCSTRING - added:
variable BERT_GENERATION_START_DOCSTRING - added:
variable logger
- added:
class BlenderbotForConditionalGeneration(BartForConditionalGeneration) - added:
field BlenderbotForConditionalGeneration.config_class - added:
method BlenderbotForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
variable BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable BLENDER_START_DOCSTRING
- added:
class CamembertForCausalLM(RobertaForCausalLM) - added:
field CamembertForCausalLM.config_class
- changed:
constructor EncoderLayer(d_model_size, num_heads, dff, rate=…, output_attentions=…) → constructor EncoderLayer(d_model_size, num_heads, dff, rate=…) - changed:
constructor MultiHeadAttention(d_model_size, num_heads, output_attentions=…) → constructor MultiHeadAttention(d_model_size, num_heads) - changed:
method CTRLLMHeadModel.forward(input_ids=…, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…) → 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) - changed:
method CTRLLMHeadModel.prepare_inputs_for_generation(input_ids, past, **kwargs) → method CTRLLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, **kwargs) - changed:
method CTRLModel.forward(input_ids=…, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…) → 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) - changed:
method EncoderLayer.forward(x, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…) → method EncoderLayer.forward(x, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…, output_attentions=…) - changed:
method MultiHeadAttention.forward(v, k, q, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…) → method MultiHeadAttention.forward(v, k, q, mask, layer_past=…, attention_mask=…, head_mask=…, use_cache=…, output_attentions=…) - added:
method MultiHeadAttention.prune_heads(heads)
- added:
class ContextPooler(nn.Module) - added:
class DebertaAttention(nn.Module) - added:
class DebertaEmbeddings(nn.Module) - added:
class DebertaEncoder(nn.Module) - added:
class DebertaForSequenceClassification(DebertaPreTrainedModel) - added:
class DebertaIntermediate(nn.Module) - added:
class DebertaLayer(nn.Module) - added:
class DebertaLayerNorm(nn.Module) - added:
class DebertaModel(DebertaPreTrainedModel) - added:
class DebertaOutput(nn.Module) - added:
class DebertaPreTrainedModel(PreTrainedModel) - added:
class DebertaSelfOutput(nn.Module) - added:
class DisentangledSelfAttention(torch.nn.Module) - added:
class DropoutContext(object) - added:
class StableDropout(torch.nn.Module) - added:
class XDropout(torch.autograd.Function) - added:
class XSoftmax(torch.autograd.Function) - added:
constructor ContextPooler(config) - added:
constructor DebertaAttention(config) - added:
constructor DebertaEmbeddings(config) - added:
constructor DebertaEncoder(config) - added:
constructor DebertaForSequenceClassification(config) - added:
constructor DebertaIntermediate(config) - added:
constructor DebertaLayer(config) - added:
constructor DebertaLayerNorm(size, eps=…) - added:
constructor DebertaModel(config) - added:
constructor DebertaOutput(config) - added:
constructor DebertaSelfOutput(config) - added:
constructor DisentangledSelfAttention(config) - added:
constructor DropoutContext() - added:
constructor StableDropout(drop_prob) - added:
field DebertaPreTrainedModel.authorized_missing_keys - added:
field DebertaPreTrainedModel.base_model_prefix - added:
field DebertaPreTrainedModel.config_class - added:
function build_relative_position(query_size, key_size, device) - added:
function c2p_dynamic_expand(c2p_pos, query_layer, relative_pos) - added:
function get_mask(input, local_context) - added:
function p2c_dynamic_expand(c2p_pos, query_layer, key_layer) - added:
function pos_dynamic_expand(pos_index, p2c_att, key_layer) - added:
method ContextPooler.forward(hidden_states, mask=…) - added:
method DebertaAttention.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - added:
method DebertaEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, mask=…, inputs_embeds=…) - added:
method DebertaEncoder.forward(hidden_states, attention_mask, output_hidden_states=…, output_attentions=…, query_states=…, relative_pos=…, return_dict=…) - added:
method DebertaEncoder.get_attention_mask(attention_mask) - added:
method DebertaEncoder.get_rel_embedding() - added:
method DebertaEncoder.get_rel_pos(hidden_states, query_states=…, relative_pos=…) - added:
method DebertaForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method DebertaForSequenceClassification.get_input_embeddings() - added:
method DebertaForSequenceClassification.set_input_embeddings(new_embeddings) - added:
method DebertaIntermediate.forward(hidden_states) - added:
method DebertaLayer.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - added:
method DebertaLayerNorm.forward(hidden_states) - added:
method DebertaModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method DebertaModel.get_input_embeddings() - added:
method DebertaModel.set_input_embeddings(new_embeddings) - added:
method DebertaOutput.forward(hidden_states, input_tensor) - added:
method DebertaSelfOutput.forward(hidden_states, input_tensor) - added:
method DisentangledSelfAttention.disentangled_att_bias(query_layer, key_layer, relative_pos, rel_embeddings, scale_factor) - added:
method DisentangledSelfAttention.forward(hidden_states, attention_mask, return_att=…, query_states=…, relative_pos=…, rel_embeddings=…) - added:
method DisentangledSelfAttention.transpose_for_scores(x) - added:
method StableDropout.clear_context() - added:
method StableDropout.forward(x) - added:
method StableDropout.get_context() - added:
method StableDropout.init_context(reuse_mask=…, scale=…) - added:
property ContextPooler.output_dim - added:
static method XDropout.backward(ctx, grad_output) - added:
static method XDropout.forward(ctx, input, local_ctx) - added:
static method XSoftmax.backward(self, grad_output) - added:
static method XSoftmax.forward(self, input, mask, dim) - added:
variable DEBERTA_INPUTS_DOCSTRING - added:
variable DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DEBERTA_START_DOCSTRING - added:
variable logger
- changed:
method DistilBertForMaskedLM.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…) → method DistilBertForMaskedLM.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…, **kwargs) - changed:
method DistilBertForQuestionAnswering.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…) → method DistilBertForQuestionAnswering.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method DistilBertForSequenceClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…) → method DistilBertForSequenceClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method DistilBertForTokenClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…) → method DistilBertForTokenClassification.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method DistilBertModel.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…) → method DistilBertModel.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method MultiHeadSelfAttention.forward(query, key, value, mask, head_mask=…) → method MultiHeadSelfAttention.forward(query, key, value, mask, head_mask=…, output_attentions=…) - changed:
method Transformer.forward(x, attn_mask=…, head_mask=…) → method Transformer.forward(x, attn_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method TransformerBlock.forward(x, attn_mask=…, head_mask=…) → method TransformerBlock.forward(x, attn_mask=…, head_mask=…, output_attentions=…) - added:
class DistilBertForMultipleChoice(DistilBertPreTrainedModel) - added:
constructor DistilBertForMultipleChoice(config) - added:
method DistilBertForMultipleChoice.forward(input_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FFN.ff_chunk(input)
- added:
class DPRContextEncoder(DPRPretrainedContextEncoder) - added:
class DPRContextEncoderOutput(ModelOutput) - added:
class DPREncoder(PreTrainedModel) - added:
class DPRPretrainedContextEncoder(PreTrainedModel) - added:
class DPRPretrainedQuestionEncoder(PreTrainedModel) - added:
class DPRPretrainedReader(PreTrainedModel) - added:
class DPRQuestionEncoder(DPRPretrainedQuestionEncoder) - added:
class DPRQuestionEncoderOutput(ModelOutput) - added:
class DPRReader(DPRPretrainedReader) - added:
class DPRReaderOutput(ModelOutput) - added:
class DPRSpanPredictor(PreTrainedModel) - added:
constructor DPRContextEncoder(config: DPRConfig) - added:
constructor DPREncoder(config: DPRConfig) - added:
constructor DPRQuestionEncoder(config: DPRConfig) - added:
constructor DPRReader(config: DPRConfig) - added:
constructor DPRSpanPredictor(config: DPRConfig) - added:
field DPRContextEncoderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field DPRContextEncoderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field DPRContextEncoderOutput.pooler_output: torch.FloatTensor - added:
field DPREncoder.base_model_prefix - added:
field DPRPretrainedContextEncoder.authorized_missing_keys - added:
field DPRPretrainedContextEncoder.base_model_prefix - added:
field DPRPretrainedContextEncoder.config_class - added:
field DPRPretrainedContextEncoder.load_tf_weights - added:
field DPRPretrainedQuestionEncoder.authorized_missing_keys - added:
field DPRPretrainedQuestionEncoder.base_model_prefix - added:
field DPRPretrainedQuestionEncoder.config_class - added:
field DPRPretrainedQuestionEncoder.load_tf_weights - added:
field DPRPretrainedReader.authorized_missing_keys - added:
field DPRPretrainedReader.base_model_prefix - added:
field DPRPretrainedReader.config_class - added:
field DPRPretrainedReader.load_tf_weights - added:
field DPRQuestionEncoderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field DPRQuestionEncoderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field DPRQuestionEncoderOutput.pooler_output: torch.FloatTensor - added:
field DPRReaderOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field DPRReaderOutput.end_logits: torch.FloatTensor - added:
field DPRReaderOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field DPRReaderOutput.relevance_logits: torch.FloatTensor - added:
field DPRReaderOutput.start_logits: torch.FloatTensor - added:
field DPRSpanPredictor.base_model_prefix - added:
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, ...]] - added:
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, ...]] - added:
method DPREncoder.init_weights() - added:
method DPRPretrainedContextEncoder.init_weights() - added:
method DPRPretrainedQuestionEncoder.init_weights() - added:
method DPRPretrainedReader.init_weights() - added:
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, ...]] - added:
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, ...]] - added:
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, ...]] - added:
method DPRSpanPredictor.init_weights() - added:
property DPREncoder.embeddings_size -> int - added:
variable DPR_CONTEXT_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DPR_ENCODERS_INPUTS_DOCSTRING - added:
variable DPR_QUESTION_ENCODER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DPR_READER_INPUTS_DOCSTRING - added:
variable DPR_READER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DPR_START_DOCSTRING - added:
variable logger
- removed:
field ElectraModel.config_class - changed:
class ElectraEmbeddings(BertEmbeddings) → class ElectraEmbeddings(nn.Module) - changed:
class ElectraPreTrainedModel(BertPreTrainedModel) → class ElectraPreTrainedModel(PreTrainedModel) - changed:
method ElectraDiscriminatorPredictions.forward(discriminator_hidden_states, attention_mask) → method ElectraDiscriminatorPredictions.forward(discriminator_hidden_states) - changed:
method ElectraForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…) → 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) - changed:
method ElectraForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method ElectraForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method ElectraForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method ElectraForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method ElectraForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method ElectraForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method ElectraModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…) → method ElectraModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
class ElectraAttention(nn.Module) - added:
class ElectraEncoder(nn.Module) - added:
class ElectraForMultipleChoice(ElectraPreTrainedModel) - added:
class ElectraForPreTrainingOutput(ModelOutput) - added:
class ElectraForQuestionAnswering(ElectraPreTrainedModel) - added:
class ElectraIntermediate(nn.Module) - added:
class ElectraLayer(nn.Module) - added:
class ElectraOutput(nn.Module) - added:
class ElectraSelfAttention(nn.Module) - added:
class ElectraSelfOutput(nn.Module) - added:
constructor ElectraAttention(config) - added:
constructor ElectraEncoder(config) - added:
constructor ElectraForMultipleChoice(config) - added:
constructor ElectraForQuestionAnswering(config) - added:
constructor ElectraIntermediate(config) - added:
constructor ElectraLayer(config) - added:
constructor ElectraOutput(config) - added:
constructor ElectraSelfAttention(config) - added:
constructor ElectraSelfOutput(config) - added:
field ElectraForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ElectraForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ElectraForPreTrainingOutput.logits: torch.FloatTensor - added:
field ElectraForPreTrainingOutput.loss: Optional[torch.FloatTensor] - added:
field ElectraForQuestionAnswering.base_model_prefix - added:
field ElectraForQuestionAnswering.config_class - added:
field ElectraPreTrainedModel.authorized_missing_keys - added:
field ElectraPreTrainedModel.authorized_unexpected_keys - added:
method ElectraAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method ElectraAttention.prune_heads(heads) - added:
method ElectraEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method ElectraEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method ElectraForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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=…) - added:
method ElectraIntermediate.forward(hidden_states) - added:
method ElectraLayer.feed_forward_chunk(attention_output) - added:
method ElectraLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method ElectraOutput.forward(hidden_states, input_tensor) - added:
method ElectraSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method ElectraSelfAttention.transpose_for_scores(x) - added:
method ElectraSelfOutput.forward(hidden_states, input_tensor)
- changed:
method EncoderDecoderModel.forward(input_ids=…, inputs_embeds=…, attention_mask=…, head_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, decoder_head_mask=…, decoder_inputs_embeds=…, masked_lm_labels=…, lm_labels=…, **kwargs) → 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) - changed:
method EncoderDecoderModel.prepare_inputs_for_generation(input_ids, past, attention_mask, **kwargs) → method EncoderDecoderModel.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, encoder_outputs=…, **kwargs) - added:
variable ENCODER_DECODER_INPUTS_DOCSTRING - added:
variable ENCODER_DECODER_START_DOCSTRING
- changed:
method FlaubertModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…) → 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=…) - added:
class FlaubertForMultipleChoice(XLMForMultipleChoice) - added:
class FlaubertForTokenClassification(XLMForTokenClassification) - added:
constructor FlaubertForMultipleChoice(config) - added:
constructor FlaubertForTokenClassification(config) - added:
field FlaubertForMultipleChoice.config_class - added:
field FlaubertForTokenClassification.config_class
- added:
class FlaxAutoModel(object) - added:
constructor FlaxAutoModel() - added:
static method FlaxAutoModel.from_config(config) - added:
static method FlaxAutoModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable ALL_PRETRAINED_MODEL_ARCHIVE_MAP - added:
variable MODEL_MAPPING - added:
variable logger
- added:
class FlaxBertAttention(nn.Module) - added:
class FlaxBertEmbedding(nn.Module) - added:
class FlaxBertEmbeddings(nn.Module) - added:
class FlaxBertEncoder(nn.Module) - added:
class FlaxBertIntermediate(nn.Module) - added:
class FlaxBertLayer(nn.Module) - added:
class FlaxBertLayerCollection(nn.Module) - added:
class FlaxBertLayerNorm(nn.Module) - added:
class FlaxBertModel(FlaxPreTrainedModel) - added:
class FlaxBertModule(nn.Module) - added:
class FlaxBertOutput(nn.Module) - added:
class FlaxBertPooler(nn.Module) - added:
constructor FlaxBertModel(config: BertConfig, state: dict, seed: int=…, **kwargs) - added:
field FlaxBertAttention.head_size: int - added:
field FlaxBertAttention.num_heads: int - added:
field FlaxBertEmbedding.emb_init: Callable[..., np.ndarray] - added:
field FlaxBertEmbedding.hidden_size: int - added:
field FlaxBertEmbedding.vocab_size: int - added:
field FlaxBertEmbeddings.hidden_size: int - added:
field FlaxBertEmbeddings.max_length: int - added:
field FlaxBertEmbeddings.type_vocab_size: int - added:
field FlaxBertEmbeddings.vocab_size: int - added:
field FlaxBertEncoder.head_size: int - added:
field FlaxBertEncoder.intermediate_size: int - added:
field FlaxBertEncoder.num_heads: int - added:
field FlaxBertEncoder.num_layers: int - added:
field FlaxBertIntermediate.output_size: int - added:
field FlaxBertLayer.head_size: int - added:
field FlaxBertLayer.intermediate_size: int - added:
field FlaxBertLayer.num_heads: int - added:
field FlaxBertLayerCollection.head_size: int - added:
field FlaxBertLayerCollection.intermediate_size: int - added:
field FlaxBertLayerCollection.num_heads: int - added:
field FlaxBertLayerCollection.num_layers: int - added:
field FlaxBertLayerNorm.bias: bool - added:
field FlaxBertLayerNorm.bias_init: jnp.ndarray - added:
field FlaxBertLayerNorm.dtype: jnp.dtype - added:
field FlaxBertLayerNorm.epsilon: float - added:
field FlaxBertLayerNorm.scale: bool - added:
field FlaxBertLayerNorm.scale_init: jnp.ndarray - added:
field FlaxBertModel.base_model_prefix - added:
field FlaxBertModel.config_class - added:
field FlaxBertModel.model_class - added:
field FlaxBertModule.head_size: int - added:
field FlaxBertModule.hidden_size: int - added:
field FlaxBertModule.intermediate_size: int - added:
field FlaxBertModule.max_length: int - added:
field FlaxBertModule.num_encoder_layers: int - added:
field FlaxBertModule.num_heads: int - added:
field FlaxBertModule.type_vocab_size: int - added:
field FlaxBertModule.vocab_size: int - added:
property FlaxBertModel.module -> nn.Module - added:
static method FlaxBertModel.convert_from_pytorch(pt_state: Dict, config: BertConfig) -> Dict - added:
variable BERT_INPUTS_DOCSTRING - added:
variable BERT_START_DOCSTRING - added:
variable logger
- added:
class FlaxRobertaAttention(nn.Module) - added:
class FlaxRobertaEmbedding(nn.Module) - added:
class FlaxRobertaEmbeddings(nn.Module) - added:
class FlaxRobertaEncoder(nn.Module) - added:
class FlaxRobertaIntermediate(nn.Module) - added:
class FlaxRobertaLayer(nn.Module) - added:
class FlaxRobertaLayerCollection(nn.Module) - added:
class FlaxRobertaLayerNorm(nn.Module) - added:
class FlaxRobertaModel(FlaxPreTrainedModel) - added:
class FlaxRobertaModule(nn.Module) - added:
class FlaxRobertaOutput(nn.Module) - added:
class FlaxRobertaPooler(nn.Module) - added:
constructor FlaxRobertaModel(config: RobertaConfig, state: dict, seed: int=…, **kwargs) - added:
field FlaxRobertaAttention.head_size: int - added:
field FlaxRobertaAttention.num_heads: int - added:
field FlaxRobertaEmbedding.emb_init: Callable[..., np.ndarray] - added:
field FlaxRobertaEmbedding.hidden_size: int - added:
field FlaxRobertaEmbedding.vocab_size: int - added:
field FlaxRobertaEmbeddings.hidden_size: int - added:
field FlaxRobertaEmbeddings.max_length: int - added:
field FlaxRobertaEmbeddings.type_vocab_size: int - added:
field FlaxRobertaEmbeddings.vocab_size: int - added:
field FlaxRobertaEncoder.head_size: int - added:
field FlaxRobertaEncoder.intermediate_size: int - added:
field FlaxRobertaEncoder.num_heads: int - added:
field FlaxRobertaEncoder.num_layers: int - added:
field FlaxRobertaIntermediate.output_size: int - added:
field FlaxRobertaLayer.head_size: int - added:
field FlaxRobertaLayer.intermediate_size: int - added:
field FlaxRobertaLayer.num_heads: int - added:
field FlaxRobertaLayerCollection.head_size: int - added:
field FlaxRobertaLayerCollection.intermediate_size: int - added:
field FlaxRobertaLayerCollection.num_heads: int - added:
field FlaxRobertaLayerCollection.num_layers: int - added:
field FlaxRobertaLayerNorm.bias: bool - added:
field FlaxRobertaLayerNorm.bias_init: jnp.ndarray - added:
field FlaxRobertaLayerNorm.dtype: jnp.dtype - added:
field FlaxRobertaLayerNorm.epsilon: float - added:
field FlaxRobertaLayerNorm.scale: bool - added:
field FlaxRobertaLayerNorm.scale_init: jnp.ndarray - added:
field FlaxRobertaModel.base_model_prefix - added:
field FlaxRobertaModel.config_class - added:
field FlaxRobertaModel.model_class - added:
field FlaxRobertaModule.head_size: int - added:
field FlaxRobertaModule.hidden_size: int - added:
field FlaxRobertaModule.intermediate_size: int - added:
field FlaxRobertaModule.max_length: int - added:
field FlaxRobertaModule.num_encoder_layers: int - added:
field FlaxRobertaModule.num_heads: int - added:
field FlaxRobertaModule.type_vocab_size: int - added:
field FlaxRobertaModule.vocab_size: int - added:
property FlaxRobertaModel.module -> nn.Module - added:
static method FlaxRobertaModel.convert_from_pytorch(pt_state: Dict, config: RobertaConfig) -> Dict - added:
variable ROBERTA_INPUTS_DOCSTRING - added:
variable ROBERTA_START_DOCSTRING - added:
variable logger
- added:
class FlaxPreTrainedModel(ABC) - added:
constructor FlaxPreTrainedModel(config: PretrainedConfig, module: nn.Module, params: Dict, seed: int=…) - added:
field FlaxPreTrainedModel.base_model_prefix - added:
field FlaxPreTrainedModel.config_class - added:
field FlaxPreTrainedModel.model_class - added:
field FlaxPreTrainedModel.pretrained_model_archive_map - added:
function gelu(x) - added:
method FlaxPreTrainedModel.save_pretrained(folder) - added:
property FlaxPreTrainedModel.config -> PretrainedConfig - added:
static method FlaxPreTrainedModel.convert_from_pytorch(pt_state: Dict, config: PretrainedConfig) -> Dict - added:
static method FlaxPreTrainedModel.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable ACT2FN - added:
variable logger
- added:
class Attention(nn.Module) - added:
class DecoderLayer(nn.Module) - added:
class EncoderLayer(nn.Module) - added:
class FSMTDecoder(nn.Module) - added:
class FSMTEncoder(nn.Module) - added:
class FSMTForConditionalGeneration(PretrainedFSMTModel) - added:
class FSMTModel(PretrainedFSMTModel) - added:
class PretrainedFSMTModel(PreTrainedModel) - added:
class SinusoidalPositionalEmbedding(nn.Embedding) - added:
constructor Attention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…) - added:
constructor DecoderLayer(config: FSMTConfig) - added:
constructor EncoderLayer(config: FSMTConfig) - added:
constructor FSMTDecoder(config: FSMTConfig, embed_tokens: nn.Embedding) - added:
constructor FSMTEncoder(config: FSMTConfig, embed_tokens) - added:
constructor FSMTForConditionalGeneration(config: FSMTConfig) - added:
constructor FSMTModel(config: FSMTConfig) - added:
constructor SinusoidalPositionalEmbedding(num_positions, embedding_dim, padding_idx) - added:
field FSMTForConditionalGeneration.authorized_missing_keys - added:
field FSMTForConditionalGeneration.base_model_prefix - added:
field FSMTForConditionalGeneration.keys_to_never_save - added:
field PretrainedFSMTModel.base_model_prefix - added:
field PretrainedFSMTModel.config_class - added:
function fill_with_neg_inf(t) - added:
function invert_mask(attention_mask) - added:
function make_padding_mask(input_ids, padding_idx=…) - added:
function shift_tokens_right(input_ids, pad_token_id) - added:
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]] - added:
method DecoderLayer.forward(x, encoder_hidden_states, encoder_attn_mask=…, layer_state=…, causal_mask=…, decoder_padding_mask=…, output_attentions=…) - added:
method EncoderLayer.forward(x, encoder_padding_mask, output_attentions=…) - added:
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) - added:
method FSMTEncoder.forward(input_ids, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FSMTForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
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) - added:
method FSMTForConditionalGeneration.get_encoder() - added:
method FSMTForConditionalGeneration.get_output_embeddings() - added:
method FSMTForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
method FSMTForConditionalGeneration.resize_token_embeddings(new_num_tokens: int) -> nn.Embedding - added:
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) - 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:
method SinusoidalPositionalEmbedding.forward(input, incremental_state: Optional[Any]=…, timestep: Optional[Tensor]=…) - added:
method SinusoidalPositionalEmbedding.make_weight(num_positions, embedding_dim, padding_idx) - added:
property PretrainedFSMTModel.dummy_inputs - added:
static method SinusoidalPositionalEmbedding.get_embedding(num_embeddings, embedding_dim, padding_idx) - added:
static method SinusoidalPositionalEmbedding.make_positions(tensor, padding_idx: int) - added:
variable FSMT_GENERATION_EXAMPLE - added:
variable FSMT_INPUTS_DOCSTRING - added:
variable FSMT_START_DOCSTRING - added:
variable LayerNorm - added:
variable have_fused_layer_norm - added:
variable logger
- added:
class FunnelAttentionStructure(nn.Module) - added:
class FunnelBaseModel(FunnelPreTrainedModel) - added:
class FunnelClassificationHead(nn.Module) - added:
class FunnelDecoder(nn.Module) - added:
class FunnelDiscriminatorPredictions(nn.Module) - added:
class FunnelEmbeddings(nn.Module) - added:
class FunnelEncoder(nn.Module) - added:
class FunnelForMaskedLM(FunnelPreTrainedModel) - added:
class FunnelForMultipleChoice(FunnelPreTrainedModel) - added:
class FunnelForPreTraining(FunnelPreTrainedModel) - added:
class FunnelForPreTrainingOutput(ModelOutput) - added:
class FunnelForQuestionAnswering(FunnelPreTrainedModel) - added:
class FunnelForSequenceClassification(FunnelPreTrainedModel) - added:
class FunnelForTokenClassification(FunnelPreTrainedModel) - added:
class FunnelLayer(nn.Module) - added:
class FunnelModel(FunnelPreTrainedModel) - added:
class FunnelPositionwiseFFN(nn.Module) - added:
class FunnelPreTrainedModel(PreTrainedModel) - added:
class FunnelRelMultiheadAttention(nn.Module) - added:
constructor FunnelAttentionStructure(config) - added:
constructor FunnelBaseModel(config) - added:
constructor FunnelClassificationHead(config, n_labels) - added:
constructor FunnelDecoder(config) - added:
constructor FunnelDiscriminatorPredictions(config) - added:
constructor FunnelEmbeddings(config) - added:
constructor FunnelEncoder(config) - added:
constructor FunnelForMaskedLM(config) - added:
constructor FunnelForMultipleChoice(config) - added:
constructor FunnelForPreTraining(config) - added:
constructor FunnelForQuestionAnswering(config) - added:
constructor FunnelForSequenceClassification(config) - added:
constructor FunnelForTokenClassification(config) - added:
constructor FunnelLayer(config, block_index) - added:
constructor FunnelModel(config) - added:
constructor FunnelPositionwiseFFN(config) - added:
constructor FunnelRelMultiheadAttention(config, block_index) - added:
field FunnelAttentionStructure.cls_token_type_id: int - added:
field FunnelForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field FunnelForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field FunnelForPreTrainingOutput.logits: torch.FloatTensor - added:
field FunnelForPreTrainingOutput.loss: Optional[torch.FloatTensor] - added:
field FunnelPreTrainedModel.base_model_prefix - added:
field FunnelPreTrainedModel.config_class - added:
field FunnelPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_funnel(model, config, tf_checkpoint_path) - added:
function upsample(x, stride, target_len, separate_cls=…, truncate_seq=…) - added:
method FunnelAttentionStructure.get_position_embeds(seq_len, dtype, device) - added:
method FunnelAttentionStructure.init_attention_inputs(inputs_embeds, attention_mask=…, token_type_ids=…) - added:
method FunnelAttentionStructure.pool_tensor(tensor, mode=…, stride=…) - added:
method FunnelAttentionStructure.post_attention_pooling(attention_inputs) - added:
method FunnelAttentionStructure.pre_attention_pooling(output, attention_inputs) - added:
method FunnelAttentionStructure.relative_pos(pos, stride, pooled_pos=…, shift=…) - added:
method FunnelAttentionStructure.stride_pool(tensor, axis) - added:
method FunnelAttentionStructure.stride_pool_pos(pos_id, block_index) - added:
method FunnelAttentionStructure.token_type_ids_to_mat(token_type_ids) - added:
method FunnelBaseModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelBaseModel.get_input_embeddings() - added:
method FunnelBaseModel.set_input_embeddings(new_embeddings) - added:
method FunnelClassificationHead.forward(hidden) - added:
method FunnelDecoder.forward(final_hidden, first_block_hidden, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelDiscriminatorPredictions.forward(discriminator_hidden_states) - added:
method FunnelEmbeddings.forward(input_ids=…, inputs_embeds=…) - added:
method FunnelEncoder.forward(inputs_embeds, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForMaskedLM.get_output_embeddings() - added:
method FunnelForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForPreTraining.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForQuestionAnswering.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, start_positions=…, end_positions=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelLayer.forward(query, key, value, attention_inputs, output_attentions=…) - added:
method FunnelModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method FunnelModel.get_input_embeddings() - added:
method FunnelModel.set_input_embeddings(new_embeddings) - added:
method FunnelPositionwiseFFN.forward(hidden) - added:
method FunnelRelMultiheadAttention.forward(query, key, value, attention_inputs, output_attentions=…) - added:
method FunnelRelMultiheadAttention.relative_positional_attention(position_embeds, q_head, context_len, cls_mask=…) - added:
method FunnelRelMultiheadAttention.relative_token_type_attention(token_type_mat, q_head, cls_mask=…) - added:
variable FUNNEL_INPUTS_DOCSTRING - added:
variable FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable FUNNEL_START_DOCSTRING - added:
variable INF - added:
variable logger
- changed:
constructor Attention(nx, n_ctx, config, scale=…) → constructor Attention(nx, n_ctx, config, scale=…, is_cross_attention=…) - changed:
method Attention.forward(x, layer_past=…, attention_mask=…, head_mask=…, use_cache=…) → method Attention.forward(hidden_states, layer_past=…, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, use_cache=…, output_attentions=…) - changed:
method Block.forward(x, layer_past=…, attention_mask=…, head_mask=…, use_cache=…) → method Block.forward(hidden_states, layer_past=…, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, use_cache=…, output_attentions=…) - changed:
method GPT2DoubleHeadsModel.forward(input_ids=…, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, lm_labels=…, mc_labels=…, use_cache=…) → 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) - changed:
method GPT2LMHeadModel.forward(input_ids=…, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, use_cache=…) → 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) - changed:
method GPT2LMHeadModel.prepare_inputs_for_generation(input_ids, past, **kwargs) → method GPT2LMHeadModel.prepare_inputs_for_generation(input_ids, past=…, **kwargs) - changed:
method GPT2Model.forward(input_ids=…, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…) → 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) - added:
class GPT2DoubleHeadsModelOutput(ModelOutput) - added:
class GPT2ForSequenceClassification(GPT2PreTrainedModel) - added:
constructor GPT2ForSequenceClassification(config) - added:
field GPT2DoubleHeadsModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field GPT2DoubleHeadsModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field GPT2DoubleHeadsModelOutput.logits: torch.FloatTensor - added:
field GPT2DoubleHeadsModelOutput.loss: Optional[torch.FloatTensor] - added:
field GPT2DoubleHeadsModelOutput.mc_logits: torch.FloatTensor - added:
field GPT2DoubleHeadsModelOutput.mc_loss: Optional[torch.FloatTensor] - added:
field GPT2DoubleHeadsModelOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field GPT2ForSequenceClassification.authorized_missing_keys - added:
field GPT2LMHeadModel.authorized_missing_keys - added:
method GPT2DoubleHeadsModel.prepare_inputs_for_generation(input_ids, past=…, **kwargs) - added:
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=…)
- added:
class LayoutLMAttention(nn.Module) - added:
class LayoutLMEmbeddings(nn.Module) - added:
class LayoutLMEncoder(nn.Module) - added:
class LayoutLMForMaskedLM(LayoutLMPreTrainedModel) - added:
class LayoutLMForTokenClassification(LayoutLMPreTrainedModel) - added:
class LayoutLMIntermediate(nn.Module) - added:
class LayoutLMLMPredictionHead(nn.Module) - added:
class LayoutLMLayer(nn.Module) - added:
class LayoutLMModel(LayoutLMPreTrainedModel) - added:
class LayoutLMOnlyMLMHead(nn.Module) - added:
class LayoutLMOutput(nn.Module) - added:
class LayoutLMPooler(nn.Module) - added:
class LayoutLMPreTrainedModel(PreTrainedModel) - added:
class LayoutLMPredictionHeadTransform(nn.Module) - added:
class LayoutLMSelfAttention(nn.Module) - added:
class LayoutLMSelfOutput(nn.Module) - added:
constructor LayoutLMAttention(config) - added:
constructor LayoutLMEmbeddings(config) - added:
constructor LayoutLMEncoder(config) - added:
constructor LayoutLMForMaskedLM(config) - added:
constructor LayoutLMForTokenClassification(config) - added:
constructor LayoutLMIntermediate(config) - added:
constructor LayoutLMLMPredictionHead(config) - added:
constructor LayoutLMLayer(config) - added:
constructor LayoutLMModel(config) - added:
constructor LayoutLMOnlyMLMHead(config) - added:
constructor LayoutLMOutput(config) - added:
constructor LayoutLMPooler(config) - added:
constructor LayoutLMPredictionHeadTransform(config) - added:
constructor LayoutLMSelfAttention(config) - added:
constructor LayoutLMSelfOutput(config) - added:
field LayoutLMForMaskedLM.base_model_prefix - added:
field LayoutLMForMaskedLM.config_class - added:
field LayoutLMForMaskedLM.pretrained_model_archive_map - added:
field LayoutLMForTokenClassification.base_model_prefix - added:
field LayoutLMForTokenClassification.config_class - added:
field LayoutLMForTokenClassification.pretrained_model_archive_map - added:
field LayoutLMModel.base_model_prefix - added:
field LayoutLMModel.config_class - added:
field LayoutLMModel.pretrained_model_archive_map - added:
field LayoutLMPreTrainedModel.authorized_missing_keys - added:
field LayoutLMPreTrainedModel.base_model_prefix - added:
field LayoutLMPreTrainedModel.config_class - added:
method LayoutLMAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method LayoutLMAttention.prune_heads(heads) - added:
method LayoutLMEmbeddings.forward(input_ids=…, bbox=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method LayoutLMEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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=…) - added:
method LayoutLMForMaskedLM.get_input_embeddings() - added:
method LayoutLMForMaskedLM.get_output_embeddings() - added:
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=…) - added:
method LayoutLMForTokenClassification.get_input_embeddings() - added:
method LayoutLMIntermediate.forward(hidden_states) - added:
method LayoutLMLMPredictionHead.forward(hidden_states) - added:
method LayoutLMLayer.feed_forward_chunk(attention_output) - added:
method LayoutLMLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
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=…) - added:
method LayoutLMModel.get_input_embeddings() - added:
method LayoutLMModel.set_input_embeddings(value) - added:
method LayoutLMOnlyMLMHead.forward(sequence_output) - added:
method LayoutLMOutput.forward(hidden_states, input_tensor) - added:
method LayoutLMPooler.forward(hidden_states) - added:
method LayoutLMPredictionHeadTransform.forward(hidden_states) - added:
method LayoutLMSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method LayoutLMSelfAttention.transpose_for_scores(x) - added:
method LayoutLMSelfOutput.forward(hidden_states, input_tensor) - added:
variable LAYOUTLM_INPUTS_DOCSTRING - added:
variable LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable LAYOUTLM_START_DOCSTRING - added:
variable LayoutLMLayerNorm - added:
variable logger
- removed:
field LongformerForMaskedLM.base_model_prefix - removed:
field LongformerForMaskedLM.config_class - removed:
field LongformerForMultipleChoice.base_model_prefix - removed:
field LongformerForMultipleChoice.config_class - removed:
field LongformerForQuestionAnswering.base_model_prefix - removed:
field LongformerForQuestionAnswering.config_class - removed:
field LongformerForSequenceClassification.base_model_prefix - removed:
field LongformerForSequenceClassification.config_class - removed:
field LongformerForTokenClassification.base_model_prefix - removed:
field LongformerForTokenClassification.config_class - removed:
field LongformerModel.base_model_prefix - removed:
field LongformerModel.config_class - changed:
class LongformerForMaskedLM(BertPreTrainedModel) → class LongformerForMaskedLM(LongformerPreTrainedModel) - changed:
class LongformerForMultipleChoice(BertPreTrainedModel) → class LongformerForMultipleChoice(LongformerPreTrainedModel) - changed:
class LongformerForQuestionAnswering(BertPreTrainedModel) → class LongformerForQuestionAnswering(LongformerPreTrainedModel) - changed:
class LongformerForSequenceClassification(BertPreTrainedModel) → class LongformerForSequenceClassification(LongformerPreTrainedModel) - changed:
class LongformerForTokenClassification(BertPreTrainedModel) → class LongformerForTokenClassification(LongformerPreTrainedModel) - changed:
class LongformerModel(RobertaModel) → class LongformerModel(LongformerPreTrainedModel) - changed:
constructor LongformerModel(config) → constructor LongformerModel(config, add_pooling_layer=…) - changed:
method LongformerForMaskedLM.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, masked_lm_labels=…) → 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) - changed:
method LongformerForMultipleChoice.forward(input_ids=…, token_type_ids=…, attention_mask=…, global_attention_mask=…, labels=…, position_ids=…, inputs_embeds=…) → 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=…) - changed:
method LongformerForQuestionAnswering.forward(input_ids, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, start_positions=…, end_positions=…) → 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=…) - changed:
method LongformerForSequenceClassification.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…) → 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=…) - changed:
method LongformerForTokenClassification.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, labels=…) → 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=…) - changed:
method LongformerModel.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, masked_lm_labels=…) → method LongformerModel.forward(input_ids=…, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method LongformerSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…) → method LongformerSelfAttention.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - added:
class LongformerAttention(nn.Module) - added:
class LongformerBaseModelOutput(ModelOutput) - added:
class LongformerBaseModelOutputWithPooling(ModelOutput) - added:
class LongformerEmbeddings(nn.Module) - added:
class LongformerEncoder(nn.Module) - added:
class LongformerIntermediate(nn.Module) - added:
class LongformerLMHead(nn.Module) - added:
class LongformerLayer(nn.Module) - added:
class LongformerMultipleChoiceModelOutput(ModelOutput) - added:
class LongformerOutput(nn.Module) - added:
class LongformerPooler(nn.Module) - added:
class LongformerPreTrainedModel(PreTrainedModel) - added:
class LongformerQuestionAnsweringModelOutput(ModelOutput) - added:
class LongformerSelfOutput(nn.Module) - added:
constructor LongformerAttention(config, layer_id=…) - added:
constructor LongformerEmbeddings(config) - added:
constructor LongformerEncoder(config) - added:
constructor LongformerIntermediate(config) - added:
constructor LongformerLMHead(config) - added:
constructor LongformerLayer(config, layer_id=…) - added:
constructor LongformerOutput(config) - added:
constructor LongformerPooler(config) - added:
constructor LongformerSelfOutput(config) - added:
field LongformerBaseModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutput.last_hidden_state: torch.FloatTensor - added:
field LongformerBaseModelOutputWithPooling.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutputWithPooling.global_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutputWithPooling.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerBaseModelOutputWithPooling.last_hidden_state: torch.FloatTensor - added:
field LongformerBaseModelOutputWithPooling.pooler_output: torch.FloatTensor - added:
field LongformerForMaskedLM.authorized_unexpected_keys - added:
field LongformerForQuestionAnswering.authorized_unexpected_keys - added:
field LongformerForSequenceClassification.authorized_unexpected_keys - added:
field LongformerForTokenClassification.authorized_unexpected_keys - added:
field LongformerMultipleChoiceModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerMultipleChoiceModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerMultipleChoiceModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerMultipleChoiceModelOutput.logits: torch.FloatTensor - added:
field LongformerMultipleChoiceModelOutput.loss: Optional[torch.FloatTensor] - added:
field LongformerPreTrainedModel.authorized_missing_keys - added:
field LongformerPreTrainedModel.base_model_prefix - added:
field LongformerPreTrainedModel.config_class - added:
field LongformerQuestionAnsweringModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerQuestionAnsweringModelOutput.end_logits: torch.FloatTensor - added:
field LongformerQuestionAnsweringModelOutput.global_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LongformerQuestionAnsweringModelOutput.loss: Optional[torch.FloatTensor] - added:
field LongformerQuestionAnsweringModelOutput.start_logits: torch.FloatTensor - added:
function create_position_ids_from_input_ids(input_ids, padding_idx) - added:
method LongformerAttention.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - added:
method LongformerAttention.prune_heads(heads) - added:
method LongformerEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - added:
method LongformerEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method LongformerEncoder.forward(hidden_states, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method LongformerForMaskedLM.get_output_embeddings() - added:
method LongformerIntermediate.forward(hidden_states) - added:
method LongformerLMHead.forward(features, **kwargs) - added:
method LongformerLayer.ff_chunk(attn_output) - added:
method LongformerLayer.forward(hidden_states, attention_mask=…, is_index_masked=…, is_index_global_attn=…, is_global_attn=…) - added:
method LongformerModel.get_input_embeddings() - added:
method LongformerModel.set_input_embeddings(value) - added:
method LongformerOutput.forward(hidden_states, input_tensor) - added:
method LongformerPooler.forward(hidden_states) - added:
method LongformerSelfOutput.forward(hidden_states, input_tensor)
- added:
class GeLU(nn.Module) - added:
class LxmertAttention(nn.Module) - added:
class LxmertAttentionOutput(nn.Module) - added:
class LxmertCrossAttentionLayer(nn.Module) - added:
class LxmertEmbeddings(nn.Module) - added:
class LxmertEncoder(nn.Module) - added:
class LxmertForPreTraining(LxmertPreTrainedModel) - added:
class LxmertForPreTrainingOutput(ModelOutput) - added:
class LxmertForQuestionAnswering(LxmertPreTrainedModel) - added:
class LxmertForQuestionAnsweringOutput(ModelOutput) - added:
class LxmertIntermediate(nn.Module) - added:
class LxmertLMPredictionHead(nn.Module) - added:
class LxmertLayer(nn.Module) - added:
class LxmertModel(LxmertPreTrainedModel) - added:
class LxmertModelOutput(ModelOutput) - added:
class LxmertOutput(nn.Module) - added:
class LxmertPooler(nn.Module) - added:
class LxmertPreTrainedModel(PreTrainedModel) - added:
class LxmertPreTrainingHeads(nn.Module) - added:
class LxmertPredictionHeadTransform(nn.Module) - added:
class LxmertSelfAttentionLayer(nn.Module) - added:
class LxmertVisualAnswerHead(nn.Module) - added:
class LxmertVisualFeatureEncoder(nn.Module) - added:
class LxmertVisualObjHead(nn.Module) - added:
class LxmertXLayer(nn.Module) - added:
constructor GeLU() - added:
constructor LxmertAttention(config, ctx_dim=…) - added:
constructor LxmertAttentionOutput(config) - added:
constructor LxmertCrossAttentionLayer(config) - added:
constructor LxmertEmbeddings(config) - added:
constructor LxmertEncoder(config) - added:
constructor LxmertForPreTraining(config) - added:
constructor LxmertForQuestionAnswering(config) - added:
constructor LxmertIntermediate(config) - added:
constructor LxmertLMPredictionHead(config, lxmert_model_embedding_weights) - added:
constructor LxmertLayer(config) - added:
constructor LxmertModel(config) - added:
constructor LxmertOutput(config) - added:
constructor LxmertPooler(config) - added:
constructor LxmertPreTrainingHeads(config, lxmert_model_embedding_weights) - added:
constructor LxmertPredictionHeadTransform(config) - added:
constructor LxmertSelfAttentionLayer(config) - added:
constructor LxmertVisualAnswerHead(config, num_labels) - added:
constructor LxmertVisualFeatureEncoder(config) - added:
constructor LxmertVisualObjHead(config) - added:
constructor LxmertXLayer(config) - added:
field LxmertForPreTrainingOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForPreTrainingOutput.cross_relationship_score: Optional[torch.FloatTensor] - added:
field LxmertForPreTrainingOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForPreTrainingOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForPreTrainingOutput.loss: [torch.FloatTensor] - added:
field LxmertForPreTrainingOutput.prediction_logits: Optional[torch.FloatTensor] - added:
field LxmertForPreTrainingOutput.question_answering_score: Optional[torch.FloatTensor] - added:
field LxmertForPreTrainingOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForPreTrainingOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForQuestionAnsweringOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForQuestionAnsweringOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForQuestionAnsweringOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - added:
field LxmertForQuestionAnsweringOutput.question_answering_score: Optional[torch.FloatTensor] - added:
field LxmertForQuestionAnsweringOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertForQuestionAnsweringOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.cross_encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.language_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.language_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.language_output: Optional[torch.FloatTensor] - added:
field LxmertModelOutput.pooled_output: Optional[torch.FloatTensor] - added:
field LxmertModelOutput.vision_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.vision_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field LxmertModelOutput.vision_output: Optional[torch.FloatTensor] - added:
field LxmertPreTrainedModel.base_model_prefix - added:
field LxmertPreTrainedModel.config_class - added:
field LxmertPreTrainedModel.load_tf_weights - added:
function load_tf_weights_in_lxmert(model, config, tf_checkpoint_path) - added:
method GeLU.forward(x) - added:
method LxmertAttention.forward(hidden_states, context, attention_mask=…, output_attentions=…) - added:
method LxmertAttention.transpose_for_scores(x) - added:
method LxmertAttentionOutput.forward(hidden_states, input_tensor) - added:
method LxmertCrossAttentionLayer.forward(input_tensor, ctx_tensor, ctx_att_mask=…, output_attentions=…) - added:
method LxmertEmbeddings.forward(input_ids, token_type_ids=…, inputs_embeds=…) - added:
method LxmertEncoder.forward(lang_feats, lang_attention_mask, visual_feats, visual_pos, visual_attention_mask=…, output_attentions=…) - added:
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) - added:
method LxmertForPreTraining.get_qa_logit_layer() -> nn.Module - added:
method LxmertForPreTraining.resize_num_qa_labels(num_labels) - added:
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=…) - added:
method LxmertForQuestionAnswering.get_qa_logit_layer() -> nn.Module - added:
method LxmertForQuestionAnswering.resize_num_qa_labels(num_labels) - added:
method LxmertIntermediate.forward(hidden_states) - added:
method LxmertLMPredictionHead.forward(hidden_states) - added:
method LxmertLayer.forward(hidden_states, attention_mask=…, output_attentions=…) - added:
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=…) - added:
method LxmertModel.get_input_embeddings() - added:
method LxmertModel.set_input_embeddings(new_embeddings) - added:
method LxmertOutput.forward(hidden_states, input_tensor) - added:
method LxmertPooler.forward(hidden_states) - added:
method LxmertPreTrainingHeads.forward(sequence_output, pooled_output) - added:
method LxmertPredictionHeadTransform.forward(hidden_states) - added:
method LxmertSelfAttentionLayer.forward(input_tensor, attention_mask, output_attentions=…) - added:
method LxmertVisualAnswerHead.forward(hidden_states) - added:
method LxmertVisualFeatureEncoder.forward(visual_feats, visual_pos) - added:
method LxmertVisualObjHead.forward(hidden_states) - 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:
variable LXMERT_INPUTS_DOCSTRING - added:
variable LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable LXMERT_START_DOCSTRING - added:
variable logger
- removed:
method MarianMTModel.prepare_logits_for_generation(logits, cur_len, max_length) - removed:
variable MARIAN_PRETRAINED_MODEL_ARCHIVE_LIST - added:
field MarianMTModel.authorized_missing_keys - added:
field MarianMTModel.config_class - added:
field MarianMTModel.keys_to_never_save - added:
method MarianMTModel.adjust_logits_during_generation(logits, cur_len, max_length)
- added:
class MBartForConditionalGeneration(BartForConditionalGeneration) - added:
field MBartForConditionalGeneration.authorized_missing_keys - added:
field MBartForConditionalGeneration.config_class - added:
field MBartForConditionalGeneration.keys_to_never_save - added:
field MBartForConditionalGeneration.model_type - added:
variable MBART_PRETRAINED_MODEL_ARCHIVE_LIST
- changed:
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=…) → 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=…) - changed:
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=…) → 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:
class Bottleneck(nn.Module) - added:
class BottleneckLayer(nn.Module) - added:
class FFNLayer(nn.Module) - added:
class FFNOutput(nn.Module) - added:
class MobileBertAttention(nn.Module) - added:
class MobileBertEmbeddings(nn.Module) - added:
class MobileBertEncoder(nn.Module) - added:
class MobileBertForMaskedLM(MobileBertPreTrainedModel) - added:
class MobileBertForMultipleChoice(MobileBertPreTrainedModel) - added:
class MobileBertForNextSentencePrediction(MobileBertPreTrainedModel) - added:
class MobileBertForPreTraining(MobileBertPreTrainedModel) - added:
class MobileBertForPreTrainingOutput(ModelOutput) - added:
class MobileBertForQuestionAnswering(MobileBertPreTrainedModel) - added:
class MobileBertForSequenceClassification(MobileBertPreTrainedModel) - added:
class MobileBertForTokenClassification(MobileBertPreTrainedModel) - added:
class MobileBertIntermediate(nn.Module) - added:
class MobileBertLMPredictionHead(nn.Module) - added:
class MobileBertLayer(nn.Module) - added:
class MobileBertModel(MobileBertPreTrainedModel) - added:
class MobileBertOnlyMLMHead(nn.Module) - added:
class MobileBertOnlyNSPHead(nn.Module) - added:
class MobileBertOutput(nn.Module) - added:
class MobileBertPooler(nn.Module) - added:
class MobileBertPreTrainedModel(PreTrainedModel) - added:
class MobileBertPreTrainingHeads(nn.Module) - added:
class MobileBertPredictionHeadTransform(nn.Module) - added:
class MobileBertSelfAttention(nn.Module) - added:
class MobileBertSelfOutput(nn.Module) - added:
class NoNorm(nn.Module) - added:
class OutputBottleneck(nn.Module) - added:
constructor Bottleneck(config) - added:
constructor BottleneckLayer(config) - added:
constructor FFNLayer(config) - added:
constructor FFNOutput(config) - added:
constructor MobileBertAttention(config) - added:
constructor MobileBertEmbeddings(config) - added:
constructor MobileBertEncoder(config) - 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 MobileBertIntermediate(config) - added:
constructor MobileBertLMPredictionHead(config) - added:
constructor MobileBertLayer(config) - added:
constructor MobileBertModel(config, add_pooling_layer=…) - added:
constructor MobileBertOnlyMLMHead(config) - added:
constructor MobileBertOnlyNSPHead(config) - added:
constructor MobileBertOutput(config) - added:
constructor MobileBertPooler(config) - added:
constructor MobileBertPreTrainingHeads(config) - added:
constructor MobileBertPredictionHeadTransform(config) - added:
constructor MobileBertSelfAttention(config) - added:
constructor MobileBertSelfOutput(config) - added:
constructor NoNorm(feat_size, eps=…) - added:
constructor OutputBottleneck(config) - added:
field MobileBertForMaskedLM.authorized_unexpected_keys - added:
field MobileBertForPreTrainingOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field MobileBertForPreTrainingOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field MobileBertForPreTrainingOutput.loss: Optional[torch.FloatTensor] - added:
field MobileBertForPreTrainingOutput.prediction_logits: torch.FloatTensor - added:
field MobileBertForPreTrainingOutput.seq_relationship_logits: torch.FloatTensor - added:
field MobileBertForQuestionAnswering.authorized_unexpected_keys - added:
field MobileBertForTokenClassification.authorized_unexpected_keys - added:
field MobileBertPreTrainedModel.authorized_missing_keys - added:
field MobileBertPreTrainedModel.base_model_prefix - added:
field MobileBertPreTrainedModel.config_class - added:
field MobileBertPreTrainedModel.load_tf_weights - added:
field MobileBertPreTrainedModel.pretrained_model_archive_map - added:
function load_tf_weights_in_mobilebert(model, config, tf_checkpoint_path) - added:
function mish(x) - added:
method Bottleneck.forward(hidden_states) - added:
method BottleneckLayer.forward(hidden_states) - added:
method FFNLayer.forward(hidden_states) - added:
method FFNOutput.forward(hidden_states, residual_tensor) - added:
method MobileBertAttention.forward(query_tensor, key_tensor, value_tensor, layer_input, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method MobileBertAttention.prune_heads(heads) - added:
method MobileBertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method MobileBertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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) - added:
method MobileBertForMaskedLM.get_output_embeddings() - added:
method MobileBertForMaskedLM.tie_weights() - added:
method MobileBertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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) - added:
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=…) - added:
method MobileBertForPreTraining.get_output_embeddings() - added:
method MobileBertForPreTraining.tie_weights() - added:
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=…) - added:
method MobileBertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method MobileBertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method MobileBertIntermediate.forward(hidden_states) - added:
method MobileBertLMPredictionHead.forward(hidden_states) - added:
method MobileBertLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
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=…) - added:
method MobileBertModel.get_input_embeddings() - added:
method MobileBertModel.set_input_embeddings(value) - added:
method MobileBertOnlyMLMHead.forward(sequence_output) - added:
method MobileBertOnlyNSPHead.forward(pooled_output) - added:
method MobileBertOutput.forward(intermediate_states, residual_tensor_1, residual_tensor_2) - added:
method MobileBertPooler.forward(hidden_states) - added:
method MobileBertPreTrainingHeads.forward(sequence_output, pooled_output) - added:
method MobileBertPredictionHeadTransform.forward(hidden_states) - added:
method MobileBertSelfAttention.forward(query_tensor, key_tensor, value_tensor, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method MobileBertSelfAttention.transpose_for_scores(x) - added:
method MobileBertSelfOutput.forward(hidden_states, residual_tensor) - added:
method NoNorm.forward(input_tensor) - added:
method OutputBottleneck.forward(hidden_states, residual_tensor) - added:
variable MOBILEBERT_INPUTS_DOCSTRING - added:
variable MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable MOBILEBERT_START_DOCSTRING - added:
variable NORM2FN - added:
variable logger
- changed:
method Attention.forward(x, attention_mask=…, head_mask=…) → method Attention.forward(x, attention_mask=…, head_mask=…, output_attentions=…) - changed:
method Block.forward(x, attention_mask=…, head_mask=…) → method Block.forward(x, attention_mask=…, head_mask=…, output_attentions=…) - changed:
method OpenAIGPTDoubleHeadsModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, lm_labels=…, mc_labels=…) → 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) - changed:
method OpenAIGPTLMHeadModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method OpenAIGPTLMHeadModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method OpenAIGPTModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…) → method OpenAIGPTModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
class OpenAIGPTDoubleHeadsModelOutput(ModelOutput) - added:
class OpenAIGPTForSequenceClassification(OpenAIGPTPreTrainedModel) - added:
constructor OpenAIGPTForSequenceClassification(config) - added:
field OpenAIGPTDoubleHeadsModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field OpenAIGPTDoubleHeadsModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field OpenAIGPTDoubleHeadsModelOutput.logits: torch.FloatTensor - added:
field OpenAIGPTDoubleHeadsModelOutput.loss: Optional[torch.FloatTensor] - added:
field OpenAIGPTDoubleHeadsModelOutput.mc_logits: torch.FloatTensor - added:
field OpenAIGPTDoubleHeadsModelOutput.mc_loss: Optional[torch.FloatTensor] - added:
field OpenAIGPTPreTrainedModel.authorized_missing_keys - added:
method OpenAIGPTForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…)
- added:
class BaseModelOutput(ModelOutput) - added:
class BaseModelOutputWithCrossAttentions(ModelOutput) - added:
class BaseModelOutputWithPast(ModelOutput) - added:
class BaseModelOutputWithPastAndCrossAttentions(ModelOutput) - added:
class BaseModelOutputWithPooling(ModelOutput) - added:
class BaseModelOutputWithPoolingAndCrossAttentions(ModelOutput) - added:
class CausalLMOutput(ModelOutput) - added:
class CausalLMOutputWithCrossAttentions(ModelOutput) - added:
class CausalLMOutputWithPast(ModelOutput) - added:
class CausalLMOutputWithPastAndCrossAttentions(ModelOutput) - added:
class MaskedLMOutput(ModelOutput) - added:
class MultipleChoiceModelOutput(ModelOutput) - added:
class NextSentencePredictorOutput(ModelOutput) - added:
class QuestionAnsweringModelOutput(ModelOutput) - added:
class Seq2SeqLMOutput(ModelOutput) - added:
class Seq2SeqModelOutput(ModelOutput) - added:
class Seq2SeqQuestionAnsweringModelOutput(ModelOutput) - added:
class Seq2SeqSequenceClassifierOutput(ModelOutput) - added:
class SequenceClassifierOutput(ModelOutput) - added:
class SequenceClassifierOutputWithPast(ModelOutput) - added:
class TokenClassifierOutput(ModelOutput) - added:
field BaseModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutput.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithCrossAttentions.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPast.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] - added:
field BaseModelOutputWithPastAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPastAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPastAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPastAndCrossAttentions.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithPastAndCrossAttentions.past_key_values: Optional[List[torch.FloatTensor]] - added:
field BaseModelOutputWithPooling.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPooling.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPooling.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithPooling.pooler_output: torch.FloatTensor - added:
field BaseModelOutputWithPoolingAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPoolingAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPoolingAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field BaseModelOutputWithPoolingAndCrossAttentions.last_hidden_state: torch.FloatTensor - added:
field BaseModelOutputWithPoolingAndCrossAttentions.pooler_output: torch.FloatTensor - added:
field CausalLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutput.logits: torch.FloatTensor - added:
field CausalLMOutput.loss: Optional[torch.FloatTensor] - added:
field CausalLMOutputWithCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithCrossAttentions.logits: torch.FloatTensor - added:
field CausalLMOutputWithCrossAttentions.loss: Optional[torch.FloatTensor] - added:
field CausalLMOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithPast.logits: torch.FloatTensor - added:
field CausalLMOutputWithPast.loss: Optional[torch.FloatTensor] - added:
field CausalLMOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] - added:
field CausalLMOutputWithPastAndCrossAttentions.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithPastAndCrossAttentions.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithPastAndCrossAttentions.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field CausalLMOutputWithPastAndCrossAttentions.logits: torch.FloatTensor - added:
field CausalLMOutputWithPastAndCrossAttentions.loss: Optional[torch.FloatTensor] - added:
field CausalLMOutputWithPastAndCrossAttentions.past_key_values: Optional[List[torch.FloatTensor]] - added:
field MaskedLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field MaskedLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field MaskedLMOutput.logits: torch.FloatTensor - added:
field MaskedLMOutput.loss: Optional[torch.FloatTensor] - added:
field MultipleChoiceModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field MultipleChoiceModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field MultipleChoiceModelOutput.logits: torch.FloatTensor - added:
field MultipleChoiceModelOutput.loss: Optional[torch.FloatTensor] - added:
field NextSentencePredictorOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field NextSentencePredictorOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field NextSentencePredictorOutput.logits: torch.FloatTensor - added:
field NextSentencePredictorOutput.loss: Optional[torch.FloatTensor] - added:
field QuestionAnsweringModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field QuestionAnsweringModelOutput.end_logits: torch.FloatTensor - added:
field QuestionAnsweringModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field QuestionAnsweringModelOutput.loss: Optional[torch.FloatTensor] - added:
field QuestionAnsweringModelOutput.start_logits: torch.FloatTensor - added:
field Seq2SeqLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqLMOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqLMOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqLMOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqLMOutput.logits: torch.FloatTensor - added:
field Seq2SeqLMOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqLMOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqModelOutput.last_hidden_state: torch.FloatTensor - added:
field Seq2SeqModelOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqQuestionAnsweringModelOutput.end_logits: torch.FloatTensor - added:
field Seq2SeqQuestionAnsweringModelOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field Seq2SeqQuestionAnsweringModelOutput.start_logits: torch.FloatTensor - added:
field Seq2SeqSequenceClassifierOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqSequenceClassifierOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqSequenceClassifierOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field Seq2SeqSequenceClassifierOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field Seq2SeqSequenceClassifierOutput.logits: torch.FloatTensor - added:
field Seq2SeqSequenceClassifierOutput.loss: Optional[torch.FloatTensor] - added:
field Seq2SeqSequenceClassifierOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field SequenceClassifierOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field SequenceClassifierOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field SequenceClassifierOutput.logits: torch.FloatTensor - added:
field SequenceClassifierOutput.loss: Optional[torch.FloatTensor] - added:
field SequenceClassifierOutputWithPast.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field SequenceClassifierOutputWithPast.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field SequenceClassifierOutputWithPast.logits: torch.FloatTensor - added:
field SequenceClassifierOutputWithPast.loss: Optional[torch.FloatTensor] - added:
field SequenceClassifierOutputWithPast.past_key_values: Optional[List[torch.FloatTensor]] - added:
field TokenClassifierOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field TokenClassifierOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field TokenClassifierOutput.logits: torch.FloatTensor - added:
field TokenClassifierOutput.loss: Optional[torch.FloatTensor]
- added:
class PegasusForConditionalGeneration(BartForConditionalGeneration) - added:
field PegasusForConditionalGeneration.authorized_missing_keys - added:
field PegasusForConditionalGeneration.config_class - added:
field PegasusForConditionalGeneration.keys_to_never_save
- added:
class ProhpetNetFeedForward(nn.Module) - added:
class ProhpetNetPositionalEmbeddings(nn.Embedding) - added:
class ProphetNetDecoder(ProphetNetPreTrainedModel) - added:
class ProphetNetDecoderLMOutput(ModelOutput) - added:
class ProphetNetDecoderLayer(nn.Module) - added:
class ProphetNetDecoderModelOutput(ModelOutput) - added:
class ProphetNetEncoder(ProphetNetPreTrainedModel) - added:
class ProphetNetEncoderLayer(nn.Module) - added:
class ProphetNetForCausalLM(ProphetNetPreTrainedModel) - added:
class ProphetNetForConditionalGeneration(ProphetNetPreTrainedModel) - added:
class ProphetNetModel(ProphetNetPreTrainedModel) - added:
class ProphetNetNgramProphetNetSelfAttention(nn.Module) - added:
class ProphetNetPreTrainedModel(PreTrainedModel) - added:
class ProphetNetSelfAttention(nn.Module) - added:
class ProphetNetSeq2SeqLMOutput(ModelOutput) - added:
class ProphetNetSeq2SeqModelOutput(ModelOutput) - added:
constructor ProhpetNetFeedForward(config: ProphetNetConfig, ffn_dim: int) - added:
constructor ProhpetNetPositionalEmbeddings(config: ProphetNetConfig) - added:
constructor ProphetNetDecoder(config: ProphetNetConfig, word_embeddings: nn.Embedding=…) - added:
constructor ProphetNetDecoderLayer(config: ProphetNetConfig) - added:
constructor ProphetNetEncoder(config: ProphetNetConfig, word_embeddings: nn.Embedding=…) - added:
constructor ProphetNetEncoderLayer(config: ProphetNetConfig) - added:
constructor ProphetNetForCausalLM(config) - added:
constructor ProphetNetForConditionalGeneration(config: ProphetNetConfig) - added:
constructor ProphetNetModel(config) - added:
constructor ProphetNetNgramProphetNetSelfAttention(config: ProphetNetConfig) - added:
constructor ProphetNetSelfAttention(config: ProphetNetConfig, num_attn_heads: int) - added:
field ProphetNetDecoderLMOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderLMOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderLMOutput.hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderLMOutput.logits: torch.FloatTensor - added:
field ProphetNetDecoderLMOutput.logits_ngram: Optional[torch.FloatTensor] - added:
field ProphetNetDecoderLMOutput.loss: Optional[torch.FloatTensor] - added:
field ProphetNetDecoderLMOutput.ngram_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderLMOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.hidden_states_ngram: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.last_hidden_state: torch.FloatTensor - added:
field ProphetNetDecoderModelOutput.last_hidden_state_ngram: Optional[torch.FloatTensor] - added:
field ProphetNetDecoderModelOutput.ngram_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetDecoderModelOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetPreTrainedModel.base_model_prefix - added:
field ProphetNetPreTrainedModel.config_class - added:
field ProphetNetSeq2SeqLMOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqLMOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field ProphetNetSeq2SeqLMOutput.logits: torch.FloatTensor - added:
field ProphetNetSeq2SeqLMOutput.logits_ngram: Optional[torch.FloatTensor] - added:
field ProphetNetSeq2SeqLMOutput.loss: Optional[torch.FloatTensor] - added:
field ProphetNetSeq2SeqLMOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.cross_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.decoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.decoder_ngram_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.decoder_ngram_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.encoder_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ProphetNetSeq2SeqModelOutput.encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field ProphetNetSeq2SeqModelOutput.last_hidden_state: torch.FloatTensor - added:
field ProphetNetSeq2SeqModelOutput.last_hidden_state_ngram: Optional[torch.FloatTensor] - added:
field ProphetNetSeq2SeqModelOutput.past_key_values: Optional[Tuple[torch.FloatTensor]] - added:
function ProphetNetLayerNorm(normalized_shape, eps=…, elementwise_affine=…) - added:
function compute_all_stream_relative_buckets(num_buckets, max_distance, position_ids) - added:
function compute_relative_buckets(num_buckets, max_distance, relative_positions, is_bidirectional=…) - added:
function ngram_attention_bias(sequence_length, ngram, device, dtype) - added:
function softmax(hidden_state, dim, onnx_trace=…) - added:
method ProhpetNetFeedForward.forward(hidden_states) - added:
method ProhpetNetPositionalEmbeddings.forward(inputs_shape, device, attention_mask=…, past_key_values=…, position_ids=…) - added:
method ProphetNetDecoder.compute_buffered_relative_buckets(position_ids) - added:
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=…) - 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 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=…) - added:
method ProphetNetEncoder.forward(input_ids=…, attention_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method ProphetNetEncoder.get_input_embeddings() - added:
method ProphetNetEncoder.set_input_embeddings(value) - added:
method ProphetNetEncoderLayer.forward(hidden_states, attention_mask) - added:
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=…) - added:
method ProphetNetForCausalLM.get_decoder() - added:
method ProphetNetForCausalLM.get_input_embeddings() - added:
method ProphetNetForCausalLM.get_output_embeddings() - added:
method ProphetNetForCausalLM.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, use_cache=…, **kwargs) - added:
method ProphetNetForCausalLM.set_decoder(decoder) - added:
method ProphetNetForCausalLM.set_input_embeddings(value) - added:
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=…) - added:
method ProphetNetForConditionalGeneration.get_decoder() - added:
method ProphetNetForConditionalGeneration.get_encoder() - added:
method ProphetNetForConditionalGeneration.get_input_embeddings() - added:
method ProphetNetForConditionalGeneration.get_output_embeddings() - added:
method ProphetNetForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - added:
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=…) - added:
method ProphetNetModel.get_decoder() - added:
method ProphetNetModel.get_encoder() - added:
method ProphetNetModel.get_input_embeddings() - added:
method ProphetNetModel.set_input_embeddings(value) - added:
method ProphetNetNgramProphetNetSelfAttention.forward(hidden_states, layer_state=…, attention_mask=…, extended_predict_attention_mask=…, main_relative_position_buckets=…, predict_relative_position_buckets=…, position_ids=…) - added:
method ProphetNetNgramProphetNetSelfAttention.get_main_relative_pos_embeddings(hidden_states, attn_weights, position_ids, main_relative_position_buckets) - added:
method ProphetNetNgramProphetNetSelfAttention.get_predict_relative_pos_embeddings(hidden_states, attn_weights, position_ids, predict_relative_position_buckets) - added:
method ProphetNetNgramProphetNetSelfAttention.prepare_for_onnx_export_() - added:
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]] - added:
property ProphetNetSeq2SeqLMOutput.decoder_cross_attentions - added:
property ProphetNetSeq2SeqModelOutput.decoder_cross_attentions - added:
variable PROPHETNET_INPUTS_DOCSTRING - added:
variable PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable PROPHETNET_STANDALONE_INPUTS_DOCSTRING - added:
variable PROPHETNET_START_DOCSTRING - added:
variable logger
- added:
class RagModel(RagPreTrainedModel) - added:
class RagPreTrainedModel(PreTrainedModel) - added:
class RagSequenceForGeneration(RagPreTrainedModel) - added:
class RagTokenForGeneration(RagPreTrainedModel) - added:
class RetrievAugLMMarginOutput(ModelOutput) - added:
class RetrievAugLMOutput(ModelOutput) - added:
constructor RagModel(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - added:
constructor RagSequenceForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - added:
constructor RagTokenForGeneration(config: Optional[PretrainedConfig]=…, question_encoder: Optional[PreTrainedModel]=…, generator: Optional[PreTrainedModel]=…, retriever: Optional=…, **kwargs) - added:
field RagPreTrainedModel.authorized_missing_keys - added:
field RagPreTrainedModel.base_model_prefix - added:
field RagPreTrainedModel.config_class - added:
field RetrievAugLMMarginOutput.context_attention_mask: Optional[torch.LongTensor] - added:
field RetrievAugLMMarginOutput.context_input_ids: Optional[torch.LongTensor] - added:
field RetrievAugLMMarginOutput.doc_scores: torch.FloatTensor - added:
field RetrievAugLMMarginOutput.generator_dec_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.generator_dec_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.generator_enc_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.generator_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.generator_enc_last_hidden_state: Optional[torch.FloatTensor] - added:
field RetrievAugLMMarginOutput.logits: torch.FloatTensor - added:
field RetrievAugLMMarginOutput.loss: Optional[torch.FloatTensor] - added:
field RetrievAugLMMarginOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.question_enc_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.question_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMMarginOutput.question_encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field RetrievAugLMMarginOutput.retrieved_doc_embeds: Optional[torch.FloatTensor] - added:
field RetrievAugLMMarginOutput.retrieved_doc_ids: Optional[torch.LongTensor] - added:
field RetrievAugLMOutput.context_attention_mask: Optional[torch.LongTensor] - added:
field RetrievAugLMOutput.context_input_ids: Optional[torch.LongTensor] - added:
field RetrievAugLMOutput.doc_scores: torch.FloatTensor - added:
field RetrievAugLMOutput.generator_dec_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.generator_dec_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.generator_enc_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.generator_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.generator_enc_last_hidden_state: Optional[torch.FloatTensor] - added:
field RetrievAugLMOutput.logits: torch.FloatTensor - added:
field RetrievAugLMOutput.past_key_values: Optional[List[torch.FloatTensor]] - added:
field RetrievAugLMOutput.question_enc_attentions: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.question_enc_hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field RetrievAugLMOutput.question_encoder_last_hidden_state: Optional[torch.FloatTensor] - added:
field RetrievAugLMOutput.retrieved_doc_embeds: Optional[torch.FloatTensor] - added:
field RetrievAugLMOutput.retrieved_doc_ids: Optional[torch.LongTensor] - added:
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=…) - added:
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) - added:
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) - added:
method RagSequenceForGeneration.get_nll(seq_logits, doc_scores, target, reduce_loss=…, epsilon=…, exclude_bos_score=…, n_docs=…) - added:
method RagSequenceForGeneration.set_retriever(retriever: RagRetriever) - added:
method RagTokenForGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
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) - added:
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… - 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=…, attention_mask=…, use_cache=…, encoder_outputs=…, doc_scores=…, n_docs=…, **kwargs) - 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: str=…, generator_pretrained_model_name_or_path: str=…, retriever: RagRetriever=…, *model_args, **kwargs) -> PreTrainedModel - added:
variable RAG_FORWARD_INPUTS_DOCSTRING - added:
variable RAG_START_DOCSTRING - added:
variable logger
- removed:
function mish(x) - removed:
method ReformerModelWithLMHead.tie_weights() - removed:
variable ACT2FN - changed:
method LSHSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, do_output_attentions=…, buckets=…, **kwargs) → method LSHSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, buckets=…, past_buckets_states=…, use_cache=…, output_attentions=…, **kwargs) - changed:
method LocalSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, do_output_attentions=…, **kwargs) → method LocalSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, past_buckets_states=…, use_cache=…, output_attentions=…, **kwargs) - changed:
method ReformerAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, do_output_attentions=…, buckets=…) → method ReformerAttention.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…, buckets=…) - changed:
method ReformerEmbeddings.forward(input_ids=…, position_ids=…, inputs_embeds=…) → method ReformerEmbeddings.forward(input_ids=…, position_ids=…, inputs_embeds=…, start_idx_pos_encodings=…) - changed:
method ReformerEncoder.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, do_output_hidden_states=…, do_output_attentions=…) → method ReformerEncoder.forward(hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_hidden_states=…, output_attentions=…) - changed:
method ReformerLayer.forward(prev_attn_output, hidden_states, attention_mask=…, head_mask=…, num_hashes=…, do_output_attentions=…) → method ReformerLayer.forward(prev_attn_output, hidden_states, attention_mask=…, head_mask=…, num_hashes=…, past_buckets_states=…, use_cache=…, orig_sequence_length=…, output_attentions=…) - changed:
method ReformerModel.forward(input_ids=…, attention_mask=…, position_ids=…, head_mask=…, inputs_embeds=…, num_hashes=…, do_output_hidden_states=…, do_output_attentions=…) → 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=…) - changed:
method ReformerModelWithLMHead.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, labels=…, do_output_hidden_states=…, do_output_attentions=…) → 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=…) - changed:
method ReformerModelWithLMHead.prepare_inputs_for_generation(input_ids, past, **kwargs) → method ReformerModelWithLMHead.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, num_hashes=…, **kwargs) - changed:
static method ReverseSort.forward(ctx, out_vectors, logits, sorted_bucket_idx, undo_sorted_bucket_idx, num_hashes) → static method ReverseSort.forward(ctx, out_vectors, logits, sorted_bucket_idx, undo_sorted_bucket_idx) - added:
class ReformerClassificationHead(nn.Module) - added:
class ReformerForMaskedLM(ReformerPreTrainedModel) - added:
class ReformerForQuestionAnswering(ReformerPreTrainedModel) - added:
class ReformerForSequenceClassification(ReformerPreTrainedModel) - added:
class ReformerModelOutput(ModelOutput) - added:
class ReformerModelWithLMHeadOutput(ModelOutput) - added:
constructor ReformerClassificationHead(config) - added:
constructor ReformerForMaskedLM(config) - added:
constructor ReformerForQuestionAnswering(config) - added:
constructor ReformerForSequenceClassification(config) - added:
field ReformerModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ReformerModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ReformerModelOutput.last_hidden_state: torch.FloatTensor - added:
field ReformerModelOutput.past_buckets_states: Optional[List[Tuple[torch.LongTensor, torch.FloatTensor]]] - added:
field ReformerModelWithLMHeadOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field ReformerModelWithLMHeadOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field ReformerModelWithLMHeadOutput.logits: torch.FloatTensor - added:
field ReformerModelWithLMHeadOutput.loss: Optional[torch.FloatTensor] - added:
field ReformerModelWithLMHeadOutput.past_buckets_states: Optional[List[Tuple[torch.LongTensor, torch.FloatTensor]]] - added:
method ReformerClassificationHead.forward(hidden_states, **kwargs) - added:
method ReformerForMaskedLM.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, labels=…, output_hidden_states=…, output_attentions=…, return_dict=…) - added:
method ReformerForMaskedLM.get_output_embeddings() - added:
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=…) - added:
method ReformerForSequenceClassification.forward(input_ids=…, position_ids=…, attention_mask=…, head_mask=…, inputs_embeds=…, num_hashes=…, labels=…, output_hidden_states=…, output_attentions=…, return_dict=…)
- added:
class RetriBertModel(RetriBertPreTrainedModel) - added:
class RetriBertPreTrainedModel(PreTrainedModel) - added:
constructor RetriBertModel(config) - added:
field RetriBertPreTrainedModel.base_model_prefix - added:
field RetriBertPreTrainedModel.config_class - 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, attention_mask_query, input_ids_doc, attention_mask_doc, checkpoint_batch_size=…) - added:
variable RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable RETRIBERT_START_DOCSTRING - added:
variable logger
- removed:
field RobertaForMaskedLM.base_model_prefix - removed:
field RobertaForMaskedLM.config_class - removed:
field RobertaForMultipleChoice.base_model_prefix - removed:
field RobertaForMultipleChoice.config_class - removed:
field RobertaForQuestionAnswering.base_model_prefix - removed:
field RobertaForQuestionAnswering.config_class - removed:
field RobertaForSequenceClassification.base_model_prefix - removed:
field RobertaForSequenceClassification.config_class - removed:
field RobertaForTokenClassification.base_model_prefix - removed:
field RobertaForTokenClassification.config_class - removed:
field RobertaModel.base_model_prefix - removed:
field RobertaModel.config_class - changed:
class RobertaEmbeddings(BertEmbeddings) → class RobertaEmbeddings(nn.Module) - changed:
class RobertaForMaskedLM(BertPreTrainedModel) → class RobertaForMaskedLM(RobertaPreTrainedModel) - changed:
class RobertaForMultipleChoice(BertPreTrainedModel) → class RobertaForMultipleChoice(RobertaPreTrainedModel) - changed:
class RobertaForQuestionAnswering(BertPreTrainedModel) → class RobertaForQuestionAnswering(RobertaPreTrainedModel) - changed:
class RobertaForSequenceClassification(BertPreTrainedModel) → class RobertaForSequenceClassification(RobertaPreTrainedModel) - changed:
class RobertaForTokenClassification(BertPreTrainedModel) → class RobertaForTokenClassification(RobertaPreTrainedModel) - changed:
