Improve mismatched sizes management when loading a pretrained model (#17257)

- Add --ignore_mismatched_sizes argument to classification examples

- Expand the error message when loading a model whose head dimensions are different from expected dimensions

examples/pytorch/image-classification/README.md

@@ -62,9 +62,11 @@ python run_image_classification.py \
 
 Note that you can replace the model and dataset by simply setting the `model_name_or_path` and `dataset_name` arguments respectively, with any model or dataset from the [hub](https://huggingface.co/). For an overview of all possible arguments, we refer to the [docs](https://huggingface.co/docs/transformers/main_classes/trainer#transformers.TrainingArguments) of the `TrainingArguments`, which can be passed as flags.
 
+> If your model classification head dimensions do not fit the number of labels in the dataset, you can specify `--ignore_mismatched_sizes` to adapt it.
+
 ### Using your own data
 
-To use your own dataset, there are 2 ways: 
+To use your own dataset, there are 2 ways:
 - you can either provide your own folders as `--train_dir` and/or `--validation_dir` arguments
 - you can upload your dataset to the hub (possibly as a private repo, if you prefer so), and simply pass the `--dataset_name` argument.
 

examples/pytorch/image-classification/run_image_classification.py

@@ -150,6 +150,10 @@ class ModelArguments:
             )
         },
     )
+    ignore_mismatched_sizes: bool = field(
+        default=False,
+        metadata={"help": "Will enable to load a pretrained model whose head dimensions are different."},
+    )
 
 
 def collate_fn(examples):
@@ -269,6 +273,7 @@ def main():
         cache_dir=model_args.cache_dir,
         revision=model_args.model_revision,
         use_auth_token=True if model_args.use_auth_token else None,
+        ignore_mismatched_sizes=model_args.ignore_mismatched_sizes,
     )
     feature_extractor = AutoFeatureExtractor.from_pretrained(
         model_args.feature_extractor_name or model_args.model_name_or_path,

examples/pytorch/image-classification/run_image_classification_no_trainer.py

@@ -165,6 +165,11 @@ def parse_args():
         action="store_true",
         help="Whether to load in all available experiment trackers from the environment and use them for logging.",
     )
+    parser.add_argument(
+        "--ignore_mismatched_sizes",
+        action="store_true",
+        help="Whether or not to enable to load a pretrained model whose head dimensions are different.",
+    )
     args = parser.parse_args()
 
     # Sanity checks
@@ -278,6 +283,7 @@ def main():
         args.model_name_or_path,
         from_tf=bool(".ckpt" in args.model_name_or_path),
         config=config,
+        ignore_mismatched_sizes=args.ignore_mismatched_sizes,
     )
 
     # Preprocessing the datasets