Add AlbertForMultipleChoice (#4959)

* Add AlbertForMultipleChoice

* Make up to date and add all models to common tests

docs/source/model_doc/albert.rst

@@ -67,6 +67,14 @@ AlbertForSequenceClassification
 .. autoclass:: transformers.AlbertForSequenceClassification
     :members:
 
+
+AlbertForMultipleChoice
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.AlbertForMultipleChoice
+    :members:
+
+
 AlbertForTokenClassification
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 

src/transformers/__init__.py

@@ -294,6 +294,7 @@ if is_torch_available():
         AlbertModel,
         AlbertForPreTraining,
         AlbertForMaskedLM,
+        AlbertForMultipleChoice,
         AlbertForSequenceClassification,
         AlbertForQuestionAnswering,
         AlbertForTokenClassification,

src/transformers/modeling_albert.py

@@ -1135,3 +1135,112 @@ class AlbertForQuestionAnswering(AlbertPreTrainedModel):
             outputs = (total_loss,) + outputs
 
         return outputs  # (loss), start_logits, end_logits, (hidden_states), (attentions)
+
+
+@add_start_docstrings(
+    """Albert Model with a multiple choice classification head on top (a linear layer on top of
+    the pooled output and a softmax) e.g. for RocStories/SWAG tasks. """,
+    ALBERT_START_DOCSTRING,
+)
+class AlbertForMultipleChoice(AlbertPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+
+        self.albert = AlbertModel(config)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob)
+        self.classifier = nn.Linear(config.hidden_size, 1)
+
+        self.init_weights()
+
+    @add_start_docstrings_to_callable(ALBERT_INPUTS_DOCSTRING.format("(batch_size, num_choices, sequence_length)"))
+    def forward(
+        self,
+        input_ids=None,
+        attention_mask=None,
+        token_type_ids=None,
+        position_ids=None,
+        head_mask=None,
+        inputs_embeds=None,
+        labels=None,
+        output_attentions=None,
+    ):
+        r"""
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`, defaults to :obj:`None`):
+            Labels for computing the multiple choice classification loss.
+            Indices should be in ``[0, ..., num_choices-1]`` where `num_choices` is the size of the second dimension
+            of the input tensors. (see `input_ids` above)
+
+    Returns:
+        :obj:`tuple(torch.FloatTensor)` comprising various elements depending on the configuration (:class:`~transformers.BertConfig`) and inputs:
+        loss (:obj:`torch.FloatTensor` of shape `(1,)`, `optional`, returned when :obj:`labels` is provided):
+            Classification loss.
+        classification_scores (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, num_choices)`):
+            `num_choices` is the second dimension of the input tensors. (see `input_ids` above).
+
+            Classification 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 AlbertTokenizer, AlbertForMultipleChoice
+        import torch
+
+        tokenizer = AlbertTokenizer.from_pretrained('albert-base-v2')
+        model = AlbertForMultipleChoice.from_pretrained('albert-base-v2')
+
+        prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
+        choice0 = "It is eaten with a fork and a knife."
+        choice1 = "It is eaten while held in the hand."
+        labels = torch.tensor(0).unsqueeze(0)  # choice0 is correct (according to Wikipedia ;)), batch size 1
+
+        encoding = tokenizer.batch_encode_plus([[prompt, choice0], [prompt, choice1]], return_tensors='pt', pad_to_max_length=True)
+        outputs = model(**{k: v.unsqueeze(0) for k,v in encoding.items()}, labels=labels) # batch size is 1
+
+        # the linear classifier still needs to be trained
+        loss, logits = outputs[:2]
+        """
+        num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1]
+
+        input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None
+        attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None
+        token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None
+        position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None
+        inputs_embeds = (
+            inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1))
+            if inputs_embeds is not None
+            else None
+        )
+        outputs = self.albert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+        )
+
+        pooled_output = outputs[1]
+
+        pooled_output = self.dropout(pooled_output)
+        logits = self.classifier(pooled_output)
+        reshaped_logits = logits.view(-1, num_choices)
+
+        outputs = (reshaped_logits,) + outputs[2:]  # add hidden states and attention if they are here
+
+        if labels is not None:
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(reshaped_logits, labels)
+            outputs = (loss,) + outputs
+
+        return outputs  # (loss), reshaped_logits, (hidden_states), (attentions)