class RobertaModel(BertModel) → class RobertaModel(RobertaPreTrainedModel) - changed:
constructor RobertaModel(config) → constructor RobertaModel(config, add_pooling_layer=…) - changed:
method RobertaForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, masked_lm_labels=…) → 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) - changed:
method RobertaForMultipleChoice.forward(input_ids=…, token_type_ids=…, attention_mask=…, labels=…, position_ids=…, head_mask=…, inputs_embeds=…) → method RobertaForMultipleChoice.forward(input_ids=…, token_type_ids=…, attention_mask=…, labels=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method RobertaForQuestionAnswering.forward(input_ids, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…) → 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=…) - changed:
method RobertaForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method RobertaForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method RobertaForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…) → method RobertaForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
class RobertaAttention(nn.Module) - added:
class RobertaEncoder(nn.Module) - added:
class RobertaForCausalLM(RobertaPreTrainedModel) - added:
class RobertaIntermediate(nn.Module) - added:
class RobertaLayer(nn.Module) - added:
class RobertaOutput(nn.Module) - added:
class RobertaPooler(nn.Module) - added:
class RobertaPreTrainedModel(PreTrainedModel) - added:
class RobertaSelfAttention(nn.Module) - added:
class RobertaSelfOutput(nn.Module) - added:
constructor RobertaAttention(config) - added:
constructor RobertaEncoder(config) - added:
constructor RobertaForCausalLM(config) - added:
constructor RobertaIntermediate(config) - added:
constructor RobertaLayer(config) - added:
constructor RobertaOutput(config) - added:
constructor RobertaPooler(config) - added:
constructor RobertaSelfAttention(config) - added:
constructor RobertaSelfOutput(config) - added:
field RobertaForCausalLM.authorized_missing_keys - added:
field RobertaForCausalLM.authorized_unexpected_keys - added:
field RobertaForMaskedLM.authorized_missing_keys - added:
field RobertaForMaskedLM.authorized_unexpected_keys - added:
field RobertaForMultipleChoice.authorized_missing_keys - added:
field RobertaForQuestionAnswering.authorized_missing_keys - added:
field RobertaForQuestionAnswering.authorized_unexpected_keys - added:
field RobertaForSequenceClassification.authorized_missing_keys - added:
field RobertaForTokenClassification.authorized_missing_keys - added:
field RobertaForTokenClassification.authorized_unexpected_keys - added:
field RobertaModel.authorized_missing_keys - added:
field RobertaPreTrainedModel.base_model_prefix - added:
field RobertaPreTrainedModel.config_class - added:
function create_position_ids_from_input_ids(input_ids, padding_idx) - added:
method RobertaAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method RobertaAttention.prune_heads(heads) - added:
method RobertaEncoder.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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=…) - added:
method RobertaForCausalLM.get_output_embeddings() - added:
method RobertaForCausalLM.prepare_inputs_for_generation(input_ids, attention_mask=…, **model_kwargs) - added:
method RobertaIntermediate.forward(hidden_states) - added:
method RobertaLayer.feed_forward_chunk(attention_output) - added:
method RobertaLayer.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
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=…) - added:
method RobertaOutput.forward(hidden_states, input_tensor) - added:
method RobertaPooler.forward(hidden_states) - added:
method RobertaSelfAttention.forward(hidden_states, attention_mask=…, head_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, output_attentions=…) - added:
method RobertaSelfAttention.transpose_for_scores(x) - added:
method RobertaSelfOutput.forward(hidden_states, input_tensor)
- added:
class ConvActivation(nn.Module) - added:
class ConvDropoutLayerNorm(nn.Module) - added:
class MatMulWrapper(torch.nn.Module) - added:
class SqueezeBertEmbeddings(nn.Module) - added:
class SqueezeBertEncoder(nn.Module) - added:
class SqueezeBertForMaskedLM(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForMultipleChoice(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForQuestionAnswering(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForSequenceClassification(SqueezeBertPreTrainedModel) - added:
class SqueezeBertForTokenClassification(SqueezeBertPreTrainedModel) - added:
class SqueezeBertLayerNorm(nn.LayerNorm) - added:
class SqueezeBertModel(SqueezeBertPreTrainedModel) - added:
class SqueezeBertModule(nn.Module) - added:
class SqueezeBertPooler(nn.Module) - added:
class SqueezeBertPreTrainedModel(PreTrainedModel) - added:
class SqueezeBertSelfAttention(nn.Module) - added:
constructor ConvActivation(cin, cout, groups, act) - added:
constructor ConvDropoutLayerNorm(cin, cout, groups, dropout_prob) - added:
constructor MatMulWrapper() - added:
constructor SqueezeBertEmbeddings(config) - added:
constructor SqueezeBertEncoder(config) - added:
constructor SqueezeBertForMaskedLM(config) - added:
constructor SqueezeBertForMultipleChoice(config) - added:
constructor SqueezeBertForQuestionAnswering(config) - added:
constructor SqueezeBertForSequenceClassification(config) - added:
constructor SqueezeBertForTokenClassification(config) - added:
constructor SqueezeBertLayerNorm(hidden_size, eps=…) - added:
constructor SqueezeBertModel(config) - added:
constructor SqueezeBertModule(config) - added:
constructor SqueezeBertPooler(config) - added:
constructor SqueezeBertSelfAttention(config, cin, q_groups=…, k_groups=…, v_groups=…) - added:
field SqueezeBertPreTrainedModel.authorized_missing_keys - added:
field SqueezeBertPreTrainedModel.base_model_prefix - added:
field SqueezeBertPreTrainedModel.config_class - added:
method ConvActivation.forward(x) - added:
method ConvDropoutLayerNorm.forward(hidden_states, input_tensor) - added:
method MatMulWrapper.forward(mat1, mat2) - added:
method SqueezeBertEmbeddings.forward(input_ids=…, token_type_ids=…, position_ids=…, inputs_embeds=…) - added:
method SqueezeBertEncoder.forward(hidden_states, attention_mask=…, head_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method SqueezeBertForMaskedLM.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method SqueezeBertForMaskedLM.get_output_embeddings() - added:
method SqueezeBertForMultipleChoice.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
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=…) - added:
method SqueezeBertForSequenceClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method SqueezeBertForTokenClassification.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
method SqueezeBertLayerNorm.forward(x) - added:
method SqueezeBertModel.forward(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - 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:
method SqueezeBertPooler.forward(hidden_states) - added:
method SqueezeBertSelfAttention.forward(hidden_states, attention_mask, output_attentions) - added:
method SqueezeBertSelfAttention.transpose_for_scores(x) - added:
method SqueezeBertSelfAttention.transpose_key_for_scores(x) - added:
method SqueezeBertSelfAttention.transpose_output(x) - added:
variable SQUEEZEBERT_INPUTS_DOCSTRING - added:
variable SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable SQUEEZEBERT_START_DOCSTRING - added:
variable logger
- changed:
constructor T5Attention(config: T5Config, has_relative_attention_bias=…) → constructor T5Attention(config: T5Config, has_relative_attention_bias=…, is_bidirectional=…) - changed:
constructor T5Model(config) → constructor T5Model(config: T5Config) - changed:
method T5Attention.forward(input, mask=…, kv=…, position_bias=…, past_key_value_state=…, head_mask=…, query_length=…, use_cache=…) → method T5Attention.forward(input, mask=…, kv=…, position_bias=…, past_key_value=…, head_mask=…, query_length=…, use_cache=…, output_attentions=…) - changed:
method T5Block.forward(hidden_states, attention_mask=…, position_bias=…, encoder_hidden_states=…, encoder_attention_mask=…, encoder_decoder_position_bias=…, head_mask=…, past_key_value_state=…, use_cache=…) → 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=…) - changed:
method T5ForConditionalGeneration.forward(input_ids=…, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, decoder_past_key_value_states=…, use_cache=…, lm_labels=…, inputs_embeds=…, decoder_inputs_embeds=…, head_mask=…) → 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) - changed:
method T5ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past, attention_mask, use_cache, **kwargs) → method T5ForConditionalGeneration.prepare_inputs_for_generation(input_ids, past=…, attention_mask=…, use_cache=…, encoder_outputs=…, **kwargs) - changed:
method T5LayerCrossAttention.forward(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value_state=…, use_cache=…, query_length=…) → method T5LayerCrossAttention.forward(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, query_length=…, output_attentions=…) - changed:
method T5LayerSelfAttention.forward(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value_state=…, use_cache=…) → method T5LayerSelfAttention.forward(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…) - changed:
method T5Model.forward(input_ids=…, attention_mask=…, encoder_outputs=…, decoder_input_ids=…, decoder_attention_mask=…, decoder_past_key_value_states=…, use_cache=…, inputs_embeds=…, decoder_inputs_embeds=…, head_mask=…) → 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) - changed:
method T5Stack.forward(input_ids=…, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, inputs_embeds=…, head_mask=…, past_key_value_states=…, use_cache=…) → 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=…) - added:
field T5ForConditionalGeneration.authorized_missing_keys
- removed:
class TFAlbertSelfAttention(tf.keras.layers.Layer) - removed:
constructor TFAlbertSelfAttention(config, **kwargs) - removed:
method TFAlbertSelfAttention.call(inputs, training=…) - removed:
method TFAlbertSelfAttention.transpose_for_scores(x, batch_size) - changed:
class TFAlbertAttention(TFBertSelfAttention) → class TFAlbertAttention(tf.keras.layers.Layer) - changed:
class TFAlbertForMaskedLM(TFAlbertPreTrainedModel) → class TFAlbertForMaskedLM(TFAlbertPreTrainedModel, TFMaskedLanguageModelingLoss) - changed:
class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel) → class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss) - changed:
class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel) → class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel) → class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClassificationLoss) - changed:
method TFAlbertAttention.call(inputs, training=…) → method TFAlbertAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFAlbertEmbeddings.call(inputs, mode=…, training=…) → method TFAlbertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - changed:
method TFAlbertForMaskedLM.call(inputs, **kwargs) → method TFAlbertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFAlbertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFAlbertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFAlbertForQuestionAnswering.call(inputs, **kwargs) → 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=…) - changed:
method TFAlbertForSequenceClassification.call(inputs, **kwargs) → method TFAlbertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFAlbertLayer.call(inputs, training=…) → method TFAlbertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFAlbertLayerGroup.call(inputs, training=…) → method TFAlbertLayerGroup.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, training=…) - changed:
method TFAlbertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFAlbertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - changed:
method TFAlbertSelfOutput.call(inputs, training=…) → method TFAlbertSelfOutput.call(hidden_states, input_tensor, training=…) - changed:
method TFAlbertTransformer.call(inputs, training=…) → method TFAlbertTransformer.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - added:
class TFAlbertForPreTrainingOutput(ModelOutput) - added:
class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificationLoss) - added:
constructor TFAlbertForTokenClassification(config, *inputs, **kwargs) - added:
field TFAlbertForMaskedLM.authorized_missing_keys - added:
field TFAlbertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFAlbertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFAlbertForPreTrainingOutput.prediction_logits: tf.Tensor - added:
field TFAlbertForPreTrainingOutput.sop_logits: tf.Tensor - added:
field TFAlbertForQuestionAnswering.authorized_missing_keys - added:
field TFAlbertForTokenClassification.authorized_missing_keys - added:
method TFAlbertAttention.transpose_for_scores(x, batch_size) - added:
method TFAlbertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFAlbertMainLayer.set_input_embeddings(value)
- added:
class TFAutoModelForCausalLM - added:
class TFAutoModelForMaskedLM - added:
class TFAutoModelForSeq2SeqLM - added:
constructor TFAutoModelForCausalLM() - added:
constructor TFAutoModelForMaskedLM() - added:
constructor TFAutoModelForSeq2SeqLM() - added:
static method TFAutoModelForCausalLM.from_config(config) - added:
static method TFAutoModelForCausalLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method TFAutoModelForMaskedLM.from_config(config) - added:
static method TFAutoModelForMaskedLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
static method TFAutoModelForSeq2SeqLM.from_config(config) - added:
static method TFAutoModelForSeq2SeqLM.from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs) - added:
variable TF_AUTO_MODEL_PRETRAINED_DOCSTRING - added:
variable TF_MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable TF_MODEL_FOR_MASKED_LM_MAPPING - added:
variable TF_MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING
- added:
class TFAttention(Layer) - added:
class TFBartDecoder(Layer) - added:
class TFBartEncoder(Layer) - added:
class TFBartForConditionalGeneration(TFPretrainedBartModel) - added:
class TFBartModel(TFPretrainedBartModel) - added:
class TFDecoderLayer(Layer) - added:
class TFEncoderLayer(Layer) - added:
class TFLearnedPositionalEmbedding(TFSharedEmbeddings) - added:
class TFPretrainedBartModel(TFPreTrainedModel) - added:
class TFSinusoidalPositionalEmbedding(tf.keras.layers.Embedding) - added:
constructor TFAttention(embed_dim, num_heads, dropout=…, bias=…, encoder_decoder_attention=…, **kwargs) - added:
constructor TFBartDecoder(config: BartConfig, embed_tokens, **kwargs) - added:
constructor TFBartEncoder(config: BartConfig, embed_tokens: TFSharedEmbeddings, **kwargs) - added:
constructor TFBartForConditionalGeneration(config: BartConfig, *args, **kwargs) - added:
constructor TFBartModel(config: BartConfig, *inputs, **kwargs) - added:
constructor TFDecoderLayer(config: BartConfig, **kwargs) - added:
constructor TFEncoderLayer(config: BartConfig, **kwargs) - added:
constructor TFLearnedPositionalEmbedding(num_embeddings: int, embedding_dim: int, padding_idx: int, offset, **kwargs) - added:
constructor TFSinusoidalPositionalEmbedding(num_positions, embedding_dim, **kwargs) - added:
field TFBartForConditionalGeneration.authorized_missing_keys - added:
field TFBartForConditionalGeneration.authorized_unexpected_keys - added:
field TFBartForConditionalGeneration.base_model_prefix - added:
field TFPretrainedBartModel.base_model_prefix - added:
field TFPretrainedBartModel.config_class - added:
function causal_attention_mask(nd, ns, dtype) - added:
function create_position_ids_from_input_ids(input_ids, padding_idx) - added:
function invert_mask(attention_mask: tf.Tensor) - added:
function make_padding_mask(input_ids, padding_idx=…) - added:
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]] - added:
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=…) - added:
method TFBartEncoder.call(input_ids=…, attention_mask=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFBartForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
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) - added:
method TFBartForConditionalGeneration.compute_loss(labels, logits) - added:
method TFBartForConditionalGeneration.get_encoder() - added:
method TFBartForConditionalGeneration.get_output_embeddings() - added:
method TFBartForConditionalGeneration.prepare_inputs_for_generation(decoder_input_ids, past, attention_mask, use_cache=…, **kwargs) -> Dict - added:
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) - added:
method TFBartModel.get_input_embeddings() - added:
method TFBartModel.get_output_embeddings() - added:
method TFBartModel.set_input_embeddings(value) - added:
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]] - added:
method TFEncoderLayer.call(x, encoder_padding_mask, training=…) - added:
method TFLearnedPositionalEmbedding.call(input_ids: tf.Tensor, use_cache=…) - added:
method TFSinusoidalPositionalEmbedding.build(input_shape) - added:
method TFSinusoidalPositionalEmbedding.call(input_ids, use_cache=…) - added:
property TFPretrainedBartModel.dummy_inputs - added:
variable BART_INPUTS_DOCSTRING - added:
variable BART_START_DOCSTRING - added:
variable LARGE_NEGATIVE - added:
variable PAST_KV_DEPRECATION_WARNING - added:
variable logger
- removed:
function gelu(x) - removed:
function gelu_new(x) - removed:
function swish(x) - removed:
variable ACT2FN - changed:
class TFBertForMaskedLM(TFBertPreTrainedModel) → class TFBertForMaskedLM(TFBertPreTrainedModel, TFMaskedLanguageModelingLoss) - changed:
class TFBertForMultipleChoice(TFBertPreTrainedModel) → class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss) - changed:
class TFBertForQuestionAnswering(TFBertPreTrainedModel) → class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFBertForSequenceClassification(TFBertPreTrainedModel) → class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassificationLoss) - changed:
class TFBertForTokenClassification(TFBertPreTrainedModel) → class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationLoss) - changed:
method TFBertAttention.call(inputs, training=…) → method TFBertAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFBertEmbeddings.call(inputs, mode=…, training=…) → method TFBertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - changed:
method TFBertEncoder.call(inputs, training=…) → method TFBertEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - changed:
method TFBertForMaskedLM.call(inputs, **kwargs) → method TFBertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFBertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFBertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFBertForQuestionAnswering.call(inputs, **kwargs) → 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=…) - changed:
method TFBertForSequenceClassification.call(inputs, **kwargs) → method TFBertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFBertForTokenClassification.call(inputs, **kwargs) → method TFBertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFBertLayer.call(inputs, training=…) → method TFBertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFBertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFBertMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - changed:
method TFBertOutput.call(inputs, training=…) → method TFBertOutput.call(hidden_states, input_tensor, training=…) - changed:
method TFBertSelfAttention.call(inputs, training=…) → method TFBertSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFBertSelfOutput.call(inputs, training=…) → method TFBertSelfOutput.call(hidden_states, input_tensor, training=…) - added:
class TFBertForPreTrainingOutput(ModelOutput) - added:
class TFBertLMHeadModel(TFBertPreTrainedModel, TFCausalLanguageModelingLoss) - added:
constructor TFBertLMHeadModel(config, *inputs, **kwargs) - added:
field TFBertForMaskedLM.authorized_missing_keys - added:
field TFBertForMaskedLM.authorized_unexpected_keys - added:
field TFBertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFBertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFBertForPreTrainingOutput.prediction_logits: tf.Tensor - added:
field TFBertForPreTrainingOutput.seq_relationship_logits: tf.Tensor - added:
field TFBertForQuestionAnswering.authorized_missing_keys - added:
field TFBertForQuestionAnswering.authorized_unexpected_keys - added:
field TFBertForTokenClassification.authorized_missing_keys - added:
field TFBertForTokenClassification.authorized_unexpected_keys - added:
field TFBertLMHeadModel.authorized_missing_keys - added:
field TFBertLMHeadModel.authorized_unexpected_keys - added:
method TFBertLMHeadModel.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFBertLMHeadModel.get_output_embeddings() - added:
method TFBertMainLayer.set_input_embeddings(value)
- added:
class TFBlenderbotForConditionalGeneration(TFBartForConditionalGeneration) - added:
field TFBlenderbotForConditionalGeneration.config_class - added:
method TFBlenderbotForConditionalGeneration.adjust_logits_during_generation(logits, cur_len, max_length) - added:
variable START_DOCSTRING - added:
variable logger
- added:
class TFCamembertForMultipleChoice(TFRobertaForMultipleChoice) - added:
class TFCamembertForQuestionAnswering(TFRobertaForQuestionAnswering) - added:
field TFCamembertForMultipleChoice.config_class - added:
field TFCamembertForQuestionAnswering.config_class
- removed:
function point_wise_feed_forward_network(d_model_size, dff, name=…) - changed:
class TFCTRLLMHeadModel(TFCTRLPreTrainedModel) → class TFCTRLLMHeadModel(TFCTRLPreTrainedModel, TFCausalLanguageModelingLoss) - changed:
method TFCTRLLMHeadModel.call(inputs, **kwargs) → 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=…) - changed:
method TFCTRLMainLayer.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, training=…) → 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=…) - changed:
method TFEncoderLayer.call(inputs, training=…) → method TFEncoderLayer.call(x, mask, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - changed:
method TFMultiHeadAttention.call(inputs, training=…) → method TFMultiHeadAttention.call(v, k, q, mask, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - added:
class TFPointWiseFeedForwardLayer(tf.keras.layers.Layer) - added:
constructor TFPointWiseFeedForwardLayer(d_model_size, dff, **kwargs) - added:
method TFCTRLMainLayer.set_input_embeddings(value) - added:
method TFPointWiseFeedForwardLayer.call(inputs, trainable=…)
- removed:
function gelu(x) - removed:
function gelu_new(x) - changed:
class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel) → class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel, TFMaskedLanguageModelingLoss) - changed:
class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel) → class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel) → class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSequenceClassificationLoss) - changed:
class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel) → class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenClassificationLoss) - changed:
method TFDistilBertForMaskedLM.call(inputs, **kwargs) → method TFDistilBertForMaskedLM.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFDistilBertForQuestionAnswering.call(inputs, **kwargs) → method TFDistilBertForQuestionAnswering.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - changed:
method TFDistilBertForSequenceClassification.call(inputs, **kwargs) → method TFDistilBertForSequenceClassification.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFDistilBertForTokenClassification.call(inputs, **kwargs) → method TFDistilBertForTokenClassification.call(inputs=…, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFDistilBertMainLayer.call(inputs, attention_mask=…, head_mask=…, inputs_embeds=…, training=…) → method TFDistilBertMainLayer.call(inputs, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - changed:
method TFEmbeddings.call(inputs, inputs_embeds=…, mode=…, training=…) → method TFEmbeddings.call(input_ids=…, position_ids=…, inputs_embeds=…, mode=…, training=…) - changed:
method TFMultiHeadSelfAttention.call(inputs, training=…) → method TFMultiHeadSelfAttention.call(query, key, value, mask, head_mask, output_attentions, training=…) - changed:
method TFTransformer.call(inputs, training=…) → method TFTransformer.call(x, attn_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - changed:
method TFTransformerBlock.call(inputs, training=…) → method TFTransformerBlock.call(x, attn_mask, head_mask, output_attentions, training=…) - added:
class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoiceLoss) - added:
constructor TFDistilBertForMultipleChoice(config, *inputs, **kwargs) - added:
field TFDistilBertMainLayer.config_class - added:
method TFDistilBertForMultipleChoice.call(inputs, attention_mask=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFDistilBertMainLayer.set_input_embeddings(value) - added:
property TFDistilBertForMultipleChoice.dummy_inputs
- removed:
method TFElectraForMaskedLM.get_input_embeddings() - removed:
method TFElectraForPreTraining.get_input_embeddings() - removed:
method TFElectraModel.get_input_embeddings() - removed:
method TFElectraPreTrainedModel.get_extended_attention_mask(attention_mask, input_shape) - removed:
method TFElectraPreTrainedModel.get_head_mask(head_mask) - changed:
class TFElectraForMaskedLM(TFElectraPreTrainedModel) → class TFElectraForMaskedLM(TFElectraPreTrainedModel, TFMaskedLanguageModelingLoss) - changed:
class TFElectraForTokenClassification(TFElectraPreTrainedModel) → class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassificationLoss) - changed:
class TFElectraMainLayer(TFElectraPreTrainedModel) → class TFElectraMainLayer(tf.keras.layers.Layer) - changed:
class TFElectraPreTrainedModel(TFBertPreTrainedModel) → class TFElectraPreTrainedModel(TFPreTrainedModel) - changed:
method TFElectraEmbeddings.call(inputs, mode=…, training=…) → method TFElectraEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - changed:
method TFElectraForMaskedLM.call(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → 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) - changed:
method TFElectraForPreTraining.call(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFElectraForPreTraining.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - changed:
method TFElectraForTokenClassification.call(input_ids=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFElectraForTokenClassification.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFElectraMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFElectraMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
class TFElectraAttention(tf.keras.layers.Layer) - added:
class TFElectraClassificationHead(tf.keras.layers.Layer) - added:
class TFElectraEncoder(tf.keras.layers.Layer) - added:
class TFElectraForMultipleChoice(TFElectraPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFElectraForPreTrainingOutput(ModelOutput) - added:
class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFElectraForSequenceClassification(TFElectraPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFElectraIntermediate(tf.keras.layers.Layer) - added:
class TFElectraLayer(tf.keras.layers.Layer) - added:
class TFElectraOutput(tf.keras.layers.Layer) - added:
class TFElectraPooler(tf.keras.layers.Layer) - added:
class TFElectraSelfAttention(tf.keras.layers.Layer) - added:
class TFElectraSelfOutput(tf.keras.layers.Layer) - added:
constructor TFElectraAttention(config, **kwargs) - added:
constructor TFElectraClassificationHead(config, **kwargs) - added:
constructor TFElectraEncoder(config, **kwargs) - added:
constructor TFElectraForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFElectraForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFElectraForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFElectraIntermediate(config, **kwargs) - added:
constructor TFElectraLayer(config, **kwargs) - added:
constructor TFElectraOutput(config, **kwargs) - added:
constructor TFElectraPooler(config, **kwargs) - added:
constructor TFElectraSelfAttention(config, **kwargs) - added:
constructor TFElectraSelfOutput(config, **kwargs) - added:
field TFElectraForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFElectraForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFElectraForPreTrainingOutput.logits: tf.Tensor - added:
method TFElectraAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - added:
method TFElectraAttention.prune_heads(heads) - added:
method TFElectraClassificationHead.call(inputs, **kwargs) - added:
method TFElectraEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - added:
method TFElectraForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
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=…) - added:
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) - added:
method TFElectraIntermediate.call(hidden_states) - added:
method TFElectraLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFElectraMainLayer.get_extended_attention_mask(attention_mask, input_shape, dtype) - added:
method TFElectraMainLayer.get_head_mask(head_mask) - added:
method TFElectraMainLayer.set_input_embeddings(value) - added:
method TFElectraOutput.call(hidden_states, input_tensor, training=…) - added:
method TFElectraPooler.call(hidden_states) - added:
method TFElectraSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFElectraSelfAttention.transpose_for_scores(x, batch_size) - added:
method TFElectraSelfOutput.call(hidden_states, input_tensor, training=…) - added:
property TFElectraForMultipleChoice.dummy_inputs
- removed:
field TFFlaubertModel.config_class - removed:
field TFFlaubertWithLMHeadModel.config_class - changed:
class TFFlaubertMainLayer(TFXLMMainLayer) → class TFFlaubertMainLayer(tf.keras.layers.Layer) - changed:
class TFFlaubertModel(TFXLMModel) → class TFFlaubertModel(TFFlaubertPreTrainedModel) - changed:
class TFFlaubertWithLMHeadModel(TFXLMWithLMHeadModel) → class TFFlaubertWithLMHeadModel(TFFlaubertPreTrainedModel) - changed:
method TFFlaubertMainLayer.call(inputs, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, training=…) → 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=…) - added:
class TFFlaubertForMultipleChoice(TFXLMForMultipleChoice) - added:
class TFFlaubertForQuestionAnsweringSimple(TFXLMForQuestionAnsweringSimple) - added:
class TFFlaubertForTokenClassification(TFXLMForTokenClassification) - added:
class TFFlaubertMultiHeadAttention(tf.keras.layers.Layer) - added:
class TFFlaubertPreTrainedModel(TFPreTrainedModel) - added:
class TFFlaubertPredLayer(tf.keras.layers.Layer) - added:
class TFFlaubertTransformerFFN(tf.keras.layers.Layer) - added:
class TFFlaubertWithLMHeadModelOutput(ModelOutput) - added:
constructor TFFlaubertForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFFlaubertForQuestionAnsweringSimple(config, *inputs, **kwargs) - added:
constructor TFFlaubertForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFFlaubertMultiHeadAttention(n_heads, dim, config, **kwargs) - added:
constructor TFFlaubertPredLayer(config, input_embeddings, **kwargs) - added:
constructor TFFlaubertTransformerFFN(in_dim, dim_hidden, out_dim, config, **kwargs) - added:
field TFFlaubertForQuestionAnsweringSimple.config_class - added:
field TFFlaubertMainLayer.config_class - added:
field TFFlaubertMultiHeadAttention.NEW_ID - added:
field TFFlaubertPreTrainedModel.base_model_prefix - added:
field TFFlaubertPreTrainedModel.config_class - added:
field TFFlaubertWithLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFFlaubertWithLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFFlaubertWithLMHeadModelOutput.logits: tf.Tensor - added:
function get_masks(slen, lengths, causal, padding_mask=…, dtype=…) - added:
method TFFlaubertMainLayer.get_input_embeddings() - added:
method TFFlaubertModel.call(inputs, **kwargs) - added:
method TFFlaubertMultiHeadAttention.call(input, mask, kv, cache, head_mask, output_attentions, training=…) - added:
method TFFlaubertMultiHeadAttention.prune_heads(heads) - added:
method TFFlaubertPredLayer.build(input_shape) - added:
method TFFlaubertPredLayer.call(hidden_states) - added:
method TFFlaubertTransformerFFN.call(input, training=…) - added:
method TFFlaubertWithLMHeadModel.call(inputs, **kwargs) - added:
method TFFlaubertWithLMHeadModel.get_output_embeddings() - added:
method TFFlaubertWithLMHeadModel.prepare_inputs_for_generation(inputs, **kwargs) - added:
property TFFlaubertPreTrainedModel.dummy_inputs
- added:
class TFFunnelAttentionStructure - added:
class TFFunnelBaseLayer(tf.keras.layers.Layer) - added:
class TFFunnelBaseModel(TFFunnelPreTrainedModel) - added:
class TFFunnelClassificationHead(tf.keras.layers.Layer) - added:
class TFFunnelDecoder(tf.keras.layers.Layer) - added:
class TFFunnelDiscriminatorPredictions(tf.keras.layers.Layer) - added:
class TFFunnelEmbeddings(tf.keras.layers.Layer) - added:
class TFFunnelEncoder(tf.keras.layers.Layer) - added:
class TFFunnelForMaskedLM(TFFunnelPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFFunnelForMultipleChoice(TFFunnelPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFFunnelForPreTraining(TFFunnelPreTrainedModel) - added:
class TFFunnelForPreTrainingOutput(ModelOutput) - added:
class TFFunnelForQuestionAnswering(TFFunnelPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFFunnelForSequenceClassification(TFFunnelPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFFunnelForTokenClassification(TFFunnelPreTrainedModel, TFTokenClassificationLoss) - added:
class TFFunnelLayer(tf.keras.layers.Layer) - added:
class TFFunnelMainLayer(tf.keras.layers.Layer) - added:
class TFFunnelMaskedLMHead(tf.keras.layers.Layer) - added:
class TFFunnelModel(TFFunnelPreTrainedModel) - added:
class TFFunnelPositionwiseFFN(tf.keras.layers.Layer) - added:
class TFFunnelPreTrainedModel(TFPreTrainedModel) - added:
class TFFunnelRelMultiheadAttention(tf.keras.layers.Layer) - added:
constructor TFFunnelAttentionStructure(config) - added:
constructor TFFunnelBaseLayer(config, **kwargs) - added:
constructor TFFunnelBaseModel(config, *inputs, **kwargs) - added:
constructor TFFunnelClassificationHead(config, n_labels, **kwargs) - added:
constructor TFFunnelDecoder(config, **kwargs) - added:
constructor TFFunnelDiscriminatorPredictions(config, **kwargs) - added:
constructor TFFunnelEmbeddings(config, **kwargs) - added:
constructor TFFunnelEncoder(config, **kwargs) - added:
constructor TFFunnelForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFFunnelForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFFunnelForPreTraining(config, **kwargs) - added:
constructor TFFunnelForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFFunnelForSequenceClassification(config, *inputs, **kwargs) - added:
constructor TFFunnelForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFFunnelLayer(config, block_index, **kwargs) - added:
constructor TFFunnelMainLayer(config, **kwargs) - added:
constructor TFFunnelMaskedLMHead(config, input_embeddings, **kwargs) - added:
constructor TFFunnelModel(config, *inputs, **kwargs) - added:
constructor TFFunnelPositionwiseFFN(config, **kwargs) - added:
constructor TFFunnelRelMultiheadAttention(config, block_index, **kwargs) - added:
field TFFunnelAttentionStructure.cls_token_type_id: int - added:
field TFFunnelBaseLayer.config_class - added:
field TFFunnelForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFFunnelForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFFunnelForPreTrainingOutput.logits: tf.Tensor - added:
field TFFunnelMainLayer.config_class - added:
field TFFunnelPreTrainedModel.base_model_prefix - added:
field TFFunnelPreTrainedModel.config_class - added:
function upsample(x, stride, target_len, separate_cls=…, truncate_seq=…) - added:
method TFFunnelAttentionStructure.get_position_embeds(seq_len, dtype=…, training=…) - added:
method TFFunnelAttentionStructure.init_attention_inputs(inputs_embeds, attention_mask=…, token_type_ids=…, training=…) - added:
method TFFunnelAttentionStructure.pool_tensor(tensor, mode=…, stride=…) - added:
method TFFunnelAttentionStructure.post_attention_pooling(attention_inputs) - added:
method TFFunnelAttentionStructure.pre_attention_pooling(output, attention_inputs) - added:
method TFFunnelAttentionStructure.relative_pos(pos, stride, pooled_pos=…, shift=…) - added:
method TFFunnelAttentionStructure.stride_pool(tensor, axis) - added:
method TFFunnelAttentionStructure.stride_pool_pos(pos_id, block_index) - added:
method TFFunnelAttentionStructure.token_type_ids_to_mat(token_type_ids) - added:
method TFFunnelBaseLayer.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFFunnelBaseLayer.get_input_embeddings() - added:
method TFFunnelBaseLayer.set_input_embeddings(value) - added:
method TFFunnelBaseModel.call(inputs, **kwargs) - added:
method TFFunnelClassificationHead.call(hidden, training=…) - added:
method TFFunnelDecoder.call(final_hidden, first_block_hidden, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFFunnelDiscriminatorPredictions.call(discriminator_hidden_states) - added:
method TFFunnelEmbeddings.build(input_shape) - added:
method TFFunnelEmbeddings.call(input_ids=…, inputs_embeds=…, mode=…, training=…) - added:
method TFFunnelEncoder.call(inputs_embeds, attention_mask=…, token_type_ids=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFFunnelForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFFunnelForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFFunnelForPreTraining.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…, **kwargs) - added:
method TFFunnelForQuestionAnswering.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, start_positions=…, end_positions=…, training=…) - added:
method TFFunnelForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFFunnelForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFFunnelLayer.call(query, key, value, attention_inputs, output_attentions=…, training=…) - added:
method TFFunnelMainLayer.call(inputs, attention_mask=…, token_type_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFFunnelMainLayer.get_input_embeddings() - added:
method TFFunnelMainLayer.set_input_embeddings(value) - added:
method TFFunnelMaskedLMHead.build(input_shape) - added:
method TFFunnelMaskedLMHead.call(hidden_states, training=…) - added:
method TFFunnelModel.call(inputs, **kwargs) - added:
method TFFunnelPositionwiseFFN.call(hidden, training=…) - added:
method TFFunnelRelMultiheadAttention.build(input_shape) - added:
method TFFunnelRelMultiheadAttention.call(query, key, value, attention_inputs, output_attentions=…, training=…) - added:
method TFFunnelRelMultiheadAttention.relative_positional_attention(position_embeds, q_head, context_len, cls_mask=…) - added:
method TFFunnelRelMultiheadAttention.relative_token_type_attention(token_type_mat, q_head, cls_mask=…) - added:
property TFFunnelForMultipleChoice.dummy_inputs - added:
variable FUNNEL_INPUTS_DOCSTRING - added:
variable FUNNEL_START_DOCSTRING - added:
variable INF - added:
variable TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable logger
- removed:
function gelu(x) - changed:
class TFGPT2LMHeadModel(TFGPT2PreTrainedModel) → class TFGPT2LMHeadModel(TFGPT2PreTrainedModel, TFCausalLanguageModelingLoss) - changed:
method TFAttention.call(inputs, training=…) → method TFAttention.call(x, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - changed:
method TFBlock.call(inputs, training=…) → method TFBlock.call(x, layer_past, attention_mask, head_mask, use_cache, output_attentions, training=…) - changed:
method TFGPT2DoubleHeadsModel.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, use_cache=…, training=…) → 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=…) - changed:
method TFGPT2LMHeadModel.call(inputs, **kwargs) → 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=…) - changed:
method TFGPT2MainLayer.call(inputs, past=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, use_cache=…, training=…) → 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=…) - added:
class TFGPT2DoubleHeadsModelOutput(ModelOutput) - added:
field TFGPT2DoubleHeadsModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFGPT2DoubleHeadsModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFGPT2DoubleHeadsModelOutput.logits: tf.Tensor - added:
field TFGPT2DoubleHeadsModelOutput.mc_logits: tf.Tensor - added:
field TFGPT2DoubleHeadsModelOutput.past_key_values: Optional[List[tf.Tensor]] - added:
method TFGPT2MainLayer.set_input_embeddings(value)
- added:
class TFLongformerAttention(tf.keras.layers.Layer) - added:
class TFLongformerBaseModelOutput(ModelOutput) - added:
class TFLongformerBaseModelOutputWithPooling(ModelOutput) - added:
class TFLongformerEmbeddings(tf.keras.layers.Layer) - added:
class TFLongformerEncoder(tf.keras.layers.Layer) - added:
class TFLongformerForMaskedLM(TFLongformerPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFLongformerForQuestionAnswering(TFLongformerPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFLongformerIntermediate(tf.keras.layers.Layer) - added:
class TFLongformerLMHead(tf.keras.layers.Layer) - added:
class TFLongformerLayer(tf.keras.layers.Layer) - added:
class TFLongformerMainLayer(tf.keras.layers.Layer) - added:
class TFLongformerModel(TFLongformerPreTrainedModel) - added:
class TFLongformerOutput(tf.keras.layers.Layer) - added:
class TFLongformerPooler(tf.keras.layers.Layer) - added:
class TFLongformerPreTrainedModel(TFPreTrainedModel) - added:
class TFLongformerQuestionAnsweringModelOutput(ModelOutput) - added:
class TFLongformerSelfAttention(tf.keras.layers.Layer) - added:
class TFLongformerSelfOutput(tf.keras.layers.Layer) - added:
constructor TFLongformerAttention(config, layer_id=…, **kwargs) - added:
constructor TFLongformerEmbeddings(config, **kwargs) - added:
constructor TFLongformerEncoder(config, **kwargs) - added:
constructor TFLongformerForMaskedLM(config, *inputs, **kwargs) - added:
constructor TFLongformerForQuestionAnswering(config, *inputs, **kwargs) - added:
constructor TFLongformerIntermediate(config, **kwargs) - added:
constructor TFLongformerLMHead(config, input_embeddings, **kwargs) - added:
constructor TFLongformerLayer(config, layer_id=…, **kwargs) - added:
