Add option to load a pretrained model with mismatched shapes (#12664)

* Add option to load a pretrained model with mismatched shapes

* Fail at loading when mismatched shapes in Flax

* Fix tests

* Update src/transformers/modeling_flax_utils.py

Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Address review comments

Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

src/transformers/modeling_flax_utils.py

@@ -199,6 +199,10 @@ class FlaxPreTrainedModel(PushToHubMixin, FlaxGenerationMixin):
             from_pt (:obj:`bool`, `optional`, defaults to :obj:`False`):
                 Load the model weights from a PyTorch checkpoint save file (see docstring of
                 ``pretrained_model_name_or_path`` argument).
+            ignore_mismatched_size (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to raise an error if some of the weights from the checkpoint do not have the same size
+                as the weights of the model (if for instance, you are instantiating a model with 10 labels from a
+                checkpoint with 3 labels).
             force_download (:obj:`bool`, `optional`, defaults to :obj:`False`):
                 Whether or not to force the (re-)download of the model weights and configuration files, overriding the
                 cached versions if they exist.
@@ -242,6 +246,7 @@ class FlaxPreTrainedModel(PushToHubMixin, FlaxGenerationMixin):
         config = kwargs.pop("config", None)
         cache_dir = kwargs.pop("cache_dir", None)
         from_pt = kwargs.pop("from_pt", False)
+        ignore_mismatched_sizes = kwargs.pop("ignore_mismatched_sizes", False)
         force_download = kwargs.pop("force_download", False)
         resume_download = kwargs.pop("resume_download", False)
         proxies = kwargs.pop("proxies", None)
@@ -367,6 +372,22 @@ class FlaxPreTrainedModel(PushToHubMixin, FlaxGenerationMixin):
         missing_keys = model.required_params - set(state.keys())
         unexpected_keys = set(state.keys()) - model.required_params
 
+        # Mistmatched keys contains tuples key/shape1/shape2 of weights in the checkpoint that have a shape not
+        # matching the weights in the model.
+        mismatched_keys = []
+        for key in state.keys():
+            if key in random_state and state[key].shape != random_state[key].shape:
+                if ignore_mismatched_sizes:
+                    mismatched_keys.append((key, state[key].shape, random_state[key].shape))
+                    state[key] = random_state[key]
+                else:
+                    raise ValueError(
+                        f"Trying to load the pretrained weight for {key} failed: checkpoint has shape "
+                        f"{state[key].shape} which is incompatible with the model shape {random_state[key].shape}. "
+                        "Using `ignore_mismatched_sizes=True` if you really want to load this checkpoint inside this "
+                        "model."
+                    )
+
         # add missing keys as random parameters
         for missing_key in missing_keys:
             state[missing_key] = random_state[missing_key]
@@ -393,12 +414,24 @@ class FlaxPreTrainedModel(PushToHubMixin, FlaxGenerationMixin):
                 f"and are newly initialized: {missing_keys}\n"
                 f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
             )
-        else:
+        elif len(mismatched_keys) == 0:
             logger.info(
                 f"All the weights of {model.__class__.__name__} were initialized from the model checkpoint at {pretrained_model_name_or_path}.\n"
                 f"If your task is similar to the task the model of the checkpoint was trained on, "
                 f"you can already use {model.__class__.__name__} for predictions without further training."
             )
+        if len(mismatched_keys) > 0:
+            mismatched_warning = "\n".join(
+                [
+                    f"- {key}: found shape {shape1} in the checkpoint and {shape2} in the model instantiated"
+                    for key, shape1, shape2 in mismatched_keys
+                ]
+            )
+            logger.warning(
+                f"Some weights of {model.__class__.__name__} were not initialized from the model checkpoint at {pretrained_model_name_or_path} "
+                f"and are newly initialized because the shapes did not match:\n{mismatched_warning}\n"
+                f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
+            )
 
         # set correct parameters
         model.params = unflatten_dict(state)

src/transformers/modeling_tf_utils.py

@@ -450,7 +450,7 @@ def input_processing(func, config, input_ids, **kwargs):
     return output
 
 
-def load_tf_weights(model, resolved_archive_file, _prefix=None):
+def load_tf_weights(model, resolved_archive_file, ignore_mismatched_sizes=False, _prefix=None):
     """
     Detect missing and unexpected layers and load the TF weights accordingly to their names and shapes.
 
