Add TF auto model to the docs + fix sphinx warnings (#5187)

src/transformers/modeling_tf_distilbert.py

@@ -581,17 +581,17 @@ class TFDistilBertModel(TFDistilBertPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers,DistilBertConfig`) and inputs:
         last_hidden_state (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
             Sequence of hidden-states at the output of the last layer of the model.
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -654,17 +654,17 @@ class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers,DistilBertConfig`) and inputs:
         prediction_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -735,17 +735,17 @@ class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSeque
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers,DistilBertConfig`) and inputs:
         logits (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, config.num_labels)`):
             Classification (or regression if config.num_labels==1) scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -822,17 +822,17 @@ class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenCla
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers,DistilBertConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -925,17 +925,17 @@ class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoic
             `num_choices` is the size of the second dimension of the input tensors. (see `input_ids` above).
 
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1052,17 +1052,17 @@ class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAn
             Span-start scores (before SoftMax).
         end_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length,)`):
             Span-end scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 

src/transformers/modeling_tf_electra.py

@@ -389,17 +389,17 @@ class TFElectraModel(TFElectraPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.ElectraConfig`) and inputs:
         last_hidden_state (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
             Sequence of hidden-states at the output of the last layer of the model.
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -449,17 +449,17 @@ class TFElectraForPreTraining(TFElectraPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.ElectraConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Prediction scores of the head (scores for each token before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -549,17 +549,17 @@ class TFElectraForMaskedLM(TFElectraPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.ElectraConfig`) and inputs:
         prediction_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -633,17 +633,17 @@ class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassific
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.ElectraConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -732,17 +732,17 @@ class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnswerin
             Span-start scores (before SoftMax).
         end_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length,)`):
             Span-end scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 

src/transformers/modeling_tf_gpt2.py

@@ -497,14 +497,14 @@ class TFGPT2Model(TFGPT2PreTrainedModel):
             Contains pre-computed hidden-states (key and values in the attention blocks).
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
-        hidden_states (:obj:`tuple(tf.Tensor)` `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -557,13 +557,13 @@ class TFGPT2LMHeadModel(TFGPT2PreTrainedModel):
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -643,18 +643,17 @@ class TFGPT2DoubleHeadsModel(TFGPT2PreTrainedModel):
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as `input_ids` as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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::
 
         # For example purposes. Not runnable.

src/transformers/modeling_tf_mobilebert.py

@@ -858,16 +858,17 @@ class TFMobileBertModel(TFMobileBertPreTrainedModel):
             objective during the original Bert pretraining. This output is usually *not* a good summary
             of the semantic content of the input, you're often better with averaging or pooling
             the sequence of hidden-states for the whole input sequence.
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when :obj:`config.output_hidden_states=True`):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+            :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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
 
     Examples::
@@ -908,16 +909,17 @@ class TFMobileBertForPreTraining(TFMobileBertPreTrainedModel):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
         seq_relationship_scores (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, 2)`):
             Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when :obj:`config.output_hidden_states=True`):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -960,16 +962,17 @@ class TFMobileBertForMaskedLM(TFMobileBertPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.MobileBertConfig`) and inputs:
         prediction_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when :obj:`config.output_hidden_states=True`):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1021,16 +1024,17 @@ class TFMobileBertForNextSentencePrediction(TFMobileBertPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.MobileBertConfig`) and inputs:
         seq_relationship_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, 2)`)
             Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when :obj:`config.output_hidden_states=True`):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1098,16 +1102,17 @@ class TFMobileBertForSequenceClassification(TFMobileBertPreTrainedModel, TFSeque
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.MobileBertConfig`) and inputs:
         logits (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, config.num_labels)`):
             Classification (or regression if config.num_labels==1) scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1198,16 +1203,17 @@ class TFMobileBertForQuestionAnswering(TFMobileBertPreTrainedModel, TFQuestionAn
             Span-start scores (before SoftMax).
         end_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length,)`):
             Span-end scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1305,16 +1311,17 @@ class TFMobileBertForMultipleChoice(TFMobileBertPreTrainedModel, TFMultipleChoic
             `num_choices` is the size of the second dimension of the input tensors. (see `input_ids` above).
 
