vit_l_32 — Torchvision 0.16 documentation (original) (raw)

torchvision.models.vit_l_32(*, weights: Optional[ViT_L_32_Weights] = None, progress: bool = True, **kwargs: Any) → VisionTransformer[source]¶

Constructs a vit_l_32 architecture fromAn Image is Worth 16x16 Words: Transformers for Image Recognition at Scale.

Parameters:

• weights (ViT_L_32_Weights, optional) – The pretrained weights to use. See ViT_L_32_Weightsbelow for more details and possible values. By default, no pre-trained weights are used.
• progress (bool, optional) – If True, displays a progress bar of the download to stderr. Default is True.
• **kwargs – parameters passed to the torchvision.models.vision_transformer.VisionTransformerbase class. Please refer to the source codefor more details about this class.

class torchvision.models.ViT_L_32_Weights(value)[source]¶

The model builder above accepts the following values as the weights parameter.ViT_L_32_Weights.DEFAULT is equivalent to ViT_L_32_Weights.IMAGENET1K_V1. You can also use strings, e.g. weights='DEFAULT' or weights='IMAGENET1K_V1'.

ViT_L_32_Weights.IMAGENET1K_V1:

These weights were trained from scratch by using a modified version of DeIT’s training recipe. Also available as ViT_L_32_Weights.DEFAULT.

The inference transforms are available at ViT_L_32_Weights.IMAGENET1K_V1.transforms and perform the following preprocessing operations: Accepts PIL.Image, batched (B, C, H, W) and single (C, H, W) image torch.Tensor objects. The images are resized to resize_size=[256] using interpolation=InterpolationMode.BILINEAR, followed by a central crop of crop_size=[224]. Finally the values are first rescaled to [0.0, 1.0] and then normalized using mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].