[DINOv2] Convert more checkpoints (#26177)

* Convert checkpoints

* Update doc test

* Address comment

docs/source/en/_toctree.yml

@@ -508,7 +508,7 @@
       - local: model_doc/dinat
         title: DiNAT
       - local: model_doc/dinov2
-        title: DINO V2
+        title: DINOV2
       - local: model_doc/dit
         title: DiT
       - local: model_doc/dpt

src/transformers/models/dinov2/convert_dinov2_to_hf.py

@@ -19,14 +19,17 @@ URL: https://github.com/facebookresearch/dinov2/tree/main
 
 
 import argparse
+import json
 from pathlib import Path
 
 import requests
 import torch
+import torch.nn as nn
+from huggingface_hub import hf_hub_download
 from PIL import Image
 from torchvision import transforms
 
-from transformers import BitImageProcessor, Dinov2Config, Dinov2Model
+from transformers import BitImageProcessor, Dinov2Config, Dinov2ForImageClassification, Dinov2Model
 from transformers.image_utils import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD, PILImageResampling
 from transformers.utils import logging
 
@@ -35,7 +38,7 @@ logging.set_verbosity_info()
 logger = logging.get_logger(__name__)
 
 
-def get_dinov2_config(model_name):
+def get_dinov2_config(model_name, image_classifier=False):
     config = Dinov2Config(image_size=518, patch_size=14)
 
     # size of the architecture
@@ -56,6 +59,13 @@ def get_dinov2_config(model_name):
     else:
         raise ValueError("Model not supported")
 
+    if image_classifier:
+        repo_id = "huggingface/label-files"
+        filename = "imagenet-1k-id2label.json"
+        config.num_labels = 1000
+        config.id2label = json.load(open(hf_hub_download(repo_id, filename, repo_type="dataset"), "r"))
+        config.id2label = {int(k): v for k, v in config.id2label.items()}
+
     return config
 
 
@@ -140,10 +150,11 @@ def convert_dinov2_checkpoint(model_name, pytorch_dump_folder_path, push_to_hub=
     """
 
     # define default Dinov2 configuration
-    config = get_dinov2_config(model_name)
+    image_classifier = "1layer" in model_name
+    config = get_dinov2_config(model_name, image_classifier=image_classifier)
 
     # load original model from torch hub
-    original_model = torch.hub.load("facebookresearch/dinov2", model_name)
+    original_model = torch.hub.load("facebookresearch/dinov2", model_name.replace("_1layer", ""))
     original_model.eval()
 
     # load state_dict of original model, remove and rename some keys
@@ -162,7 +173,21 @@ def convert_dinov2_checkpoint(model_name, pytorch_dump_folder_path, push_to_hub=
         state_dict[key] = val
 
     # load HuggingFace model
-    model = Dinov2Model(config, add_pooling_layer=False).eval()
+    if image_classifier:
+        model = Dinov2ForImageClassification(config).eval()
+        model.dinov2.load_state_dict(state_dict)
+        model_name_to_classifier_dict_url = {
+            "dinov2_vits14_1layer": "https://dl.fbaipublicfiles.com/dinov2/dinov2_vits14/dinov2_vits14_linear_head.pth",
+            "dinov2_vitb14_1layer": "https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_linear_head.pth",
+            "dinov2_vitl14_1layer": "https://dl.fbaipublicfiles.com/dinov2/dinov2_vitl14/dinov2_vitl14_linear_head.pth",
+            "dinov2_vitg14_1layer": "https://dl.fbaipublicfiles.com/dinov2/dinov2_vitg14/dinov2_vitg14_linear_head.pth",
+        }
+        url = model_name_to_classifier_dict_url[model_name]
+        classifier_state_dict = torch.hub.load_state_dict_from_url(url, map_location="cpu")
+        model.classifier.weight = nn.Parameter(classifier_state_dict["weight"])
+        model.classifier.bias = nn.Parameter(classifier_state_dict["bias"])
+    else:
+        model = Dinov2Model(config).eval()
         model.load_state_dict(state_dict)
 
     # load image
@@ -195,10 +220,15 @@ def convert_dinov2_checkpoint(model_name, pytorch_dump_folder_path, push_to_hub=
     assert torch.allclose(original_pixel_values, pixel_values)
 
     with torch.no_grad():
-        outputs = model(pixel_values)
+        outputs = model(pixel_values, output_hidden_states=True)
         original_outputs = original_model(pixel_values)
 
     # assert values
+    if image_classifier:
+        print("Predicted class:")
+        class_idx = outputs.logits.argmax(-1).item()
+        print(model.config.id2label[class_idx])
+    else:
         assert outputs.last_hidden_state[:, 0].shape == original_outputs.shape
         assert torch.allclose(outputs.last_hidden_state[:, 0], original_outputs, atol=1e-3)
     print("Looks ok!")
@@ -216,6 +246,10 @@ def convert_dinov2_checkpoint(model_name, pytorch_dump_folder_path, push_to_hub=
             "dinov2_vitb14": "dinov2-base",
             "dinov2_vitl14": "dinov2-large",
             "dinov2_vitg14": "dinov2-giant",
+            "dinov2_vits14_1layer": "dinov2-small-imagenet1k-1-layer",
+            "dinov2_vitb14_1layer": "dinov2-base-imagenet1k-1-layer",
+            "dinov2_vitl14_1layer": "dinov2-large-imagenet1k-1-layer",
+            "dinov2_vitg14_1layer": "dinov2-giant-imagenet1k-1-layer",
         }
 
         name = model_name_to_hf_name[model_name]
@@ -230,7 +264,16 @@ if __name__ == "__main__":
         "--model_name",
         default="dinov2_vitb14",
         type=str,
-        choices=["dinov2_vits14", "dinov2_vitb14", "dinov2_vitl14", "dinov2_vitg14"],
+        choices=[
+            "dinov2_vits14",
+            "dinov2_vitb14",
+            "dinov2_vitl14",
+            "dinov2_vitg14",
+            "dinov2_vits14_1layer",
+            "dinov2_vitb14_1layer",
+            "dinov2_vitl14_1layer",
+            "dinov2_vitg14_1layer",
+        ],
         help="Name of the model you'd like to convert.",
     )
     parser.add_argument(

src/transformers/models/dinov2/modeling_dinov2.py

@@ -54,7 +54,8 @@ _CHECKPOINT_FOR_DOC = "facebook/dinov2-base"
 _EXPECTED_OUTPUT_SHAPE = [1, 257, 768]
 
 # Image classification docstring
-_IMAGE_CLASS_CHECKPOINT = "facebook/dinov2-base"
+_IMAGE_CLASS_CHECKPOINT = "facebook/dinov2-small-imagenet1k-1-layer"
+_IMAGE_CLASS_EXPECTED_OUTPUT = "tabby, tabby cat"
 
 
 DINOV2_PRETRAINED_MODEL_ARCHIVE_LIST = [
@@ -693,6 +694,7 @@ class Dinov2ForImageClassification(Dinov2PreTrainedModel):
         checkpoint=_IMAGE_CLASS_CHECKPOINT,
         output_type=ImageClassifierOutput,
         config_class=_CONFIG_FOR_DOC,
+        expected_output=_IMAGE_CLASS_EXPECTED_OUTPUT,
     )
     def forward(
         self,