src/transformers/modeling_auto.py

@@ -45,6 +45,7 @@ from .configuration_marian import MarianConfig
 from .configuration_utils import PretrainedConfig
 from .modeling_albert import (
     AlbertForMaskedLM,
+    AlbertForMultipleChoice,
     AlbertForPreTraining,
     AlbertForQuestionAnswering,
     AlbertForSequenceClassification,
@@ -308,6 +309,7 @@ MODEL_FOR_MULTIPLE_CHOICE_MAPPING = OrderedDict(
         (RobertaConfig, RobertaForMultipleChoice),
         (BertConfig, BertForMultipleChoice),
         (XLNetConfig, XLNetForMultipleChoice),
+        (AlbertConfig, AlbertForMultipleChoice),
     ]
 )
 

src/transformers/modeling_bert.py

@@ -1354,7 +1354,7 @@ class BertForMultipleChoice(BertPreTrainedModel):
         prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
         choice0 = "It is eaten with a fork and a knife."
         choice1 = "It is eaten while held in the hand."
-        labels = torch.tensor(0) # choice0 is correct (according to Wikipedia ;))
+        labels = torch.tensor(0).unsqueeze(0)  # choice0 is correct (according to Wikipedia ;)), batch size 1
 
         encoding = tokenizer.batch_encode_plus([[prompt, choice0], [prompt, choice1]], return_tensors='pt', pad_to_max_length=True)
         outputs = model(**{k: v.unsqueeze(0) for k,v in encoding.items()}, labels=labels) # batch size is 1

tests/test_modeling_albert.py

@@ -29,6 +29,7 @@ if is_torch_available():
         AlbertModel,
         AlbertForPreTraining,
         AlbertForMaskedLM,
+        AlbertForMultipleChoice,
         AlbertForSequenceClassification,
         AlbertForTokenClassification,
         AlbertForQuestionAnswering,
@@ -39,7 +40,19 @@ if is_torch_available():
 @require_torch
 class AlbertModelTest(ModelTesterMixin, unittest.TestCase):
 
-    all_model_classes = (AlbertModel, AlbertForPreTraining, AlbertForMaskedLM) if is_torch_available() else ()
+    all_model_classes = (
+        (
+            AlbertModel,
+            AlbertForPreTraining,
+            AlbertForMaskedLM,
+            AlbertForMultipleChoice,
+            AlbertForSequenceClassification,
+            AlbertForTokenClassification,
+            AlbertForQuestionAnswering,
+        )
+        if is_torch_available()
+        else ()
+    )
 
     class AlbertModelTester(object):
         def __init__(
@@ -252,6 +265,29 @@ class AlbertModelTest(ModelTesterMixin, unittest.TestCase):
             )
             self.check_loss_output(result)
 
+        def create_and_check_albert_for_multiple_choice(
+            self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+        ):
+            config.num_choices = self.num_choices
+            model = AlbertForMultipleChoice(config=config)
+            model.to(torch_device)
+            model.eval()
+            multiple_choice_inputs_ids = input_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+            multiple_choice_token_type_ids = token_type_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+            multiple_choice_input_mask = input_mask.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
+            loss, logits = model(
+                multiple_choice_inputs_ids,
+                attention_mask=multiple_choice_input_mask,
+                token_type_ids=multiple_choice_token_type_ids,
+                labels=choice_labels,
+            )
+            result = {
+                "loss": loss,
+                "logits": logits,
+            }
+            self.parent.assertListEqual(list(result["logits"].size()), [self.batch_size, self.num_choices])
+            self.check_loss_output(result)
+
         def prepare_config_and_inputs_for_common(self):
             config_and_inputs = self.prepare_config_and_inputs()
             (
@@ -285,6 +321,10 @@ class AlbertModelTest(ModelTesterMixin, unittest.TestCase):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_albert_for_masked_lm(*config_and_inputs)
 
+    def test_for_multiple_choice(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_albert_for_multiple_choice(*config_and_inputs)
+
     def test_for_question_answering(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_albert_for_question_answering(*config_and_inputs)