LongformerForSequenceClassification (#4580)

* LongformerForSequenceClassification

* better naming x=>hidden_states, fix typo in doc

* Update src/transformers/modeling_longformer.py

* Update src/transformers/modeling_longformer.py

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

src/transformers/__init__.py

@@ -338,6 +338,7 @@ if is_torch_available():
     from .modeling_longformer import (
         LongformerModel,
         LongformerForMaskedLM,
+        LongformerForSequenceClassification,
         LongformerForQuestionAnswering,
         LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP,
     )

src/transformers/modeling_auto.py

@@ -105,6 +105,7 @@ from .modeling_longformer import (
     LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP,
     LongformerForMaskedLM,
     LongformerForQuestionAnswering,
+    LongformerForSequenceClassification,
     LongformerModel,
 )
 from .modeling_marian import MarianMTModel
@@ -252,6 +253,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING = OrderedDict(
         (CamembertConfig, CamembertForSequenceClassification),
         (XLMRobertaConfig, XLMRobertaForSequenceClassification),
         (BartConfig, BartForSequenceClassification),
+        (LongformerConfig, LongformerForSequenceClassification),
         (RobertaConfig, RobertaForSequenceClassification),
         (BertConfig, BertForSequenceClassification),
         (XLNetConfig, XLNetForSequenceClassification),

src/transformers/modeling_longformer.py

@@ -19,7 +19,7 @@ import math
 
 import torch
 import torch.nn as nn
-from torch.nn import CrossEntropyLoss
+from torch.nn import CrossEntropyLoss, MSELoss
 from torch.nn import functional as F
 
 from .configuration_longformer import LongformerConfig
@@ -710,6 +710,121 @@ class LongformerForMaskedLM(BertPreTrainedModel):
         return outputs  # (masked_lm_loss), prediction_scores, (hidden_states), (attentions)
 
 
+@add_start_docstrings(
+    """Longformer Model transformer with a sequence classification/regression head on top (a linear layer
+    on top of the pooled output) e.g. for GLUE tasks. """,
+    LONGFORMER_START_DOCSTRING,
+)
+class LongformerForSequenceClassification(BertPreTrainedModel):
+    config_class = LongformerConfig
+    pretrained_model_archive_map = LONGFORMER_PRETRAINED_MODEL_ARCHIVE_MAP
+    base_model_prefix = "longformer"
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.longformer = LongformerModel(config)
+        self.classifier = LongformerClassificationHead(config)
+
+    @add_start_docstrings_to_callable(LONGFORMER_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        inputs_embeds=None,
+        labels=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`, defaults to :obj:`None`):
+            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).
+
+    Returns:
+        :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.LongformerConfig`) and inputs:
+        loss (:obj:`torch.FloatTensor` of shape :obj:`(1,)`, `optional`, returned when :obj:`label` is provided):
+            Classification (or regression if config.num_labels==1) loss.
+        logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, config.num_labels)`):
+            Classification (or regression if config.num_labels==1) scores (before SoftMax).
+        hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_hidden_states=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
+            of shape :obj:`(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the model at the output of each layer plus the initial embedding outputs.
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``config.output_attentions=True``):
+            Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
+            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+
+    Examples::
+
+        from transformers import LongformerTokenizer, LongformerForSequenceClassification
+        import torch
+
+        tokenizer = LongformerTokenizer.from_pretrained('longformer-base-4096')
+        model = LongformerForSequenceClassification.from_pretrained('longformer-base-4096')
+        input_ids = torch.tensor(tokenizer.encode("Hello, my dog is cute", add_special_tokens=True)).unsqueeze(0)  # Batch size 1
+        labels = torch.tensor([1]).unsqueeze(0)  # Batch size 1
+        outputs = model(input_ids, labels=labels)
+        loss, logits = outputs[:2]
+
+        """
+
+        if attention_mask is None:
+            attention_mask = torch.ones_like(input_ids)
+
+        # global attention on cls token
+        attention_mask[:, 0] = 2
+
+        outputs = self.longformer(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            inputs_embeds=inputs_embeds,
+        )
+        sequence_output = outputs[0]
+        logits = self.classifier(sequence_output)
+
+        outputs = (logits,) + outputs[2:]
+        if labels is not None:
+            if self.num_labels == 1:
+                #  We are doing regression
+                loss_fct = MSELoss()
+                loss = loss_fct(logits.view(-1), labels.view(-1))
+            else:
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
+            outputs = (loss,) + outputs
+
+        return outputs  # (loss), logits, (hidden_states), (attentions)
+
+
+class LongformerClassificationHead(nn.Module):
+    """Head for sentence-level classification tasks."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.dense = nn.Linear(config.hidden_size, config.hidden_size)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        self.out_proj = nn.Linear(config.hidden_size, config.num_labels)
+
+    def forward(self, hidden_states, **kwargs):
+        hidden_states = hidden_states[:, 0, :]  # take <s> token (equiv. to [CLS])
+        hidden_states = self.dropout(hidden_states)
+        hidden_states = self.dense(hidden_states)
+        hidden_states = torch.tanh(hidden_states)
+        hidden_states = self.dropout(hidden_states)
+        output = self.out_proj(hidden_states)
+        return output
+
+
 @add_start_docstrings(
     """Longformer Model with a span classification head on top for extractive question-answering tasks like SQuAD / TriviaQA (a linear layers on top of
     the hidden-states output to compute `span start logits` and `span end logits`). """,

tests/test_modeling_longformer.py

@@ -29,6 +29,7 @@ if is_torch_available():
         LongformerConfig,
         LongformerModel,
         LongformerForMaskedLM,
+        LongformerForSequenceClassification,
         LongformerForQuestionAnswering,
     )
 
@@ -194,6 +195,23 @@ class LongformerModelTester(object):
         self.parent.assertListEqual(list(result["end_logits"].size()), [self.batch_size, self.seq_length])
         self.check_loss_output(result)
 
+    def create_and_check_longformer_for_sequence_classification(
+        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        config.num_labels = self.num_labels
+        model = LongformerForSequenceClassification(config)
+        model.to(torch_device)
+        model.eval()
+        loss, logits = model(
+            input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels
+        )
+        result = {
+            "loss": loss,
+            "logits": logits,
+        }
+        self.parent.assertListEqual(list(result["logits"].size()), [self.batch_size, self.num_labels])
+        self.check_loss_output(result)
+
     def prepare_config_and_inputs_for_common(self):
         config_and_inputs = self.prepare_config_and_inputs()
         (
@@ -256,6 +274,10 @@ class LongformerModelTest(ModelTesterMixin, unittest.TestCase):
         config_and_inputs = self.model_tester.prepare_config_and_inputs_for_question_answering()
         self.model_tester.create_and_check_longformer_for_question_answering(*config_and_inputs)
 
+    def test_for_sequence_classification(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_longformer_for_sequence_classification(*config_and_inputs)
+
 
 class LongformerModelIntegrationTest(unittest.TestCase):
     @slow