Upgrading torchvision 0.13 to 0.14
Unknown0.13.1 → 0.14.1
Fetching 2 releases from GitHub…0 / 2
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.
Installs 16 packages in all (was 9).
Also installs nvidia-cublas-cu11, nvidia-cuda-nvrtc-cu11, nvidia-cuda-runtime-cu11, nvidia-cudnn-cu11, packaging, setuptools, wheel.
Moves torch 1.12.1 → 1.13.1. Your build may pick other versions.
API changes
Generated by DepDelta from published Python source. Not mixed with prose notes.
- removed:
export Kinetics400 - added:
export CREStereo - added:
export CarlaStereo - added:
export ETH3DStereo - added:
export FallingThingsStereo - added:
export InStereo2k - added:
export Kitti2012Stereo - added:
export Kitti2015Stereo - added:
export Middlebury2014Stereo - added:
export SceneFlowStereo - added:
export SintelStereo
- removed:
class Kinetics400(Kinetics) - removed:
constructor Kinetics400(root: str, frames_per_clip: int, num_classes: Any=…, split: Any=…, download: Any=…, num_download_workers: Any=…, **kwargs: Any) -> None
- changed:
function verify_str_arg(value: T, arg: Optional[str]=…, valid_values: Iterable[T]=…, custom_msg: Optional[str]=…) -> T → function verify_str_arg(value: T, arg: Optional[str]=…, valid_values: Optional[Iterable[T]]=…, custom_msg: Optional[str]=…) -> T
- added:
variable env_path - added:
variable path_arr
- added:
export get_model - added:
export get_model_builder - added:
export get_model_weights - added:
export list_models
- added:
class MaxVit(nn.Module) - added:
class MaxVit_T_Weights(WeightsEnum) - added:
constructor MaxVit(input_size: Tuple[int, int], stem_channels: int, partition_size: int, block_channels: List[int], block_layers: List[int], head_dim: int, stochastic_depth_prob: float, norm_layer: Optional[Callable[..., nn.Module]]=…, activation_layer: Callable[..., nn.Module]=…, squeeze_ratio: float=…, expansion_ratio: float=…, mlp_ratio: int=…, mlp_dropout: float=…, attention_dropout: float=…,… - added:
field MaxVit_T_Weights.DEFAULT - added:
field MaxVit_T_Weights.IMAGENET1K_V1 - added:
function maxvit_t(*, weights: Optional[MaxVit_T_Weights]=…, progress: bool=…, **kwargs: Any) -> MaxVit - added:
method MaxVit.forward(x: Tensor) -> Tensor
- changed:
constructor QuantizableInception3(num_classes: int=…, aux_logits: bool=…, transform_input: bool=…) -> None → constructor QuantizableInception3(*args: Any, **kwargs: Any) -> None
- changed:
constructor SwinTransformer(patch_size: List[int], embed_dim: int, depths: List[int], num_heads: List[int], window_size: List[int], mlp_ratio: float=…, dropout: float=…, attention_dropout: float=…, stochastic_depth_prob: float=…, num_classes: int=…, norm_layer: Optional[Callable[..., nn.Module]]=…, block: Optional[Callable[..., nn.Module]]=…) → constructor SwinTransformer(patch_size: List[int], embed_dim: int, depths: List[int], num_heads: List[int], window_size: List[int], mlp_ratio: float=…, dropout: float=…, attention_dropout: float=…, stochastic_depth_prob: float=…, num_classes: int=…, norm_layer: Optional[Callable[..., nn.Module]]=…, block: Optional[Callable[..., nn.Module]]=…, downsample_layer: Callable[..., nn.Module]=…) - added:
class Swin_V2_B_Weights(WeightsEnum) - added:
class Swin_V2_S_Weights(WeightsEnum) - added:
class Swin_V2_T_Weights(WeightsEnum) - added:
field Swin_V2_B_Weights.DEFAULT - added:
field Swin_V2_B_Weights.IMAGENET1K_V1 - added:
field Swin_V2_S_Weights.DEFAULT - added:
field Swin_V2_S_Weights.IMAGENET1K_V1 - added:
field Swin_V2_T_Weights.DEFAULT - added:
field Swin_V2_T_Weights.IMAGENET1K_V1 - added:
function swin_v2_b(*, weights: Optional[Swin_V2_B_Weights]=…, progress: bool=…, **kwargs: Any) -> SwinTransformer - added:
function swin_v2_s(*, weights: Optional[Swin_V2_S_Weights]=…, progress: bool=…, **kwargs: Any) -> SwinTransformer - added:
function swin_v2_t(*, weights: Optional[Swin_V2_T_Weights]=…, progress: bool=…, **kwargs: Any) -> SwinTransformer
- added:
class MViT(nn.Module) - added:
class MViT_V1_B_Weights(WeightsEnum) - added:
class MViT_V2_S_Weights(WeightsEnum) - added:
constructor MViT(spatial_size: Tuple[int, int], temporal_size: int, block_setting: Sequence[MSBlockConfig], residual_pool: bool, residual_with_cls_embed: bool, rel_pos_embed: bool, proj_after_attn: bool, dropout: float=…, attention_dropout: float=…, stochastic_depth_prob: float=…, num_classes: int=…, block: Optional[Callable[..., nn.Module]]=…, norm_layer: Optional[Callable[..., nn.Module]]=…, pa… - added:
field MViT_V1_B_Weights.DEFAULT - added:
field MViT_V1_B_Weights.KINETICS400_V1 - added:
field MViT_V2_S_Weights.DEFAULT - added:
field MViT_V2_S_Weights.KINETICS400_V1 - added:
function mvit_v1_b(*, weights: Optional[MViT_V1_B_Weights]=…, progress: bool=…, **kwargs: Any) -> MViT - added:
function mvit_v2_s(*, weights: Optional[MViT_V2_S_Weights]=…, progress: bool=…, **kwargs: Any) -> MViT - added:
method MViT.forward(x: torch.Tensor) -> torch.Tensor
- added:
class S3D(nn.Module) - added:
class S3D_Weights(WeightsEnum) - added:
constructor S3D(num_classes: int=…, dropout: float=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…) -> None - added:
field S3D_Weights.DEFAULT - added:
field S3D_Weights.KINETICS400_V1 - added:
function s3d(*, weights: Optional[S3D_Weights]=…, progress: bool=…, **kwargs: Any) -> S3D - added:
method S3D.forward(x)
- changed:
constructor Conv2dNormActivation(in_channels: int, out_channels: int, kernel_size: int=…, stride: int=…, padding: Optional[int]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: int=…, inplace: Optional[bool]=…, bias: Optional[bool]=…) -> None → constructor Conv2dNormActivation(in_channels: int, out_channels: int, kernel_size: Union[int, Tuple[int, int]]=…, stride: Union[int, Tuple[int, int]]=…, padding: Optional[Union[int, Tuple[int, int], str]]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: Union[int, Tuple[int, int]]=…, inplace: Optional… - changed:
constructor Conv3dNormActivation(in_channels: int, out_channels: int, kernel_size: int=…, stride: int=…, padding: Optional[int]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: int=…, inplace: Optional[bool]=…, bias: Optional[bool]=…) -> None → constructor Conv3dNormActivation(in_channels: int, out_channels: int, kernel_size: Union[int, Tuple[int, int, int]]=…, stride: Union[int, Tuple[int, int, int]]=…, padding: Optional[Union[int, Tuple[int, int, int], str]]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: Union[int, Tuple[int, int, int]]=… - changed:
constructor ConvNormActivation(in_channels: int, out_channels: int, kernel_size: int=…, stride: int=…, padding: Optional[int]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: int=…, inplace: Optional[bool]=…, bias: Optional[bool]=…, conv_layer: Callable[..., torch.nn.Module]=…) -> None → constructor ConvNormActivation(in_channels: int, out_channels: int, kernel_size: Union[int, Tuple[int, ...]]=…, stride: Union[int, Tuple[int, ...]]=…, padding: Optional[Union[int, Tuple[int, ...], str]]=…, groups: int=…, norm_layer: Optional[Callable[..., torch.nn.Module]]=…, activation_layer: Optional[Callable[..., torch.nn.Module]]=…, dilation: Union[int, Tuple[int, ...]]=…, inplace: Optional[b…
- removed:
method MultiScaleRoIAlign.convert_to_roi_format(boxes: List[Tensor]) -> Tensor - removed:
method MultiScaleRoIAlign.infer_scale(feature: Tensor, original_size: List[int]) -> float - removed:
method MultiScaleRoIAlign.setup_setup_scales(features: List[Tensor], image_shapes: List[Tuple[int, int]]) -> None
- changed:
method RoIAlign.forward(input: Tensor, rois: Tensor) -> Tensor → method RoIAlign.forward(input: Tensor, rois: Union[Tensor, List[Tensor]]) -> Tensor
- changed:
method RoIPool.forward(input: Tensor, rois: Tensor) -> Tensor → method RoIPool.forward(input: Tensor, rois: Union[Tensor, List[Tensor]]) -> Tensor
- changed:
function affine(img: Tensor, angle: float, translate: List[int], scale: float, shear: List[float], interpolation: InterpolationMode=…, fill: Optional[List[float]]=…, resample: Optional[int]=…, fillcolor: Optional[List[float]]=…, center: Optional[List[int]]=…) -> Tensor → function affine(img: Tensor, angle: float, translate: List[int], scale: float, shear: List[float], interpolation: InterpolationMode=…, fill: Optional[List[float]]=…, center: Optional[List[int]]=…) -> Tensor - changed:
function pad(img: Tensor, padding: List[int], fill: int=…, padding_mode: str=…) -> Tensor → function pad(img: Tensor, padding: List[int], fill: Union[int, float]=…, padding_mode: str=…) -> Tensor - changed:
function pil_to_tensor(pic) → function pil_to_tensor(pic: Any) -> Tensor - changed:
function resized_crop(img: Tensor, top: int, left: int, height: int, width: int, size: List[int], interpolation: InterpolationMode=…) -> Tensor → function resized_crop(img: Tensor, top: int, left: int, height: int, width: int, size: List[int], interpolation: InterpolationMode=…, antialias: Optional[bool]=…) -> Tensor - changed:
