Fix doc mistakes (#14874)

* Remove double returns

* Last fixes

* Quality

* Last fix for Lxmert

src/transformers/models/lxmert/modeling_lxmert.py

@@ -823,12 +823,12 @@ LXMERT_INPUTS_DOCSTRING = r"""
             details.
 
             [What are input IDs?](../glossary#input-ids)
-        visual_feats: (`torch.FloatTensor` of shape :obj:՝(batch_size, num_visual_features, visual_feat_dim)՝):
+        visual_feats: (`torch.FloatTensor` of shape `(batch_size, num_visual_features, visual_feat_dim)`):
             This input represents visual features. They ROI pooled object features from bounding boxes using a
             faster-RCNN model)
 
             These are currently not provided by the transformers library.
-        visual_pos: (`torch.FloatTensor` of shape :obj:՝(batch_size, num_visual_features, visual_pos_dim)՝):
+        visual_pos: (`torch.FloatTensor` of shape `(batch_size, num_visual_features, visual_pos_dim)`):
             This input represents spacial features corresponding to their relative (via index) visual features. The
             pre-trained LXMERT model expects these spacial features to be normalized bounding boxes on a scale of 0 to
             1.
@@ -1176,7 +1176,7 @@ class LxmertForPreTraining(LxmertPreTrainedModel):
 
             - 0 indicates that the sentence does not match the image,
             - 1 indicates that the sentence does match the image.
-        ans: (`Torch.Tensor` of shape `(batch_size)`, *optional*):
+        ans (`Torch.Tensor` of shape `(batch_size)`, *optional*):
             a one hot representation hof the correct answer *optional*
 
         Returns:
@@ -1397,8 +1397,6 @@ class LxmertForQuestionAnswering(LxmertPreTrainedModel):
         r"""
         labels: (`Torch.Tensor` of shape `(batch_size)`, *optional*):
             A one-hot representation of the correct answer
-
-        Returns:
         """
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 

src/transformers/models/lxmert/modeling_tf_lxmert.py

@@ -878,12 +878,12 @@ LXMERT_INPUTS_DOCSTRING = r"""
             details.
 
             [What are input IDs?](../glossary#input-ids)
-        visual_feats: (`tf.Tensor` of shape :obj:՝(batch_size, num_visual_features, visual_feat_dim)՝):
+        visual_feats: (`tf.Tensor` of shape `(batch_size, num_visual_features, visual_feat_dim)`):
             This input represents visual features. They ROI pooled object features from bounding boxes using a
             faster-RCNN model)
 
             These are currently not provided by the transformers library.
-        visual_pos: (`tf.Tensor` of shape :obj:՝(batch_size, num_visual_features, visual_feat_dim)՝):
+        visual_pos: (`tf.Tensor` of shape `(batch_size, num_visual_features, visual_feat_dim)`):
             This input represents spacial features corresponding to their relative (via index) visual features. The
             pre-trained LXMERT model expects these spacial features to be normalized bounding boxes on a scale of 0 to
             1.
@@ -1327,7 +1327,7 @@ class TFLxmertForPreTraining(TFLxmertPreTrainedModel):
 
             - 0 indicates that the sentence does not match the image,
             - 1 indicates that the sentence does match the image.
-        ans: (`Torch.Tensor` of shape `(batch_size)`, *optional*, defaults to :obj: *None*):
+        ans (`Torch.Tensor` of shape `(batch_size)`, *optional*, defaults to :obj: *None*):
             a one hot representation hof the correct answer *optional*
 
         Returns:

src/transformers/models/perceiver/modeling_perceiver.py

@@ -30,7 +30,6 @@ from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
 from ...activations import ACT2FN
 from ...file_utils import (
     ModelOutput,
-    add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     replace_return_docstrings,
@@ -1087,12 +1086,7 @@ class PerceiverForSequenceClassification(PerceiverPreTrainedModel):
         self.post_init()
 
