[Tests] Add Common Test for Training + Fix a couple of bugs

docs/source/model_doc/auto.rst

@@ -81,6 +81,13 @@ AutoModelForMultipleChoice
     :members:
 
 
+AutoModelForNextSentencePrediction
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: transformers.AutoModelForNextSentencePrediction
+    :members:
+
+
 AutoModelForTokenClassification
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 

examples/lxmert/modeling_frcnn.py

@@ -1801,7 +1801,7 @@ def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
                 f"Some weights of the model checkpoint at {pretrained_model_name_or_path} were not used when "
                 f"initializing {model.__class__.__name__}: {unexpected_keys}\n"
                 f"- This IS expected if you are initializing {model.__class__.__name__} from the checkpoint of a model trained on another task "
-                f"or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPretraining model).\n"
+                f"or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n"
                 f"- This IS NOT expected if you are initializing {model.__class__.__name__} from the checkpoint of a model that you expect "
                 f"to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model)."
             )

model_cards/neuralmind/bert-base-portuguese-cased/README.md

@@ -29,7 +29,7 @@ For further information or requests, please go to [BERTimbau repository](https:/
 
 ```python
 from transformers import AutoTokenizer  # Or BertTokenizer
-from transformers import AutoModelForPretraining  # Or BertForPreTraining for loading pretraining heads
+from transformers import AutoModelForPreTraining  # Or BertForPreTraining for loading pretraining heads
 from transformers import AutoModel  # or BertModel, for BERT without pretraining heads
 
 model = AutoModelForPreTraining.from_pretrained('neuralmind/bert-base-portuguese-cased')

src/transformers/__init__.py

@@ -329,6 +329,7 @@
         MODEL_FOR_CAUSAL_LM_MAPPING,
         MODEL_FOR_MASKED_LM_MAPPING,
         MODEL_FOR_MULTIPLE_CHOICE_MAPPING,
+        MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING,
         MODEL_FOR_PRETRAINING_MAPPING,
         MODEL_FOR_QUESTION_ANSWERING_MAPPING,
         MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING,
@@ -340,6 +341,7 @@
         AutoModelForCausalLM,
         AutoModelForMaskedLM,
         AutoModelForMultipleChoice,
+        AutoModelForNextSentencePrediction,
         AutoModelForPreTraining,
         AutoModelForQuestionAnswering,
         AutoModelForSeq2SeqLM,

src/transformers/modeling_albert.py

@@ -738,7 +738,7 @@ def forward(
         head_mask=None,
         inputs_embeds=None,
         labels=None,
-        sentence_order_label=None,
+        next_sentence_label=None,
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
@@ -749,7 +749,7 @@ def forward(
             Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ...,
             config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored
             (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]``
-        sentence_order_label (``torch.LongTensor`` of shape ``(batch_size,)``, `optional`):
+        next_sentence_label (``torch.LongTensor`` of shape ``(batch_size,)``, `optional`):
             Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair
             (see :obj:`input_ids` docstring) Indices should be in ``[0, 1]``. ``0`` indicates original order (sequence
             A, then sequence B), ``1`` indicates switched order (sequence B, then sequence A).
@@ -780,6 +780,12 @@ def forward(
                 FutureWarning,
             )
             labels = kwargs.pop("masked_lm_labels")
+        if "sentence_order_label" in kwargs:
+            warnings.warn(
+                "The `sentence_order_label` argument is deprecated and will be removed in a future version, use `next_sentence_label` instead.",
+                FutureWarning,
+            )
+            next_sentence_label = kwargs.pop("sentence_order_label")
         assert kwargs == {}, f"Unexpected keyword arguments: {list(kwargs.keys())}."
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
@@ -801,10 +807,10 @@ def forward(
         sop_scores = self.sop_classifier(pooled_output)
 
         total_loss = None
-        if labels is not None and sentence_order_label is not None:
+        if labels is not None and next_sentence_label is not None:
             loss_fct = CrossEntropyLoss()
             masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1))
-            sentence_order_loss = loss_fct(sop_scores.view(-1, 2), sentence_order_label.view(-1))
+            sentence_order_loss = loss_fct(sop_scores.view(-1, 2), next_sentence_label.view(-1))
             total_loss = masked_lm_loss + sentence_order_loss
 
         if not return_dict:

src/transformers/modeling_auto.py

@@ -77,6 +77,7 @@
 from .modeling_bert import (
     BertForMaskedLM,
     BertForMultipleChoice,
+    BertForNextSentencePrediction,
     BertForPreTraining,
     BertForQuestionAnswering,
     BertForSequenceClassification,
@@ -128,6 +129,7 @@
 from .modeling_funnel import (
     FunnelForMaskedLM,