function rotate(img: Tensor, angle: float, interpolation: InterpolationMode=…, expand: bool=…, center: Optional[List[int]]=…, fill: Optional[List[float]]=…, resample: Optional[int]=…) -> Tensor → function rotate(img: Tensor, angle: float, interpolation: InterpolationMode=…, expand: bool=…, center: Optional[List[int]]=…, fill: Optional[List[float]]=…) -> Tensor - added:
function elastic_transform(img: Tensor, displacement: Tensor, interpolation: InterpolationMode=…, fill: Optional[List[float]]=…) -> Tensor
- changed:
function affine(img: Image.Image, matrix: List[float], interpolation: int=…, fill: Optional[Union[float, List[float], Tuple[float, ...]]]=…) -> Image.Image → function affine(img: Image.Image, matrix: List[float], interpolation: int=…, fill: Optional[Union[int, float, Sequence[int], Sequence[float]]]=…) -> Image.Image - changed:
function perspective(img: Image.Image, perspective_coeffs: float, interpolation: int=…, fill: Optional[Union[float, List[float], Tuple[float, ...]]]=…) -> Image.Image → function perspective(img: Image.Image, perspective_coeffs: List[float], interpolation: int=…, fill: Optional[Union[int, float, Sequence[int], Sequence[float]]]=…) -> Image.Image - changed:
function resize(img: Image.Image, size: Union[Sequence[int], int], interpolation: int=…, max_size: Optional[int]=…) -> Image.Image → function resize(img: Image.Image, size: Union[List[int], int], interpolation: int=…) -> Image.Image - changed:
function rotate(img: Image.Image, angle: float, interpolation: int=…, expand: bool=…, center: Optional[Tuple[int, int]]=…, fill: Optional[Union[float, List[float], Tuple[float, ...]]]=…) -> Image.Image → function rotate(img: Image.Image, angle: float, interpolation: int=…, expand: bool=…, center: Optional[Tuple[int, int]]=…, fill: Optional[Union[int, float, Sequence[int], Sequence[float]]]=…) -> Image.Image
- changed:
function affine(img: Tensor, matrix: List[float], interpolation: str=…, fill: Optional[List[float]]=…) -> Tensor → function affine(img: Tensor, matrix: List[float], interpolation: str=…, fill: Optional[Union[int, float, List[float]]]=…) -> Tensor - changed:
function pad(img: Tensor, padding: List[int], fill: int=…, padding_mode: str=…) -> Tensor → function pad(img: Tensor, padding: Union[int, List[int]], fill: Optional[Union[int, float]]=…, padding_mode: str=…) -> Tensor - changed:
function perspective(img: Tensor, perspective_coeffs: List[float], interpolation: str=…, fill: Optional[List[float]]=…) -> Tensor → function perspective(img: Tensor, perspective_coeffs: List[float], interpolation: str=…, fill: Optional[Union[int, float, List[float]]]=…) -> Tensor - changed:
function resize(img: Tensor, size: List[int], interpolation: str=…, max_size: Optional[int]=…, antialias: Optional[bool]=…) -> Tensor → function resize(img: Tensor, size: List[int], interpolation: str=…, antialias: Optional[bool]=…) -> Tensor - changed:
function rotate(img: Tensor, matrix: List[float], interpolation: str=…, expand: bool=…, fill: Optional[List[float]]=…) -> Tensor → function rotate(img: Tensor, matrix: List[float], interpolation: str=…, expand: bool=…, fill: Optional[Union[int, float, List[float]]]=…) -> Tensor - added:
function elastic_transform(img: Tensor, displacement: Tensor, interpolation: str=…, fill: Optional[Union[int, float, List[float]]]=…) -> Tensor
- changed:
constructor RandomAffine(degrees, translate=…, scale=…, shear=…, interpolation=…, fill=…, fillcolor=…, resample=…, center=…) → constructor RandomAffine(degrees, translate=…, scale=…, shear=…, interpolation=…, fill=…, center=…) - changed:
constructor RandomResizedCrop(size, scale=…, ratio=…, interpolation=…) → constructor RandomResizedCrop(size, scale=…, ratio=…, interpolation=…, antialias: Optional[bool]=…) - changed:
constructor RandomRotation(degrees, interpolation=…, expand=…, center=…, fill=…, resample=…) → constructor RandomRotation(degrees, interpolation=…, expand=…, center=…, fill=…) - added:
class ElasticTransform(torch.nn.Module) - added:
constructor ElasticTransform(alpha=…, sigma=…, interpolation=…, fill=…) - added:
method ElasticTransform.forward(tensor: Tensor) -> Tensor - added:
static method ElasticTransform.get_params(alpha: List[float], sigma: List[float], size: List[int]) -> Tensor