Add GPT2ForTokenClassification (#13290)

* Add GPT2ForTokenClassification

* Fix dropout exception for GPT2 NER

* Remove sequence label in test

* Change TokenClassifierOutput to TokenClassifierOutputWithPast

* Fix for black formatter

* Remove dummy

* Update docs for GPT2ForTokenClassification

* Fix check_inits ci fail

* Update dummy_pt_objects after make fix-copies

* Remove TokenClassifierOutputWithPast

* Fix tuple input issue
Co-authored-by: danielsejong55@gmail.com <danielsejong55@gmail.com>

docs/source/model_doc/gpt2.rst

@@ -108,6 +108,13 @@ GPT2ForSequenceClassification
     :members: forward
 
 
+GPT2ForTokenClassification
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.GPT2ForTokenClassification
+    :members: forward
+
+
 TFGPT2Model
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 

src/transformers/__init__.py

@@ -807,6 +807,7 @@ if is_torch_available():
             "GPT2_PRETRAINED_MODEL_ARCHIVE_LIST",
             "GPT2DoubleHeadsModel",
             "GPT2ForSequenceClassification",
+            "GPT2ForTokenClassification",
             "GPT2LMHeadModel",
             "GPT2Model",
             "GPT2PreTrainedModel",
@@ -2460,6 +2461,7 @@ if TYPE_CHECKING:
             GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
             GPT2DoubleHeadsModel,
             GPT2ForSequenceClassification,
+            GPT2ForTokenClassification,
             GPT2LMHeadModel,
             GPT2Model,
             GPT2PreTrainedModel,

src/transformers/models/auto/modeling_auto.py

@@ -399,6 +399,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES = OrderedDict(
         ("mpnet", "MPNetForTokenClassification"),
         ("deberta", "DebertaForTokenClassification"),
         ("deberta-v2", "DebertaV2ForTokenClassification"),
+        ("gpt2", "GPT2ForTokenClassification"),
         ("ibert", "IBertForTokenClassification"),
     ]
 )

src/transformers/models/gpt2/__init__.py

@@ -34,6 +34,7 @@ if is_torch_available():
         "GPT2_PRETRAINED_MODEL_ARCHIVE_LIST",
         "GPT2DoubleHeadsModel",
         "GPT2ForSequenceClassification",
+        "GPT2ForTokenClassification",
         "GPT2LMHeadModel",
         "GPT2Model",
         "GPT2PreTrainedModel",
@@ -66,6 +67,7 @@ if TYPE_CHECKING:
             GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
             GPT2DoubleHeadsModel,
             GPT2ForSequenceClassification,
+            GPT2ForTokenClassification,
             GPT2LMHeadModel,
             GPT2Model,
             GPT2PreTrainedModel,

src/transformers/models/gpt2/modeling_gpt2.py

@@ -36,6 +36,7 @@ from ...modeling_outputs import (
     BaseModelOutputWithPastAndCrossAttentions,
     CausalLMOutputWithCrossAttentions,
     SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
 )
 from ...modeling_utils import (
     Conv1D,
@@ -1331,3 +1332,105 @@ class GPT2ForSequenceClassification(GPT2PreTrainedModel):
             hidden_states=transformer_outputs.hidden_states,
             attentions=transformer_outputs.attentions,
         )
+
+
+@add_start_docstrings(
+    """
+    GPT2 Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
+    Named-Entity-Recognition (NER) tasks.
+    """,
+    GPT2_START_DOCSTRING,
+)
+class GPT2ForTokenClassification(GPT2PreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.transformer = GPT2Model(config)
+        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
+            classifier_dropout = config.classifier_dropout
+        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        self.init_weights()
+
+        # Model parallel
+        self.model_parallel = False
+        self.device_map = None
+
+    @add_start_docstrings_to_model_forward(GPT2_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        tokenizer_class=_TOKENIZER_FOR_DOC,
+        checkpoint="microsoft/DialogRPT-updown",
+        output_type=TokenClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids=None,
+        past_key_values=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        head_mask=None,
+        inputs_embeds=None,
+        labels=None,
+        use_cache=None,
+        output_attentions=None,
+        output_hidden_states=None,
+        return_dict=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+            Labels for computing the sequence classification/regression loss. Indices should be in :obj:`[0, ...,
+            config.num_labels - 1]`. If :obj:`config.num_labels == 1` a regression loss is computed (Mean-Square loss),
+            If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        transformer_outputs = self.transformer(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = transformer_outputs[0]
+        hidden_states = self.dropout(hidden_states)
+        logits = self.classifier(hidden_states)
+
+        loss = None
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            # Only keep active parts of the loss
+            if attention_mask is not None:
+                active_loss = attention_mask.view(-1) == 1
+                active_logits = logits.view(-1, self.num_labels)
+                active_labels = torch.where(
+                    active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels)
+                )
+                loss = loss_fct(active_logits, active_labels)
+            else:
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + transformer_outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )

src/transformers/utils/dummy_pt_objects.py

@@ -1781,6 +1781,15 @@ class GPT2ForSequenceClassification:
         requires_backends(cls, ["torch"])
 
 
+class GPT2ForTokenClassification:
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch"])
+
+
 class GPT2LMHeadModel:
     def __init__(self, *args, **kwargs):
         requires_backends(self, ["torch"])

tests/test_modeling_gpt2.py

@@ -32,6 +32,7 @@ if is_torch_available():
         GPT2_PRETRAINED_MODEL_ARCHIVE_LIST,
         GPT2DoubleHeadsModel,
         GPT2ForSequenceClassification,
+        GPT2ForTokenClassification,
         GPT2LMHeadModel,
         GPT2Model,
         GPT2Tokenizer,
@@ -366,6 +367,16 @@ class GPT2ModelTester:
         result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels)
         self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
 
+    def create_and_check_gpt2_for_token_classification(
+        self, config, input_ids, input_mask, head_mask, token_type_ids, mc_token_ids, sequence_labels, *args
+    ):
+        config.num_labels = self.num_labels
+        model = GPT2ForTokenClassification(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
+
     def prepare_config_and_inputs_for_common(self):
         config_and_inputs = self.prepare_config_and_inputs()
 
@@ -394,7 +405,7 @@ class GPT2ModelTester:
 class GPT2ModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
 
     all_model_classes = (
-        (GPT2Model, GPT2LMHeadModel, GPT2DoubleHeadsModel, GPT2ForSequenceClassification)
+        (GPT2Model, GPT2LMHeadModel, GPT2DoubleHeadsModel, GPT2ForSequenceClassification, GPT2ForTokenClassification)
         if is_torch_available()
         else ()
     )
@@ -462,6 +473,10 @@ class GPT2ModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.TestCase):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_gpt2_for_sequence_classification(*config_and_inputs)
 
+    def test_gpt2_token_classification_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_gpt2_for_token_classification(*config_and_inputs)
+
     def test_gpt2_gradient_checkpointing(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs(gradient_checkpointing=True)
         self.model_tester.create_and_check_forward_and_backwards(*config_and_inputs)