@@ -459,12 +459,16 @@ def load_tf_weights(model, resolved_archive_file, _prefix=None):
             The model to load the weights into.
         resolved_archive_file (:obj:`str`):
             The location of the H5 file.
+        ignore_mismatched_sizes (:obj:`bool`, `optional`, defaults to :obj:`False`):
+            Whether or not to ignore weights with shapes that don't match between the checkpoint of the model.
 
     Returns:
-        Two lists, one for the missing layers, and another one for the unexpected layers.
+        Three lists, one for the missing layers, another one for the unexpected layers, and a last one for the
+        mismatched layers.
     """
     missing_layers = []
     unexpected_layers = []
+    mismatched_layers = []
 
     # Read the H5 file
     with h5py.File(resolved_archive_file, "r") as f:
@@ -533,9 +537,14 @@ def load_tf_weights(model, resolved_archive_file, _prefix=None):
                             # If the two shapes are not compatible we raise an issue
                             try:
                                 array = np.reshape(saved_weight_value, K.int_shape(symbolic_weight))
-                            except AssertionError as e:
-                                e.args += (K.int_shape(symbolic_weight), saved_weight_value.shape)
-                                raise e
+                            except ValueError as e:
+                                if ignore_mismatched_sizes:
+                                    mismatched_layers.append(
+                                        (symbolic_weight_name, saved_weight_value.shape, K.int_shape(symbolic_weight))
+                                    )
+                                    continue
+                                else:
+                                    raise e
                         else:
                             array = saved_weight_value
 
@@ -549,7 +558,7 @@ def load_tf_weights(model, resolved_archive_file, _prefix=None):
     missing_layers.extend(list(symbolic_weights_names - saved_weight_names_set))
     unexpected_layers.extend(list(saved_weight_names_set - symbolic_weights_names))
 
-    return missing_layers, unexpected_layers
+    return missing_layers, unexpected_layers, mismatched_layers
 
 
 def init_copy_embeddings(old_embeddings, new_num_tokens):
@@ -1123,6 +1132,10 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin, Pu
             from_pt: (:obj:`bool`, `optional`, defaults to :obj:`False`):
                 Load the model weights from a PyTorch state_dict save file (see docstring of
                 ``pretrained_model_name_or_path`` argument).
+            ignore_mismatched_size (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to raise an error if some of the weights from the checkpoint do not have the same size
+                as the weights of the model (if for instance, you are instantiating a model with 10 labels from a
+                checkpoint with 3 labels).
             cache_dir (:obj:`str`, `optional`):
                 Path to a directory in which a downloaded pretrained model configuration should be cached if the
                 standard cache should not be used.
@@ -1186,6 +1199,7 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin, Pu
         config = kwargs.pop("config", None)
         cache_dir = kwargs.pop("cache_dir", None)
         from_pt = kwargs.pop("from_pt", False)
+        ignore_mismatched_sizes = kwargs.pop("ignore_mismatched_sizes", False)
         force_download = kwargs.pop("force_download", False)
         resume_download = kwargs.pop("resume_download", False)
         proxies = kwargs.pop("proxies", None)
@@ -1307,7 +1321,12 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin, Pu
         # 'by_name' allow us to do transfer learning by skipping/adding layers
         # see https://github.com/tensorflow/tensorflow/blob/00fad90125b18b80fe054de1055770cfb8fe4ba3/tensorflow/python/keras/engine/network.py#L1339-L1357
         try:
-            missing_keys, unexpected_keys = load_tf_weights(model, resolved_archive_file, load_weight_prefix)
+            missing_keys, unexpected_keys, mismatched_keys = load_tf_weights(
+                model,
+                resolved_archive_file,
+                ignore_mismatched_sizes=ignore_mismatched_sizes,
+                _prefix=load_weight_prefix,
+            )
         except OSError:
             raise OSError(
                 "Unable to load weights from h5 file. "
@@ -1342,15 +1361,31 @@ class TFPreTrainedModel(tf.keras.Model, TFModelUtilsMixin, TFGenerationMixin, Pu
                 f"and are newly initialized: {missing_keys}\n"
                 f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
             )
-        else:
+        elif len(mismatched_keys) == 0:
             logger.warning(
                 f"All the layers of {model.__class__.__name__} were initialized from the model checkpoint at {pretrained_model_name_or_path}.\n"
                 f"If your task is similar to the task the model of the checkpoint was trained on, "
                 f"you can already use {model.__class__.__name__} for predictions without further training."
             )
+        if len(mismatched_keys) > 0:
+            mismatched_warning = "\n".join(
+                [
+                    f"- {key}: found shape {shape1} in the checkpoint and {shape2} in the model instantiated"
+                    for key, shape1, shape2 in mismatched_keys
+                ]
+            )
+            logger.warning(
+                f"Some weights of {model.__class__.__name__} were not initialized from the model checkpoint at {pretrained_model_name_or_path} "
+                f"and are newly initialized because the shapes did not match:\n{mismatched_warning}\n"
+                f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
+            )
 
