[fbsync] Adding missing named param check on ViT (#5196)

Reviewed By: datumbox, NicolasHug

Differential Revision: D33655255

fbshipit-source-id: a6f5fe7ddd30850236d2b82a8847a2fa1b499a9b

torchvision/prototype/models/vision_transformer.py

@@ -11,7 +11,7 @@
 from ...models.vision_transformer import VisionTransformer, interpolate_embeddings  # noqa: F401
 from ._api import WeightsEnum, Weights
 from ._meta import _IMAGENET_CATEGORIES
-from ._utils import handle_legacy_interface
+from ._utils import handle_legacy_interface, _ovewrite_named_param
 
 __all__ = [
     "VisionTransformer",
@@ -111,6 +111,9 @@ def _vision_transformer(
 ) -> VisionTransformer:
     image_size = kwargs.pop("image_size", 224)
 
+    if weights is not None:
+        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))
+
     model = VisionTransformer(
         image_size=image_size,
         patch_size=patch_size,