constructor TFLongformerMainLayer(config, **kwargs) - added:
constructor TFLongformerModel(config, *inputs, **kwargs) - added:
constructor TFLongformerOutput(config, **kwargs) - added:
constructor TFLongformerPooler(config, **kwargs) - added:
constructor TFLongformerSelfAttention(config, layer_id, **kwargs) - added:
constructor TFLongformerSelfOutput(config, **kwargs) - added:
field TFLongformerBaseModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutput.global_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutput.last_hidden_state: tf.Tensor - added:
field TFLongformerBaseModelOutputWithPooling.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutputWithPooling.global_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutputWithPooling.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerBaseModelOutputWithPooling.last_hidden_state: tf.Tensor - added:
field TFLongformerBaseModelOutputWithPooling.pooler_output: tf.Tensor - added:
field TFLongformerForMaskedLM.authorized_missing_keys - added:
field TFLongformerForQuestionAnswering.authorized_missing_keys - added:
field TFLongformerMainLayer.config_class - added:
field TFLongformerPreTrainedModel.base_model_prefix - added:
field TFLongformerPreTrainedModel.config_class - added:
field TFLongformerQuestionAnsweringModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerQuestionAnsweringModelOutput.end_logits: tf.Tensor - added:
field TFLongformerQuestionAnsweringModelOutput.global_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLongformerQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] - added:
field TFLongformerQuestionAnsweringModelOutput.start_logits: tf.Tensor - added:
method TFLongformerAttention.call(inputs, training=…) - added:
method TFLongformerAttention.prune_heads(heads) - added:
method TFLongformerEmbeddings.build(input_shape) - added:
method TFLongformerEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - added:
method TFLongformerEmbeddings.create_position_ids_from_input_ids(x) - added:
method TFLongformerEmbeddings.create_position_ids_from_inputs_embeds(inputs_embeds) - added:
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=…) - added:
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=…) - added:
method TFLongformerForMaskedLM.get_output_embeddings() - added:
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=…) - added:
method TFLongformerIntermediate.call(hidden_states) - added:
method TFLongformerLMHead.build(input_shape) - added:
method TFLongformerLMHead.call(features) - added:
method TFLongformerLayer.call(inputs, training=…) - added:
method TFLongformerMainLayer.call(inputs, attention_mask=…, global_attention_mask=…, token_type_ids=…, position_ids=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFLongformerMainLayer.get_input_embeddings() - added:
method TFLongformerMainLayer.set_input_embeddings(value) - added:
method TFLongformerModel.call(inputs, **kwargs) - added:
method TFLongformerOutput.call(hidden_states, input_tensor, training=…) - added:
method TFLongformerPooler.call(hidden_states) - added:
method TFLongformerSelfAttention.call(inputs, training=…) - added:
method TFLongformerSelfAttention.reshape_and_transpose(vector, batch_size) - added:
method TFLongformerSelfOutput.call(hidden_states, input_tensor, training=…) - added:
property TFLongformerPreTrainedModel.dummy_inputs - added:
variable LONGFORMER_INPUTS_DOCSTRING - added:
variable LONGFORMER_START_DOCSTRING - added:
variable TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable logger
- added:
class TFLxmertAttention(tf.keras.layers.Layer) - added:
class TFLxmertAttentionOutput(tf.keras.layers.Layer) - added:
class TFLxmertCrossAttentionLayer(tf.keras.layers.Layer) - added:
class TFLxmertEmbeddings(tf.keras.layers.Layer) - added:
class TFLxmertEncoder(tf.keras.layers.Layer) - added:
class TFLxmertForPreTraining(TFLxmertPreTrainedModel) - added:
class TFLxmertForPreTrainingOutput(ModelOutput) - added:
class TFLxmertIntermediate(tf.keras.layers.Layer) - added:
class TFLxmertLMPredictionHead(tf.keras.layers.Layer) - added:
class TFLxmertLayer(tf.keras.layers.Layer) - added:
class TFLxmertMLMHead(tf.keras.layers.Layer) - added:
class TFLxmertMainLayer(tf.keras.layers.Layer) - added:
class TFLxmertModel(TFLxmertPreTrainedModel) - added:
class TFLxmertModelOutput(ModelOutput) - added:
class TFLxmertOutput(tf.keras.layers.Layer) - added:
class TFLxmertPooler(tf.keras.layers.Layer) - added:
class TFLxmertPreTrainedModel(TFPreTrainedModel) - added:
class TFLxmertPreTrainingHeads(tf.keras.layers.Layer) - added:
class TFLxmertPredictionHeadTransform(tf.keras.layers.Layer) - added:
class TFLxmertSelfAttentionLayer(tf.keras.layers.Layer) - added:
class TFLxmertVisualAnswerHead(tf.keras.layers.Layer) - added:
class TFLxmertVisualFeatureEncoder(tf.keras.layers.Layer) - added:
class TFLxmertVisualObjHead(tf.keras.layers.Layer) - added:
class TFLxmertXLayer(tf.keras.layers.Layer) - added:
constructor TFLxmertAttention(config, **kwargs) - added:
constructor TFLxmertAttentionOutput(config, **kwargs) - added:
constructor TFLxmertCrossAttentionLayer(config, **kwargs) - added:
constructor TFLxmertEmbeddings(config, **kwargs) - added:
constructor TFLxmertEncoder(config, **kwargs) - added:
constructor TFLxmertForPreTraining(config, *inputs, **kwargs) - added:
constructor TFLxmertIntermediate(config, **kwargs) - added:
constructor TFLxmertLMPredictionHead(config, input_embeddings, **kwargs) - added:
constructor TFLxmertLayer(config, **kwargs) - added:
constructor TFLxmertMLMHead(config, input_embeddings, **kwargs) - added:
constructor TFLxmertMainLayer(config, **kwargs) - added:
constructor TFLxmertModel(config, *inputs, **kwargs) - added:
constructor TFLxmertOutput(config, **kwargs) - added:
constructor TFLxmertPooler(config, **kwargs) - added:
constructor TFLxmertPreTrainingHeads(config, input_embeddings, **kwargs) - added:
constructor TFLxmertPredictionHeadTransform(config, **kwargs) - added:
constructor TFLxmertSelfAttentionLayer(config, **kwargs) - added:
constructor TFLxmertVisualAnswerHead(config, num_labels, **kwargs) - added:
constructor TFLxmertVisualFeatureEncoder(config, **kwargs) - added:
constructor TFLxmertVisualObjHead(config, **kwargs) - added:
constructor TFLxmertXLayer(config, **kwargs) - added:
field TFLxmertForPreTrainingOutput.cross_encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertForPreTrainingOutput.cross_relationship_score: Optional[tf.Tensor] - added:
field TFLxmertForPreTrainingOutput.language_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertForPreTrainingOutput.language_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertForPreTrainingOutput.loss: [tf.Tensor] - added:
field TFLxmertForPreTrainingOutput.prediction_logits: Optional[tf.Tensor] - added:
field TFLxmertForPreTrainingOutput.question_answering_score: Optional[tf.Tensor] - added:
field TFLxmertForPreTrainingOutput.vision_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertForPreTrainingOutput.vision_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertMainLayer.config_class - added:
field TFLxmertModelOutput.cross_encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertModelOutput.language_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertModelOutput.language_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertModelOutput.language_output: Optional[tf.Tensor] - added:
field TFLxmertModelOutput.pooled_output: Optional[tf.Tensor] - added:
field TFLxmertModelOutput.vision_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertModelOutput.vision_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFLxmertModelOutput.vision_output: Optional[tf.Tensor] - added:
field TFLxmertPreTrainedModel.base_model_prefix - added:
field TFLxmertPreTrainedModel.config_class - added:
method TFLxmertAttention.call(hidden_states, context, attention_mask, output_attentions, training=…) - added:
method TFLxmertAttention.transpose_for_scores(x, batch_size) - added:
method TFLxmertAttentionOutput.call(hidden_states, input_tensor, training=…) - added:
method TFLxmertCrossAttentionLayer.call(input_tensor, ctx_tensor, ctx_att_mask, output_attentions=…, training=…) - added:
method TFLxmertEmbeddings.build(input_shape) - added:
method TFLxmertEmbeddings.call(inputs, mode=…, training=…) - added:
method TFLxmertEncoder.call(lang_feats=…, lang_attention_mask=…, visual_feats=…, visual_pos=…, visual_attention_mask=…, output_attentions=…, training=…) - added:
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=…) - added:
method TFLxmertIntermediate.call(hidden_states) - added:
method TFLxmertLMPredictionHead.build(input_shape) - added:
method TFLxmertLMPredictionHead.call(hidden_states) - added:
method TFLxmertLayer.call(hidden_states, attention_mask, output_attentions, training=…) - added:
method TFLxmertMLMHead.call(sequence_output) - added:
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=…) - added:
method TFLxmertMainLayer.get_input_embeddings() - added:
method TFLxmertMainLayer.set_input_embeddings(value) - added:
method TFLxmertModel.call(inputs, *args, **kwargs) - added:
method TFLxmertOutput.call(hidden_states, input_tensor, training=…) - added:
method TFLxmertPooler.call(hidden_states) - added:
method TFLxmertPreTrainingHeads.call(sequence_output, pooled_output) - added:
method TFLxmertPredictionHeadTransform.call(hidden_states) - added:
method TFLxmertSelfAttentionLayer.call(input_tensor, attention_mask, output_attentions, training=…) - added:
method TFLxmertVisualAnswerHead.call(hidden_states) - added:
method TFLxmertVisualFeatureEncoder.call(visn_input, training=…) - added:
method TFLxmertVisualObjHead.call(hidden_states) - added:
method TFLxmertXLayer.call(lang_feats, lang_attention_mask, visn_feats, visn_attention_mask, output_attentions, training=…) - added:
method TFLxmertXLayer.cross_att(lang_input, lang_attention_mask, visn_input, visn_attention_mask, output_attentions, training=…) - added:
method TFLxmertXLayer.output_fc(lang_input, visn_input, training=…) - added:
method TFLxmertXLayer.self_att(lang_input, lang_attention_mask, visn_input, visn_attention_mask, training=…) - added:
property TFLxmertForPreTraining.dummy_inputs - added:
property TFLxmertMainLayer.dummy_inputs - added:
property TFLxmertPreTrainedModel.dummy_inputs -> Dict[str, tf.Tensor] - added:
variable LXMERT_INPUTS_DOCSTRING - added:
variable LXMERT_START_DOCSTRING - added:
variable TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable logger
- added:
class TFMarianMTModel(TFBartForConditionalGeneration) - added:
field TFMarianMTModel.authorized_missing_keys - added:
field TFMarianMTModel.config_class - added:
method TFMarianMTModel.adjust_logits_during_generation(logits, cur_len, max_length) - added:
variable START_DOCSTRING - added:
variable logger
- added:
class TFMBartForConditionalGeneration(TFBartForConditionalGeneration) - added:
field TFMBartForConditionalGeneration.config_class - added:
variable START_DOCSTRING - added:
variable logger
- added:
class TFBottleneck(tf.keras.layers.Layer) - added:
class TFBottleneckLayer(tf.keras.layers.Layer) - added:
class TFFFNLayer(tf.keras.layers.Layer) - added:
class TFFFNOutput(tf.keras.layers.Layer) - added:
class TFLayerNorm(tf.keras.layers.LayerNormalization) - added:
class TFMobileBertAttention(tf.keras.layers.Layer) - added:
class TFMobileBertEmbeddings(tf.keras.layers.Layer) - added:
class TFMobileBertEncoder(tf.keras.layers.Layer) - added:
class TFMobileBertForMaskedLM(TFMobileBertPreTrainedModel, TFMaskedLanguageModelingLoss) - added:
class TFMobileBertForMultipleChoice(TFMobileBertPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFMobileBertForNextSentencePrediction(TFMobileBertPreTrainedModel) - added:
class TFMobileBertForPreTraining(TFMobileBertPreTrainedModel) - added:
class TFMobileBertForPreTrainingOutput(ModelOutput) - added:
class TFMobileBertForQuestionAnswering(TFMobileBertPreTrainedModel, TFQuestionAnsweringLoss) - added:
class TFMobileBertForSequenceClassification(TFMobileBertPreTrainedModel, TFSequenceClassificationLoss) - added:
class TFMobileBertForTokenClassification(TFMobileBertPreTrainedModel, TFTokenClassificationLoss) - added:
class TFMobileBertIntermediate(tf.keras.layers.Layer) - added:
class TFMobileBertLMPredictionHead(tf.keras.layers.Layer) - added:
class TFMobileBertLayer(tf.keras.layers.Layer) - added:
class TFMobileBertMLMHead(tf.keras.layers.Layer) - added:
class TFMobileBertMainLayer(tf.keras.layers.Layer) - added:
class TFMobileBertModel(TFMobileBertPreTrainedModel) - added:
class TFMobileBertOnlyNSPHead(tf.keras.layers.Layer) - added:
class TFMobileBertOutput(tf.keras.layers.Layer) - added:
class TFMobileBertPooler(tf.keras.layers.Layer) - added:
class TFMobileBertPreTrainedModel(TFPreTrainedModel) - added:
class TFMobileBertPredictionHeadTransform(tf.keras.layers.Layer) - added:
class TFMobileBertSelfAttention(tf.keras.layers.Layer) - added:
class TFMobileBertSelfOutput(tf.keras.layers.Layer) - added:
class TFNoNorm(tf.keras.layers.Layer) - added:
class TFOutputBottleneck(tf.keras.layers.Layer) - added:
constructor TFBottleneck(config, **kwargs) - added:
constructor TFBottleneckLayer(config, **kwargs) - added:
constructor TFFFNLayer(config, **kwargs) - added:
constructor TFFFNOutput(config, **kwargs) - added:
constructor TFLayerNorm(feat_size, *args, **kwargs) - added:
constructor TFMobileBertAttention(config, **kwargs) - added:
constructor TFMobileBertEmbeddings(config, **kwargs) - added:
constructor TFMobileBertEncoder(config, **kwargs) - 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 TFMobileBertIntermediate(config, **kwargs) - added:
constructor TFMobileBertLMPredictionHead(config, **kwargs) - added:
constructor TFMobileBertLayer(config, **kwargs) - added:
constructor TFMobileBertMLMHead(config, **kwargs) - added:
constructor TFMobileBertMainLayer(config, **kwargs) - added:
constructor TFMobileBertModel(config, *inputs, **kwargs) - added:
constructor TFMobileBertOnlyNSPHead(config, **kwargs) - added:
constructor TFMobileBertOutput(config, **kwargs) - added:
constructor TFMobileBertPooler(config, **kwargs) - added:
constructor TFMobileBertPredictionHeadTransform(config, **kwargs) - added:
constructor TFMobileBertSelfAttention(config, **kwargs) - added:
constructor TFMobileBertSelfOutput(config, **kwargs) - added:
constructor TFNoNorm(feat_size, epsilon=…, **kwargs) - added:
constructor TFOutputBottleneck(config, **kwargs) - added:
field TFMobileBertForMaskedLM.authorized_missing_keys - added:
field TFMobileBertForPreTrainingOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFMobileBertForPreTrainingOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFMobileBertForPreTrainingOutput.loss: Optional[tf.Tensor] - added:
field TFMobileBertForPreTrainingOutput.prediction_logits: tf.Tensor - added:
field TFMobileBertForPreTrainingOutput.seq_relationship_logits: tf.Tensor - added:
field TFMobileBertForQuestionAnswering.authorized_missing_keys - added:
field TFMobileBertForTokenClassification.authorized_missing_keys - added:
field TFMobileBertMainLayer.config_class - added:
field TFMobileBertPreTrainedModel.base_model_prefix - added:
field TFMobileBertPreTrainedModel.config_class - added:
method TFBottleneck.call(hidden_states) - added:
method TFBottleneckLayer.call(inputs) - added:
method TFFFNLayer.call(hidden_states) - added:
method TFFFNOutput.call(hidden_states, residual_tensor) - added:
method TFMobileBertAttention.call(query_tensor, key_tensor, value_tensor, layer_input, attention_mask, head_mask, output_attentions, training=…) - added:
method TFMobileBertAttention.prune_heads(heads) - added:
method TFMobileBertEmbeddings.build(input_shape) - added:
method TFMobileBertEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - added:
method TFMobileBertEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - added:
method TFMobileBertForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFMobileBertForMaskedLM.get_output_embeddings() - added:
method TFMobileBertForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFMobileBertForNextSentencePrediction.call(inputs, **kwargs) - added:
method TFMobileBertForPreTraining.call(inputs, **kwargs) - added:
method TFMobileBertForPreTraining.get_output_embeddings() - added:
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=…) - added:
method TFMobileBertForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFMobileBertForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFMobileBertIntermediate.call(hidden_states) - added:
method TFMobileBertLMPredictionHead.build(input_shape) - added:
method TFMobileBertLMPredictionHead.call(hidden_states) - added:
method TFMobileBertLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFMobileBertMLMHead.call(sequence_output) - added:
method TFMobileBertMainLayer.call(inputs, 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 TFMobileBertModel.call(inputs, **kwargs) - added:
method TFMobileBertOnlyNSPHead.call(pooled_output) - added:
method TFMobileBertOutput.call(hidden_states, residual_tensor_1, residual_tensor_2, training=…) - added:
method TFMobileBertPooler.call(hidden_states) - added:
method TFMobileBertPredictionHeadTransform.call(hidden_states) - added:
method TFMobileBertSelfAttention.call(query_tensor, key_tensor, value_tensor, attention_mask, head_mask, output_attentions, training=…) - added:
method TFMobileBertSelfAttention.transpose_for_scores(x, batch_size) - added:
method TFMobileBertSelfOutput.call(hidden_states, residual_tensor, training=…) - added:
method TFNoNorm.build(input_shape) - added:
method TFNoNorm.call(inputs: tf.Tensor) - added:
method TFOutputBottleneck.call(hidden_states, residual_tensor, training=…) - added:
property TFMobileBertForMultipleChoice.dummy_inputs - added:
variable MOBILEBERT_INPUTS_DOCSTRING - added:
variable MOBILEBERT_START_DOCSTRING - added:
variable NORM2FN - added:
variable TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable logger
- removed:
function gelu(x) - removed:
function swish(x) - removed:
variable ACT_FNS - changed:
class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel) → class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel, TFCausalLanguageModelingLoss) - changed:
method TFAttention.call(inputs, training=…) → method TFAttention.call(x, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFBlock.call(inputs, training=…) → method TFBlock.call(x, attention_mask, head_mask, output_attentions, training=…) - changed:
method TFOpenAIGPTDoubleHeadsModel.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, mc_token_ids=…, training=…) → 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=…) - changed:
method TFOpenAIGPTLMHeadModel.call(inputs, **kwargs) → method TFOpenAIGPTLMHeadModel.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFOpenAIGPTMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, training=…) → method TFOpenAIGPTMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
class TFOpenAIGPTDoubleHeadsModelOutput(ModelOutput) - added:
field TFOpenAIGPTDoubleHeadsModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFOpenAIGPTDoubleHeadsModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFOpenAIGPTDoubleHeadsModelOutput.logits: tf.Tensor - added:
field TFOpenAIGPTDoubleHeadsModelOutput.mc_logits: tf.Tensor - added:
field TFOpenAIGPTMainLayer.config_class - added:
method TFOpenAIGPTMainLayer.set_input_embeddings(value)
- added:
class TFBaseModelOutput(ModelOutput) - added:
class TFBaseModelOutputWithPast(ModelOutput) - added:
class TFBaseModelOutputWithPooling(ModelOutput) - added:
class TFCausalLMOutput(ModelOutput) - added:
class TFCausalLMOutputWithPast(ModelOutput) - added:
class TFMaskedLMOutput(ModelOutput) - added:
class TFMultipleChoiceModelOutput(ModelOutput) - added:
class TFNextSentencePredictorOutput(ModelOutput) - added:
class TFQuestionAnsweringModelOutput(ModelOutput) - added:
class TFSeq2SeqLMOutput(ModelOutput) - added:
class TFSeq2SeqModelOutput(ModelOutput) - added:
class TFSeq2SeqQuestionAnsweringModelOutput(ModelOutput) - added:
class TFSeq2SeqSequenceClassifierOutput(ModelOutput) - added:
class TFSequenceClassifierOutput(ModelOutput) - added:
class TFTokenClassifierOutput(ModelOutput) - added:
field TFBaseModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutput.last_hidden_state: tf.Tensor - added:
field TFBaseModelOutputWithPast.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutputWithPast.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutputWithPast.last_hidden_state: tf.Tensor - added:
field TFBaseModelOutputWithPast.past_key_values: Optional[List[tf.Tensor]] - added:
field TFBaseModelOutputWithPooling.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutputWithPooling.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFBaseModelOutputWithPooling.last_hidden_state: tf.Tensor - added:
field TFBaseModelOutputWithPooling.pooler_output: tf.Tensor - added:
field TFCausalLMOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFCausalLMOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFCausalLMOutput.logits: tf.Tensor - added:
field TFCausalLMOutput.loss: Optional[tf.Tensor] - added:
field TFCausalLMOutputWithPast.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFCausalLMOutputWithPast.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFCausalLMOutputWithPast.logits: tf.Tensor - added:
field TFCausalLMOutputWithPast.loss: Optional[tf.Tensor] - added:
field TFCausalLMOutputWithPast.past_key_values: Optional[List[tf.Tensor]] - added:
field TFMaskedLMOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFMaskedLMOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFMaskedLMOutput.logits: tf.Tensor - added:
field TFMaskedLMOutput.loss: Optional[tf.Tensor] - added:
field TFMultipleChoiceModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFMultipleChoiceModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFMultipleChoiceModelOutput.logits: tf.Tensor - added:
field TFMultipleChoiceModelOutput.loss: Optional[tf.Tensor] - added:
field TFNextSentencePredictorOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFNextSentencePredictorOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFNextSentencePredictorOutput.logits: tf.Tensor - added:
field TFQuestionAnsweringModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFQuestionAnsweringModelOutput.end_logits: tf.Tensor - added:
field TFQuestionAnsweringModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] - added:
field TFQuestionAnsweringModelOutput.start_logits: tf.Tensor - added:
field TFSeq2SeqLMOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqLMOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqLMOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqLMOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqLMOutput.encoder_last_hidden_state: Optional[tf.Tensor] - added:
field TFSeq2SeqLMOutput.logits: tf.Tensor - added:
field TFSeq2SeqLMOutput.loss: Optional[tf.Tensor] - added:
field TFSeq2SeqLMOutput.past_key_values: Optional[List[tf.Tensor]] - added:
field TFSeq2SeqModelOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqModelOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqModelOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqModelOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqModelOutput.encoder_last_hidden_state: Optional[tf.Tensor] - added:
field TFSeq2SeqModelOutput.last_hidden_state: tf.Tensor - added:
field TFSeq2SeqModelOutput.past_key_values: Optional[List[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.encoder_last_hidden_state: Optional[tf.Tensor] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.end_logits: tf.Tensor - added:
field TFSeq2SeqQuestionAnsweringModelOutput.loss: Optional[tf.Tensor] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.past_key_values: Optional[List[tf.Tensor]] - added:
field TFSeq2SeqQuestionAnsweringModelOutput.start_logits: tf.Tensor - added:
field TFSeq2SeqSequenceClassifierOutput.decoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqSequenceClassifierOutput.decoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqSequenceClassifierOutput.encoder_attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqSequenceClassifierOutput.encoder_hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSeq2SeqSequenceClassifierOutput.encoder_last_hidden_state: Optional[tf.Tensor] - added:
field TFSeq2SeqSequenceClassifierOutput.logits: tf.Tensor - added:
field TFSeq2SeqSequenceClassifierOutput.loss: Optional[tf.Tensor] - added:
field TFSeq2SeqSequenceClassifierOutput.past_key_values: Optional[List[tf.Tensor]] - added:
field TFSequenceClassifierOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFSequenceClassifierOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFSequenceClassifierOutput.logits: tf.Tensor - added:
field TFSequenceClassifierOutput.loss: Optional[tf.Tensor] - added:
field TFTokenClassifierOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFTokenClassifierOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFTokenClassifierOutput.logits: tf.Tensor - added:
field TFTokenClassifierOutput.loss: Optional[tf.Tensor]
- added:
class TFPegasusForConditionalGeneration(TFBartForConditionalGeneration) - added:
field TFPegasusForConditionalGeneration.authorized_missing_keys - added:
field TFPegasusForConditionalGeneration.config_class - added:
variable START_DOCSTRING - added:
variable logger
- changed:
class TFRobertaEmbeddings(TFBertEmbeddings) → class TFRobertaEmbeddings(tf.keras.layers.Layer) - changed:
class TFRobertaForMaskedLM(TFRobertaPreTrainedModel) → class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLoss) - changed:
class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel) → class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel) → class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceClassificationLoss) - changed:
class TFRobertaForTokenClassification(TFRobertaPreTrainedModel) → class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassificationLoss) - changed:
class TFRobertaMainLayer(TFBertMainLayer) → class TFRobertaMainLayer(tf.keras.layers.Layer) - changed:
method TFRobertaForMaskedLM.call(inputs, **kwargs) → method TFRobertaForMaskedLM.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFRobertaForQuestionAnswering.call(inputs, **kwargs) → 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=…) - changed:
method TFRobertaForSequenceClassification.call(inputs, **kwargs) → method TFRobertaForSequenceClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFRobertaForTokenClassification.call(inputs, **kwargs) → method TFRobertaForTokenClassification.call(inputs=…, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
class TFRobertaAttention(tf.keras.layers.Layer) - added:
class TFRobertaEncoder(tf.keras.layers.Layer) - added:
class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFRobertaIntermediate(tf.keras.layers.Layer) - added:
class TFRobertaLayer(tf.keras.layers.Layer) - added:
class TFRobertaOutput(tf.keras.layers.Layer) - added:
class TFRobertaPooler(tf.keras.layers.Layer) - added:
class TFRobertaSelfAttention(tf.keras.layers.Layer) - added:
class TFRobertaSelfOutput(tf.keras.layers.Layer) - added:
constructor TFRobertaAttention(config, **kwargs) - added:
constructor TFRobertaEncoder(config, **kwargs) - added:
constructor TFRobertaForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFRobertaIntermediate(config, **kwargs) - added:
constructor TFRobertaLayer(config, **kwargs) - added:
constructor TFRobertaOutput(config, **kwargs) - added:
constructor TFRobertaPooler(config, **kwargs) - added:
constructor TFRobertaSelfAttention(config, **kwargs) - added:
constructor TFRobertaSelfOutput(config, **kwargs) - added:
field TFRobertaForMaskedLM.authorized_missing_keys - added:
field TFRobertaForQuestionAnswering.authorized_missing_keys - added:
field TFRobertaForSequenceClassification.authorized_missing_keys - added:
field TFRobertaForTokenClassification.authorized_missing_keys - added:
field TFRobertaMainLayer.config_class - added:
method TFRobertaAttention.call(input_tensor, attention_mask, head_mask, output_attentions, training=…) - added:
method TFRobertaAttention.prune_heads(heads) - added:
method TFRobertaEmbeddings.build(input_shape) - added:
method TFRobertaEmbeddings.call(input_ids=…, position_ids=…, token_type_ids=…, inputs_embeds=…, mode=…, training=…) - added:
method TFRobertaEncoder.call(hidden_states, attention_mask, head_mask, output_attentions, output_hidden_states, return_dict, training=…) - added:
method TFRobertaForMultipleChoice.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - added:
method TFRobertaIntermediate.call(hidden_states) - added:
method TFRobertaLayer.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFRobertaMainLayer.call(inputs, attention_mask=…, token_type_ids=…, position_ids=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
method TFRobertaMainLayer.set_input_embeddings(value) - added:
method TFRobertaOutput.call(hidden_states, input_tensor, training=…) - added:
method TFRobertaPooler.call(hidden_states) - added:
method TFRobertaSelfAttention.call(hidden_states, attention_mask, head_mask, output_attentions, training=…) - added:
method TFRobertaSelfAttention.transpose_for_scores(x, batch_size) - added:
method TFRobertaSelfOutput.call(hidden_states, input_tensor, training=…) - added:
property TFRobertaForMultipleChoice.dummy_inputs
- changed:
class TFT5ForConditionalGeneration(TFT5PreTrainedModel) → class TFT5ForConditionalGeneration(TFT5PreTrainedModel, TFCausalLanguageModelingLoss) - changed:
constructor TFT5Attention(config, has_relative_attention_bias=…, **kwargs) → constructor TFT5Attention(config, has_relative_attention_bias=…, is_bidirectional=…, **kwargs) - changed:
method TFT5Attention.call(input, mask=…, kv=…, position_bias=…, cache=…, past_key_value_state=…, head_mask=…, query_length=…, use_cache=…, training=…) → method TFT5Attention.call(input, mask=…, kv=…, position_bias=…, past_key_value=…, head_mask=…, query_length=…, use_cache=…, training=…, output_attentions=…) - changed:
method TFT5Block.call(hidden_states, attention_mask=…, position_bias=…, encoder_hidden_states=…, encoder_attention_mask=…, encoder_decoder_position_bias=…, head_mask=…, past_key_value_state=…, use_cache=…, training=…) → 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=…) - changed:
method TFT5ForConditionalGeneration.call(inputs, **kwargs) → 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) - changed:
method TFT5LayerCrossAttention.call(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value_state=…, query_length=…, use_cache=…, training=…) → method TFT5LayerCrossAttention.call(hidden_states, kv, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, query_length=…, use_cache=…, output_attentions=…, training=…) - changed:
method TFT5LayerSelfAttention.call(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value_state=…, use_cache=…, training=…) → method TFT5LayerSelfAttention.call(hidden_states, attention_mask=…, position_bias=…, head_mask=…, past_key_value=…, use_cache=…, output_attentions=…, training=…) - changed:
method TFT5MainLayer.call(inputs, attention_mask=…, encoder_hidden_states=…, encoder_attention_mask=…, inputs_embeds=…, head_mask=…, past_key_value_states=…, use_cache=…, training=…) → 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 - changed:
method TFT5Model.call(inputs, **kwargs) → 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) - added:
field TFT5MainLayer.config_class - added:
method TFT5ForConditionalGeneration.set_input_embeddings(new_embeddings) - added:
method TFT5Model.set_input_embeddings(new_embeddings)
- removed:
class TFTransfoXLLMHead(tf.keras.layers.Layer) - removed:
constructor TFTransfoXLLMHead(config, input_embeddings, **kwargs) - removed:
method TFTransfoXLLMHead.build(input_shape) - removed:
method TFTransfoXLLMHead.call(hidden_states) - removed:
method TFTransfoXLMainLayer.reset_length(tgt_len, ext_len, mem_len) - changed:
constructor TFRelPartialLearnableDecoderLayer(n_head, d_model, d_head, d_inner, dropout, tgt_len=…, ext_len=…, mem_len=…, dropatt=…, pre_lnorm=…, r_w_bias=…, r_r_bias=…, output_attentions=…, layer_norm_epsilon=…, init_std=…, **kwargs) → 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) - changed:
constructor TFRelPartialLearnableMultiHeadAttn(n_head, d_model, d_head, dropout, dropatt=…, tgt_len=…, ext_len=…, mem_len=…, pre_lnorm=…, r_r_bias=…, r_w_bias=…, output_attentions=…, layer_norm_epsilon=…, init_std=…, **kwargs) → 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) - changed:
method TFRelPartialLearnableDecoderLayer.call(inputs, training=…) → method TFRelPartialLearnableDecoderLayer.call(dec_inp, r, dec_attn_mask, mems, head_mask, output_attentions, training=…) - changed:
method TFRelPartialLearnableMultiHeadAttn.call(inputs, training=…) → method TFRelPartialLearnableMultiHeadAttn.call(w, r, attn_mask, mems, head_mask, output_attentions, training=…) - changed:
method TFTransfoXLLMHeadModel.call(inputs, mems=…, head_mask=…, inputs_embeds=…, labels=…, training=…) → method TFTransfoXLLMHeadModel.call(inputs, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, labels=…, training=…) - changed:
method TFTransfoXLMainLayer.call(inputs, mems=…, head_mask=…, inputs_embeds=…, training=…) → method TFTransfoXLMainLayer.call(inputs, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…, training=…) - added:
class TFTransfoXLLMHeadModelOutput(ModelOutput) - added:
class TFTransfoXLMHead(tf.keras.layers.Layer) - added:
class TFTransfoXLModelOutput(ModelOutput) - added:
constructor TFTransfoXLMHead(config, input_embeddings, **kwargs) - added:
field TFTransfoXLLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFTransfoXLLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFTransfoXLLMHeadModelOutput.mems: List[tf.Tensor] - added:
field TFTransfoXLLMHeadModelOutput.prediction_scores: tf.Tensor - added:
field TFTransfoXLModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFTransfoXLModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFTransfoXLModelOutput.last_hidden_state: tf.Tensor - added:
field TFTransfoXLModelOutput.mems: List[tf.Tensor] - added:
method TFTransfoXLLMHeadModel.reset_memory_length(mem_len) - added:
method TFTransfoXLMHead.build(input_shape) - added:
method TFTransfoXLMHead.call(hidden_states) - added:
method TFTransfoXLMainLayer.reset_memory_length(mem_len) - added:
method TFTransfoXLMainLayer.set_input_embeddings(value)
- changed:
method TFAdaptiveSoftmaxMask.call(inputs, return_mean=…, training=…) → method TFAdaptiveSoftmaxMask.call(hidden, target, return_mean=…, training=…)
- removed:
class BeamHypotheses(object) - 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 scatter_values_on_batch_indices(values, batch_indices) - removed:
function set_tensor_by_indices_to_value(tensor, indices, value) - 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 TFPreTrainedModel.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 TFPreTrainedModel.prepare_inputs_for_generation(inputs, **kwargs) - removed:
property TFPreTrainedModel.dummy_inputs - changed:
class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin) → class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin) - changed:
constructor TFSequenceSummary(config, initializer_range=…, **kwargs) → constructor TFSequenceSummary(config: PretrainedConfig, initializer_range: float=…, **kwargs) - changed:
constructor TFSharedEmbeddings(vocab_size, hidden_size, initializer_range=…, **kwargs) → constructor TFSharedEmbeddings(vocab_size: int, hidden_size: int, initializer_range: Optional[float]=…, **kwargs) - changed:
function get_initializer(initializer_range=…) → function get_initializer(initializer_range: float=…) -> tf.initializers.TruncatedNormal - changed:
function shape_list(x) → function shape_list(x: tf.Tensor) -> List[int] - changed:
method TFPreTrainedModel.get_input_embeddings() → method TFPreTrainedModel.get_input_embeddings() -> tf.keras.layers.Layer - changed:
method TFPreTrainedModel.get_output_embeddings() → method TFPreTrainedModel.get_output_embeddings() -> tf.keras.layers.Layer - changed:
method TFPreTrainedModel.resize_token_embeddings(new_num_tokens=…) → method TFPreTrainedModel.resize_token_embeddings(new_num_tokens=…) -> tf.Variable - changed:
method TFSequenceSummary.call(inputs, training=…) → method TFSequenceSummary.call(inputs, cls_index=…, training=…) - changed:
method TFSharedEmbeddings.call(inputs, mode=…) → method TFSharedEmbeddings.call(inputs: tf.Tensor, mode: str=…) -> tf.Tensor - added:
class TFCausalLanguageModelingLoss - added:
class TFMaskedLanguageModelingLoss(TFCausalLanguageModelingLoss) - added:
class TFMultipleChoiceLoss(TFSequenceClassificationLoss) - added:
class TFQuestionAnsweringLoss - added:
class TFSequenceClassificationLoss - added:
class TFTokenClassificationLoss - added:
class TFWrappedEmbeddings - added:
constructor TFWrappedEmbeddings(layer, abs_scope_name=…) - added:
field TFPreTrainedModel.authorized_missing_keys - added:
field TFPreTrainedModel.authorized_unexpected_keys - added:
function cast_bool_to_primitive(bool_variable: Union[tf.Tensor, bool], default_tensor_to_true=…) -> bool - added:
function detect_tf_missing_unexpected_layers(model, resolved_archive_file) - added:
function load_tf_weights(model, resolved_archive_file) - added:
method TFCausalLanguageModelingLoss.compute_loss(labels, logits) - added:
method TFPreTrainedModel.set_input_embeddings(value) - added:
method TFQuestionAnsweringLoss.compute_loss(labels, logits) - added:
method TFSequenceClassificationLoss.compute_loss(labels, logits) - added:
method TFSharedEmbeddings.get_config() - added:
method TFTokenClassificationLoss.compute_loss(labels, logits) - added:
method TFWrappedEmbeddings.call(inputs, mode=…) - added:
property TFPreTrainedModel.dummy_inputs -> Dict[str, tf.Tensor]
- removed:
class TFMultiHeadAttention(tf.keras.layers.Layer) - removed:
class TFTransformerFFN(tf.keras.layers.Layer) - removed:
constructor TFMultiHeadAttention(n_heads, dim, config, **kwargs) - removed:
constructor TFTransformerFFN(in_dim, dim_hidden, out_dim, config, **kwargs) - removed:
field TFMultiHeadAttention.NEW_ID - removed:
function gelu(x) - removed:
method TFMultiHeadAttention.call(inputs, training=…) - removed:
method TFMultiHeadAttention.prune_heads(heads) - removed:
method TFTransformerFFN.call(input, training=…) - changed:
class TFXLMForQuestionAnsweringSimple(TFXLMPreTrainedModel) → class TFXLMForQuestionAnsweringSimple(TFXLMPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFXLMForSequenceClassification(TFXLMPreTrainedModel) → class TFXLMForSequenceClassification(TFXLMPreTrainedModel, TFSequenceClassificationLoss) - changed:
method TFXLMForQuestionAnsweringSimple.call(inputs, **kwargs) → 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=…) - changed:
method TFXLMForSequenceClassification.call(inputs, **kwargs) → 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=…) - changed:
method TFXLMMainLayer.call(inputs, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, training=…) → 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=…) - added:
class TFXLMForMultipleChoice(TFXLMPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFXLMForTokenClassification(TFXLMPreTrainedModel, TFTokenClassificationLoss) - added:
class TFXLMMultiHeadAttention(tf.keras.layers.Layer) - added:
class TFXLMTransformerFFN(tf.keras.layers.Layer) - added:
class TFXLMWithLMHeadModelOutput(ModelOutput) - added:
constructor TFXLMForMultipleChoice(config, *inputs, **kwargs) - added:
constructor TFXLMForTokenClassification(config, *inputs, **kwargs) - added:
constructor TFXLMMultiHeadAttention(n_heads, dim, config, **kwargs) - added:
constructor TFXLMTransformerFFN(in_dim, dim_hidden, out_dim, config, **kwargs) - added:
field TFXLMMainLayer.config_class - added:
field TFXLMMultiHeadAttention.NEW_ID - added:
field TFXLMWithLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLMWithLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLMWithLMHeadModelOutput.logits: tf.Tensor - added:
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=…) - added:
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=…) - added:
method TFXLMMainLayer.set_input_embeddings(value) - added:
method TFXLMMultiHeadAttention.call(input, mask, kv, cache, head_mask, output_attentions, training=…) - added:
method TFXLMMultiHeadAttention.prune_heads(heads) - added:
method TFXLMTransformerFFN.call(input, training=…) - added:
property TFXLMForMultipleChoice.dummy_inputs
- added:
class TFXLMRobertaForMultipleChoice(TFRobertaForMultipleChoice) - added:
class TFXLMRobertaForQuestionAnswering(TFRobertaForQuestionAnswering) - added:
field TFXLMRobertaForMultipleChoice.config_class - added:
field TFXLMRobertaForQuestionAnswering.config_class
- removed:
function gelu(x) - removed:
function swish(x) - removed:
variable ACT2FN - changed:
class TFXLNetForQuestionAnsweringSimple(TFXLNetPreTrainedModel) → class TFXLNetForQuestionAnsweringSimple(TFXLNetPreTrainedModel, TFQuestionAnsweringLoss) - changed:
class TFXLNetForSequenceClassification(TFXLNetPreTrainedModel) → class TFXLNetForSequenceClassification(TFXLNetPreTrainedModel, TFSequenceClassificationLoss) - changed:
class TFXLNetForTokenClassification(TFXLNetPreTrainedModel) → class TFXLNetForTokenClassification(TFXLNetPreTrainedModel, TFTokenClassificationLoss) - changed:
class TFXLNetLMHeadModel(TFXLNetPreTrainedModel) → class TFXLNetLMHeadModel(TFXLNetPreTrainedModel, TFCausalLanguageModelingLoss) - changed:
method TFXLNetForQuestionAnsweringSimple.call(inputs, **kwargs) → 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=…) - changed:
method TFXLNetForSequenceClassification.call(inputs, **kwargs) → 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=…) - changed:
method TFXLNetForTokenClassification.call(inputs, **kwargs) → 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=…) - changed:
method TFXLNetLMHeadModel.call(inputs, **kwargs) → 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=…) - changed:
method TFXLNetLayer.call(inputs, training=…) → method TFXLNetLayer.call(output_h, output_g, non_tgt_mask, attn_mask, pos_emb, seg_mat, mems, target_mapping, head_mask, output_attentions, training=…) - changed:
method TFXLNetMainLayer.call(inputs, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, training=…) → 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=…) - changed:
method TFXLNetRelativeAttention.call(inputs, training=…) → method TFXLNetRelativeAttention.call(h, g, attn_mask_h, attn_mask_g, r, seg_mat, mems, target_mapping, head_mask, output_attentions, training=…) - changed:
method TFXLNetRelativeAttention.post_attention(inputs, residual=…, training=…) → method TFXLNetRelativeAttention.post_attention(h, attn_vec, residual=…, training=…) - changed:
method TFXLNetRelativeAttention.rel_attn_core(inputs, training=…) → 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=…) - added:
class TFXLNetForMultipleChoice(TFXLNetPreTrainedModel, TFMultipleChoiceLoss) - added:
class TFXLNetForMultipleChoiceOutput(ModelOutput) - added:
class TFXLNetForQuestionAnsweringSimpleOutput(ModelOutput) - added:
class TFXLNetForSequenceClassificationOutput(ModelOutput) - added:
class TFXLNetForTokenClassificationOutput(ModelOutput) - added:
class TFXLNetLMHeadModelOutput(ModelOutput) - added:
class TFXLNetModelOutput(ModelOutput) - added:
constructor TFXLNetForMultipleChoice(config, *inputs, **kwargs) - added:
field TFXLNetForMultipleChoiceOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForMultipleChoiceOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForMultipleChoiceOutput.logits: tf.Tensor - added:
field TFXLNetForMultipleChoiceOutput.loss: Optional[tf.Tensor] - added:
field TFXLNetForMultipleChoiceOutput.mems: Optional[List[tf.Tensor]] - added:
field TFXLNetForQuestionAnsweringSimpleOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForQuestionAnsweringSimpleOutput.end_logits: tf.Tensor - added:
field TFXLNetForQuestionAnsweringSimpleOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForQuestionAnsweringSimpleOutput.loss: Optional[tf.Tensor] - added:
field TFXLNetForQuestionAnsweringSimpleOutput.mems: Optional[List[tf.Tensor]] - added:
field TFXLNetForQuestionAnsweringSimpleOutput.start_logits: tf.Tensor - added:
field TFXLNetForSequenceClassificationOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForSequenceClassificationOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForSequenceClassificationOutput.logits: tf.Tensor - added:
field TFXLNetForSequenceClassificationOutput.loss: Optional[tf.Tensor] - added:
field TFXLNetForSequenceClassificationOutput.mems: Optional[List[tf.Tensor]] - added:
field TFXLNetForTokenClassificationOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForTokenClassificationOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetForTokenClassificationOutput.logits: tf.Tensor - added:
field TFXLNetForTokenClassificationOutput.loss: Optional[tf.Tensor] - added:
field TFXLNetForTokenClassificationOutput.mems: Optional[List[tf.Tensor]] - added:
field TFXLNetLMHeadModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetLMHeadModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetLMHeadModelOutput.logits: tf.Tensor - added:
field TFXLNetLMHeadModelOutput.loss: Optional[tf.Tensor] - added:
field TFXLNetLMHeadModelOutput.mems: Optional[List[tf.Tensor]] - added:
field TFXLNetModelOutput.attentions: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetModelOutput.hidden_states: Optional[Tuple[tf.Tensor]] - added:
field TFXLNetModelOutput.last_hidden_state: tf.Tensor - added:
field TFXLNetModelOutput.mems: Optional[List[tf.Tensor]] - added:
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=…) - added:
method TFXLNetMainLayer.set_input_embeddings(value) - added:
property TFXLNetForMultipleChoice.dummy_inputs
- removed:
method TransfoXLModel.reset_length(tgt_len, ext_len, mem_len) - changed:
constructor RelPartialLearnableMultiHeadAttn(n_head, d_model, d_head, dropout, dropatt=…, tgt_len=…, ext_len=…, mem_len=…, pre_lnorm=…, r_r_bias=…, r_w_bias=…, output_attentions=…, layer_norm_epsilon=…) → constructor RelPartialLearnableMultiHeadAttn(n_head, d_model, d_head, dropout, dropatt=…, pre_lnorm=…, r_r_bias=…, r_w_bias=…, layer_norm_epsilon=…) - changed:
method RelPartialLearnableDecoderLayer.forward(dec_inp, r, dec_attn_mask=…, mems=…, head_mask=…) → method RelPartialLearnableDecoderLayer.forward(dec_inp, r, dec_attn_mask=…, mems=…, head_mask=…, output_attentions=…) - changed:
method RelPartialLearnableMultiHeadAttn.forward(w, r, attn_mask=…, mems=…, head_mask=…) → method RelPartialLearnableMultiHeadAttn.forward(w, r, attn_mask=…, mems=…, head_mask=…, output_attentions=…) - changed:
method TransfoXLLMHeadModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…, labels=…) → method TransfoXLLMHeadModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…, labels=…, output_attentions=…, output_hidden_states=…, return_dict=…) - changed:
method TransfoXLLMHeadModel.prepare_inputs_for_generation(input_ids, past, **model_kwargs) → method TransfoXLLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, **model_kwargs) - changed:
method TransfoXLModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…) → method TransfoXLModel.forward(input_ids=…, mems=…, head_mask=…, inputs_embeds=…, output_attentions=…, output_hidden_states=…, return_dict=…) - added:
class TransfoXLLMHeadModelOutput(ModelOutput) - added:
class TransfoXLModelOutput(ModelOutput) - added:
field TransfoXLLMHeadModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field TransfoXLLMHeadModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field TransfoXLLMHeadModelOutput.losses: Optional[torch.FloatTensor] - added:
field TransfoXLLMHeadModelOutput.mems: List[torch.FloatTensor] - added:
field TransfoXLLMHeadModelOutput.prediction_scores: torch.FloatTensor - added:
field TransfoXLModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field TransfoXLModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field TransfoXLModelOutput.last_hidden_state: torch.FloatTensor - added:
field TransfoXLModelOutput.mems: List[torch.FloatTensor] - added:
method TransfoXLLMHeadModel.reset_memory_length(mem_len) - added:
method TransfoXLModel.reset_memory_length(mem_len) - added:
method TransfoXLPreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…, layer: Optional[int]=…) - added:
property TransfoXLLMHeadModelOutput.logits
- removed:
class BeamHypotheses(object) - removed:
constructor BeamHypotheses(num_beams, max_length, length_penalty, early_stopping) - removed:
function calc_banned_bad_words_ids(prev_input_ids: Iterable[int], bad_words_ids: Iterable[int]) -> Iterable[int] - removed:
function calc_banned_ngram_tokens(prev_input_ids: Tensor, num_hypos: int, no_repeat_ngram_size: int, cur_len: int) -> None - removed:
function create_position_ids_from_input_ids(input_ids, padding_idx) - removed:
function top_k_top_p_filtering(logits: Tensor, top_k: int=…, top_p: float=…, filter_value: float=…, min_tokens_to_keep: int=…) -> Tensor - removed:
method BeamHypotheses.add(hyp, sum_logprobs) - removed:
method BeamHypotheses.is_done(best_sum_logprobs, cur_len=…) - removed:
method PreTrainedModel.enforce_repetition_penalty_(lprobs, batch_size, num_beams, prev_output_tokens, repetition_penalty) - removed:
method PreTrainedModel.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 PreTrainedModel.prepare_inputs_for_generation(input_ids, **kwargs) - removed:
method PreTrainedModel.prepare_logits_for_generation(logits, **kwargs) - removed:
property PreTrainedModel.base_model - removed:
property PreTrainedModel.dummy_inputs - changed:
class PreTrainedModel(nn.Module, ModuleUtilsMixin) → class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin) - changed:
constructor PoolerEndLogits(config) → constructor PoolerEndLogits(config: PretrainedConfig) - changed:
constructor PoolerStartLogits(config) → constructor PoolerStartLogits(config: PretrainedConfig) - changed:
constructor PreTrainedModel(config, *inputs, **kwargs) → constructor PreTrainedModel(config: PretrainedConfig, *inputs, **kwargs) - changed:
function apply_chunking_to_forward(chunk_size: int, chunk_dim: int, forward_fn: Callable[..., torch.Tensor], *input_tensors) -> torch.Tensor → function apply_chunking_to_forward(forward_fn: Callable[..., torch.Tensor], chunk_size: int, chunk_dim: int, *input_tensors) -> torch.Tensor - changed:
function prune_conv1d_layer(layer, index, dim=…) → function prune_conv1d_layer(layer: Conv1D, index: torch.LongTensor, dim: int=…) -> Conv1D - changed:
function prune_layer(layer, index, dim=…) → function prune_layer(layer: Union[torch.nn.Linear, Conv1D], index: torch.LongTensor, dim: Optional[int]=…) -> Union[torch.nn.Linear, Conv1D] - changed:
function prune_linear_layer(layer, index, dim=…) → function prune_linear_layer(layer: torch.nn.Linear, index: torch.LongTensor, dim: int=…) -> torch.nn.Linear - changed:
method ModuleUtilsMixin.get_extended_attention_mask(attention_mask: Tensor, input_shape: Tuple, device: device) -> Tensor → method ModuleUtilsMixin.get_extended_attention_mask(attention_mask: Tensor, input_shape: Tuple[int], device: device) -> Tensor - changed:
method ModuleUtilsMixin.get_head_mask(head_mask: Tensor, num_hidden_layers: int, is_attention_chunked: bool=…) -> Tensor → method ModuleUtilsMixin.get_head_mask(head_mask: Optional[Tensor], num_hidden_layers: int, is_attention_chunked: bool=…) -> Tensor - changed:
method ModuleUtilsMixin.num_parameters(only_trainable: bool=…) -> int → method ModuleUtilsMixin.num_parameters(only_trainable: bool=…, exclude_embeddings: bool=…) -> int - changed:
method PoolerAnswerClass.forward(hidden_states, start_states=…, start_positions=…, cls_index=…) → method PoolerAnswerClass.forward(hidden_states: torch.FloatTensor, start_states: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, cls_index: Optional[torch.LongTensor]=…) -> torch.FloatTensor - changed:
method PoolerEndLogits.forward(hidden_states, start_states=…, start_positions=…, p_mask=…) → method PoolerEndLogits.forward(hidden_states: torch.FloatTensor, start_states: Optional[torch.FloatTensor]=…, start_positions: Optional[torch.LongTensor]=…, p_mask: Optional[torch.FloatTensor]=…) -> torch.FloatTensor - changed:
method PoolerStartLogits.forward(hidden_states, p_mask=…) → method PoolerStartLogits.forward(hidden_states: torch.FloatTensor, p_mask: Optional[torch.FloatTensor]=…) -> torch.FloatTensor - changed:
method PreTrainedModel.get_input_embeddings() → method PreTrainedModel.get_input_embeddings() -> nn.Module - changed:
method PreTrainedModel.get_output_embeddings() → method PreTrainedModel.get_output_embeddings() -> nn.Module - changed:
method PreTrainedModel.prune_heads(heads_to_prune: Dict) → method PreTrainedModel.prune_heads(heads_to_prune: Dict[int, List[int]]) - changed:
method PreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…) → method PreTrainedModel.resize_token_embeddings(new_num_tokens: Optional[int]=…) -> torch.nn.Embedding - changed:
method SQuADHead.forward(hidden_states, start_positions=…, end_positions=…, cls_index=…, is_impossible=…, p_mask=…) → 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]] - changed:
method SequenceSummary.forward(hidden_states, cls_index=…) → method SequenceSummary.forward(hidden_states: torch.FloatTensor, cls_index: Optional[torch.LongTensor]=…) -> torch.FloatTensor - added:
class SquadHeadOutput(ModelOutput) - added:
field PreTrainedModel.authorized_missing_keys - added:
field PreTrainedModel.authorized_unexpected_keys - added:
field PreTrainedModel.keys_to_never_save - added:
field SquadHeadOutput.cls_logits: Optional[torch.FloatTensor] - added:
field SquadHeadOutput.end_top_index: Optional[torch.LongTensor] - added:
field SquadHeadOutput.end_top_log_probs: Optional[torch.FloatTensor] - added:
field SquadHeadOutput.loss: Optional[torch.FloatTensor] - added:
field SquadHeadOutput.start_top_index: Optional[torch.LongTensor] - added:
field SquadHeadOutput.start_top_log_probs: Optional[torch.FloatTensor] - 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:
method ModuleUtilsMixin.estimate_tokens(input_dict: Dict[str, Union[torch.Tensor, Any]]) -> int - added:
method ModuleUtilsMixin.floating_point_ops(input_dict: Dict[str, Union[torch.Tensor, Any]], exclude_embeddings: bool=…) -> int - added:
property PreTrainedModel.base_model -> nn.Module - added:
property PreTrainedModel.dummy_inputs -> Dict[str, torch.Tensor]
- changed:
method MultiHeadAttention.forward(input, mask, kv=…, cache=…, head_mask=…) → method MultiHeadAttention.forward(input, mask, kv=…, cache=…, head_mask=…, output_attentions=…) - changed:
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=…) → 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=…) - changed:
method XLMForQuestionAnsweringSimple.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, start_positions=…, end_positions=…) → 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=…) - changed:
method XLMForSequenceClassification.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…) → 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=…) - changed:
method XLMForTokenClassification.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, head_mask=…, labels=…) → 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=…) - changed:
method XLMModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…) → 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=…) - changed:
method XLMWithLMHeadModel.forward(input_ids=…, attention_mask=…, langs=…, token_type_ids=…, position_ids=…, lengths=…, cache=…, head_mask=…, inputs_embeds=…, labels=…) → 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=…) - added:
class XLMForMultipleChoice(XLMPreTrainedModel) - added:
class XLMForQuestionAnsweringOutput(ModelOutput) - added:
constructor XLMForMultipleChoice(config, *inputs, **kwargs) - added:
field XLMForQuestionAnsweringOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLMForQuestionAnsweringOutput.cls_logits: Optional[torch.FloatTensor] - added:
field XLMForQuestionAnsweringOutput.end_top_index: Optional[torch.LongTensor] - added:
field XLMForQuestionAnsweringOutput.end_top_log_probs: Optional[torch.FloatTensor] - added:
field XLMForQuestionAnsweringOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLMForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - added:
field XLMForQuestionAnsweringOutput.start_top_index: Optional[torch.LongTensor] - added:
field XLMForQuestionAnsweringOutput.start_top_log_probs: Optional[torch.FloatTensor] - added:
field XLMModel.authorized_missing_keys - added:
method TransformerFFN.ff_chunk(input) - added:
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=…)
- added:
class XLMProphetNetDecoder(ProphetNetDecoder) - added:
class XLMProphetNetEncoder(ProphetNetEncoder) - added:
class XLMProphetNetForCausalLM(ProphetNetForCausalLM) - added:
class XLMProphetNetForConditionalGeneration(ProphetNetForConditionalGeneration) - added:
class XLMProphetNetModel(ProphetNetModel) - added:
field XLMProphetNetDecoder.config_class - added:
field XLMProphetNetEncoder.config_class - added:
field XLMProphetNetForCausalLM.config_class - added:
field XLMProphetNetForConditionalGeneration.config_class - added:
field XLMProphetNetModel.config_class - added:
variable XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable logger
- added:
class XLMRobertaForCausalLM(RobertaForCausalLM) - added:
class XLMRobertaForQuestionAnswering(RobertaForQuestionAnswering) - added:
field XLMRobertaForCausalLM.config_class - added:
field XLMRobertaForQuestionAnswering.config_class
- removed:
variable ACT2FN - removed:
variable XLNetLayerNorm - changed:
method XLNetForMultipleChoice.forward(input_ids=…, token_type_ids=…, input_mask=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, head_mask=…, inputs_embeds=…, use_cache=…, labels=…) → 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=…) - changed:
method XLNetForQuestionAnswering.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, start_positions=…, end_positions=…, is_impossible=…, cls_index=…, p_mask=…) → 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=…) - changed:
method XLNetForQuestionAnsweringSimple.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, start_positions=…, end_positions=…) → 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=…) - changed:
method XLNetForSequenceClassification.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, labels=…) → 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=…) - changed:
method XLNetForTokenClassification.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, labels=…) → 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=…) - changed:
method XLNetLMHeadModel.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…, labels=…) → 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=…) - changed:
method XLNetLMHeadModel.prepare_inputs_for_generation(input_ids, past, **kwargs) → method XLNetLMHeadModel.prepare_inputs_for_generation(input_ids, past=…, use_cache=…, **kwargs) - changed:
method XLNetLayer.forward(output_h, output_g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…) → method XLNetLayer.forward(output_h, output_g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…, output_attentions=…) - changed:
method XLNetModel.forward(input_ids=…, attention_mask=…, mems=…, perm_mask=…, target_mapping=…, token_type_ids=…, input_mask=…, head_mask=…, inputs_embeds=…, use_cache=…) → 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=…) - changed:
method XLNetRelativeAttention.forward(h, g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…) → method XLNetRelativeAttention.forward(h, g, attn_mask_h, attn_mask_g, r, seg_mat, mems=…, target_mapping=…, head_mask=…, output_attentions=…) - changed:
method XLNetRelativeAttention.rel_attn_core(q_head, k_head_h, v_head_h, k_head_r, seg_mat=…, attn_mask=…, head_mask=…) → method XLNetRelativeAttention.rel_attn_core(q_head, k_head_h, v_head_h, k_head_r, seg_mat=…, attn_mask=…, head_mask=…, output_attentions=…) - added:
class XLNetForMultipleChoiceOutput(ModelOutput) - added:
class XLNetForQuestionAnsweringOutput(ModelOutput) - added:
class XLNetForQuestionAnsweringSimpleOutput(ModelOutput) - added:
class XLNetForSequenceClassificationOutput(ModelOutput) - added:
class XLNetForTokenClassificationOutput(ModelOutput) - added:
class XLNetLMHeadModelOutput(ModelOutput) - added:
class XLNetModelOutput(ModelOutput) - added:
field XLNetForMultipleChoiceOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForMultipleChoiceOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForMultipleChoiceOutput.logits: torch.FloatTensor - added:
field XLNetForMultipleChoiceOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetForMultipleChoiceOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringOutput.cls_logits: Optional[torch.FloatTensor] - added:
field XLNetForQuestionAnsweringOutput.end_top_index: Optional[torch.LongTensor] - added:
field XLNetForQuestionAnsweringOutput.end_top_log_probs: Optional[torch.FloatTensor] - added:
field XLNetForQuestionAnsweringOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetForQuestionAnsweringOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringOutput.start_top_index: Optional[torch.LongTensor] - added:
field XLNetForQuestionAnsweringOutput.start_top_log_probs: Optional[torch.FloatTensor] - added:
field XLNetForQuestionAnsweringSimpleOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringSimpleOutput.end_logits: torch.FloatTensor - added:
field XLNetForQuestionAnsweringSimpleOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringSimpleOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetForQuestionAnsweringSimpleOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetForQuestionAnsweringSimpleOutput.start_logits: torch.FloatTensor - added:
field XLNetForSequenceClassificationOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForSequenceClassificationOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForSequenceClassificationOutput.logits: torch.FloatTensor - added:
field XLNetForSequenceClassificationOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetForSequenceClassificationOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetForTokenClassificationOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForTokenClassificationOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetForTokenClassificationOutput.logits: torch.FloatTensor - added:
field XLNetForTokenClassificationOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetForTokenClassificationOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetLMHeadModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetLMHeadModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetLMHeadModelOutput.logits: torch.FloatTensor - added:
field XLNetLMHeadModelOutput.loss: Optional[torch.FloatTensor] - added:
field XLNetLMHeadModelOutput.mems: Optional[List[torch.FloatTensor]] - added:
field XLNetModelOutput.attentions: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetModelOutput.hidden_states: Optional[Tuple[torch.FloatTensor]] - added:
field XLNetModelOutput.last_hidden_state: torch.FloatTensor - added:
field XLNetModelOutput.mems: Optional[List[torch.FloatTensor]] - added:
method XLNetLayer.ff_chunk(output_x)
- changed:
constructor AdamW(params, lr=…, betas=…, eps=…, weight_decay=…, correct_bias=…) → constructor AdamW(params: Iterable[torch.nn.parameter.Parameter], lr: float=…, betas: Tuple[float, float]=…, eps: float=…, weight_decay: float=…, correct_bias: bool=…) - changed:
function get_constant_schedule(optimizer, last_epoch=…) → function get_constant_schedule(optimizer: Optimizer, last_epoch: int=…) - changed:
function get_constant_schedule_with_warmup(optimizer, num_warmup_steps, last_epoch=…) → function get_constant_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, last_epoch: int=…) - changed:
function get_cosine_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps, num_cycles=…, last_epoch=…) → function get_cosine_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: float=…, last_epoch: int=…) - changed:
function get_cosine_with_hard_restarts_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps, num_cycles=…, last_epoch=…) → function get_cosine_with_hard_restarts_schedule_with_warmup(optimizer: Optimizer, num_warmup_steps: int, num_training_steps: int, num_cycles: int=…, last_epoch: int=…) - changed:
method AdamW.step(closure=…) → method AdamW.step(closure: Callable=…) - added:
class Adafactor(Optimizer) - added:
constructor Adafactor(params, lr=…, eps=…, clip_threshold=…, decay_rate=…, beta1=…, weight_decay=…, scale_parameter=…, relative_step=…, warmup_init=…) - added:
function get_polynomial_decay_schedule_with_warmup(optimizer, num_warmup_steps, num_training_steps, lr_end=…, power=…, last_epoch=…) - added:
method Adafactor.step(closure=…)
- changed:
constructor AdamWeightDecay(learning_rate=…, beta_1=…, beta_2=…, epsilon=…, amsgrad=…, weight_decay_rate=…, include_in_weight_decay=…, exclude_from_weight_decay=…, name=…, **kwargs) → 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) - changed:
constructor WarmUp(initial_learning_rate, decay_schedule_fn, warmup_steps, power=…, name=…) → constructor WarmUp(initial_learning_rate: float, decay_schedule_fn: Callable, warmup_steps: int, power: float=…, name: str=…) - changed:
function create_optimizer(init_lr, num_train_steps, num_warmup_steps, end_lr=…, optimizer_type=…) → 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]]=…) - changed:
method AdamWeightDecay.apply_gradients(grads_and_vars, name=…) → method AdamWeightDecay.apply_gradients(grads_and_vars, name=…, **kwargs)
- removed:
class DefaultArgumentHandler(ArgumentHandler) - removed:
class NerPipeline(Pipeline) - removed:
constructor NerPipeline(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=…) - removed:
field NerPipeline.default_input_names - removed:
field TextGenerationPipeline.PADDING_TEXT - removed:
method NerPipeline.group_entities(entities) - removed:
static method DefaultArgumentHandler.handle_args(args: Sequence[Any]) -> List[str] - removed:
static method DefaultArgumentHandler.handle_kwargs(kwargs: Dict) -> List - removed:
variable TokenClassificationPipeline - changed:
constructor FillMaskPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, args_parser: ArgumentHandler=…, device: int=…, topk=…, task: str=…) → constructor FillMaskPipeline(model: Union[PreTrainedModel, TFPreTrainedModel], tokenizer: PreTrainedTokenizer, modelcard: Optional[ModelCard]=…, framework: Optional[str]=…, args_parser: ArgumentHandler=…, device: int=…, top_k=…, task: str=…, **kwargs) - changed:
constructor PipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) → constructor PipelineDataFormat(output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite: bool=…) - changed:
function get_framework(model=…) → function get_framework(model, revision: Optional[str]=…) - changed:
function pipeline(task: str, model: Optional=…, config: Optional[Union[str, PretrainedConfig]]=…, tokenizer: Optional[Union[str, PreTrainedTokenizer]]=…, framework: Optional[str]=…, **kwargs) -> Pipeline → 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 - changed:
method Pipeline.save_pretrained(save_directory) → method Pipeline.save_pretrained(save_directory: str) - changed:
method PipelineDataFormat.save(data: dict) → method PipelineDataFormat.save(data: Union[dict, List[dict]]) - changed:
method QuestionAnsweringPipeline.span_to_answer(text: str, start: int, end: int) → method QuestionAnsweringPipeline.span_to_answer(text: str, start: int, end: int) -> Dict[str, Union[str, int]] - changed:
static method PipelineDataFormat.from_str(format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) → static method PipelineDataFormat.from_str(format: str, output_path: Optional[str], input_path: Optional[str], column: Optional[str], overwrite=…) -> PipelineDataFormat - added:
class Conversation - added:
class ConversationalPipeline(Pipeline) - added:
class PipelineException(Exception) - added:
class Text2TextGenerationPipeline(Pipeline) - added:
class TokenClassificationArgumentHandler(ArgumentHandler) - added:
class TokenClassificationPipeline(Pipeline) - added:
class ZeroShotClassificationArgumentHandler(ArgumentHandler) - added:
class ZeroShotClassificationPipeline(Pipeline) - added:
constructor Conversation(text: str=…, conversation_id: UUID=…) - added:
constructor ConversationalPipeline(min_length_for_response=…, *args, **kwargs) - added:
constructor PipelineException(task: str, model: str, reason: str) - added:
constructor SummarizationPipeline(*args, **kwargs) - added:
constructor Text2TextGenerationPipeline(*args, **kwargs) - added:
constructor TextClassificationPipeline(return_all_scores: bool=…, **kwargs) - added:
constructor TextGenerationPipeline(*args, **kwargs) - added:
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=…) - added:
constructor TranslationPipeline(*args, **kwargs) - added:
constructor ZeroShotClassificationPipeline(args_parser=…, *args, **kwargs) - added:
field TextGenerationPipeline.XL_PREFIX - added:
field TokenClassificationPipeline.default_input_names - added:
function check_task(task: str) -> Tuple[Dict, Any] - added:
function get_default_model(targeted_task: Dict, framework: Optional[str], task_options: Optional[Any]) -> str - added:
method Conversation.add_user_input(text: str, overwrite: bool=…) - added:
method Conversation.append_response(response: str) - added:
method Conversation.mark_processed() - added:
method Conversation.set_history(history: List[int]) - added:
method FillMaskPipeline.ensure_exactly_one_mask_token(masked_index: np.ndarray) - added:
method Pipeline.check_model_type(supported_models: Union[List[str], dict]) - added:
method QuestionAnsweringArgumentHandler.normalize(item) - added:
method TokenClassificationPipeline.group_entities(entities: List[dict]) -> List[dict] - added:
method TokenClassificationPipeline.group_sub_entities(entities: List[dict]) -> dict - added:
property ZeroShotClassificationPipeline.entailment_id - added:
variable NerPipeline - added:
variable PIPELINE_INIT_ARGS
- added:
class CanonicalHFIndex(HFIndexBase) - added:
class CustomHFIndex(HFIndexBase) - added:
class HFIndexBase(Index) - added:
class Index - added:
class LegacyIndex(Index) - added:
class RagRetriever - added:
constructor CanonicalHFIndex(vector_size: int, dataset_name: str=…, dataset_split: str=…, index_name: Optional[str]=…, index_path: Optional[str]=…, use_dummy_dataset=…) - added:
constructor CustomHFIndex(vector_size: int, dataset, index_path=…) - added:
constructor HFIndexBase(vector_size, dataset, index_initialized=…) - added:
constructor LegacyIndex(vector_size, index_path) - added:
constructor RagRetriever(config, question_encoder_tokenizer, generator_tokenizer, index=…) - added:
field LegacyIndex.INDEX_FILENAME - added:
field LegacyIndex.PASSAGE_FILENAME - added:
method CanonicalHFIndex.init_index() - added:
method CustomHFIndex.init_index() - added:
method HFIndexBase.get_doc_dicts(doc_ids: np.ndarray) -> List[dict] - added:
method HFIndexBase.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - added:
method HFIndexBase.init_index() - added:
method HFIndexBase.is_initialized() - added:
method Index.get_doc_dicts(doc_ids: np.ndarray) -> List[dict] - added:
method Index.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - added:
method Index.init_index() - added:
method Index.is_initialized() - added:
method LegacyIndex.get_doc_dicts(doc_ids: np.array) - added:
method LegacyIndex.get_top_docs(question_hidden_states: np.ndarray, n_docs=…) -> Tuple[np.ndarray, np.ndarray] - added:
method LegacyIndex.init_index() - added:
method LegacyIndex.is_initialized() - 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, List[dict]] - added:
method RagRetriever.save_pretrained(save_directory) - added:
static method CustomHFIndex.load_from_disk(vector_size, dataset_path, index_path) - added:
static method RagRetriever.from_pretrained(retriever_name_or_path, indexed_dataset=…, **kwargs) - added:
variable LEGACY_INDEX_PATH - added:
variable logger
- added:
class CaptureLogger - added:
class CaptureStd - added:
class CaptureStderr(CaptureStd) - added:
class CaptureStdout(CaptureStd) - added:
class TestCasePlus(unittest.TestCase) - added:
constructor CaptureLogger(logger) - added:
constructor CaptureStd(out=…, err=…) - added:
constructor CaptureStderr() - added:
constructor CaptureStdout() - added:
function apply_print_resets(buf) - added:
function assert_screenout(out, what) - added:
function custom_tokenizers(test_case) - added:
function execute_subprocess_async(cmd, env=…, stdin=…, timeout=…, quiet=…, echo=…) -> _RunOutput - added:
function get_gpu_count() - added:
function get_tests_dir(append_path=…) - added:
function is_pipeline_test(test_case) - added:
function is_pt_tf_cross_test(test_case) - added:
function mockenv(**kwargs) - added:
function parse_flag_from_env(key, default=…) - added:
function parse_int_from_env(key, default=…) - added:
function pytest_addoption_shared(parser) - added:
function pytest_terminal_summary_main(tr, id) - added:
function require_datasets(test_case) - added:
function require_faiss(test_case) - added:
function require_flax(test_case) - added:
function require_optuna(test_case) - added:
function require_ray(test_case) - added:
function require_sentencepiece(test_case) - added:
function require_tf(test_case) - added:
function require_tokenizers(test_case) - added:
function require_torch(test_case) - added:
function require_torch_gpu(test_case) - added:
function require_torch_multigpu(test_case) - added:
function require_torch_non_multigpu(test_case) - added:
function require_torch_tpu(test_case) - added:
function slow(test_case) - added:
method TestCasePlus.get_auto_remove_tmp_dir(tmp_dir=…, after=…, before=…) - added:
method TestCasePlus.get_env() - added:
method TestCasePlus.setUp() - added:
method TestCasePlus.tearDown() - added:
property TestCasePlus.examples_dir - added:
property TestCasePlus.examples_dir_str - added:
property TestCasePlus.repo_root_dir - added:
property TestCasePlus.repo_root_dir_str - added:
property TestCasePlus.src_dir - added:
property TestCasePlus.src_dir_str - added:
property TestCasePlus.test_file_dir - added:
property TestCasePlus.test_file_dir_str - added:
property TestCasePlus.test_file_path - added:
property TestCasePlus.test_file_path_str - added:
property TestCasePlus.tests_dir - added:
property TestCasePlus.tests_dir_str - added:
variable DUMMY_DIFF_TOKENIZER_IDENTIFIER - added:
variable DUMMY_UNKWOWN_IDENTIFIER - added:
variable SMALL_MODEL_IDENTIFIER - added:
variable pytest_opt_registered - added:
variable require_torch_non_multigpu_but_fix_me - added:
variable torch_device
- changed:
method AlbertTokenizer.save_vocabulary(save_directory) → method AlbertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str]
- 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.max_model_input_sizes - added:
field AlbertTokenizerFast.pretrained_vocab_files_map - 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.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
method AlbertTokenizerFast.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method AlbertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable AlbertTokenizer - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
variable AlbertTokenizer - added:
variable AlbertTokenizerFast - added:
variable BartTokenizerFast - added:
variable BertGenerationTokenizer - added:
variable BertTokenizerFast - added:
variable CamembertTokenizer - added:
variable CamembertTokenizerFast - added:
variable DPRQuestionEncoderTokenizerFast - added:
variable DistilBertTokenizerFast - added:
variable ElectraTokenizerFast - added:
variable FunnelTokenizerFast - added:
variable GPT2TokenizerFast - added:
variable HerbertTokenizerFast - added:
variable LayoutLMTokenizerFast - added:
variable LongformerTokenizerFast - added:
variable LxmertTokenizerFast - added:
variable MBartTokenizer - added:
variable MBartTokenizerFast - added:
variable MarianTokenizer - added:
variable MobileBertTokenizerFast - added:
variable OpenAIGPTTokenizerFast - added:
variable PegasusTokenizer - added:
variable PegasusTokenizerFast - added:
variable ReformerTokenizer - added:
variable ReformerTokenizerFast - added:
variable RetriBertTokenizerFast - added:
variable RobertaTokenizerFast - added:
variable SLOW_TOKENIZER_MAPPING - added:
variable SqueezeBertTokenizerFast - added:
variable T5Tokenizer - added:
variable T5TokenizerFast - added:
variable XLMProphetNetTokenizer - added:
variable XLMRobertaTokenizer - added:
variable XLMRobertaTokenizerFast - added:
variable XLNetTokenizer - added:
variable XLNetTokenizerFast
- removed:
class MBartTokenizer(XLMRobertaTokenizer) - removed:
field MBartTokenizer.max_model_input_sizes - removed:
field MBartTokenizer.pretrained_vocab_files_map - removed:
field MBartTokenizer.vocab_files_names - removed:
variable SPM_URL - added:
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
- added:
class BartTokenizerFast(RobertaTokenizerFast) - added:
field BartTokenizerFast.max_model_input_sizes - added:
field BartTokenizerFast.pretrained_vocab_files_map - added:
field BartTokenizerFast.slow_tokenizer_class - added:
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 - added:
variable logger - added:
variable merges_url - added:
variable tokenizer_url - added:
variable vocab_url
- removed:
class BertTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor BertTokenizerFast(vocab_file, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, clean_text=…, tokenize_chinese_chars=…, strip_accents=…, wordpieces_prefix=…, **kwargs) - removed:
field BertTokenizerFast.max_model_input_sizes - removed:
field BertTokenizerFast.pretrained_init_configuration - removed:
field BertTokenizerFast.pretrained_vocab_files_map - 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] - changed:
constructor BasicTokenizer(do_lower_case=…, never_split=…, tokenize_chinese_chars=…) → constructor BasicTokenizer(do_lower_case=…, never_split=…, tokenize_chinese_chars=…, strip_accents=…) - changed:
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=…, **kwargs) → 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) - changed:
method BertTokenizer.save_vocabulary(vocab_path) → method BertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
property BertTokenizer.do_lower_case
- 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.max_model_input_sizes - added:
field BertTokenizerFast.pretrained_init_configuration - added:
field BertTokenizerFast.pretrained_vocab_files_map - 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.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
method BertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class BertGenerationTokenizer(PreTrainedTokenizer) - added:
constructor BertGenerationTokenizer(vocab_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, sep_token=…, **kwargs) - added:
field BertGenerationTokenizer.max_model_input_sizes - added:
field BertGenerationTokenizer.prefix_tokens: List[int] - added:
field BertGenerationTokenizer.pretrained_vocab_files_map - 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:
variable VOCAB_FILES_NAMES - added:
variable logger - added:
variable tokenizer_url
- changed:
class CharacterTokenizer(object) → class CharacterTokenizer - changed:
constructor MecabTokenizer(do_lower_case=…, never_split=…, normalize_text=…, mecab_option: Optional[str]=…) → constructor MecabTokenizer(do_lower_case=…, never_split=…, normalize_text=…, mecab_dic: Optional[str]=…, mecab_option: Optional[str]=…) - added:
property BertJapaneseTokenizer.do_lower_case
- added:
class BertweetTokenizer(PreTrainedTokenizer) - added:
class TweetTokenizer - added:
constructor BertweetTokenizer(vocab_file, merges_file, normalization=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) - added:
constructor TweetTokenizer(preserve_case=…, reduce_len=…, strip_handles=…) - added:
field BertweetTokenizer.max_model_input_sizes - added:
field BertweetTokenizer.pretrained_vocab_files_map - added:
field BertweetTokenizer.vocab_files_names - added:
function casual_tokenize(text, preserve_case=…, reduce_len=…, strip_handles=…) - added:
function get_pairs(word) - added:
function reduce_lengthening(text) - added:
function remove_handles(text) - 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:
method TweetTokenizer.tokenize(text) - added:
property BertweetTokenizer.vocab_size - added:
variable EMOTICONS - added:
variable EMOTICON_RE - added:
variable ENT_RE - added:
variable HANG_RE - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable REGEXPS - added:
variable URLS - added:
variable VOCAB_FILES_NAMES - added:
variable WORD_RE - added:
variable logger
- added:
class BlenderbotSmallTokenizer(PreTrainedTokenizer) - added:
class BlenderbotTokenizer(RobertaTokenizer) - added:
constructor BlenderbotSmallTokenizer(vocab_file, merges_file, bos_token=…, eos_token=…, unk_token=…, pad_token=…, **kwargs) - added:
field BlenderbotSmallTokenizer.max_model_input_sizes - added:
field BlenderbotSmallTokenizer.pretrained_vocab_files_map - added:
field BlenderbotSmallTokenizer.vocab_files_names - added:
field BlenderbotTokenizer.max_model_input_sizes - added:
field BlenderbotTokenizer.pretrained_vocab_files_map - added:
field BlenderbotTokenizer.vocab_files_names - added:
function get_pairs(word) - 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:
method BlenderbotTokenizer.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: List[int]=…) - added:
property BlenderbotSmallTokenizer.vocab_size -> int - added:
variable CKPT_3B - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- changed:
method CamembertTokenizer.save_vocabulary(save_directory) → method CamembertTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
method CamembertTokenizer.get_vocab() - added:
variable SPIECE_UNDERLINE
- 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.max_model_input_sizes - added:
field CamembertTokenizerFast.model_input_names - added:
field CamembertTokenizerFast.pretrained_vocab_files_map - 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.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method CamembertTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable CamembertTokenizer - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable SHARED_MODEL_IDENTIFIERS - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- changed:
method CTRLTokenizer.save_vocabulary(save_directory) → method CTRLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str]
- added:
class DebertaTokenizer(PreTrainedTokenizer) - added:
constructor DebertaTokenizer(vocab_file, do_lower_case=…, unk_token=…, sep_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field DebertaTokenizer.max_model_input_sizes - added:
field DebertaTokenizer.pretrained_init_configuration - added:
field DebertaTokenizer.pretrained_vocab_files_map - added:
field DebertaTokenizer.vocab_files_names - added:
method DebertaTokenizer.build_inputs_with_special_tokens(token_ids_0, token_ids_1=…) - added:
method DebertaTokenizer.convert_tokens_to_string(tokens) - added:
method DebertaTokenizer.create_token_type_ids_from_sequences(token_ids_0, token_ids_1=…) - added:
method DebertaTokenizer.get_special_tokens_mask(token_ids_0, token_ids_1=…, already_has_special_tokens=…) - 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 - added:
property DebertaTokenizer.vocab_size
- 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.vocab_files_names
- added:
class DistilBertTokenizerFast(BertTokenizerFast) - added:
field DistilBertTokenizerFast.max_model_input_sizes - added:
field DistilBertTokenizerFast.model_input_names - added:
field DistilBertTokenizerFast.pretrained_init_configuration - added:
field DistilBertTokenizerFast.pretrained_vocab_files_map - added:
field DistilBertTokenizerFast.slow_tokenizer_class - added:
field DistilBertTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class CustomDPRReaderTokenizerMixin - added:
class DPRContextEncoderTokenizer(BertTokenizer) - added:
class DPRQuestionEncoderTokenizer(BertTokenizer) - added:
class DPRReaderTokenizer(CustomDPRReaderTokenizerMixin, BertTokenizer) - added:
field DPRContextEncoderTokenizer.max_model_input_sizes - added:
field DPRContextEncoderTokenizer.pretrained_init_configuration - added:
field DPRContextEncoderTokenizer.pretrained_vocab_files_map - added:
field DPRContextEncoderTokenizer.vocab_files_names - added:
field DPRQuestionEncoderTokenizer.max_model_input_sizes - added:
field DPRQuestionEncoderTokenizer.pretrained_init_configuration - added:
field DPRQuestionEncoderTokenizer.pretrained_vocab_files_map - added:
field DPRQuestionEncoderTokenizer.vocab_files_names - added:
field DPRReaderTokenizer.max_model_input_sizes - added:
field DPRReaderTokenizer.model_input_names - added:
field DPRReaderTokenizer.pretrained_init_configuration - added:
field DPRReaderTokenizer.pretrained_vocab_files_map - added:
field DPRReaderTokenizer.vocab_files_names - added:
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] - added:
variable CONTEXT_ENCODER_PRETRAINED_INIT_CONFIGURATION - added:
variable CONTEXT_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable CONTEXT_ENCODER_PRETRAINED_VOCAB_FILES_MAP - added:
variable CUSTOM_DPR_READER_DOCSTRING - added:
variable DPRReaderOutput - added:
variable DPRSpanPrediction - added:
variable QUESTION_ENCODER_PRETRAINED_INIT_CONFIGURATION - added:
variable QUESTION_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable QUESTION_ENCODER_PRETRAINED_VOCAB_FILES_MAP - added:
variable READER_PRETRAINED_INIT_CONFIGURATION - added:
variable READER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable READER_PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class CustomDPRReaderTokenizerMixin - added:
class DPRContextEncoderTokenizerFast(BertTokenizerFast) - added:
class DPRQuestionEncoderTokenizerFast(BertTokenizerFast) - added:
class DPRReaderTokenizerFast(CustomDPRReaderTokenizerMixin, BertTokenizerFast) - added:
field DPRContextEncoderTokenizerFast.max_model_input_sizes - added:
field DPRContextEncoderTokenizerFast.pretrained_init_configuration - added:
field DPRContextEncoderTokenizerFast.pretrained_vocab_files_map - added:
field DPRContextEncoderTokenizerFast.slow_tokenizer_class - added:
field DPRContextEncoderTokenizerFast.vocab_files_names - added:
field DPRQuestionEncoderTokenizerFast.max_model_input_sizes - added:
field DPRQuestionEncoderTokenizerFast.pretrained_init_configuration - added:
field DPRQuestionEncoderTokenizerFast.pretrained_vocab_files_map - added:
field DPRQuestionEncoderTokenizerFast.slow_tokenizer_class - added:
field DPRQuestionEncoderTokenizerFast.vocab_files_names - added:
field DPRReaderTokenizerFast.max_model_input_sizes - added:
field DPRReaderTokenizerFast.model_input_names - added:
field DPRReaderTokenizerFast.pretrained_init_configuration - added:
field DPRReaderTokenizerFast.pretrained_vocab_files_map - added:
field DPRReaderTokenizerFast.slow_tokenizer_class - added:
field DPRReaderTokenizerFast.vocab_files_names - added:
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] - added:
variable CONTEXT_ENCODER_PRETRAINED_INIT_CONFIGURATION - added:
variable CONTEXT_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable CONTEXT_ENCODER_PRETRAINED_VOCAB_FILES_MAP - added:
variable CUSTOM_DPR_READER_DOCSTRING - added:
variable DPRReaderOutput - added:
variable DPRSpanPrediction - added:
variable QUESTION_ENCODER_PRETRAINED_INIT_CONFIGURATION - added:
variable QUESTION_ENCODER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable QUESTION_ENCODER_PRETRAINED_VOCAB_FILES_MAP - added:
variable READER_PRETRAINED_INIT_CONFIGURATION - added:
variable READER_PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable READER_PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- 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.vocab_files_names
- added:
class ElectraTokenizerFast(BertTokenizerFast) - added:
field ElectraTokenizerFast.max_model_input_sizes - added:
field ElectraTokenizerFast.pretrained_init_configuration - added:
field ElectraTokenizerFast.pretrained_vocab_files_map - added:
field ElectraTokenizerFast.slow_tokenizer_class - added:
field ElectraTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES
- 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.max_model_input_sizes - added:
field FSMTTokenizer.pretrained_init_configuration - added:
field FSMTTokenizer.pretrained_vocab_files_map - added:
field FSMTTokenizer.vocab_files_names - added:
function get_pairs(word) - added:
function remove_non_printing_char(text) - added:
function replace_unicode_punct(text) - 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.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - 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.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 - 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:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class FunnelTokenizer(BertTokenizer) - 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=…, **kwargs) - added:
field FunnelTokenizer.cls_token_type_id: int - added:
field FunnelTokenizer.max_model_input_sizes - added:
field FunnelTokenizer.pretrained_init_configuration - added:
field FunnelTokenizer.pretrained_vocab_files_map - added:
field FunnelTokenizer.vocab_files_names - added:
method FunnelTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class FunnelTokenizerFast(BertTokenizerFast) - 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.max_model_input_sizes - added:
field FunnelTokenizerFast.pretrained_init_configuration - added:
field FunnelTokenizerFast.pretrained_vocab_files_map - added:
field FunnelTokenizerFast.slow_tokenizer_class - added:
field FunnelTokenizerFast.vocab_files_names - added:
method FunnelTokenizerFast.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
class GPT2TokenizerFast(PreTrainedTokenizerFast) - removed:
constructor GPT2TokenizerFast(vocab_file, merges_file, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - removed:
field GPT2TokenizerFast.max_model_input_sizes - removed:
field GPT2TokenizerFast.pretrained_vocab_files_map - removed:
field GPT2TokenizerFast.vocab_files_names - changed:
constructor GPT2Tokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, **kwargs) → constructor GPT2Tokenizer(vocab_file, merges_file, errors=…, unk_token=…, bos_token=…, eos_token=…, add_prefix_space=…, **kwargs) - changed:
method GPT2Tokenizer.prepare_for_tokenization(text, **kwargs) → method GPT2Tokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs) - changed:
method GPT2Tokenizer.save_vocabulary(save_directory) → method GPT2Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
field GPT2Tokenizer.model_input_names
- 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.max_model_input_sizes - added:
field GPT2TokenizerFast.model_input_names - added:
field GPT2TokenizerFast.pretrained_vocab_files_map - 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:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class HerbertTokenizer(XLMTokenizer) - added:
constructor HerbertTokenizer(**kwargs) - added:
field HerbertTokenizer.max_model_input_sizes - added:
field HerbertTokenizer.pretrained_init_configuration - added:
field HerbertTokenizer.pretrained_vocab_files_map - added:
field HerbertTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class HerbertTokenizerFast(PreTrainedTokenizerFast) - added:
constructor HerbertTokenizerFast(vocab_file, merges_file, tokenizer_file=…, **kwargs) - added:
field HerbertTokenizerFast.max_model_input_sizes - added:
field HerbertTokenizerFast.pretrained_init_configuration - added:
field HerbertTokenizerFast.pretrained_vocab_files_map - 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.create_token_type_ids_from_sequences(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:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class LayoutLMTokenizer(BertTokenizer) - added:
field LayoutLMTokenizer.max_model_input_sizes - added:
field LayoutLMTokenizer.pretrained_init_configuration - added:
field LayoutLMTokenizer.pretrained_vocab_files_map - added:
field LayoutLMTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class LayoutLMTokenizerFast(BertTokenizerFast) - added:
field LayoutLMTokenizerFast.max_model_input_sizes - added:
field LayoutLMTokenizerFast.pretrained_init_configuration - added:
field LayoutLMTokenizerFast.pretrained_vocab_files_map - added:
field LayoutLMTokenizerFast.slow_tokenizer_class - added:
field LayoutLMTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
class LongformerTokenizerFast(RobertaTokenizerFast) - removed:
field LongformerTokenizerFast.max_model_input_sizes - removed:
field LongformerTokenizerFast.pretrained_vocab_files_map
- added:
class LongformerTokenizerFast(RobertaTokenizerFast) - added:
field LongformerTokenizerFast.max_model_input_sizes - added:
field LongformerTokenizerFast.pretrained_vocab_files_map - added:
field LongformerTokenizerFast.slow_tokenizer_class - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable logger - added:
variable merges_url - added:
variable tokenizer_url - added:
variable vocab_url
- added:
class LxmertTokenizer(BertTokenizer) - added:
field LxmertTokenizer.max_model_input_sizes - added:
field LxmertTokenizer.pretrained_init_configuration - added:
field LxmertTokenizer.pretrained_vocab_files_map - added:
field LxmertTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES
- added:
class LxmertTokenizerFast(BertTokenizerFast) - added:
field LxmertTokenizerFast.max_model_input_sizes - added:
field LxmertTokenizerFast.pretrained_init_configuration - added:
field LxmertTokenizerFast.pretrained_vocab_files_map - added:
field LxmertTokenizerFast.slow_tokenizer_class - added:
field LxmertTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES
- removed:
method MarianTokenizer.prepare_translation_batch(src_texts: List[str], tgt_texts: Optional[List[str]]=…, max_length: Optional[int]=…, pad_to_max_length: bool=…, return_tensors: str=…) -> BatchEncoding - removed:
variable MODEL_NAMES - changed:
constructor MarianTokenizer(vocab=…, source_spm=…, target_spm=…, source_lang=…, target_lang=…, unk_token=…, eos_token=…, pad_token=…, max_len=…, **kwargs) → constructor MarianTokenizer(vocab, source_spm, target_spm, source_lang=…, target_lang=…, unk_token=…, eos_token=…, pad_token=…, model_max_length=…, **kwargs) - changed:
method MarianTokenizer.save_vocabulary(save_directory: str) -> Tuple[str] → method MarianTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
field MarianTokenizer.pretrained_init_configuration - added:
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 - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
- added:
class MBartTokenizer(XLMRobertaTokenizer) - added:
constructor MBartTokenizer(*args, tokenizer_file=…, **kwargs) - added:
field MBartTokenizer.max_model_input_sizes - added:
field MBartTokenizer.prefix_tokens: List[int] - added:
field MBartTokenizer.pretrained_vocab_files_map - 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.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.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 - 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.vocab_size - added:
variable FAIRSEQ_LANGUAGE_CODES - added:
variable SPM_URL - added:
variable logger
- added:
class MBartTokenizerFast(XLMRobertaTokenizerFast) - added:
constructor MBartTokenizerFast(*args, tokenizer_file=…, **kwargs) - added:
field MBartTokenizerFast.max_model_input_sizes - added:
field MBartTokenizerFast.prefix_tokens: List[int] - added:
field MBartTokenizerFast.pretrained_vocab_files_map - 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.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
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 - added:
method MBartTokenizerFast.set_src_lang_special_tokens(src_lang) -> None - added:
method MBartTokenizerFast.set_tgt_lang_special_tokens(lang: str) -> None - added:
variable FAIRSEQ_LANGUAGE_CODES - added:
variable MBartTokenizer - added:
variable SPM_URL - added:
variable logger - added:
variable tokenizer_URL
- added:
class MobileBertTokenizer(BertTokenizer) - added:
field MobileBertTokenizer.max_model_input_sizes - added:
field MobileBertTokenizer.pretrained_init_configuration - added:
field MobileBertTokenizer.pretrained_vocab_files_map - added:
field MobileBertTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class MobileBertTokenizerFast(BertTokenizerFast) - added:
field MobileBertTokenizerFast.max_model_input_sizes - added:
field MobileBertTokenizerFast.pretrained_init_configuration - added:
field MobileBertTokenizerFast.pretrained_vocab_files_map - added:
field MobileBertTokenizerFast.slow_tokenizer_class - added:
field MobileBertTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
class OpenAIGPTTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor OpenAIGPTTokenizerFast(vocab_file, merges_file, unk_token=…, **kwargs) - removed:
field OpenAIGPTTokenizerFast.max_model_input_sizes - removed:
field OpenAIGPTTokenizerFast.pretrained_vocab_files_map - removed:
field OpenAIGPTTokenizerFast.vocab_files_names - changed:
method OpenAIGPTTokenizer.save_vocabulary(save_directory) → method OpenAIGPTTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
field OpenAIGPTTokenizer.model_input_names - added:
property OpenAIGPTTokenizer.do_lower_case
- added:
class OpenAIGPTTokenizerFast(PreTrainedTokenizerFast) - added:
constructor OpenAIGPTTokenizerFast(vocab_file, merges_file, tokenizer_file=…, unk_token=…, **kwargs) - added:
field OpenAIGPTTokenizerFast.max_model_input_sizes - added:
field OpenAIGPTTokenizerFast.model_input_names - added:
field OpenAIGPTTokenizerFast.pretrained_vocab_files_map - 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:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class PegasusTokenizer(ReformerTokenizer) - added:
constructor PegasusTokenizer(*args, pad_token=…, **kwargs) - added:
field PegasusTokenizer.max_model_input_sizes - added:
field PegasusTokenizer.offset - added:
field PegasusTokenizer.pretrained_vocab_files_map - 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.get_special_tokens_mask(token_ids_0: List, token_ids_1: Optional[List]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method PegasusTokenizer.num_special_tokens_to_add(pair=…) - added:
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 - added:
property PegasusTokenizer.vocab_size -> int - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES
- added:
class PegasusTokenizerFast(ReformerTokenizerFast) - added:
field PegasusTokenizerFast.max_model_input_sizes - added:
field PegasusTokenizerFast.offset - added:
field PegasusTokenizerFast.pretrained_vocab_files_map - 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.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 - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable PegasusTokenizer - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES
- 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.max_model_input_sizes - added:
field PhobertTokenizer.pretrained_vocab_files_map - added:
field PhobertTokenizer.vocab_files_names - added:
function get_pairs(word) - 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:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class ProphetNetTokenizer(PreTrainedTokenizer) - added:
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) - added:
field ProphetNetTokenizer.max_model_input_sizes - added:
field ProphetNetTokenizer.pretrained_init_configuration - added:
field ProphetNetTokenizer.pretrained_vocab_files_map - added:
field ProphetNetTokenizer.vocab_files_names - added:
function load_vocab(vocab_file) - 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) - added:
method ProphetNetTokenizer.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
method ProphetNetTokenizer.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: 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:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class RagTokenizer - added:
constructor RagTokenizer(question_encoder, generator) - added:
method RagTokenizer.batch_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: str=…, truncation=…, **kwargs) -> BatchEncoding - added:
method RagTokenizer.save_pretrained(save_directory) - added:
static method RagTokenizer.from_pretrained(pretrained_model_name_or_path, **kwargs) - added:
variable logger
- changed:
constructor ReformerTokenizer(vocab_file, eos_token=…, unk_token=…, pad_token=…, additional_special_tokens=…, **kwargs) → constructor ReformerTokenizer(vocab_file, eos_token=…, unk_token=…, additional_special_tokens=…, **kwargs) - changed:
method ReformerTokenizer.get_vocab() → method ReformerTokenizer.get_vocab() -> Dict[str, int] - changed:
method ReformerTokenizer.save_vocabulary(save_directory) → method ReformerTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
field ReformerTokenizer.model_input_names
- added:
class ReformerTokenizerFast(PreTrainedTokenizerFast) - added:
constructor ReformerTokenizerFast(vocab_file, tokenizer_file=…, eos_token=…, unk_token=…, additional_special_tokens=…, **kwargs) - added:
field ReformerTokenizerFast.max_model_input_sizes - added:
field ReformerTokenizerFast.model_input_names - added:
field ReformerTokenizerFast.pretrained_vocab_files_map - 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:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable ReformerTokenizer - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class RetriBertTokenizer(BertTokenizer) - added:
field RetriBertTokenizer.max_model_input_sizes - added:
field RetriBertTokenizer.model_input_names - added:
field RetriBertTokenizer.pretrained_init_configuration - added:
field RetriBertTokenizer.pretrained_vocab_files_map - added:
field RetriBertTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class RetriBertTokenizerFast(BertTokenizerFast) - added:
field RetriBertTokenizerFast.max_model_input_sizes - added:
field RetriBertTokenizerFast.model_input_names - added:
field RetriBertTokenizerFast.pretrained_init_configuration - added:
field RetriBertTokenizerFast.pretrained_vocab_files_map - added:
field RetriBertTokenizerFast.slow_tokenizer_class - added:
field RetriBertTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
class RobertaTokenizerFast(GPT2TokenizerFast) - removed:
constructor RobertaTokenizerFast(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, add_prefix_space=…, trim_offsets=…, **kwargs) - removed:
field RobertaTokenizerFast.max_model_input_sizes - removed:
field RobertaTokenizerFast.model_input_names - removed:
field RobertaTokenizerFast.pretrained_vocab_files_map - 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:
method RobertaTokenizerFast.mask_token(value) - changed:
constructor RobertaTokenizer(vocab_file, merges_file, errors=…, bos_token=…, eos_token=…, sep_token=…, cls_token=…, unk_token=…, pad_token=…, mask_token=…, **kwargs) → 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) - changed:
method RobertaTokenizer.prepare_for_tokenization(text, add_special_tokens=…, **kwargs) → method RobertaTokenizer.prepare_for_tokenization(text, is_split_into_words=…, **kwargs)
- added:
class RobertaTokenizerFast(GPT2TokenizerFast) - 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=…, **kwargs) - added:
field RobertaTokenizerFast.max_model_input_sizes - added:
field RobertaTokenizerFast.model_input_names - added:
field RobertaTokenizerFast.pretrained_vocab_files_map - 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:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class SqueezeBertTokenizer(BertTokenizer) - added:
field SqueezeBertTokenizer.max_model_input_sizes - added:
field SqueezeBertTokenizer.pretrained_init_configuration - added:
field SqueezeBertTokenizer.pretrained_vocab_files_map - added:
field SqueezeBertTokenizer.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- added:
class SqueezeBertTokenizerFast(BertTokenizerFast) - added:
field SqueezeBertTokenizerFast.max_model_input_sizes - added:
field SqueezeBertTokenizerFast.pretrained_init_configuration - added:
field SqueezeBertTokenizerFast.pretrained_vocab_files_map - added:
field SqueezeBertTokenizerFast.slow_tokenizer_class - added:
field SqueezeBertTokenizerFast.vocab_files_names - added:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
variable SPIECE_UNDERLINE - changed:
method T5Tokenizer.save_vocabulary(save_directory) → method T5Tokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
field T5Tokenizer.model_input_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.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.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
- added:
class T5TokenizerFast(PreTrainedTokenizerFast) - added:
constructor T5TokenizerFast(vocab_file, tokenizer_file=…, eos_token=…, unk_token=…, pad_token=…, extra_ids=…, additional_special_tokens=…, **kwargs) - added:
field T5TokenizerFast.max_model_input_sizes - added:
field T5TokenizerFast.model_input_names - added:
field T5TokenizerFast.prefix_tokens: List[int] - added:
field T5TokenizerFast.pretrained_vocab_files_map - 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.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 - added:
method T5TokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable T5Tokenizer - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- removed:
class TransfoXLTokenizerFast(PreTrainedTokenizerFast) - removed:
constructor TransfoXLTokenizerFast(special=…, min_freq=…, max_size=…, lower_case=…, delimiter=…, vocab_file=…, pretrained_vocab_file=…, never_split=…, unk_token=…, eos_token=…, additional_special_tokens=…, add_eos=…, add_double_eos=…, normalization=…, **kwargs) - removed:
field TransfoXLTokenizerFast.max_model_input_sizes - removed:
field TransfoXLTokenizerFast.model_input_names - removed:
field TransfoXLTokenizerFast.pretrained_vocab_files_map - removed:
field TransfoXLTokenizerFast.vocab_files_names - removed:
method TransfoXLTokenizer.prepare_for_tokenization(text, **kwargs) - removed:
method TransfoXLTokenizerFast.save_pretrained(save_directory) - removed:
variable PRETRAINED_VOCAB_FILES_MAP_FAST - removed:
variable VOCAB_FILES_NAMES_FAST - changed:
constructor TransfoXLTokenizer(special=…, min_freq=…, max_size=…, lower_case=…, delimiter=…, vocab_file=…, pretrained_vocab_file=…, never_split=…, unk_token=…, eos_token=…, additional_special_tokens=…, **kwargs) → 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) - changed:
method TransfoXLTokenizer.save_vocabulary(vocab_path) → method TransfoXLTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
function detokenize_numbers(text: str) -> str - added:
function tokenize_numbers(text_array: List[str]) -> List[str] - 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:
property TransfoXLTokenizer.do_lower_case - added:
variable DETOKENIZE_NUMBERS - added:
variable MATCH_NUMBERS
- removed:
class BatchEncoding(UserDict) - removed:
class CharSpan(NamedTuple) - removed:
class PreTrainedTokenizerFast(PreTrainedTokenizer) - removed:
class SpecialTokensMixin - removed:
class TokenSpan(NamedTuple) - removed:
constructor BatchEncoding(data: Optional[Dict[str, Any]]=…, encoding: Optional[Union[EncodingFast, Sequence[EncodingFast]]]=…) - removed:
constructor PreTrainedTokenizerFast(tokenizer: BaseTokenizerFast, **kwargs) - removed:
constructor SpecialTokensMixin(**kwargs) - removed:
field CharSpan.end: int - removed:
field CharSpan.start: int - removed:
field PreTrainedTokenizer.NO_PAD_TOKEN_FOR_BATCH_MSG - removed:
field PreTrainedTokenizer.UNEVEN_SEQUENCES_FOR_BATCH_MSG - removed:
field PreTrainedTokenizer.max_model_input_sizes: Dict[str, int] - removed:
field PreTrainedTokenizer.model_input_names: List[str] - removed:
field PreTrainedTokenizer.padding_side: str - removed:
field PreTrainedTokenizer.pretrained_init_configuration: Dict[str, Dict[str, Any]] - removed:
field PreTrainedTokenizer.pretrained_vocab_files_map: Dict[str, Dict[str, str]] - removed:
field PreTrainedTokenizer.vocab_files_names: Dict[str, str] - removed:
field SpecialTokensMixin.SPECIAL_TOKENS_ATTRIBUTES - removed:
field TokenSpan.end: int - removed:
field TokenSpan.start: int - removed:
function flatten(x: Sequence) - removed:
function trim_batch(input_ids, pad_token_id, attention_mask=…) - removed:
function truncate_and_pad(tokenizer: BaseTokenizerFast, max_length: int, stride: int, strategy: str, pad_to_max_length: bool, padding_side: str, pad_token_id: int, pad_token_type_id: int, pad_token: str) - removed:
method BatchEncoding.char_to_token(batch_or_char_index: int, char_index: Optional[int]=…) -> int - removed:
method BatchEncoding.char_to_word(batch_or_char_index: int, char_index: Optional[int]=…) -> int - removed:
method BatchEncoding.items() - removed:
method BatchEncoding.keys() - removed:
method BatchEncoding.to(device: str) - removed:
method BatchEncoding.token_to_chars(batch_or_token_index: int, token_index: Optional[int]=…) -> CharSpan - removed:
method BatchEncoding.token_to_word(batch_or_token_index: int, token_index: Optional[int]=…) -> int - removed:
method BatchEncoding.tokens(batch_index: int=…) -> List[int] - removed:
method BatchEncoding.values() - removed:
method BatchEncoding.word_to_chars(batch_or_word_index: int, word_index: Optional[int]=…) -> CharSpan - removed:
method BatchEncoding.word_to_tokens(batch_or_word_index: int, word_index: Optional[int]=…) -> TokenSpan - removed:
method BatchEncoding.words(batch_index: int=…) -> List[Optional[int]] - removed:
method PreTrainedTokenizer.add_special_tokens(special_tokens_dict) - removed:
method PreTrainedTokenizer.add_tokens(new_tokens: Union[str, List[str]]) -> int - removed:
method PreTrainedTokenizer.batch_decode(sequences: List[List[int]], **kwargs) -> List[str] - removed:
method PreTrainedTokenizer.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=…, max_length: Optional[int]=…, stride: int=…, truncation_strategy: str=…, pad_to_max_length: bool=…, is_pretokenized: bool=…, return_tensors: Optional[str]=…,… - removed:
method PreTrainedTokenizer.build_inputs_with_special_tokens(token_ids_0: List, token_ids_1: Optional[List]=…) -> List - removed:
method PreTrainedTokenizer.convert_to_tensors_(batch_outputs: dict, return_tensors: str) -> None - removed:
method PreTrainedTokenizer.create_token_type_ids_from_sequences(token_ids_0: List, token_ids_1: Optional[List]=…) -> List[int] - removed:
method PreTrainedTokenizer.decode(token_ids: List[int], skip_special_tokens: bool=…, clean_up_tokenization_spaces: bool=…) -> str - removed:
method PreTrainedTokenizer.encode(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, max_length: Optional[int]=…, stride: int=…, truncation_strategy: str=…, pad_to_max_length: bool=…, return_tensors: Optional[str]=…, **kwargs) - removed:
method PreTrainedTokenizer.encode_plus(text: Union[TextInput, PreTokenizedInput, EncodedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput, EncodedInput]]=…, add_special_tokens: bool=…, max_length: Optional[int]=…, stride: int=…, truncation_strategy: str=…, pad_to_max_length: bool=…, is_pretokenized: bool=…, return_tensors: Optional[str]=…, return_token_type_ids: Optional[bool]=…, ret… - removed:
method PreTrainedTokenizer.get_vocab() - removed:
method PreTrainedTokenizer.prepare_for_model(ids: List[int], pair_ids: Optional[List[int]]=…, max_length: Optional[int]=…, add_special_tokens: bool=…, stride: int=…, truncation_strategy: str=…, pad_to_max_length: bool=…, return_tensors: Optional[str]=…, return_token_type_ids: Optional[bool]=…, return_attention_mask: Optional[bool]=…, return_overflowing_tokens: bool=…, return_special_tokens_mask:… - removed:
method PreTrainedTokenizer.save_pretrained(save_directory) - removed:
method PreTrainedTokenizer.save_vocabulary(save_directory) -> Tuple[str] - removed:
method PreTrainedTokenizer.truncate_sequences(ids: List[int], pair_ids: Optional[List[int]]=…, num_tokens_to_remove: int=…, truncation_strategy: str=…, stride: int=…) -> Tuple[List[int], List[int], List[int]] - removed:
method PreTrainedTokenizerFast.add_special_tokens(special_tokens_dict: dict) -> int - removed:
method PreTrainedTokenizerFast.add_tokens(new_tokens: List[Union[str, AddedTokenFast]]) -> int - removed:
method PreTrainedTokenizerFast.batch_encode_plus(batch_text_or_text_pairs: Union[List[TextInput], List[TextInputPair], List[PreTokenizedInput], List[PreTokenizedInputPair]], add_special_tokens: bool=…, max_length: Optional[int]=…, stride: int=…, truncation_strategy: str=…, pad_to_max_length: bool=…, is_pretokenized: bool=…, return_tensors: Optional[str]=…, return_token_type_ids: Optional[bool]=…,… - removed:
method PreTrainedTokenizerFast.convert_tokens_to_string(tokens: List[int], skip_special_tokens: bool=…) -> str - removed:
method PreTrainedTokenizerFast.decode(token_ids: List[int], skip_special_tokens: bool=…, clean_up_tokenization_spaces: bool=…) -> str - removed:
method PreTrainedTokenizerFast.encode_plus(text: Union[TextInput, PreTokenizedInput], text_pair: Optional[Union[TextInput, PreTokenizedInput]]=…, add_special_tokens: bool=…, max_length: Optional[int]=…, pad_to_max_length: bool=…, stride: int=…, truncation_strategy: str=…, is_pretokenized: bool=…, return_tensors: Optional[bool]=…, return_token_type_ids: Optional[bool]=…, return_attention_mask: Opt… - removed:
method PreTrainedTokenizerFast.get_vocab() - removed:
method PreTrainedTokenizerFast.num_special_tokens_to_add(pair: bool=…) -> int - removed:
method PreTrainedTokenizerFast.save_vocabulary(save_directory: str) -> Tuple[str] - removed:
method PreTrainedTokenizerFast.tokenize(text: TextInput, pair: Optional[TextInput]=…, add_special_tokens: bool=…) -> List[str] - removed:
property BatchEncoding.encodings -> Optional[List[EncodingFast]] - removed:
property PreTrainedTokenizer.max_len -> int - removed:
property PreTrainedTokenizer.max_len_sentences_pair -> int - removed:
property PreTrainedTokenizer.max_len_single_sentence -> int - removed:
property PreTrainedTokenizerFast.backend_tokenizer -> BaseTokenizerFast - removed:
property PreTrainedTokenizerFast.decoder -> DecoderFast - removed:
property PreTrainedTokenizerFast.is_fast -> bool - removed:
property PreTrainedTokenizerFast.vocab_size -> int - removed:
property SpecialTokensMixin.additional_special_tokens - removed:
property SpecialTokensMixin.additional_special_tokens_ids - removed:
property SpecialTokensMixin.all_special_ids - removed:
property SpecialTokensMixin.all_special_tokens - removed:
property SpecialTokensMixin.bos_token - removed:
property SpecialTokensMixin.bos_token_id - removed:
property SpecialTokensMixin.cls_token - removed:
property SpecialTokensMixin.cls_token_id - removed:
property SpecialTokensMixin.eos_token - removed:
property SpecialTokensMixin.eos_token_id - removed:
property SpecialTokensMixin.mask_token - removed:
property SpecialTokensMixin.mask_token_id - removed:
property SpecialTokensMixin.pad_token - removed:
property SpecialTokensMixin.pad_token_id - removed:
property SpecialTokensMixin.pad_token_type_id - removed:
property SpecialTokensMixin.sep_token - removed:
property SpecialTokensMixin.sep_token_id - removed:
property SpecialTokensMixin.special_tokens_map - removed:
property SpecialTokensMixin.unk_token - removed:
property SpecialTokensMixin.unk_token_id - removed:
static method PreTrainedTokenizer.clean_up_tokenization(out_string: str) -> str - removed:
static method PreTrainedTokenizer.from_pretrained(*inputs, **kwargs) - removed:
variable EncodedInput - removed:
variable EncodedInputPair - removed:
variable LARGE_INTEGER - removed:
variable PreTokenizedInput - removed:
variable PreTokenizedInputPair - removed:
variable TextInput - removed:
variable TextInputPair - removed:
variable VERY_LARGE_INTEGER - changed:
class PreTrainedTokenizer(SpecialTokensMixin) → class PreTrainedTokenizer(PreTrainedTokenizerBase) - changed:
constructor PreTrainedTokenizer(model_max_length=…, **kwargs) → constructor PreTrainedTokenizer(**kwargs) - changed:
method PreTrainedTokenizer.convert_ids_to_tokens(ids: Union[int, List[int]], skip_special_tokens: bool=…) -> Union[int, List[int]] → 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) → method PreTrainedTokenizer.convert_tokens_to_ids(tokens: Union[str, List[str]]) -> Union[int, List[int]] - changed:
method PreTrainedTokenizer.num_special_tokens_to_add(pair=…) → method PreTrainedTokenizer.num_special_tokens_to_add(pair: bool=…) -> int - changed:
method PreTrainedTokenizer.prepare_for_tokenization(text: str, **kwargs) -> str → 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) → method PreTrainedTokenizer.tokenize(text: TextInput, **kwargs) -> List[str] - added:
method PreTrainedTokenizer.get_added_vocab() -> Dict[str, int] - added:
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
- added:
class AddedToken - added:
class BatchEncoding(UserDict) - added:
class CharSpan(NamedTuple) - added:
class EncodingFast - added:
class ExplicitEnum(Enum) - added:
class PaddingStrategy(ExplicitEnum) - added:
class PreTrainedTokenizerBase(SpecialTokensMixin) - added:
class SpecialTokensMixin - added:
class TensorType(ExplicitEnum) - added:
class TokenSpan(NamedTuple) - added:
class TruncationStrategy(ExplicitEnum) - added:
constructor BatchEncoding(data: Optional[Dict[str, Any]]=…, encoding: Optional[Union[EncodingFast, Sequence[EncodingFast]]]=…, tensor_type: Union[None, str, TensorType]=…, prepend_batch_axis: bool=…) - added:
constructor PreTrainedTokenizerBase(**kwargs) - added:
constructor SpecialTokensMixin(verbose=…, **kwargs) - added:
field AddedToken.content: str - added:
field AddedToken.lstrip: bool - added:
field AddedToken.normalized: bool - added:
field AddedToken.rstrip: bool - added:
field AddedToken.single_word: bool - added:
field CharSpan.end: int - added:
field CharSpan.start: int - added:
field PaddingStrategy.DO_NOT_PAD - added:
field PaddingStrategy.LONGEST - added:
field PaddingStrategy.MAX_LENGTH - added:
field PreTrainedTokenizerBase.max_model_input_sizes: Dict[str, Optional[int]] - added:
field PreTrainedTokenizerBase.model_input_names: List[str] - added:
field PreTrainedTokenizerBase.padding_side: str - added:
field PreTrainedTokenizerBase.pretrained_init_configuration: Dict[str, Dict[str, Any]] - added:
field PreTrainedTokenizerBase.pretrained_vocab_files_map: Dict[str, Dict[str, str]] - added:
field PreTrainedTokenizerBase.slow_tokenizer_class - added:
field PreTrainedTokenizerBase.vocab_files_names: Dict[str, str] - added:
field SpecialTokensMixin.SPECIAL_TOKENS_ATTRIBUTES - added:
field TensorType.JAX - added:
field TensorType.NUMPY - added:
field TensorType.PYTORCH - added:
field TensorType.TENSORFLOW - added:
field TokenSpan.end: int - added:
field TokenSpan.start: int - added:
field TruncationStrategy.DO_NOT_TRUNCATE - added:
field TruncationStrategy.LONGEST_FIRST - added:
field TruncationStrategy.ONLY_FIRST - added:
field TruncationStrategy.ONLY_SECOND - added:
function to_py_obj(obj) - added:
method BatchEncoding.char_to_token(batch_or_char_index: int, char_index: Optional[int]=…) -> int - added:
method BatchEncoding.char_to_word(batch_or_char_index: int, char_index: Optional[int]=…) -> int - added:
method BatchEncoding.convert_to_tensors(tensor_type: Optional[Union[str, TensorType]]=…, prepend_batch_axis: bool=…) - added:
method BatchEncoding.items() - added:
method BatchEncoding.keys() - added:
method BatchEncoding.to(device: Union[str, torch.device]) -> BatchEncoding - added:
method BatchEncoding.token_to_chars(batch_or_token_index: int, token_index: Optional[int]=…) -> CharSpan - added:
method BatchEncoding.token_to_word(batch_or_token_index: int, token_index: Optional[int]=…) -> int - added:
method BatchEncoding.tokens(batch_index: int=…) -> List[str] - added:
method BatchEncoding.values() - added:
method BatchEncoding.word_to_chars(batch_or_word_index: int, word_index: Optional[int]=…) -> CharSpan - added:
method BatchEncoding.word_to_tokens(batch_or_word_index: int, word_index: Optional[int]=…) -> Optional[TokenSpan] - added:
method BatchEncoding.words(batch_index: int=…) -> List[Optional[int]] - added:
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] - added:
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… - added:
method PreTrainedTokenizerBase.build_inputs_with_special_tokens(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
method PreTrainedTokenizerBase.convert_tokens_to_string(tokens: List[str]) -> str - added:
method PreTrainedTokenizerBase.create_token_type_ids_from_sequences(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…) -> List[int] - added:
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 - added:
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… - added:
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… - added:
method PreTrainedTokenizerBase.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method PreTrainedTokenizerBase.get_vocab() -> Dict[str, int] - added:
method PreTrainedTokenizerBase.num_special_tokens_to_add(pair: bool=…) -> int - added:
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… - added:
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… - added:
method PreTrainedTokenizerBase.save_pretrained(save_directory: str, legacy_format: bool=…, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
method PreTrainedTokenizerBase.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
method PreTrainedTokenizerBase.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> List[str] - added:
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]] - added:
method SpecialTokensMixin.add_special_tokens(special_tokens_dict: Dict[str, Union[str, AddedToken]]) -> int - added:
method SpecialTokensMixin.add_tokens(new_tokens: Union[str, AddedToken, List[Union[str, AddedToken]]], special_tokens: bool=…) -> int - added:
method SpecialTokensMixin.sanitize_special_tokens() -> int - added:
property BatchEncoding.encodings -> Optional[List[EncodingFast]] - added:
property BatchEncoding.is_fast -> bool - added:
property PreTrainedTokenizerBase.max_len -> int - added:
property PreTrainedTokenizerBase.max_len_sentences_pair -> int - added:
property PreTrainedTokenizerBase.max_len_single_sentence -> int - added:
property SpecialTokensMixin.additional_special_tokens -> List[str] - added:
property SpecialTokensMixin.additional_special_tokens_ids -> List[int] - 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.bos_token -> str - added:
property SpecialTokensMixin.bos_token_id -> Optional[int] - added:
property SpecialTokensMixin.cls_token -> str - added:
property SpecialTokensMixin.cls_token_id -> Optional[int] - added:
property SpecialTokensMixin.eos_token -> str - added:
property SpecialTokensMixin.eos_token_id -> Optional[int] - added:
property SpecialTokensMixin.mask_token -> str - added:
property SpecialTokensMixin.mask_token_id -> Optional[int] - added:
property SpecialTokensMixin.pad_token -> str - added:
property SpecialTokensMixin.pad_token_id -> Optional[int] - added:
property SpecialTokensMixin.pad_token_type_id -> int - added:
property SpecialTokensMixin.sep_token -> str - added:
property SpecialTokensMixin.sep_token_id -> Optional[int] - 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:
property SpecialTokensMixin.unk_token -> str - added:
property SpecialTokensMixin.unk_token_id -> Optional[int] - added:
static method PreTrainedTokenizerBase.clean_up_tokenization(out_string: str) -> str - added:
static method PreTrainedTokenizerBase.from_pretrained(pretrained_model_name_or_path, *init_inputs, **kwargs) - added:
variable ADDED_TOKENS_FILE - added:
variable ENCODE_KWARGS_DOCSTRING - added:
variable ENCODE_PLUS_ADDITIONAL_KWARGS_DOCSTRING - added:
variable EncodedInput - added:
variable EncodedInputPair - added:
variable FULL_TOKENIZER_FILE - added:
variable INIT_TOKENIZER_DOCSTRING - added:
variable LARGE_INTEGER - added:
variable PREPARE_SEQ2SEQ_BATCH_DOCSTRING - added:
variable PreTokenizedInput - added:
variable PreTokenizedInputPair - added:
variable SPECIAL_TOKENS_MAP_FILE - added:
variable TOKENIZER_CONFIG_FILE - added:
variable TextInput - added:
variable TextInputPair - added:
variable VERY_LARGE_INTEGER - added:
variable logger
- added:
class PreTrainedTokenizerFast(PreTrainedTokenizerBase) - added:
constructor PreTrainedTokenizerFast(*args, **kwargs) - added:
field PreTrainedTokenizerFast.slow_tokenizer_class: PreTrainedTokenizer - added:
method PreTrainedTokenizerFast.convert_ids_to_tokens(ids: Union[int, List[int]], skip_special_tokens: bool=…) -> Union[str, List[str]] - added:
method PreTrainedTokenizerFast.convert_tokens_to_ids(tokens: Union[str, List[str]]) -> Union[int, List[int]] - added:
method PreTrainedTokenizerFast.convert_tokens_to_string(tokens: List[str]) -> str - added:
method PreTrainedTokenizerFast.get_added_vocab() -> Dict[str, int] - added:
method PreTrainedTokenizerFast.get_vocab() -> Dict[str, int] - added:
method PreTrainedTokenizerFast.num_special_tokens_to_add(pair: bool=…) -> int - added:
method PreTrainedTokenizerFast.set_truncation_and_padding(padding_strategy: PaddingStrategy, truncation_strategy: TruncationStrategy, max_length: int, stride: int, pad_to_multiple_of: Optional[int]) - added:
method PreTrainedTokenizerFast.tokenize(text: str, pair: Optional[str]=…, add_special_tokens: bool=…, **kwargs) -> List[str] - added:
property PreTrainedTokenizerFast.backend_tokenizer -> TokenizerFast - added:
property PreTrainedTokenizerFast.decoder -> DecoderFast - added:
property PreTrainedTokenizerFast.is_fast -> bool - added:
property PreTrainedTokenizerFast.vocab -> Dict[str, int] - added:
property PreTrainedTokenizerFast.vocab_size -> int - added:
variable ADDED_TOKENS_FILE - added:
variable SPECIAL_TOKENS_MAP_FILE - added:
variable TOKENIZER_CONFIG_FILE - added:
variable TOKENIZER_FILE - added:
variable logger
- changed:
method XLMTokenizer.save_vocabulary(save_directory) → method XLMTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
property XLMTokenizer.do_lower_case
- added:
class XLMProphetNetTokenizer(PreTrainedTokenizer) - added:
constructor XLMProphetNetTokenizer(vocab_file, bos_token=…, eos_token=…, sep_token=…, unk_token=…, pad_token=…, cls_token=…, mask_token=…, **kwargs) - added:
field XLMProphetNetTokenizer.max_model_input_sizes - added:
field XLMProphetNetTokenizer.model_input_names - added:
field XLMProphetNetTokenizer.pretrained_vocab_files_map - added:
field XLMProphetNetTokenizer.vocab_files_names - added:
function load_vocab(vocab_file) - 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:
variable PRETRAINED_INIT_CONFIGURATION - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES - added:
variable logger
- changed:
method XLMRobertaTokenizer.save_vocabulary(save_directory) → method XLMRobertaTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable SPIECE_UNDERLINE
- 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.max_model_input_sizes - added:
field XLMRobertaTokenizerFast.model_input_names - added:
field XLMRobertaTokenizerFast.pretrained_vocab_files_map - 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.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method XLMRobertaTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable VOCAB_FILES_NAMES - added:
variable XLMRobertaTokenizer - added:
variable logger
- removed:
variable SPIECE_UNDERLINE - changed:
method XLNetTokenizer.save_vocabulary(save_directory) → method XLNetTokenizer.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str]
- 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.max_model_input_sizes - added:
field XLNetTokenizerFast.padding_side - added:
field XLNetTokenizerFast.pretrained_vocab_files_map - 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.get_special_tokens_mask(token_ids_0: List[int], token_ids_1: Optional[List[int]]=…, already_has_special_tokens: bool=…) -> List[int] - added:
method XLNetTokenizerFast.save_vocabulary(save_directory: str, filename_prefix: Optional[str]=…) -> Tuple[str] - added:
variable PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES - added:
variable PRETRAINED_VOCAB_FILES_MAP - added:
variable SEG_ID_A - added:
variable SEG_ID_B - added:
variable SEG_ID_CLS - added:
variable SEG_ID_PAD - added:
variable SEG_ID_SEP - added:
variable SPIECE_UNDERLINE - added:
variable VOCAB_FILES_NAMES - added:
variable XLNetTokenizer - added:
variable logger
- removed:
class SequentialDistributedSampler(Sampler) - removed:
constructor SequentialDistributedSampler(dataset, num_replicas=…, rank=…) - removed:
field Trainer.args: TrainingArguments - removed:
field Trainer.compute_metrics: Optional[Callable[[EvalPrediction], Dict]] - removed:
field Trainer.data_collator: DataCollator - removed:
field Trainer.epoch: Optional[float] - removed:
field Trainer.eval_dataset: Optional[Dataset] - removed:
field Trainer.global_step: Optional[int] - removed:
field Trainer.model: PreTrainedModel - removed:
field Trainer.optimizers: Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR] - removed:
field Trainer.prediction_loss_only: bool - removed:
field Trainer.tb_writer: Optional[SummaryWriter] - removed:
field Trainer.train_dataset: Optional[Dataset] - removed:
function get_tpu_sampler(dataset: Dataset) - removed:
function is_apex_available() - removed:
function is_tensorboard_available() - removed:
function is_wandb_available() - removed:
function set_seed(seed: int) - removed:
function torch_distributed_zero_first(local_rank: int) - removed:
method Trainer.distributed_concat(tensor: torch.Tensor, num_total_examples: int) -> torch.Tensor - removed:
method Trainer.get_optimizers(num_training_steps: int) -> Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR] - changed:
constructor Trainer(model: PreTrainedModel, args: TrainingArguments, data_collator: Optional[DataCollator]=…, train_dataset: Optional[Dataset]=…, eval_dataset: Optional[Dataset]=…, compute_metrics: Optional[Callable[[EvalPrediction], Dict]]=…, prediction_loss_only=…, tb_writer: Optional[SummaryWriter]=…, optimizers: Tuple[torch.optim.Optimizer, torch.optim.lr_scheduler.LambdaLR]=…) → 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… - changed:
method Trainer.evaluate(eval_dataset: Optional[Dataset]=…, prediction_loss_only: Optional[bool]=…) -> Dict[str, float] → method Trainer.evaluate(eval_dataset: Optional[Dataset]=…) -> Dict[str, float] - changed:
method Trainer.train(model_path: Optional[str]=…) → method Trainer.train(model_path: Optional[str]=…, trial: Union[optuna.Trial, Dict[str, Any]]=…) - added:
method Trainer.add_callback(callback) - added:
method Trainer.call_model_init(trial=…) - added:
method Trainer.compute_loss(model, inputs) - added:
method Trainer.create_optimizer_and_scheduler(num_training_steps: int) - added:
method Trainer.floating_point_ops(inputs: Dict[str, Union[torch.Tensor, Any]]) - added:
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 - added:
method Trainer.is_local_process_zero() -> bool - added:
method Trainer.is_world_process_zero() -> bool - added:
method Trainer.log(logs: Dict[str, float]) -> None - added:
method Trainer.pop_callback(callback) - added:
method Trainer.prediction_loop(dataloader: DataLoader, description: str, prediction_loss_only: Optional[bool]=…) -> PredictionOutput - added:
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]] - added:
method Trainer.remove_callback(callback) - added:
method Trainer.store_flos() - added:
method Trainer.training_step(model: nn.Module, inputs: Dict[str, Union[torch.Tensor, Any]]) -> torch.Tensor - added:
variable DEFAULT_CALLBACKS - added:
variable DEFAULT_PROGRESS_CALLBACK
- added:
class CallbackHandler(TrainerCallback) - added:
class DefaultFlowCallback(TrainerCallback) - added:
class PrinterCallback(TrainerCallback) - added:
class ProgressCallback(TrainerCallback) - added:
class TrainerCallback - added:
class TrainerControl - added:
class TrainerState - added:
constructor CallbackHandler(callbacks, model, optimizer, lr_scheduler) - added:
constructor ProgressCallback() - added:
field TrainerControl.should_epoch_stop: bool - added:
field TrainerControl.should_evaluate: bool - added:
field TrainerControl.should_log: bool - added:
field TrainerControl.should_save: bool - added:
field TrainerControl.should_training_stop: bool - added:
field TrainerState.best_metric: Optional[float] - added:
field TrainerState.best_model_checkpoint: Optional[str] - added:
field TrainerState.epoch: Optional[float] - added:
field TrainerState.global_step: int - added:
field TrainerState.is_hyper_param_search: bool - added:
field TrainerState.is_local_process_zero: bool - added:
field TrainerState.is_world_process_zero: bool - added:
field TrainerState.log_history: List[Dict[str, float]] - added:
field TrainerState.max_steps: int - added:
field TrainerState.num_train_epochs: int - added:
field TrainerState.total_flos: int - added:
field TrainerState.trial_name: str - added:
field TrainerState.trial_params: Dict[str, Union[str, float, int, bool]] - added:
method CallbackHandler.add_callback(callback) - added:
method CallbackHandler.call_event(event, args, state, control, **kwargs) - added:
method CallbackHandler.on_epoch_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_epoch_end(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_evaluate(args: TrainingArguments, state: TrainerState, control: TrainerControl, metrics) - added:
method CallbackHandler.on_init_end(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_log(args: TrainingArguments, state: TrainerState, control: TrainerControl, logs) - added:
method CallbackHandler.on_prediction_step(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_save(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_step_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_step_end(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_train_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.on_train_end(args: TrainingArguments, state: TrainerState, control: TrainerControl) - added:
method CallbackHandler.pop_callback(callback) - added:
method CallbackHandler.remove_callback(callback) - added:
method DefaultFlowCallback.on_epoch_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method DefaultFlowCallback.on_step_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method PrinterCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method ProgressCallback.on_evaluate(args, state, control, **kwargs) - added:
method ProgressCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method ProgressCallback.on_prediction_step(args, state, control, eval_dataloader=…, **kwargs) - added:
method ProgressCallback.on_step_end(args, state, control, **kwargs) - added:
method ProgressCallback.on_train_begin(args, state, control, **kwargs) - added:
method ProgressCallback.on_train_end(args, state, control, **kwargs) - added:
method TrainerCallback.on_epoch_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_epoch_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_evaluate(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_init_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_log(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_prediction_step(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_save(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_step_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_step_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_train_begin(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerCallback.on_train_end(args: TrainingArguments, state: TrainerState, control: TrainerControl, **kwargs) - added:
method TrainerState.save_to_json(json_path: str) - added:
property CallbackHandler.callback_list - added:
static method TrainerState.load_from_json(json_path: str) - added:
variable logger
- added:
class DistributedTensorGatherer - added:
class SequentialDistributedSampler(Sampler) - added:
constructor DistributedTensorGatherer(world_size, num_samples, make_multiple_of=…, padding_index=…) - added:
constructor SequentialDistributedSampler(dataset, num_replicas=…, rank=…) - added:
function distributed_broadcast_scalars(scalars: List[Union[int, float]], num_total_examples: Optional[int]=…) -> torch.Tensor - added:
function distributed_concat(tensor: torch.Tensor, num_total_examples: Optional[int]=…) -> torch.Tensor - added:
function get_tpu_sampler(dataset: torch.utils.data.dataset.Dataset) - added:
function nested_concat(tensors, new_tensors, padding_index=…) - added:
function nested_detach(tensors) - added:
function nested_expand_like(arrays, new_seq_length, padding_index=…) - added:
function nested_new_like(arrays, num_samples, padding_index=…) - added:
function nested_numpify(tensors) - added:
function nested_truncate(tensors, limit) - added:
function nested_xla_mesh_reduce(tensors, name) - added:
function numpy_pad_and_concatenate(array1, array2, padding_index=…) - added:
function reissue_pt_warnings(caught_warnings) - added:
function torch_distributed_zero_first(local_rank: int) - added:
function torch_pad_and_concatenate(tensor1, tensor2, padding_index=…) - added:
method DistributedTensorGatherer.add_arrays(arrays) - added:
method DistributedTensorGatherer.finalize() - added:
variable SAVE_STATE_WARNING - added:
variable logger
- removed:
field TFTrainer.args: TFTrainingArguments - removed:
field TFTrainer.compute_metrics: Optional[Callable[[EvalPrediction], Dict]] - removed:
field TFTrainer.eval_dataset: Optional[tf.data.Dataset] - removed:
field TFTrainer.model: TFPreTrainedModel - removed:
field TFTrainer.prediction_loss_only: bool - removed:
field TFTrainer.train_dataset: Optional[tf.data.Dataset] - changed:
constructor TFTrainer(model: TFPreTrainedModel, args: TFTrainingArguments, train_dataset: Optional[tf.data.Dataset]=…, eval_dataset: Optional[tf.data.Dataset]=…, compute_metrics: Optional[Callable[[EvalPrediction], Dict]]=…, prediction_loss_only=…) → 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) - changed:
method TFTrainer.evaluate(eval_dataset: Optional[tf.data.Dataset]=…, prediction_loss_only: Optional[bool]=…) -> Dict[str, float] → method TFTrainer.evaluate(eval_dataset: Optional[tf.data.Dataset]=…) -> Dict[str, float] - changed:
method TFTrainer.save_model() -> None → method TFTrainer.save_model(output_dir: Optional[str]=…) - added:
method TFTrainer.apply_gradients(features, labels, nb_instances_in_global_batch) - added:
method TFTrainer.create_optimizer_and_scheduler(num_training_steps: int) - added:
method TFTrainer.distributed_prediction_steps(batch) - added:
method TFTrainer.distributed_training_steps(batch) - added:
method TFTrainer.get_eval_tfdataset(eval_dataset: Optional[tf.data.Dataset]=…) -> tf.data.Dataset - added:
method TFTrainer.get_test_tfdataset(test_dataset: tf.data.Dataset) -> tf.data.Dataset - added:
method TFTrainer.get_train_tfdataset() -> tf.data.Dataset - added:
method TFTrainer.log(logs: Dict[str, float]) -> None - added:
method TFTrainer.prediction_loop(dataset: tf.data.Dataset, steps: int, num_examples: int, description: str, prediction_loss_only: Optional[bool]=…) -> PredictionOutput - added:
method TFTrainer.prediction_step(features: tf.Tensor, labels: tf.Tensor, nb_instances_in_global_batch: tf.Tensor) -> tf.Tensor - added:
method TFTrainer.run_model(features, labels, training) - added:
method TFTrainer.setup_comet() - added:
method TFTrainer.setup_wandb() - added:
method TFTrainer.training_step(features, labels, nb_instances_in_global_batch)
- changed:
field EvalPrediction.predictions: np.ndarray → field EvalPrediction.predictions: Union[np.ndarray, Tuple[np.ndarray]] - changed:
field PredictionOutput.predictions: np.ndarray → field PredictionOutput.predictions: Union[np.ndarray, Tuple[np.ndarray]] - added:
class BestRun(NamedTuple) - added:
class EvaluationStrategy(ExplicitEnum) - added:
class HPSearchBackend(ExplicitEnum) - added:
field BestRun.hyperparameters: Dict[str, Any] - added:
field BestRun.objective: float - added:
field BestRun.run_id: str - added:
field EvaluationStrategy.EPOCH - added:
field EvaluationStrategy.NO - added:
field EvaluationStrategy.STEPS - added:
field HPSearchBackend.OPTUNA - added:
field HPSearchBackend.RAY - added:
function default_compute_objective(metrics: Dict[str, float]) -> float - added:
function default_hp_space_optuna(trial) -> Dict[str, float] - added:
function default_hp_space_ray(trial) -> Dict[str, float] - added:
function is_main_process(local_rank) - added:
function set_seed(seed: int) - added:
function total_processes_number(local_rank) - added:
variable default_hp_space
- removed:
function is_tpu_available() - added:
field TrainingArguments.adam_beta1: float - added:
field TrainingArguments.adam_beta2: float - added:
field TrainingArguments.dataloader_drop_last: bool - added:
field TrainingArguments.dataloader_num_workers: int - added:
field TrainingArguments.debug: bool - added:
field TrainingArguments.disable_tqdm: Optional[bool] - added:
field TrainingArguments.eval_accumulation_steps: Optional[int] - added:
field TrainingArguments.eval_steps: int - added:
field TrainingArguments.evaluation_strategy: EvaluationStrategy - added:
field TrainingArguments.greater_is_better: Optional[bool] - added:
field TrainingArguments.label_names: Optional[List[str]] - added:
field TrainingArguments.load_best_model_at_end: Optional[bool] - added:
field TrainingArguments.metric_for_best_model: Optional[str] - added:
field TrainingArguments.past_index: int - added:
field TrainingArguments.prediction_loss_only: bool - added:
field TrainingArguments.remove_unused_columns: Optional[bool] - added:
field TrainingArguments.run_name: Optional[str] - added:
function default_logdir() -> str - added:
method TrainingArguments.to_dict()
- removed:
field TFTrainingArguments.debug: bool - removed:
field TFTrainingArguments.end_lr: float - removed:
field TFTrainingArguments.eval_steps: int - removed:
field TFTrainingArguments.loss_name: str - removed:
field TFTrainingArguments.mode: str - removed:
field TFTrainingArguments.optimizer_name: str - added:
field TFTrainingArguments.poly_power: float - added:
field TFTrainingArguments.xla: bool - added:
property TFTrainingArguments.eval_batch_size -> int - added:
property TFTrainingArguments.n_replicas -> int - added:
property TFTrainingArguments.train_batch_size -> int
- removed:
variable checkpoint - removed:
variable model
- added:
class FlaxBertModel - added:
class FlaxRobertaModel - added:
constructor FlaxBertModel(*args, **kwargs) - added:
constructor FlaxRobertaModel(*args, **kwargs) - added:
static method FlaxBertModel.from_pretrained(*args, **kwargs) - added:
static method FlaxRobertaModel.from_pretrained(*args, **kwargs)
- added:
class Adafactor - added:
class AdamW - added:
class AdaptiveEmbedding - added:
class AlbertForMaskedLM - added:
class AlbertForMultipleChoice - added:
class AlbertForPreTraining - added:
class AlbertForQuestionAnswering - added:
class AlbertForSequenceClassification - added:
class AlbertForTokenClassification - added:
class AlbertModel - added:
class AlbertPreTrainedModel - added:
class AutoModel - added:
class AutoModelForCausalLM - added:
class AutoModelForMaskedLM - added:
class AutoModelForMultipleChoice - added:
class AutoModelForNextSentencePrediction - added:
class AutoModelForPreTraining - added:
class AutoModelForQuestionAnswering - added:
class AutoModelForSeq2SeqLM - added:
class AutoModelForSequenceClassification - added:
class AutoModelForTokenClassification - added:
class AutoModelWithLMHead - added:
class BartForConditionalGeneration - added:
class BartForQuestionAnswering - added:
class BartForSequenceClassification - added:
class BartModel - added:
class BeamScorer - added:
class BeamSearchScorer - added:
class BertForMaskedLM - added:
class BertForMultipleChoice - added:
class BertForNextSentencePrediction - added:
class BertForPreTraining - added:
class BertForQuestionAnswering - added:
class BertForSequenceClassification - added:
class BertForTokenClassification - added:
class BertGenerationDecoder - added:
class BertGenerationEncoder - added:
class BertLMHeadModel - added:
class BertLayer - added:
class BertModel - added:
class BertPreTrainedModel - added:
class BlenderbotForConditionalGeneration - added:
class CTRLLMHeadModel - added:
class CTRLModel - added:
class CTRLPreTrainedModel - added:
class CamembertForCausalLM - added:
class CamembertForMaskedLM - added:
class CamembertForMultipleChoice - added:
class CamembertForQuestionAnswering - added:
class CamembertForSequenceClassification - added:
class CamembertForTokenClassification - added:
class CamembertModel - added:
class Conv1D - added:
class DPRContextEncoder - added:
class DPRPretrainedContextEncoder - added:
class DPRPretrainedQuestionEncoder - added:
class DPRPretrainedReader - added:
class DPRQuestionEncoder - added:
class DPRReader - added:
class DataCollator - added:
class DataCollatorForLanguageModeling - added:
class DataCollatorForPermutationLanguageModeling - added:
class DataCollatorForSOP - added:
class DataCollatorForTokenClassification - added:
class DataCollatorForWholeWordMask - added:
class DataCollatorWithPadding - added:
class DebertaForSequenceClassification - added:
class DebertaModel - added:
class DebertaPreTrainedModel - added:
class DistilBertForMaskedLM - added:
class DistilBertForMultipleChoice - added:
class DistilBertForQuestionAnswering - added:
class DistilBertForSequenceClassification - added:
class DistilBertForTokenClassification - added:
class DistilBertModel - added:
class DistilBertPreTrainedModel - added:
class ElectraForMaskedLM - added:
class ElectraForMultipleChoice - added:
class ElectraForPreTraining - added:
class ElectraForQuestionAnswering - added:
class ElectraForSequenceClassification - added:
class ElectraForTokenClassification - added:
class ElectraModel - added:
class ElectraPreTrainedModel - added:
class EncoderDecoderModel - added:
class FSMTForConditionalGeneration - added:
class FSMTModel - added:
class FlaubertForMultipleChoice - added:
class FlaubertForQuestionAnswering - added:
class FlaubertForQuestionAnsweringSimple - added:
class FlaubertForSequenceClassification - added:
class FlaubertForTokenClassification - added:
class FlaubertModel - added:
class FlaubertWithLMHeadModel - added:
class FunnelBaseModel - added:
class FunnelForMaskedLM - added:
class FunnelForMultipleChoice - added:
class FunnelForPreTraining - added:
class FunnelForQuestionAnswering - added:
class FunnelForSequenceClassification - added:
class FunnelForTokenClassification - added:
class FunnelModel - added:
class GPT2DoubleHeadsModel - added:
class GPT2ForSequenceClassification - added:
class GPT2LMHeadModel - added:
class GPT2Model - added:
class GPT2PreTrainedModel - added:
class GlueDataTrainingArguments - added:
class GlueDataset - added:
class LayoutLMForMaskedLM - added:
class LayoutLMForTokenClassification - added:
class LayoutLMModel - added:
class LineByLineTextDataset - added:
class LineByLineWithRefDataset - added:
class LineByLineWithSOPTextDataset - added:
class LogitsProcessor - added:
class LogitsProcessorList - added:
class LogitsWarper - added:
class LongformerForMaskedLM - added:
class LongformerForMultipleChoice - added:
class LongformerForQuestionAnswering - added:
class LongformerForSequenceClassification - added:
class LongformerForTokenClassification - added:
class LongformerModel - added:
class LongformerSelfAttention - added:
class LxmertEncoder - added:
class LxmertForPreTraining - added:
class LxmertForQuestionAnswering - added:
class LxmertModel - added:
class LxmertPreTrainedModel - added:
class LxmertVisualFeatureEncoder - added:
class LxmertXLayer - added:
class MBartForConditionalGeneration - added:
class MMBTForClassification - added:
class MMBTModel - added:
class MarianMTModel - added:
class MinLengthLogitsProcessor - added:
class MobileBertForMaskedLM - added:
class MobileBertForMultipleChoice - added:
class MobileBertForNextSentencePrediction - added:
class MobileBertForPreTraining - added:
class MobileBertForQuestionAnswering - added:
class MobileBertForSequenceClassification - added:
class MobileBertForTokenClassification - added:
class MobileBertLayer - added:
class MobileBertModel - added:
class MobileBertPreTrainedModel - added:
class ModalEmbeddings - added:
class NoBadWordsLogitsProcessor - added:
class NoRepeatNGramLogitsProcessor - added:
class OpenAIGPTDoubleHeadsModel - added:
class OpenAIGPTForSequenceClassification - added:
class OpenAIGPTLMHeadModel - added:
class OpenAIGPTModel - added:
class OpenAIGPTPreTrainedModel - added:
class PegasusForConditionalGeneration - added:
class PreTrainedModel - added:
class PretrainedBartModel - added:
class PretrainedFSMTModel - added:
class ProphetNetDecoder - added:
class ProphetNetEncoder - added:
class ProphetNetForCausalLM - added:
class ProphetNetForConditionalGeneration - added:
class ProphetNetModel - added:
class ProphetNetPreTrainedModel - added:
class PyTorchBenchmark - added:
class PyTorchBenchmarkArguments - added:
class RagModel - added:
class RagSequenceForGeneration - added:
class RagTokenForGeneration - added:
class ReformerAttention - added:
class ReformerForMaskedLM - added:
class ReformerForQuestionAnswering - added:
class ReformerForSequenceClassification - added:
class ReformerLayer - added:
class ReformerModel - added:
class ReformerModelWithLMHead - added:
class RepetitionPenaltyLogitsProcessor - added:
class RetriBertModel - added:
class RetriBertPreTrainedModel - added:
class RobertaForCausalLM - added:
class RobertaForMaskedLM - added:
class RobertaForMultipleChoice - added:
class RobertaForQuestionAnswering - added:
class RobertaForSequenceClassification - added:
class RobertaForTokenClassification - added:
class RobertaModel - added:
class SquadDataTrainingArguments - added:
class SquadDataset - added:
class SqueezeBertForMaskedLM - added:
class SqueezeBertForMultipleChoice - added:
class SqueezeBertForQuestionAnswering - added:
class SqueezeBertForSequenceClassification - added:
class SqueezeBertForTokenClassification - added:
class SqueezeBertModel - added:
class SqueezeBertModule - added:
class SqueezeBertPreTrainedModel - added:
class T5ForConditionalGeneration - added:
class T5Model - added:
class T5PreTrainedModel - added:
class TemperatureLogitsWarper - added:
class TextDataset - added:
class TextDatasetForNextSentencePrediction - added:
class TopKLogitsWarper - added:
class TopPLogitsWarper - added:
class Trainer - added:
class TransfoXLLMHeadModel - added:
class TransfoXLModel - added:
class TransfoXLPreTrainedModel - added:
class XLMForMultipleChoice - added:
class XLMForQuestionAnswering - added:
class XLMForQuestionAnsweringSimple - added:
class XLMForSequenceClassification - added:
class XLMForTokenClassification - added:
class XLMModel - added:
class XLMPreTrainedModel - added:
class XLMProphetNetDecoder - added:
class XLMProphetNetEncoder - added:
class XLMProphetNetForCausalLM - added:
class XLMProphetNetForConditionalGeneration - added:
class XLMProphetNetModel - added:
class XLMRobertaForCausalLM - added:
class XLMRobertaForMaskedLM - added:
class XLMRobertaForMultipleChoice - added:
class XLMRobertaForQuestionAnswering - added:
class XLMRobertaForSequenceClassification - added:
class XLMRobertaForTokenClassification - added:
class XLMRobertaModel - added:
class XLMWithLMHeadModel - added:
class XLNetForMultipleChoice - added:
class XLNetForQuestionAnswering - added:
class XLNetForQuestionAnsweringSimple - added:
class XLNetForSequenceClassification - added:
class XLNetForTokenClassification - added:
class XLNetLMHeadModel - added:
class XLNetModel - added:
class XLNetPreTrainedModel - added:
constructor Adafactor(*args, **kwargs) - added:
constructor AdamW(*args, **kwargs) - added:
constructor AdaptiveEmbedding(*args, **kwargs) - added:
constructor AlbertForMaskedLM(*args, **kwargs) - added:
constructor AlbertForMultipleChoice(*args, **kwargs) - added:
constructor AlbertForPreTraining(*args, **kwargs) - added:
constructor AlbertForQuestionAnswering(*args, **kwargs) - added:
constructor AlbertForSequenceClassification(*args, **kwargs) - added:
constructor AlbertForTokenClassification(*args, **kwargs) - added:
constructor AlbertModel(*args, **kwargs) - added:
constructor AlbertPreTrainedModel(*args, **kwargs) - added:
constructor AutoModel(*args, **kwargs) - added:
constructor AutoModelForCausalLM(*args, **kwargs) - added:
constructor AutoModelForMaskedLM(*args, **kwargs) - added:
constructor AutoModelForMultipleChoice(*args, **kwargs) - added:
constructor AutoModelForNextSentencePrediction(*args, **kwargs) - added:
constructor AutoModelForPreTraining(*args, **kwargs) - added:
constructor AutoModelForQuestionAnswering(*args, **kwargs) - added:
constructor AutoModelForSeq2SeqLM(*args, **kwargs) - added:
constructor AutoModelForSequenceClassification(*args, **kwargs) - added:
constructor AutoModelForTokenClassification(*args, **kwargs) - added:
constructor AutoModelWithLMHead(*args, **kwargs) - added:
constructor BartForConditionalGeneration(*args, **kwargs) - added:
constructor BartForQuestionAnswering(*args, **kwargs) - added:
constructor BartForSequenceClassification(*args, **kwargs) - added:
constructor BartModel(*args, **kwargs) - added:
constructor BeamScorer(*args, **kwargs) - added:
constructor BeamSearchScorer(*args, **kwargs) - added:
constructor BertForMaskedLM(*args, **kwargs) - added:
constructor BertForMultipleChoice(*args, **kwargs) - added:
constructor BertForNextSentencePrediction(*args, **kwargs) - added:
constructor BertForPreTraining(*args, **kwargs) - added:
constructor BertForQuestionAnswering(*args, **kwargs) - added:
constructor BertForSequenceClassification(*args, **kwargs) - added:
constructor BertForTokenClassification(*args, **kwargs) - added:
constructor BertGenerationDecoder(*args, **kwargs) - added:
constructor BertGenerationEncoder(*args, **kwargs) - added:
constructor BertLMHeadModel(*args, **kwargs) - added:
constructor BertLayer(*args, **kwargs) - added:
constructor BertModel(*args, **kwargs) - added:
constructor BertPreTrainedModel(*args, **kwargs) - added:
constructor BlenderbotForConditionalGeneration(*args, **kwargs) - added:
constructor CTRLLMHeadModel(*args, **kwargs) - added:
constructor CTRLModel(*args, **kwargs) - added:
constructor CTRLPreTrainedModel(*args, **kwargs) - added:
constructor CamembertForCausalLM(*args, **kwargs) - added:
constructor CamembertForMaskedLM(*args, **kwargs) - added:
constructor CamembertForMultipleChoice(*args, **kwargs) - added:
constructor CamembertForQuestionAnswering(*args, **kwargs) - added:
constructor CamembertForSequenceClassification(*args, **kwargs) - added:
constructor CamembertForTokenClassification(*args, **kwargs) - added:
constructor CamembertModel(*args, **kwargs) - added:
constructor Conv1D(*args, **kwargs) - added:
constructor DPRContextEncoder(*args, **kwargs) - added:
constructor DPRPretrainedContextEncoder(*args, **kwargs) - added:
constructor DPRPretrainedQuestionEncoder(*args, **kwargs) - added:
constructor DPRPretrainedReader(*args, **kwargs) - added:
constructor DPRQuestionEncoder(*args, **kwargs) - added:
constructor DPRReader(*args, **kwargs) - added:
constructor DataCollator(*args, **kwargs) - added:
constructor DataCollatorForLanguageModeling(*args, **kwargs) - added:
constructor DataCollatorForPermutationLanguageModeling(*args, **kwargs) - added:
constructor DataCollatorForSOP(*args, **kwargs) - added:
constructor DataCollatorForTokenClassification(*args, **kwargs) - added:
constructor DataCollatorForWholeWordMask(*args, **kwargs) - added:
constructor DataCollatorWithPadding(*args, **kwargs) - added:
constructor DebertaForSequenceClassification(*args, **kwargs) - added:
constructor DebertaModel(*args, **kwargs) - added:
constructor DebertaPreTrainedModel(*args, **kwargs) - added:
constructor DistilBertForMaskedLM(*args, **kwargs) - added:
constructor DistilBertForMultipleChoice(*args, **kwargs) - added:
constructor DistilBertForQuestionAnswering(*args, **kwargs) - added:
constructor DistilBertForSequenceClassification(*args, **kwargs) - added:
constructor DistilBertForTokenClassification(*args, **kwargs) - added:
constructor DistilBertModel(*args, **kwargs) - added:
constructor DistilBertPreTrainedModel(*args, **kwargs) - added:
constructor ElectraForMaskedLM(*args, **kwargs) - added:
constructor ElectraForMultipleChoice(*args, **kwargs) - added:
constructor ElectraForPreTraining(*args, **kwargs) - added:
constructor ElectraForQuestionAnswering(*args, **kwargs) - added:
constructor ElectraForSequenceClassification(*args, **kwargs) - added:
constructor ElectraForTokenClassification(*args, **kwargs) - added:
constructor ElectraModel(*args, **kwargs) - added:
constructor ElectraPreTrainedModel(*args, **kwargs) - added:
constructor EncoderDecoderModel(*args, **kwargs) - added:
constructor FSMTForConditionalGeneration(*args, **kwargs) - added:
constructor FSMTModel(*args, **kwargs) - added:
constructor FlaubertForMultipleChoice(*args, **kwargs) - added:
constructor FlaubertForQuestionAnswering(*args, **kwargs) - added:
constructor FlaubertForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor FlaubertForSequenceClassification(*args, **kwargs) - added:
constructor FlaubertForTokenClassification(*args, **kwargs) - added:
constructor FlaubertModel(*args, **kwargs) - added:
constructor FlaubertWithLMHeadModel(*args, **kwargs) - added:
constructor FunnelBaseModel(*args, **kwargs) - added:
constructor FunnelForMaskedLM(*args, **kwargs) - added:
constructor FunnelForMultipleChoice(*args, **kwargs) - added:
constructor FunnelForPreTraining(*args, **kwargs) - added:
constructor FunnelForQuestionAnswering(*args, **kwargs) - added:
constructor FunnelForSequenceClassification(*args, **kwargs) - added:
constructor FunnelForTokenClassification(*args, **kwargs) - added:
constructor FunnelModel(*args, **kwargs) - added:
constructor GPT2DoubleHeadsModel(*args, **kwargs) - added:
constructor GPT2ForSequenceClassification(*args, **kwargs) - added:
constructor GPT2LMHeadModel(*args, **kwargs) - added:
constructor GPT2Model(*args, **kwargs) - added:
constructor GPT2PreTrainedModel(*args, **kwargs) - added:
constructor GlueDataTrainingArguments(*args, **kwargs) - added:
constructor GlueDataset(*args, **kwargs) - added:
constructor LayoutLMForMaskedLM(*args, **kwargs) - added:
constructor LayoutLMForTokenClassification(*args, **kwargs) - added:
constructor LayoutLMModel(*args, **kwargs) - added:
constructor LineByLineTextDataset(*args, **kwargs) - added:
constructor LineByLineWithRefDataset(*args, **kwargs) - added:
constructor LineByLineWithSOPTextDataset(*args, **kwargs) - added:
constructor LogitsProcessor(*args, **kwargs) - added:
constructor LogitsProcessorList(*args, **kwargs) - added:
constructor LogitsWarper(*args, **kwargs) - added:
constructor LongformerForMaskedLM(*args, **kwargs) - added:
constructor LongformerForMultipleChoice(*args, **kwargs) - added:
constructor LongformerForQuestionAnswering(*args, **kwargs) - added:
constructor LongformerForSequenceClassification(*args, **kwargs) - added:
constructor LongformerForTokenClassification(*args, **kwargs) - added:
constructor LongformerModel(*args, **kwargs) - added:
constructor LongformerSelfAttention(*args, **kwargs) - added:
constructor LxmertEncoder(*args, **kwargs) - added:
constructor LxmertForPreTraining(*args, **kwargs) - added:
constructor LxmertForQuestionAnswering(*args, **kwargs) - added:
constructor LxmertModel(*args, **kwargs) - added:
constructor LxmertPreTrainedModel(*args, **kwargs) - added:
constructor LxmertVisualFeatureEncoder(*args, **kwargs) - added:
constructor LxmertXLayer(*args, **kwargs) - added:
constructor MBartForConditionalGeneration(*args, **kwargs) - added:
constructor MMBTForClassification(*args, **kwargs) - added:
constructor MMBTModel(*args, **kwargs) - added:
constructor MarianMTModel(*args, **kwargs) - added:
constructor MinLengthLogitsProcessor(*args, **kwargs) - added:
constructor MobileBertForMaskedLM(*args, **kwargs) - added:
constructor MobileBertForMultipleChoice(*args, **kwargs) - added:
constructor MobileBertForNextSentencePrediction(*args, **kwargs) - added:
constructor MobileBertForPreTraining(*args, **kwargs) - added:
constructor MobileBertForQuestionAnswering(*args, **kwargs) - added:
constructor MobileBertForSequenceClassification(*args, **kwargs) - added:
constructor MobileBertForTokenClassification(*args, **kwargs) - added:
constructor MobileBertLayer(*args, **kwargs) - added:
constructor MobileBertModel(*args, **kwargs) - added:
constructor MobileBertPreTrainedModel(*args, **kwargs) - added:
constructor ModalEmbeddings(*args, **kwargs) - added:
constructor NoBadWordsLogitsProcessor(*args, **kwargs) - added:
constructor NoRepeatNGramLogitsProcessor(*args, **kwargs) - added:
constructor OpenAIGPTDoubleHeadsModel(*args, **kwargs) - added:
constructor OpenAIGPTForSequenceClassification(*args, **kwargs) - added:
constructor OpenAIGPTLMHeadModel(*args, **kwargs) - added:
constructor OpenAIGPTModel(*args, **kwargs) - added:
constructor OpenAIGPTPreTrainedModel(*args, **kwargs) - added:
constructor PegasusForConditionalGeneration(*args, **kwargs) - added:
constructor PreTrainedModel(*args, **kwargs) - added:
constructor PretrainedBartModel(*args, **kwargs) - added:
constructor PretrainedFSMTModel(*args, **kwargs) - added:
constructor ProphetNetDecoder(*args, **kwargs) - added:
constructor ProphetNetEncoder(*args, **kwargs) - added:
constructor ProphetNetForCausalLM(*args, **kwargs) - added:
constructor ProphetNetForConditionalGeneration(*args, **kwargs) - added:
constructor ProphetNetModel(*args, **kwargs) - added:
constructor ProphetNetPreTrainedModel(*args, **kwargs) - added:
constructor PyTorchBenchmark(*args, **kwargs) - added:
constructor PyTorchBenchmarkArguments(*args, **kwargs) - added:
constructor RagModel(*args, **kwargs) - added:
constructor RagSequenceForGeneration(*args, **kwargs) - added:
constructor RagTokenForGeneration(*args, **kwargs) - added:
constructor ReformerAttention(*args, **kwargs) - added:
constructor ReformerForMaskedLM(*args, **kwargs) - added:
constructor ReformerForQuestionAnswering(*args, **kwargs) - added:
constructor ReformerForSequenceClassification(*args, **kwargs) - added:
constructor ReformerLayer(*args, **kwargs) - added:
constructor ReformerModel(*args, **kwargs) - added:
constructor ReformerModelWithLMHead(*args, **kwargs) - added:
constructor RepetitionPenaltyLogitsProcessor(*args, **kwargs) - added:
constructor RetriBertModel(*args, **kwargs) - added:
constructor RetriBertPreTrainedModel(*args, **kwargs) - added:
constructor RobertaForCausalLM(*args, **kwargs) - added:
constructor RobertaForMaskedLM(*args, **kwargs) - added:
constructor RobertaForMultipleChoice(*args, **kwargs) - added:
constructor RobertaForQuestionAnswering(*args, **kwargs) - added:
constructor RobertaForSequenceClassification(*args, **kwargs) - added:
constructor RobertaForTokenClassification(*args, **kwargs) - added:
constructor RobertaModel(*args, **kwargs) - added:
constructor SquadDataTrainingArguments(*args, **kwargs) - added:
constructor SquadDataset(*args, **kwargs) - added:
constructor SqueezeBertForMaskedLM(*args, **kwargs) - added:
constructor SqueezeBertForMultipleChoice(*args, **kwargs) - added:
constructor SqueezeBertForQuestionAnswering(*args, **kwargs) - added:
constructor SqueezeBertForSequenceClassification(*args, **kwargs) - added:
constructor SqueezeBertForTokenClassification(*args, **kwargs) - added:
constructor SqueezeBertModel(*args, **kwargs) - added:
constructor SqueezeBertModule(*args, **kwargs) - added:
constructor SqueezeBertPreTrainedModel(*args, **kwargs) - added:
constructor T5ForConditionalGeneration(*args, **kwargs) - added:
constructor T5Model(*args, **kwargs) - added:
constructor T5PreTrainedModel(*args, **kwargs) - added:
constructor TemperatureLogitsWarper(*args, **kwargs) - added:
constructor TextDataset(*args, **kwargs) - added:
constructor TextDatasetForNextSentencePrediction(*args, **kwargs) - added:
constructor TopKLogitsWarper(*args, **kwargs) - added:
constructor TopPLogitsWarper(*args, **kwargs) - added:
constructor Trainer(*args, **kwargs) - added:
constructor TransfoXLLMHeadModel(*args, **kwargs) - added:
constructor TransfoXLModel(*args, **kwargs) - added:
constructor TransfoXLPreTrainedModel(*args, **kwargs) - added:
constructor XLMForMultipleChoice(*args, **kwargs) - added:
constructor XLMForQuestionAnswering(*args, **kwargs) - added:
constructor XLMForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor XLMForSequenceClassification(*args, **kwargs) - added:
constructor XLMForTokenClassification(*args, **kwargs) - added:
constructor XLMModel(*args, **kwargs) - added:
constructor XLMPreTrainedModel(*args, **kwargs) - added:
constructor XLMProphetNetDecoder(*args, **kwargs) - added:
constructor XLMProphetNetEncoder(*args, **kwargs) - added:
constructor XLMProphetNetForCausalLM(*args, **kwargs) - added:
constructor XLMProphetNetForConditionalGeneration(*args, **kwargs) - added:
constructor XLMProphetNetModel(*args, **kwargs) - added:
constructor XLMRobertaForCausalLM(*args, **kwargs) - added:
constructor XLMRobertaForMaskedLM(*args, **kwargs) - added:
constructor XLMRobertaForMultipleChoice(*args, **kwargs) - added:
constructor XLMRobertaForQuestionAnswering(*args, **kwargs) - added:
constructor XLMRobertaForSequenceClassification(*args, **kwargs) - added:
constructor XLMRobertaForTokenClassification(*args, **kwargs) - added:
constructor XLMRobertaModel(*args, **kwargs) - added:
constructor XLMWithLMHeadModel(*args, **kwargs) - added:
constructor XLNetForMultipleChoice(*args, **kwargs) - added:
constructor XLNetForQuestionAnswering(*args, **kwargs) - added:
constructor XLNetForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor XLNetForSequenceClassification(*args, **kwargs) - added:
constructor XLNetForTokenClassification(*args, **kwargs) - added:
constructor XLNetLMHeadModel(*args, **kwargs) - added:
constructor XLNetModel(*args, **kwargs) - added:
constructor XLNetPreTrainedModel(*args, **kwargs) - added:
function apply_chunking_to_forward(*args, **kwargs) - added:
function default_data_collator(*args, **kwargs) - added:
function get_constant_schedule(*args, **kwargs) - added:
function get_constant_schedule_with_warmup(*args, **kwargs) - added:
function get_cosine_schedule_with_warmup(*args, **kwargs) - added:
function get_cosine_with_hard_restarts_schedule_with_warmup(*args, **kwargs) - added:
function get_linear_schedule_with_warmup(*args, **kwargs) - added:
function get_polynomial_decay_schedule_with_warmup(*args, **kwargs) - added:
function load_tf_weights_in_albert(*args, **kwargs) - added:
function load_tf_weights_in_bert(*args, **kwargs) - added:
function load_tf_weights_in_bert_generation(*args, **kwargs) - added:
function load_tf_weights_in_electra(*args, **kwargs) - added:
function load_tf_weights_in_funnel(*args, **kwargs) - added:
function load_tf_weights_in_gpt2(*args, **kwargs) - added:
function load_tf_weights_in_mobilebert(*args, **kwargs) - added:
function load_tf_weights_in_openai_gpt(*args, **kwargs) - added:
function load_tf_weights_in_t5(*args, **kwargs) - added:
function load_tf_weights_in_transfo_xl(*args, **kwargs) - added:
function load_tf_weights_in_xlnet(*args, **kwargs) - added:
function prune_layer(*args, **kwargs) - added:
function top_k_top_p_filtering(*args, **kwargs) - added:
function torch_distributed_zero_first(*args, **kwargs) - added:
static method AlbertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method AlbertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method AlbertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method AlbertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method AlbertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method AlbertModel.from_pretrained(*args, **kwargs) - added:
static method AlbertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method AutoModel.from_pretrained(*args, **kwargs) - added:
static method AutoModelForCausalLM.from_pretrained(*args, **kwargs) - added:
static method AutoModelForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method AutoModelForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method AutoModelForNextSentencePrediction.from_pretrained(*args, **kwargs) - added:
static method AutoModelForPreTraining.from_pretrained(*args, **kwargs) - added:
static method AutoModelForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method AutoModelForSeq2SeqLM.from_pretrained(*args, **kwargs) - added:
static method AutoModelForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method AutoModelForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method AutoModelWithLMHead.from_pretrained(*args, **kwargs) - added:
static method BartForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method BartForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method BartForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method BartModel.from_pretrained(*args, **kwargs) - added:
static method BertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method BertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method BertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method BertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method BertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method BertLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method BertModel.from_pretrained(*args, **kwargs) - added:
static method BertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method BlenderbotForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method CTRLLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method CTRLModel.from_pretrained(*args, **kwargs) - added:
static method CTRLPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method CamembertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method CamembertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method CamembertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method CamembertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method CamembertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method CamembertModel.from_pretrained(*args, **kwargs) - added:
static method DataCollatorForLanguageModeling.from_pretrained(*args, **kwargs) - added:
static method DataCollatorForPermutationLanguageModeling.from_pretrained(*args, **kwargs) - added:
static method DataCollatorForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method DebertaForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method DebertaModel.from_pretrained(*args, **kwargs) - added:
static method DebertaPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method DistilBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method DistilBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method DistilBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method DistilBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method DistilBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method DistilBertModel.from_pretrained(*args, **kwargs) - added:
static method DistilBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method ElectraForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method ElectraForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method ElectraForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method ElectraForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method ElectraForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method ElectraModel.from_pretrained(*args, **kwargs) - added:
static method ElectraPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method EncoderDecoderModel.from_pretrained(*args, **kwargs) - added:
static method FSMTForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method FSMTModel.from_pretrained(*args, **kwargs) - added:
static method FlaubertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method FlaubertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method FlaubertForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method FlaubertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method FlaubertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method FlaubertModel.from_pretrained(*args, **kwargs) - added:
static method FlaubertWithLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method FunnelBaseModel.from_pretrained(*args, **kwargs) - added:
static method FunnelForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method FunnelForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method FunnelForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method FunnelForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method FunnelForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method FunnelModel.from_pretrained(*args, **kwargs) - added:
static method GPT2DoubleHeadsModel.from_pretrained(*args, **kwargs) - added:
static method GPT2ForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method GPT2LMHeadModel.from_pretrained(*args, **kwargs) - added:
static method GPT2Model.from_pretrained(*args, **kwargs) - added:
static method GPT2PreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method LayoutLMForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method LayoutLMForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method LayoutLMModel.from_pretrained(*args, **kwargs) - added:
static method LongformerForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method LongformerForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method LongformerForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method LongformerForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method LongformerForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method LongformerModel.from_pretrained(*args, **kwargs) - added:
static method LxmertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method LxmertModel.from_pretrained(*args, **kwargs) - added:
static method LxmertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method MBartForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method MMBTModel.from_pretrained(*args, **kwargs) - added:
static method MarianMTModel.from_pretrained(*args, **kwargs) - added:
static method MobileBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method MobileBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method MobileBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method MobileBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method MobileBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method MobileBertModel.from_pretrained(*args, **kwargs) - added:
static method MobileBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTDoubleHeadsModel.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTModel.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method PegasusForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method PreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method PretrainedBartModel.from_pretrained(*args, **kwargs) - added:
static method PretrainedFSMTModel.from_pretrained(*args, **kwargs) - added:
static method ProphetNetForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method ProphetNetModel.from_pretrained(*args, **kwargs) - added:
static method ProphetNetPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method RagModel.from_pretrained(*args, **kwargs) - added:
static method ReformerForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method ReformerForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method ReformerForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method ReformerModel.from_pretrained(*args, **kwargs) - added:
static method ReformerModelWithLMHead.from_pretrained(*args, **kwargs) - added:
static method RetriBertModel.from_pretrained(*args, **kwargs) - added:
static method RetriBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method RobertaForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method RobertaForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method RobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method RobertaForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method RobertaForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method RobertaModel.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertModel.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method T5ForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method T5Model.from_pretrained(*args, **kwargs) - added:
static method T5PreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TransfoXLLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TransfoXLModel.from_pretrained(*args, **kwargs) - added:
static method TransfoXLPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method XLMForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method XLMForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method XLMForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method XLMForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method XLMForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method XLMModel.from_pretrained(*args, **kwargs) - added:
static method XLMPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method XLMProphetNetForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method XLMProphetNetModel.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaModel.from_pretrained(*args, **kwargs) - added:
static method XLMWithLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method XLNetForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method XLNetForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method XLNetForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method XLNetForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method XLNetForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method XLNetLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method XLNetModel.from_pretrained(*args, **kwargs) - added:
static method XLNetPreTrainedModel.from_pretrained(*args, **kwargs) - added:
variable ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable BART_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable BERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable BLENDERBOT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DEBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable LAYOUTLM_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable MODEL_FOR_MASKED_LM_MAPPING - added:
variable MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
variable MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING - added:
variable MODEL_FOR_PRETRAINING_MAPPING - added:
variable MODEL_FOR_QUESTION_ANSWERING_MAPPING - added:
variable MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING - added:
variable MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING - added:
variable MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING - added:
variable MODEL_MAPPING - added:
variable MODEL_WITH_LM_HEAD_MAPPING - added:
variable OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable REFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable RETRIBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable SQUEEZEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable T5_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable XLM_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable XLM_PROPHETNET_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable XLNET_PRETRAINED_MODEL_ARCHIVE_LIST
- added:
class AlbertTokenizer - added:
class BertGenerationTokenizer - added:
class CamembertTokenizer - added:
class MBartTokenizer - added:
class MarianTokenizer - added:
class PegasusTokenizer - added:
class ReformerTokenizer - added:
class T5Tokenizer - added:
class XLMProphetNetTokenizer - added:
class XLMRobertaTokenizer - added:
class XLNetTokenizer - added:
constructor AlbertTokenizer(*args, **kwargs) - added:
constructor BertGenerationTokenizer(*args, **kwargs) - added:
constructor CamembertTokenizer(*args, **kwargs) - added:
constructor MBartTokenizer(*args, **kwargs) - added:
constructor MarianTokenizer(*args, **kwargs) - added:
constructor PegasusTokenizer(*args, **kwargs) - added:
constructor ReformerTokenizer(*args, **kwargs) - added:
constructor T5Tokenizer(*args, **kwargs) - added:
constructor XLMProphetNetTokenizer(*args, **kwargs) - added:
constructor XLMRobertaTokenizer(*args, **kwargs) - added:
constructor XLNetTokenizer(*args, **kwargs) - added:
static method AlbertTokenizer.from_pretrained(*args, **kwargs) - added:
static method BertGenerationTokenizer.from_pretrained(*args, **kwargs) - added:
static method CamembertTokenizer.from_pretrained(*args, **kwargs) - added:
static method MBartTokenizer.from_pretrained(*args, **kwargs) - added:
static method MarianTokenizer.from_pretrained(*args, **kwargs) - added:
static method PegasusTokenizer.from_pretrained(*args, **kwargs) - added:
static method ReformerTokenizer.from_pretrained(*args, **kwargs) - added:
static method T5Tokenizer.from_pretrained(*args, **kwargs) - added:
static method XLMProphetNetTokenizer.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaTokenizer.from_pretrained(*args, **kwargs) - added:
static method XLNetTokenizer.from_pretrained(*args, **kwargs)
- added:
class AdamWeightDecay - added:
class GradientAccumulator - added:
class TFAdaptiveEmbedding - added:
class TFAlbertForMaskedLM - added:
class TFAlbertForMultipleChoice - added:
class TFAlbertForPreTraining - added:
class TFAlbertForQuestionAnswering - added:
class TFAlbertForSequenceClassification - added:
class TFAlbertForTokenClassification - added:
class TFAlbertMainLayer - added:
class TFAlbertModel - added:
class TFAlbertPreTrainedModel - added:
class TFAutoModel - added:
class TFAutoModelForCausalLM - added:
class TFAutoModelForMaskedLM - added:
class TFAutoModelForMultipleChoice - added:
class TFAutoModelForPreTraining - added:
class TFAutoModelForQuestionAnswering - added:
class TFAutoModelForSeq2SeqLM - added:
class TFAutoModelForSequenceClassification - added:
class TFAutoModelForTokenClassification - added:
class TFAutoModelWithLMHead - added:
class TFBartForConditionalGeneration - added:
class TFBartModel - added:
class TFBertEmbeddings - added:
class TFBertForMaskedLM - added:
class TFBertForMultipleChoice - added:
class TFBertForNextSentencePrediction - added:
class TFBertForPreTraining - added:
class TFBertForQuestionAnswering - added:
class TFBertForSequenceClassification - added:
class TFBertForTokenClassification - added:
class TFBertLMHeadModel - added:
class TFBertMainLayer - added:
class TFBertModel - added:
class TFBertPreTrainedModel - added:
class TFBlenderbotForConditionalGeneration - added:
class TFCTRLLMHeadModel - added:
class TFCTRLModel - added:
class TFCTRLPreTrainedModel - added:
class TFCamembertForMaskedLM - added:
class TFCamembertForMultipleChoice - added:
class TFCamembertForQuestionAnswering - added:
class TFCamembertForSequenceClassification - added:
class TFCamembertForTokenClassification - added:
class TFCamembertModel - added:
class TFDistilBertForMaskedLM - added:
class TFDistilBertForMultipleChoice - added:
class TFDistilBertForQuestionAnswering - added:
class TFDistilBertForSequenceClassification - added:
class TFDistilBertForTokenClassification - added:
class TFDistilBertMainLayer - added:
class TFDistilBertModel - added:
class TFDistilBertPreTrainedModel - added:
class TFElectraForMaskedLM - added:
class TFElectraForMultipleChoice - added:
class TFElectraForPreTraining - added:
class TFElectraForQuestionAnswering - added:
class TFElectraForSequenceClassification - added:
class TFElectraForTokenClassification - added:
class TFElectraModel - added:
class TFElectraPreTrainedModel - added:
class TFFlaubertForMultipleChoice - added:
class TFFlaubertForQuestionAnsweringSimple - added:
class TFFlaubertForSequenceClassification - added:
class TFFlaubertForTokenClassification - added:
class TFFlaubertModel - added:
class TFFlaubertWithLMHeadModel - added:
class TFFunnelBaseModel - added:
class TFFunnelForMaskedLM - added:
class TFFunnelForMultipleChoice - added:
class TFFunnelForPreTraining - added:
class TFFunnelForQuestionAnswering - added:
class TFFunnelForSequenceClassification - added:
class TFFunnelForTokenClassification - added:
class TFFunnelModel - added:
class TFGPT2DoubleHeadsModel - added:
class TFGPT2LMHeadModel - added:
class TFGPT2MainLayer - added:
class TFGPT2Model - added:
class TFGPT2PreTrainedModel - added:
class TFLongformerForMaskedLM - added:
class TFLongformerForQuestionAnswering - added:
class TFLongformerModel - added:
class TFLongformerSelfAttention - added:
class TFLxmertForPreTraining - added:
class TFLxmertMainLayer - added:
class TFLxmertModel - added:
class TFLxmertPreTrainedModel - added:
class TFLxmertVisualFeatureEncoder - added:
class TFMBartForConditionalGeneration - added:
class TFMarianMTModel - added:
class TFMobileBertForMaskedLM - added:
class TFMobileBertForMultipleChoice - added:
class TFMobileBertForNextSentencePrediction - added:
class TFMobileBertForPreTraining - added:
class TFMobileBertForQuestionAnswering - added:
class TFMobileBertForSequenceClassification - added:
class TFMobileBertForTokenClassification - added:
class TFMobileBertMainLayer - added:
class TFMobileBertModel - added:
class TFMobileBertPreTrainedModel - added:
class TFOpenAIGPTDoubleHeadsModel - added:
class TFOpenAIGPTLMHeadModel - added:
class TFOpenAIGPTMainLayer - added:
class TFOpenAIGPTModel - added:
class TFOpenAIGPTPreTrainedModel - added:
class TFPegasusForConditionalGeneration - added:
class TFPreTrainedModel - added:
class TFRobertaForMaskedLM - added:
class TFRobertaForMultipleChoice - added:
class TFRobertaForQuestionAnswering - added:
class TFRobertaForSequenceClassification - added:
class TFRobertaForTokenClassification - added:
class TFRobertaMainLayer - added:
class TFRobertaModel - added:
class TFRobertaPreTrainedModel - added:
class TFSequenceSummary - added:
class TFSharedEmbeddings - added:
class TFT5ForConditionalGeneration - added:
class TFT5Model - added:
class TFT5PreTrainedModel - added:
class TFTrainer - added:
class TFTransfoXLLMHeadModel - added:
class TFTransfoXLMainLayer - added:
class TFTransfoXLModel - added:
class TFTransfoXLPreTrainedModel - added:
class TFXLMForMultipleChoice - added:
class TFXLMForQuestionAnsweringSimple - added:
class TFXLMForSequenceClassification - added:
class TFXLMForTokenClassification - added:
class TFXLMMainLayer - added:
class TFXLMModel - added:
class TFXLMPreTrainedModel - added:
class TFXLMRobertaForMaskedLM - added:
class TFXLMRobertaForMultipleChoice - added:
class TFXLMRobertaForQuestionAnswering - added:
class TFXLMRobertaForSequenceClassification - added:
class TFXLMRobertaForTokenClassification - added:
class TFXLMRobertaModel - added:
class TFXLMWithLMHeadModel - added:
class TFXLNetForMultipleChoice - added:
class TFXLNetForQuestionAnsweringSimple - added:
class TFXLNetForSequenceClassification - added:
class TFXLNetForTokenClassification - added:
class TFXLNetLMHeadModel - added:
class TFXLNetMainLayer - added:
class TFXLNetModel - added:
class TFXLNetPreTrainedModel - added:
class TensorFlowBenchmark - added:
class TensorFlowBenchmarkArguments - added:
class WarmUp - added:
constructor AdamWeightDecay(*args, **kwargs) - added:
constructor GradientAccumulator(*args, **kwargs) - added:
constructor TFAdaptiveEmbedding(*args, **kwargs) - added:
constructor TFAlbertForMaskedLM(*args, **kwargs) - added:
constructor TFAlbertForMultipleChoice(*args, **kwargs) - added:
constructor TFAlbertForPreTraining(*args, **kwargs) - added:
constructor TFAlbertForQuestionAnswering(*args, **kwargs) - added:
constructor TFAlbertForSequenceClassification(*args, **kwargs) - added:
constructor TFAlbertForTokenClassification(*args, **kwargs) - added:
constructor TFAlbertMainLayer(*args, **kwargs) - added:
constructor TFAlbertModel(*args, **kwargs) - added:
constructor TFAlbertPreTrainedModel(*args, **kwargs) - added:
constructor TFAutoModel(*args, **kwargs) - added:
constructor TFAutoModelForCausalLM(*args, **kwargs) - added:
constructor TFAutoModelForMaskedLM(*args, **kwargs) - added:
constructor TFAutoModelForMultipleChoice(*args, **kwargs) - added:
constructor TFAutoModelForPreTraining(*args, **kwargs) - added:
constructor TFAutoModelForQuestionAnswering(*args, **kwargs) - added:
constructor TFAutoModelForSeq2SeqLM(*args, **kwargs) - added:
constructor TFAutoModelForSequenceClassification(*args, **kwargs) - added:
constructor TFAutoModelForTokenClassification(*args, **kwargs) - added:
constructor TFAutoModelWithLMHead(*args, **kwargs) - added:
constructor TFBartForConditionalGeneration(*args, **kwargs) - added:
constructor TFBartModel(*args, **kwargs) - added:
constructor TFBertEmbeddings(*args, **kwargs) - added:
constructor TFBertForMaskedLM(*args, **kwargs) - added:
constructor TFBertForMultipleChoice(*args, **kwargs) - added:
constructor TFBertForNextSentencePrediction(*args, **kwargs) - added:
constructor TFBertForPreTraining(*args, **kwargs) - added:
constructor TFBertForQuestionAnswering(*args, **kwargs) - added:
constructor TFBertForSequenceClassification(*args, **kwargs) - added:
constructor TFBertForTokenClassification(*args, **kwargs) - added:
constructor TFBertLMHeadModel(*args, **kwargs) - added:
constructor TFBertMainLayer(*args, **kwargs) - added:
constructor TFBertModel(*args, **kwargs) - added:
constructor TFBertPreTrainedModel(*args, **kwargs) - added:
constructor TFBlenderbotForConditionalGeneration(*args, **kwargs) - added:
constructor TFCTRLLMHeadModel(*args, **kwargs) - added:
constructor TFCTRLModel(*args, **kwargs) - added:
constructor TFCTRLPreTrainedModel(*args, **kwargs) - added:
constructor TFCamembertForMaskedLM(*args, **kwargs) - added:
constructor TFCamembertForMultipleChoice(*args, **kwargs) - added:
constructor TFCamembertForQuestionAnswering(*args, **kwargs) - added:
constructor TFCamembertForSequenceClassification(*args, **kwargs) - added:
constructor TFCamembertForTokenClassification(*args, **kwargs) - added:
constructor TFCamembertModel(*args, **kwargs) - added:
constructor TFDistilBertForMaskedLM(*args, **kwargs) - added:
constructor TFDistilBertForMultipleChoice(*args, **kwargs) - added:
constructor TFDistilBertForQuestionAnswering(*args, **kwargs) - added:
constructor TFDistilBertForSequenceClassification(*args, **kwargs) - added:
constructor TFDistilBertForTokenClassification(*args, **kwargs) - added:
constructor TFDistilBertMainLayer(*args, **kwargs) - added:
constructor TFDistilBertModel(*args, **kwargs) - added:
constructor TFDistilBertPreTrainedModel(*args, **kwargs) - added:
constructor TFElectraForMaskedLM(*args, **kwargs) - added:
constructor TFElectraForMultipleChoice(*args, **kwargs) - added:
constructor TFElectraForPreTraining(*args, **kwargs) - added:
constructor TFElectraForQuestionAnswering(*args, **kwargs) - added:
constructor TFElectraForSequenceClassification(*args, **kwargs) - added:
constructor TFElectraForTokenClassification(*args, **kwargs) - added:
constructor TFElectraModel(*args, **kwargs) - added:
constructor TFElectraPreTrainedModel(*args, **kwargs) - added:
constructor TFFlaubertForMultipleChoice(*args, **kwargs) - added:
constructor TFFlaubertForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor TFFlaubertForSequenceClassification(*args, **kwargs) - added:
constructor TFFlaubertForTokenClassification(*args, **kwargs) - added:
constructor TFFlaubertModel(*args, **kwargs) - added:
constructor TFFlaubertWithLMHeadModel(*args, **kwargs) - added:
constructor TFFunnelBaseModel(*args, **kwargs) - added:
constructor TFFunnelForMaskedLM(*args, **kwargs) - added:
constructor TFFunnelForMultipleChoice(*args, **kwargs) - added:
constructor TFFunnelForPreTraining(*args, **kwargs) - added:
constructor TFFunnelForQuestionAnswering(*args, **kwargs) - added:
constructor TFFunnelForSequenceClassification(*args, **kwargs) - added:
constructor TFFunnelForTokenClassification(*args, **kwargs) - added:
constructor TFFunnelModel(*args, **kwargs) - added:
constructor TFGPT2DoubleHeadsModel(*args, **kwargs) - added:
constructor TFGPT2LMHeadModel(*args, **kwargs) - added:
constructor TFGPT2MainLayer(*args, **kwargs) - added:
constructor TFGPT2Model(*args, **kwargs) - added:
constructor TFGPT2PreTrainedModel(*args, **kwargs) - added:
constructor TFLongformerForMaskedLM(*args, **kwargs) - added:
constructor TFLongformerForQuestionAnswering(*args, **kwargs) - added:
constructor TFLongformerModel(*args, **kwargs) - added:
constructor TFLongformerSelfAttention(*args, **kwargs) - added:
constructor TFLxmertForPreTraining(*args, **kwargs) - added:
constructor TFLxmertMainLayer(*args, **kwargs) - added:
constructor TFLxmertModel(*args, **kwargs) - added:
constructor TFLxmertPreTrainedModel(*args, **kwargs) - added:
constructor TFLxmertVisualFeatureEncoder(*args, **kwargs) - added:
constructor TFMBartForConditionalGeneration(*args, **kwargs) - added:
constructor TFMarianMTModel(*args, **kwargs) - added:
constructor TFMobileBertForMaskedLM(*args, **kwargs) - added:
constructor TFMobileBertForMultipleChoice(*args, **kwargs) - added:
constructor TFMobileBertForNextSentencePrediction(*args, **kwargs) - added:
constructor TFMobileBertForPreTraining(*args, **kwargs) - added:
constructor TFMobileBertForQuestionAnswering(*args, **kwargs) - added:
constructor TFMobileBertForSequenceClassification(*args, **kwargs) - added:
constructor TFMobileBertForTokenClassification(*args, **kwargs) - added:
constructor TFMobileBertMainLayer(*args, **kwargs) - added:
constructor TFMobileBertModel(*args, **kwargs) - added:
constructor TFMobileBertPreTrainedModel(*args, **kwargs) - added:
constructor TFOpenAIGPTDoubleHeadsModel(*args, **kwargs) - added:
constructor TFOpenAIGPTLMHeadModel(*args, **kwargs) - added:
constructor TFOpenAIGPTMainLayer(*args, **kwargs) - added:
constructor TFOpenAIGPTModel(*args, **kwargs) - added:
constructor TFOpenAIGPTPreTrainedModel(*args, **kwargs) - added:
constructor TFPegasusForConditionalGeneration(*args, **kwargs) - added:
constructor TFPreTrainedModel(*args, **kwargs) - added:
constructor TFRobertaForMaskedLM(*args, **kwargs) - added:
constructor TFRobertaForMultipleChoice(*args, **kwargs) - added:
constructor TFRobertaForQuestionAnswering(*args, **kwargs) - added:
constructor TFRobertaForSequenceClassification(*args, **kwargs) - added:
constructor TFRobertaForTokenClassification(*args, **kwargs) - added:
constructor TFRobertaMainLayer(*args, **kwargs) - added:
constructor TFRobertaModel(*args, **kwargs) - added:
constructor TFRobertaPreTrainedModel(*args, **kwargs) - added:
constructor TFSequenceSummary(*args, **kwargs) - added:
constructor TFSharedEmbeddings(*args, **kwargs) - added:
constructor TFT5ForConditionalGeneration(*args, **kwargs) - added:
constructor TFT5Model(*args, **kwargs) - added:
constructor TFT5PreTrainedModel(*args, **kwargs) - added:
constructor TFTrainer(*args, **kwargs) - added:
constructor TFTransfoXLLMHeadModel(*args, **kwargs) - added:
constructor TFTransfoXLMainLayer(*args, **kwargs) - added:
constructor TFTransfoXLModel(*args, **kwargs) - added:
constructor TFTransfoXLPreTrainedModel(*args, **kwargs) - added:
constructor TFXLMForMultipleChoice(*args, **kwargs) - added:
constructor TFXLMForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor TFXLMForSequenceClassification(*args, **kwargs) - added:
constructor TFXLMForTokenClassification(*args, **kwargs) - added:
constructor TFXLMMainLayer(*args, **kwargs) - added:
constructor TFXLMModel(*args, **kwargs) - added:
constructor TFXLMPreTrainedModel(*args, **kwargs) - added:
constructor TFXLMRobertaForMaskedLM(*args, **kwargs) - added:
constructor TFXLMRobertaForMultipleChoice(*args, **kwargs) - added:
constructor TFXLMRobertaForQuestionAnswering(*args, **kwargs) - added:
constructor TFXLMRobertaForSequenceClassification(*args, **kwargs) - added:
constructor TFXLMRobertaForTokenClassification(*args, **kwargs) - added:
constructor TFXLMRobertaModel(*args, **kwargs) - added:
constructor TFXLMWithLMHeadModel(*args, **kwargs) - added:
constructor TFXLNetForMultipleChoice(*args, **kwargs) - added:
constructor TFXLNetForQuestionAnsweringSimple(*args, **kwargs) - added:
constructor TFXLNetForSequenceClassification(*args, **kwargs) - added:
constructor TFXLNetForTokenClassification(*args, **kwargs) - added:
constructor TFXLNetLMHeadModel(*args, **kwargs) - added:
constructor TFXLNetMainLayer(*args, **kwargs) - added:
constructor TFXLNetModel(*args, **kwargs) - added:
constructor TFXLNetPreTrainedModel(*args, **kwargs) - added:
constructor TensorFlowBenchmark(*args, **kwargs) - added:
constructor TensorFlowBenchmarkArguments(*args, **kwargs) - added:
constructor WarmUp(*args, **kwargs) - added:
function create_optimizer(*args, **kwargs) - added:
function shape_list(*args, **kwargs) - added:
function tf_top_k_top_p_filtering(*args, **kwargs) - added:
static method TFAlbertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFAlbertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFAlbertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFAlbertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFAlbertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFAlbertModel.from_pretrained(*args, **kwargs) - added:
static method TFAlbertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFAutoModel.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForCausalLM.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForPreTraining.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForSeq2SeqLM.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFAutoModelWithLMHead.from_pretrained(*args, **kwargs) - added:
static method TFBartForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method TFBartModel.from_pretrained(*args, **kwargs) - added:
static method TFBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFBertLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFBertModel.from_pretrained(*args, **kwargs) - added:
static method TFBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFBlenderbotForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method TFCTRLLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFCTRLModel.from_pretrained(*args, **kwargs) - added:
static method TFCTRLPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFCamembertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFCamembertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFCamembertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFCamembertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFCamembertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFCamembertModel.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertModel.from_pretrained(*args, **kwargs) - added:
static method TFDistilBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFElectraForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFElectraForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFElectraForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFElectraForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFElectraForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFElectraModel.from_pretrained(*args, **kwargs) - added:
static method TFElectraPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertModel.from_pretrained(*args, **kwargs) - added:
static method TFFlaubertWithLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFFunnelBaseModel.from_pretrained(*args, **kwargs) - added:
static method TFFunnelForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFFunnelForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFFunnelForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFFunnelForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFFunnelForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFFunnelModel.from_pretrained(*args, **kwargs) - added:
static method TFGPT2DoubleHeadsModel.from_pretrained(*args, **kwargs) - added:
static method TFGPT2LMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFGPT2Model.from_pretrained(*args, **kwargs) - added:
static method TFGPT2PreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFLongformerForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFLongformerForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFLongformerModel.from_pretrained(*args, **kwargs) - added:
static method TFLxmertModel.from_pretrained(*args, **kwargs) - added:
static method TFLxmertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFMBartForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method TFMarianMTModel.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertModel.from_pretrained(*args, **kwargs) - added:
static method TFMobileBertPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFOpenAIGPTDoubleHeadsModel.from_pretrained(*args, **kwargs) - added:
static method TFOpenAIGPTLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFOpenAIGPTModel.from_pretrained(*args, **kwargs) - added:
static method TFOpenAIGPTPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFPegasusForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method TFPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFRobertaForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFRobertaForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFRobertaModel.from_pretrained(*args, **kwargs) - added:
static method TFRobertaPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFT5ForConditionalGeneration.from_pretrained(*args, **kwargs) - added:
static method TFT5Model.from_pretrained(*args, **kwargs) - added:
static method TFT5PreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFTransfoXLLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFTransfoXLModel.from_pretrained(*args, **kwargs) - added:
static method TFTransfoXLPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFXLMForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFXLMForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method TFXLMForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLMForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLMModel.from_pretrained(*args, **kwargs) - added:
static method TFXLMPreTrainedModel.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaForMaskedLM.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaForQuestionAnswering.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLMRobertaModel.from_pretrained(*args, **kwargs) - added:
static method TFXLMWithLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFXLNetForMultipleChoice.from_pretrained(*args, **kwargs) - added:
static method TFXLNetForQuestionAnsweringSimple.from_pretrained(*args, **kwargs) - added:
static method TFXLNetForSequenceClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLNetForTokenClassification.from_pretrained(*args, **kwargs) - added:
static method TFXLNetLMHeadModel.from_pretrained(*args, **kwargs) - added:
static method TFXLNetModel.from_pretrained(*args, **kwargs) - added:
static method TFXLNetPreTrainedModel.from_pretrained(*args, **kwargs) - added:
variable TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_BERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_CAMEMBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_CTRL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_DISTILBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_ELECTRA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_FLAUBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_FUNNEL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_GPT2_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_LONGFORMER_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_LXMERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_MOBILEBERT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_MODEL_FOR_CAUSAL_LM_MAPPING - added:
variable TF_MODEL_FOR_MASKED_LM_MAPPING - added:
variable TF_MODEL_FOR_MULTIPLE_CHOICE_MAPPING - added:
variable TF_MODEL_FOR_PRETRAINING_MAPPING - added:
variable TF_MODEL_FOR_QUESTION_ANSWERING_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_TOKEN_CLASSIFICATION_MAPPING - added:
variable TF_MODEL_MAPPING - added:
variable TF_MODEL_WITH_LM_HEAD_MAPPING - added:
variable TF_OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_T5_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_TRANSFO_XL_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_XLM_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_XLM_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST - added:
variable TF_XLNET_PRETRAINED_MODEL_ARCHIVE_LIST
- added:
class AlbertTokenizerFast - added:
class BartTokenizerFast - added:
class BertTokenizerFast - added:
class CamembertTokenizerFast - added:
class DPRContextEncoderTokenizerFast - added:
class DPRQuestionEncoderTokenizerFast - added:
class DPRReaderTokenizerFast - added:
class DistilBertTokenizerFast - added:
class ElectraTokenizerFast - added:
class FunnelTokenizerFast - added:
class GPT2TokenizerFast - added:
class HerbertTokenizerFast - added:
class LayoutLMTokenizerFast - added:
class LongformerTokenizerFast - added:
class LxmertTokenizerFast - added:
class MBartTokenizerFast - added:
class MobileBertTokenizerFast - added:
class OpenAIGPTTokenizerFast - added:
class PegasusTokenizerFast - added:
class PreTrainedTokenizerFast - added:
class ReformerTokenizerFast - added:
class RetriBertTokenizerFast - added:
class RobertaTokenizerFast - added:
class SqueezeBertTokenizerFast - added:
class T5TokenizerFast - added:
class XLMRobertaTokenizerFast - added:
class XLNetTokenizerFast - added:
constructor AlbertTokenizerFast(*args, **kwargs) - added:
constructor BartTokenizerFast(*args, **kwargs) - added:
constructor BertTokenizerFast(*args, **kwargs) - added:
constructor CamembertTokenizerFast(*args, **kwargs) - added:
constructor DPRContextEncoderTokenizerFast(*args, **kwargs) - added:
constructor DPRQuestionEncoderTokenizerFast(*args, **kwargs) - added:
constructor DPRReaderTokenizerFast(*args, **kwargs) - added:
constructor DistilBertTokenizerFast(*args, **kwargs) - added:
constructor ElectraTokenizerFast(*args, **kwargs) - added:
constructor FunnelTokenizerFast(*args, **kwargs) - added:
constructor GPT2TokenizerFast(*args, **kwargs) - added:
constructor HerbertTokenizerFast(*args, **kwargs) - added:
constructor LayoutLMTokenizerFast(*args, **kwargs) - added:
constructor LongformerTokenizerFast(*args, **kwargs) - added:
constructor LxmertTokenizerFast(*args, **kwargs) - added:
constructor MBartTokenizerFast(*args, **kwargs) - added:
constructor MobileBertTokenizerFast(*args, **kwargs) - added:
constructor OpenAIGPTTokenizerFast(*args, **kwargs) - added:
constructor PegasusTokenizerFast(*args, **kwargs) - added:
constructor PreTrainedTokenizerFast(*args, **kwargs) - added:
constructor ReformerTokenizerFast(*args, **kwargs) - added:
constructor RetriBertTokenizerFast(*args, **kwargs) - added:
constructor RobertaTokenizerFast(*args, **kwargs) - added:
constructor SqueezeBertTokenizerFast(*args, **kwargs) - added:
constructor T5TokenizerFast(*args, **kwargs) - added:
constructor XLMRobertaTokenizerFast(*args, **kwargs) - added:
constructor XLNetTokenizerFast(*args, **kwargs) - added:
function convert_slow_tokenizer(*args, **kwargs) - added:
static method AlbertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method BartTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method BertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method CamembertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method DPRContextEncoderTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method DPRQuestionEncoderTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method DPRReaderTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method DistilBertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method ElectraTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method FunnelTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method GPT2TokenizerFast.from_pretrained(*args, **kwargs) - added:
static method HerbertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method LayoutLMTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method LongformerTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method LxmertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method MBartTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method MobileBertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method OpenAIGPTTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method PegasusTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method PreTrainedTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method ReformerTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method RetriBertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method RobertaTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method SqueezeBertTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method T5TokenizerFast.from_pretrained(*args, **kwargs) - added:
static method XLMRobertaTokenizerFast.from_pretrained(*args, **kwargs) - added:
static method XLNetTokenizerFast.from_pretrained(*args, **kwargs) - added:
variable SLOW_TO_FAST_CONVERTERS
- added:
class TrialShortNamer - added:
field TrialShortNamer.DEFAULTS - added:
field TrialShortNamer.NAMING_INFO - added:
field TrialShortNamer.PREFIX - added:
static method TrialShortNamer.add_new_param_name(info, param_name) - added:
static method TrialShortNamer.build_naming_info() - added:
static method TrialShortNamer.parse_repr(repr) - added:
static method TrialShortNamer.set_defaults(prefix, defaults) - added:
static method TrialShortNamer.shortname(params) - added:
static method TrialShortNamer.shortname_for_key(info, param_name) - added:
static method TrialShortNamer.shortname_for_word(info, word)
- added:
function disable_default_handler() -> None - added:
function disable_propagation() -> None - added:
function enable_default_handler() -> None - added:
function enable_explicit_format() -> None - added:
function enable_propagation() -> None - added:
function get_logger(name: Optional[str]=…) -> logging.Logger - added:
function get_verbosity() -> int - added:
function reset_format() -> None - added:
function set_verbosity(verbosity: int) -> None - added:
function set_verbosity_debug() - added:
function set_verbosity_error() - added:
function set_verbosity_info() - added:
function set_verbosity_warning() - added:
variable log_levels
- added:
class NotebookProgressBar - added:
class NotebookProgressCallback(TrainerCallback) - added:
class NotebookTrainingTracker(NotebookProgressBar) - added:
constructor NotebookProgressBar(total: int, prefix: Optional[str]=…, leave: bool=…, parent: Optional[NotebookTrainingTracker]=…, width: int=…) - added:
constructor NotebookProgressCallback() - added:
constructor NotebookTrainingTracker(num_steps, column_names=…) - added:
field NotebookProgressBar.update_every - added:
field NotebookProgressBar.warmup - added:
function format_time(t) - added:
function html_progress_bar(value, total, prefix, label, width=…) - added:
function text_to_html_table(items) - added:
method NotebookProgressBar.close() - added:
method NotebookProgressBar.display() - added:
method NotebookProgressBar.update(value: int, force_update: bool=…, comment: str=…) - added:
method NotebookProgressBar.update_bar(value, comment=…) - added:
method NotebookProgressCallback.on_evaluate(args, state, control, metrics=…, **kwargs) - added:
method NotebookProgressCallback.on_log(args, state, control, logs=…, **kwargs) - added:
method NotebookProgressCallback.on_prediction_step(args, state, control, eval_dataloader=…, **kwargs) - added:
method NotebookProgressCallback.on_step_end(args, state, control, **kwargs) - added:
method NotebookProgressCallback.on_train_begin(args, state, control, **kwargs) - added:
method NotebookProgressCallback.on_train_end(args, state, control, **kwargs) - added:
method NotebookTrainingTracker.add_child(total, prefix=…, width=…) - added:
method NotebookTrainingTracker.display() - added:
method NotebookTrainingTracker.remove_child() - added:
method NotebookTrainingTracker.write_line(values)
- added:
variable DESCRIPTOR - added:
variable ModelProto - added:
variable NormalizerSpec - added:
variable SelfTestData - added:
variable TrainerSpec
- removed:
function prepare_encoder_decoder_model_kwargs(**kwargs)