         if output_loading_info:
-            loading_info = {"missing_keys": missing_keys, "unexpected_keys": unexpected_keys}
+            loading_info = {
+                "missing_keys": missing_keys,
+                "unexpected_keys": unexpected_keys,
+                "mismatched_keys": mismatched_keys,
+            }
 
             return model, loading_info
 

src/transformers/modeling_utils.py

@@ -1037,6 +1037,10 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             from_flax (:obj:`bool`, `optional`, defaults to :obj:`False`):
                 Load the model weights from a Flax checkpoint save file (see docstring of
                 ``pretrained_model_name_or_path`` argument).
+            ignore_mismatched_size (:obj:`bool`, `optional`, defaults to :obj:`False`):
+                Whether or not to raise an error if some of the weights from the checkpoint do not have the same size
+                as the weights of the model (if for instance, you are instantiating a model with 10 labels from a
+                checkpoint with 3 labels).
             force_download (:obj:`bool`, `optional`, defaults to :obj:`False`):
                 Whether or not to force the (re-)download of the model weights and configuration files, overriding the
                 cached versions if they exist.
@@ -1120,6 +1124,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         cache_dir = kwargs.pop("cache_dir", None)
         from_tf = kwargs.pop("from_tf", False)
         from_flax = kwargs.pop("from_flax", False)
+        ignore_mismatched_sizes = kwargs.pop("ignore_mismatched_sizes", False)
         force_download = kwargs.pop("force_download", False)
         resume_download = kwargs.pop("resume_download", False)
         proxies = kwargs.pop("proxies", None)
@@ -1315,8 +1320,12 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 )
                 raise
         elif from_pt:
-            model, missing_keys, unexpected_keys, error_msgs = cls._load_state_dict_into_model(
-                model, state_dict, pretrained_model_name_or_path, _fast_init=_fast_init
+            model, missing_keys, unexpected_keys, mismatched_keys, error_msgs = cls._load_state_dict_into_model(
+                model,
+                state_dict,
+                pretrained_model_name_or_path,
+                ignore_mismatched_sizes=ignore_mismatched_sizes,
+                _fast_init=_fast_init,
             )
 
         # make sure token embedding weights are still tied if needed
@@ -1329,6 +1338,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             loading_info = {
                 "missing_keys": missing_keys,
                 "unexpected_keys": unexpected_keys,
+                "mismatched_keys": mismatched_keys,
                 "error_msgs": error_msgs,
             }
             return model, loading_info
@@ -1336,7 +1346,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         return model
 
     @classmethod
-    def _load_state_dict_into_model(cls, model, state_dict, pretrained_model_name_or_path, _fast_init=True):
+    def _load_state_dict_into_model(
+        cls, model, state_dict, pretrained_model_name_or_path, ignore_mismatched_sizes=False, _fast_init=True
+    ):
 
         # Convert old format to new format if needed from a PyTorch state_dict
         old_keys = []
@@ -1354,7 +1366,8 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             state_dict[new_key] = state_dict.pop(old_key)
 
         # Retrieve missing & unexpected_keys
-        expected_keys = list(model.state_dict().keys())
+        model_state_dict = model.state_dict()
+        expected_keys = list(model_state_dict.keys())
         loaded_keys = list(state_dict.keys())
         prefix = model.base_model_prefix
 
@@ -1374,6 +1387,26 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         missing_keys = list(set(expected_keys) - set(loaded_keys))
         unexpected_keys = list(set(loaded_keys) - set(expected_keys))
 