     @add_start_docstrings_to_model_forward(PERCEIVER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
-    @add_code_sample_docstrings(
-        processor_class=_TOKENIZER_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=PerceiverClassifierOutput,
-        config_class=_CONFIG_FOR_DOC,
-    )
+    @replace_return_docstrings(output_type=PerceiverClassifierOutput, config_class=_CONFIG_FOR_DOC)
     def forward(
         self,
         inputs=None,

src/transformers/models/tapas/modeling_tf_tapas.py

@@ -25,10 +25,10 @@ import tensorflow as tf
 from ...activations_tf import get_tf_activation
 from ...file_utils import (
     ModelOutput,
-    add_code_sample_docstrings,
     add_start_docstrings,
     add_start_docstrings_to_model_forward,
     is_tensorflow_probability_available,
+    replace_return_docstrings,
     requires_backends,
 )
 from ...modeling_tf_outputs import (
@@ -981,12 +981,7 @@ class TFTapasModel(TFTapasPreTrainedModel):
         self.tapas = TFTapasMainLayer(config, name="tapas")
 
     @add_start_docstrings_to_model_forward(TAPAS_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
-    @add_code_sample_docstrings(
-        processor_class=_TOKENIZER_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=TFBaseModelOutputWithPooling,
-        config_class=_CONFIG_FOR_DOC,
-    )
+    @replace_return_docstrings(output_type=TFBaseModelOutputWithPooling, config_class=_CONFIG_FOR_DOC)
     def call(
         self,
         input_ids: Optional[TFModelInputType] = None,
@@ -1085,12 +1080,7 @@ class TFTapasForMaskedLM(TFTapasPreTrainedModel, TFMaskedLanguageModelingLoss):
         return self.lm_head.predictions
 
     @add_start_docstrings_to_model_forward(TAPAS_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
-    @add_code_sample_docstrings(
-        processor_class=_TOKENIZER_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=TFMaskedLMOutput,
-        config_class=_CONFIG_FOR_DOC,
-    )
+    @replace_return_docstrings(output_type=TFMaskedLMOutput, config_class=_CONFIG_FOR_DOC)
     def call(
         self,
         input_ids: Optional[TFModelInputType] = None,
@@ -1314,12 +1304,7 @@ class TFTapasForQuestionAnswering(TFTapasPreTrainedModel):
         self.config = config
 
     @add_start_docstrings_to_model_forward(TAPAS_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
-    @add_code_sample_docstrings(
-        processor_class=_TOKENIZER_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=TFTableQuestionAnsweringOutput,
-        config_class=_CONFIG_FOR_DOC,
-    )
+    @replace_return_docstrings(output_type=TFTableQuestionAnsweringOutput, config_class=_CONFIG_FOR_DOC)
     def call(
         self,
         input_ids: Optional[TFModelInputType] = None,
@@ -1664,12 +1649,7 @@ class TFTapasForSequenceClassification(TFTapasPreTrainedModel, TFSequenceClassif
         )
 
     @add_start_docstrings_to_model_forward(TAPAS_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length"))
-    @add_code_sample_docstrings(
-        processor_class=_TOKENIZER_FOR_DOC,
-        checkpoint=_CHECKPOINT_FOR_DOC,
-        output_type=TFSequenceClassifierOutput,
-        config_class=_CONFIG_FOR_DOC,
-    )
+    @replace_return_docstrings(output_type=TFSequenceClassifierOutput, config_class=_CONFIG_FOR_DOC)
     def call(
         self,
         input_ids: Optional[TFModelInputType] = None,

src/transformers/models/visual_bert/modeling_visual_bert.py

@@ -716,6 +716,7 @@ class VisualBertModel(VisualBertPreTrainedModel):
             self.encoder.layer[layer].attention.prune_heads(heads)
 
     @add_start_docstrings_to_model_forward(VISUAL_BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
+    @replace_return_docstrings(output_type=BaseModelOutputWithPooling, config_class=_CONFIG_FOR_DOC)
     def forward(
         self,
         input_ids=None,