     FunnelForMultipleChoice,
+    FunnelForPreTraining,
     FunnelForQuestionAnswering,
     FunnelForSequenceClassification,
     FunnelForTokenClassification,
@@ -143,12 +145,13 @@
     LongformerForTokenClassification,
     LongformerModel,
 )
-from .modeling_lxmert import LxmertForPreTraining, LxmertModel
+from .modeling_lxmert import LxmertForPreTraining, LxmertForQuestionAnswering, LxmertModel
 from .modeling_marian import MarianMTModel
 from .modeling_mbart import MBartForConditionalGeneration
 from .modeling_mobilebert import (
     MobileBertForMaskedLM,
     MobileBertForMultipleChoice,
+    MobileBertForNextSentencePrediction,
     MobileBertForPreTraining,
     MobileBertForQuestionAnswering,
     MobileBertForSequenceClassification,
@@ -166,6 +169,7 @@
 from .modeling_reformer import (
     ReformerForMaskedLM,
     ReformerForQuestionAnswering,
+    ReformerForSequenceClassification,
     ReformerModel,
     ReformerModelWithLMHead,
 )
@@ -285,6 +289,7 @@
         (CTRLConfig, CTRLLMHeadModel),
         (ElectraConfig, ElectraForPreTraining),
         (LxmertConfig, LxmertForPreTraining),
+        (FunnelConfig, FunnelForPreTraining),
     ]
 )
 
@@ -396,6 +401,7 @@
         (DebertaConfig, DebertaForSequenceClassification),
         (GPT2Config, GPT2ForSequenceClassification),
         (OpenAIGPTConfig, OpenAIGPTForSequenceClassification),
+        (ReformerConfig, ReformerForSequenceClassification),
     ]
 )
 
@@ -417,6 +423,7 @@
         (ElectraConfig, ElectraForQuestionAnswering),
         (ReformerConfig, ReformerForQuestionAnswering),
         (FunnelConfig, FunnelForQuestionAnswering),
+        (LxmertConfig, LxmertForQuestionAnswering),
     ]
 )
 
@@ -460,6 +467,13 @@
     ]
 )
 
+MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING = OrderedDict(
+    [
+        (BertConfig, BertForNextSentencePrediction),
+        (MobileBertConfig, MobileBertForNextSentencePrediction),
+    ]
+)
+
 AUTO_MODEL_PRETRAINED_DOCSTRING = r"""
 
         The model class to instantiate is selected based on the :obj:`model_type` property of the config object (either
@@ -1519,3 +1533,103 @@ def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
                 ", ".join(c.__name__ for c in MODEL_FOR_MULTIPLE_CHOICE_MAPPING.keys()),
             )
         )
+
+
+class AutoModelForNextSentencePrediction:
+    r"""
+    This is a generic model class that will be instantiated as one of the model classes of the library---with a
+    multiple choice classification head---when created with the when created with the
+    :meth:`~transformers.AutoModelForNextSentencePrediction.from_pretrained` class method or the
+    :meth:`~transformers.AutoModelForNextSentencePrediction.from_config` class method.
+
+    This class cannot be instantiated directly using ``__init__()`` (throws an error).
+    """
+
+    def __init__(self):
+        raise EnvironmentError(
+            "AutoModelForNextSentencePrediction is designed to be instantiated "
+            "using the `AutoModelForNextSentencePrediction.from_pretrained(pretrained_model_name_or_path)` or "
+            "`AutoModelForNextSentencePrediction.from_config(config)` methods."
+        )
+
+    @classmethod
+    @replace_list_option_in_docstrings(MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING, use_model_types=False)
+    def from_config(cls, config):
+        r"""
+        Instantiates one of the model classes of the library---with a multiple choice classification head---from a
+        configuration.
+
+        Note:
+            Loading a model from its configuration file does **not** load the model weights. It only affects the
+            model's configuration. Use :meth:`~transformers.AutoModelForNextSentencePrediction.from_pretrained` to load
+            the model weights.
+
+        Args:
+            config (:class:`~transformers.PretrainedConfig`):
+                The model class to instantiate is selected based on the configuration class:
+
+                List options
+
+        Examples::
+
+            >>> from transformers import AutoConfig, AutoModelForNextSentencePrediction
+            >>> # Download configuration from S3 and cache.
+            >>> config = AutoConfig.from_pretrained('bert-base-uncased')
+            >>> model = AutoModelForNextSentencePrediction.from_config(config)
+        """
+        if type(config) in MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING.keys():
+            return MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING[type(config)](config)
+
+        raise ValueError(
+            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
+            "Model type should be one of {}.".format(
+                config.__class__,
+                cls.__name__,
+                ", ".join(c.__name__ for c in MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING.keys()),
+            )
+        )
+
+    @classmethod
+    @replace_list_option_in_docstrings(MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING)
+    @add_start_docstrings(
+        "Instantiate one of the model classes of the library---with a multiple choice classification head---from a "