+        # Mistmatched keys contains tuples key/shape1/shape2 of weights in the checkpoint that have a shape not
+        # matching the weights in the model.
+        mismatched_keys = []
+        if ignore_mismatched_sizes:
+            for checkpoint_key in loaded_keys:
+                model_key = checkpoint_key
+                if remove_prefix and checkpoint_key.startswith(prefix):
+                    model_key = ".".join(checkpoint_key.split(".")[1:])
+                elif add_prefix:
+                    model_key = f"{prefix}.{checkpoint_key}"
+
+                if (
+                    model_key in model_state_dict
+                    and state_dict[checkpoint_key].shape != model_state_dict[model_key].shape
+                ):
+                    mismatched_keys.append(
+                        (checkpoint_key, state_dict[checkpoint_key].shape, model_state_dict[model_key].shape)
+                    )
+                    del state_dict[checkpoint_key]
+
         # Some models may have keys that are not in the state by design, removing them before needlessly warning
         # the user.
         if cls._keys_to_ignore_on_load_missing is not None:
@@ -1452,14 +1485,26 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                 f"and are newly initialized: {missing_keys}\n"
                 f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
             )
-        else:
+        elif len(mismatched_keys) == 0:
             logger.info(
                 f"All the weights of {model.__class__.__name__} were initialized from the model checkpoint at {pretrained_model_name_or_path}.\n"
                 f"If your task is similar to the task the model of the checkpoint was trained on, "
                 f"you can already use {model.__class__.__name__} for predictions without further training."
             )
+        if len(mismatched_keys) > 0:
+            mismatched_warning = "\n".join(
+                [
+                    f"- {key}: found shape {shape1} in the checkpoint and {shape2} in the model instantiated"
+                    for key, shape1, shape2 in mismatched_keys
+                ]
+            )
+            logger.warning(
+                f"Some weights of {model.__class__.__name__} were not initialized from the model checkpoint at {pretrained_model_name_or_path} "
+                f"and are newly initialized because the shapes did not match:\n{mismatched_warning}\n"
+                f"You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference."
+            )
 
-        return model, missing_keys, unexpected_keys, error_msgs
+        return model, missing_keys, unexpected_keys, mismatched_keys, error_msgs
 
     def retrieve_modules_from_names(self, names, add_prefix=False, remove_prefix=False):
         module_keys = set([".".join(key.split(".")[:-1]) for key in names])

src/transformers/models/auto/configuration_auto.py

@@ -186,7 +186,7 @@ CONFIG_MAPPING = OrderedDict(
         ("pegasus", PegasusConfig),
         ("marian", MarianConfig),
         ("mbart", MBartConfig),
-        ("megatron_bert", MegatronBertConfig),
+        ("megatron-bert", MegatronBertConfig),
         ("mpnet", MPNetConfig),
         ("bart", BartConfig),
         ("blenderbot", BlenderbotConfig),
@@ -252,7 +252,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("blenderbot", "Blenderbot"),
         ("marian", "Marian"),
         ("mbart", "mBART"),
-        ("megatron_bert", "MegatronBert"),
+        ("megatron-bert", "MegatronBert"),
         ("bart", "BART"),
         ("reformer", "Reformer"),
         ("longformer", "Longformer"),

src/transformers/models/deberta/modeling_deberta.py

@@ -760,10 +760,6 @@ class DebertaPreTrainedModel(PreTrainedModel):
     _keys_to_ignore_on_load_missing = ["position_ids"]
     _keys_to_ignore_on_load_unexpected = ["position_embeddings"]
 
-    def __init__(self, config):
-        super().__init__(config)
-        self._register_load_state_dict_pre_hook(self._pre_load_hook)
-
     def _init_weights(self, module):
         """Initialize the weights."""
         if isinstance(module, nn.Linear):
@@ -777,25 +773,6 @@ class DebertaPreTrainedModel(PreTrainedModel):
             if module.padding_idx is not None:
                 module.weight.data[module.padding_idx].zero_()
 
-    def _pre_load_hook(self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs):
-        """
-        Removes the classifier if it doesn't have the correct number of labels.
-        """
-        self_state = self.state_dict()
-        if (
-            ("classifier.weight" in self_state)
-            and ("classifier.weight" in state_dict)
-            and self_state["classifier.weight"].size() != state_dict["classifier.weight"].size()
-        ):
-            logger.warning(
-                f"The checkpoint classifier head has a shape {state_dict['classifier.weight'].size()} and this model "
-                f"classifier head has a shape {self_state['classifier.weight'].size()}. Ignoring the checkpoint "
-                f"weights. You should train your model on new data."
-            )
-            del state_dict["classifier.weight"]
-            if "classifier.bias" in state_dict:
-                del state_dict["classifier.bias"]
-
 
 DEBERTA_START_DOCSTRING = r"""
     The DeBERTa model was proposed in `DeBERTa: Decoding-enhanced BERT with Disentangled Attention

src/transformers/models/deberta_v2/modeling_deberta_v2.py

@@ -881,10 +881,6 @@ class DebertaV2PreTrainedModel(PreTrainedModel):
     _keys_to_ignore_on_load_missing = ["position_ids"]
     _keys_to_ignore_on_load_unexpected = ["position_embeddings"]
 
-    def __init__(self, config):
-        super().__init__(config)
-        self._register_load_state_dict_pre_hook(self._pre_load_hook)
-
     def _init_weights(self, module):
         """Initialize the weights."""
         if isinstance(module, nn.Linear):
@@ -898,25 +894,6 @@ class DebertaV2PreTrainedModel(PreTrainedModel):
             if module.padding_idx is not None:
                 module.weight.data[module.padding_idx].zero_()
 
-    def _pre_load_hook(self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs):
-        """
-        Removes the classifier if it doesn't have the correct number of labels.
-        """
-        self_state = self.state_dict()
-        if (
-            ("classifier.weight" in self_state)
-            and ("classifier.weight" in state_dict)
-            and self_state["classifier.weight"].size() != state_dict["classifier.weight"].size()
-        ):
-            logger.warning(
-                f"The checkpoint classifier head has a shape {state_dict['classifier.weight'].size()} and this model "
-                f"classifier head has a shape {self_state['classifier.weight'].size()}. Ignoring the checkpoint "
-                f"weights. You should train your model on new data."
-            )
-            del state_dict["classifier.weight"]
-            if "classifier.bias" in state_dict:
-                del state_dict["classifier.bias"]
-
 
 DEBERTA_START_DOCSTRING = r"""
     The DeBERTa model was proposed in `DeBERTa: Decoding-enhanced BERT with Disentangled Attention

tests/test_modeling_common.py

@@ -25,7 +25,7 @@ from typing import Dict, List, Tuple
 
 from huggingface_hub import HfApi
 from requests.exceptions import HTTPError
-from transformers import AutoModel, is_torch_available, logging
+from transformers import AutoModel, AutoModelForSequenceClassification, is_torch_available, logging
 from transformers.file_utils import WEIGHTS_NAME, is_torch_fx_available
 from transformers.models.auto import get_values
 from transformers.testing_utils import (
@@ -1532,6 +1532,35 @@ class ModelTesterMixin:
 
                     loss.backward()
 
+    def test_load_with_mismatched_shapes(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+
+        for model_class in self.all_model_classes:
+            if model_class not in get_values(MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING):
+                continue
+
+            with self.subTest(msg=f"Testing {model_class}"):
+                with tempfile.TemporaryDirectory() as tmp_dir:
+                    model = model_class(config)
+                    model.save_pretrained(tmp_dir)
+
+                    # Fails when we don't set ignore_mismatched_sizes=True
+                    with self.assertRaises(RuntimeError) as e:
+                        print(type(e))
+                        new_model = AutoModelForSequenceClassification.from_pretrained(tmp_dir, num_labels=42)
+
+                    logger = logging.get_logger("transformers.modeling_utils")
+                    with CaptureLogger(logger) as cl:
+                        new_model = AutoModelForSequenceClassification.from_pretrained(
+                            tmp_dir, num_labels=42, ignore_mismatched_sizes=True
+                        )
+                    self.assertIn("the shapes did not match", cl.out)
+
+                    new_model.to(torch_device)
+                    inputs = self._prepare_for_class(inputs_dict, model_class)
+                    logits = new_model(**inputs).logits
+                    self.assertEqual(logits.shape[1], 42)
+
 
 global_rng = random.Random()
 

tests/test_modeling_flax_common.py

@@ -24,17 +24,19 @@ import numpy as np
 import transformers
 from huggingface_hub import HfApi
 from requests.exceptions import HTTPError
-from transformers import BertConfig, FlaxBertModel, is_flax_available, is_torch_available
+from transformers import BertConfig, is_flax_available, is_torch_available
 from transformers.models.auto import get_values
 from transformers.testing_utils import (
     ENDPOINT_STAGING,
     PASS,
     USER,
+    CaptureLogger,
     is_pt_flax_cross_test,
     is_staging_test,
     require_flax,
     slow,
 )
+from transformers.utils import logging
 