+        "pretrained model.",
+        AUTO_MODEL_PRETRAINED_DOCSTRING,
+    )
+    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
+        r"""
+        Examples::
+
+            >>> from transformers import AutoConfig, AutoModelForNextSentencePrediction
+
+            >>> # Download model and configuration from S3 and cache.
+            >>> model = AutoModelForNextSentencePrediction.from_pretrained('bert-base-uncased')
+
+            >>> # Update configuration during loading
+            >>> model = AutoModelForNextSentencePrediction.from_pretrained('bert-base-uncased', output_attentions=True)
+            >>> model.config.output_attentions
+            True
+
+            >>> # Loading from a TF checkpoint file instead of a PyTorch model (slower)
+            >>> config = AutoConfig.from_json_file('./tf_model/bert_tf_model_config.json')
+            >>> model = AutoModelForNextSentencePrediction.from_pretrained('./tf_model/bert_tf_checkpoint.ckpt.index', from_tf=True, config=config)
+        """
+        config = kwargs.pop("config", None)
+        if not isinstance(config, PretrainedConfig):
+            config, kwargs = AutoConfig.from_pretrained(
+                pretrained_model_name_or_path, return_unused_kwargs=True, **kwargs
+            )
+
+        if type(config) in MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING.keys():
+            return MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING[type(config)].from_pretrained(
+                pretrained_model_name_or_path, *model_args, config=config, **kwargs
+            )
+
+        raise ValueError(
+            "Unrecognized configuration class {} for this kind of AutoModel: {}.\n"
+            "Model type should be one of {}.".format(
+                config.__class__,
+                cls.__name__,
+                ", ".join(c.__name__ for c in MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING.keys()),
+            )
+        )

src/transformers/modeling_bert.py

@@ -1228,13 +1228,14 @@ def forward(
         position_ids=None,
         head_mask=None,
         inputs_embeds=None,
-        next_sentence_label=None,
+        labels=None,
         output_attentions=None,
         output_hidden_states=None,
         return_dict=None,
+        **kwargs
     ):
         r"""
-        next_sentence_label (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
+        labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`):
             Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair
             (see ``input_ids`` docstring). Indices should be in ``[0, 1]``:
 
@@ -1255,10 +1256,18 @@ def forward(
             >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light."
             >>> encoding = tokenizer(prompt, next_sentence, return_tensors='pt')
 
-            >>> outputs = model(**encoding, next_sentence_label=torch.LongTensor([1]))
+            >>> outputs = model(**encoding, labels=torch.LongTensor([1]))
             >>> logits = outputs.logits
             >>> assert logits[0, 0] < logits[0, 1] # next sentence was random
         """
+
+        if "next_sentence_label" in kwargs:
+            warnings.warn(
+                "The `next_sentence_label` argument is deprecated and will be removed in a future version, use `labels` instead.",
+                FutureWarning,
+            )
+            labels = kwargs.pop("next_sentence_label")
+
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         outputs = self.bert(
@@ -1278,9 +1287,9 @@ def forward(
         seq_relationship_scores = self.cls(pooled_output)
 
         next_sentence_loss = None
-        if next_sentence_label is not None:
+        if labels is not None:
             loss_fct = CrossEntropyLoss()
-            next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), next_sentence_label.view(-1))
+            next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1))
 
         if not return_dict:
             output = (seq_relationship_scores,) + outputs[2:]

src/transformers/modeling_gpt2.py

@@ -1069,7 +1069,7 @@ def forward(
             if self.num_labels == 1:
                 #  We are doing regression
                 loss_fct = MSELoss()
-                loss = loss_fct(pooled_logits.view(-1), labels.view(-1))
+                loss = loss_fct(pooled_logits.view(-1), labels.to(self.dtype).view(-1))
             else:
                 loss_fct = CrossEntropyLoss()
                 loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))

src/transformers/modeling_longformer.py

@@ -1069,7 +1069,7 @@ def forward(
 
                 def create_custom_forward(module):
                     def custom_forward(*inputs):
-                        return module(*inputs, output_attentions)
+                        return module(*inputs, is_global_attn)
 
                     return custom_forward
 
@@ -1079,7 +1079,6 @@ def custom_forward(*inputs):
                     attention_mask,
                     is_index_masked,
                     is_index_global_attn,
-                    is_global_attn,
                 )
             else:
                 layer_outputs = layer_module(