 
 if is_flax_available():
@@ -45,7 +47,13 @@ if is_flax_available():
     import jaxlib.xla_extension as jax_xla
     from flax.core.frozen_dict import unfreeze
     from flax.traverse_util import flatten_dict
-    from transformers import FLAX_MODEL_FOR_QUESTION_ANSWERING_MAPPING, FLAX_MODEL_MAPPING
+    from transformers import (
+        FLAX_MODEL_FOR_QUESTION_ANSWERING_MAPPING,
+        FLAX_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
+        FLAX_MODEL_MAPPING,
+        FlaxAutoModelForSequenceClassification,
+        FlaxBertModel,
+    )
     from transformers.modeling_flax_pytorch_utils import (
         convert_pytorch_state_dict_to_flax,
         load_flax_weights_in_pytorch_model,
@@ -516,6 +524,32 @@ class FlaxModelTesterMixin:
                 [self.model_tester.num_attention_heads, encoder_seq_length, encoder_key_length],
             )
 
+    def test_load_with_mismatched_shapes(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+
+        for model_class in self.all_model_classes:
+            if model_class not in get_values(FLAX_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING):
+                continue
+
+            with self.subTest(msg=f"Testing {model_class}"):
+                with tempfile.TemporaryDirectory() as tmp_dir:
+                    model = model_class(config)
+                    model.save_pretrained(tmp_dir)
+
+                    # Fails when we don't set ignore_mismatched_sizes=True
+                    with self.assertRaises(ValueError):
+                        new_model = FlaxAutoModelForSequenceClassification.from_pretrained(tmp_dir, num_labels=42)
+
+                    logger = logging.get_logger("transformers.modeling_flax_utils")
+                    with CaptureLogger(logger) as cl:
+                        new_model = FlaxAutoModelForSequenceClassification.from_pretrained(
+                            tmp_dir, num_labels=42, ignore_mismatched_sizes=True
+                        )
+                    self.assertIn("the shapes did not match", cl.out)
+
+                    logits = new_model(**inputs_dict)["logits"]
+                    self.assertEqual(logits.shape[1], 42)
+
 
 @require_flax
 @is_staging_test

tests/test_modeling_tf_common.py

@@ -32,6 +32,7 @@ from transformers.testing_utils import (
     ENDPOINT_STAGING,
     PASS,
     USER,
+    CaptureLogger,
     _tf_gpu_memory_limit,
     is_pt_tf_cross_test,
     is_staging_test,
@@ -40,6 +41,7 @@ from transformers.testing_utils import (
     slow,
     tooslow,
 )
+from transformers.utils import logging
 
 
 if is_tf_available():
@@ -57,6 +59,7 @@ if is_tf_available():
         TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
         TF_MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING,
         BertConfig,
+        TFAutoModelForSequenceClassification,
         TFBertModel,
         TFSharedEmbeddings,
         tf_top_k_top_p_filtering,
@@ -1308,6 +1311,34 @@ class TFModelTesterMixin:
                 attn_weights = out[attn_name] if attn_name == attention_names[0] else out[attn_name][-1]
                 self.assertEqual(sum([tf.reduce_sum(w).numpy() for w in attn_weights]), 0.0)
 
+    def test_load_with_mismatched_shapes(self):
+        config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
+
+        for model_class in self.all_model_classes:
+            if model_class not in get_values(TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING):
+                continue
+
+            with self.subTest(msg=f"Testing {model_class}"):
+                with tempfile.TemporaryDirectory() as tmp_dir:
+                    model = model_class(config)
+                    inputs = self._prepare_for_class(inputs_dict, model_class)
+                    _ = model(**inputs)
+                    model.save_pretrained(tmp_dir)
+
+                    # Fails when we don't set ignore_mismatched_sizes=True
+                    with self.assertRaises(ValueError):
+                        new_model = TFAutoModelForSequenceClassification.from_pretrained(tmp_dir, num_labels=42)
+
+                    logger = logging.get_logger("transformers.modeling_tf_utils")
+                    with CaptureLogger(logger) as cl:
+                        new_model = TFAutoModelForSequenceClassification.from_pretrained(
+                            tmp_dir, num_labels=42, ignore_mismatched_sizes=True
+                        )
+                    self.assertIn("the shapes did not match", cl.out)
+
+                    logits = new_model(**inputs).logits
+                    self.assertEqual(logits.shape[1], 42)
+
     def _generate_random_bad_tokens(self, num_bad_tokens, model):
         # special tokens cannot be bad tokens
         special_tokens = []