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1439,16 +1446,17 @@ class TFMobileBertForTokenClassification(TFMobileBertPreTrainedModel, TFTokenCla
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.MobileBertConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` or ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 

src/transformers/modeling_tf_openai.py

@@ -455,14 +455,14 @@ class TFOpenAIGPTModel(TFOpenAIGPTPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.OpenAIGPTConfig`) and inputs:
         last_hidden_state (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`):
             Sequence of hidden-states at the last layer of the model.
-        hidden_states (:obj:`tuple(tf.Tensor)` `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -504,13 +504,13 @@ class TFOpenAIGPTLMHeadModel(TFOpenAIGPTPreTrainedModel):
         prediction_scores (:obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -587,13 +587,13 @@ class TFOpenAIGPTDoubleHeadsModel(TFOpenAIGPTPreTrainedModel):
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.

src/transformers/modeling_tf_roberta.py

@@ -207,17 +207,17 @@ class TFRobertaModel(TFRobertaPreTrainedModel):
             objective during Bert pretraining. This output is usually *not* a good summary
             of the semantic content of the input, you're often better with averaging or pooling
             the sequence of hidden-states for the whole input sequence.
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -284,17 +284,17 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel):
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.RobertaConfig`) and inputs:
         prediction_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -375,17 +375,17 @@ class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceCla
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.RobertaConfig`) and inputs:
         logits (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, config.num_labels)`):
             Classification (or regression if config.num_labels==1) scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -474,17 +474,17 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
             `num_choices` is the size of the second dimension of the input tensors. (see `input_ids` above).
 
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -600,17 +600,17 @@ class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassific
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.RobertaConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -699,17 +699,17 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
             Span-start scores (before SoftMax).
         end_scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length,)`):
             Span-end scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 

src/transformers/modeling_tf_t5.py

@@ -915,13 +915,13 @@ class TFT5Model(TFT5PreTrainedModel):
             Can be used to speed up sequential decoding (see `decoder_past_key_value_states` input).
             Note that when using `decoder_past_key_value_states`, the model only outputs the last `hidden-state` of the sequence of shape :obj:`(batch_size, 1, config.vocab_size)`.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-                :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -1055,15 +1055,16 @@ class TFT5ForConditionalGeneration(TFT5PreTrainedModel):
             Can be used to speed up sequential decoding (see `decoder_past_key_value_states` input).
             Note that when using `decoder_past_key_value_states`, the model only outputs the last `prediction_score` of the sequence of shape :obj:`(batch_size, 1, config.vocab_size)`.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 

src/transformers/modeling_tf_transfo_xl.py

@@ -733,13 +733,13 @@ class TFTransfoXLModel(TFTransfoXLPreTrainedModel):
             Can be used (see `mems` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -832,13 +832,13 @@ class TFTransfoXLLMHeadModel(TFTransfoXLPreTrainedModel):
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.

src/transformers/modeling_tf_xlm.py

@@ -711,13 +711,13 @@ class TFXLMWithLMHeadModel(TFXLMPreTrainedModel):
         prediction_scores (:obj:`tf.Tensor` or :obj:`Numpy array` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`):
             Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -785,13 +785,13 @@ class TFXLMForSequenceClassification(TFXLMPreTrainedModel, TFSequenceClassificat
         logits (:obj:`tf.Tensor` or :obj:`Numpy array` of shape :obj:`(batch_size, config.num_labels)`):
             Classification (or regression if config.num_labels==1) scores (before SoftMax).
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -886,17 +886,17 @@ class TFXLMForMultipleChoice(TFXLMPreTrainedModel, TFMultipleChoiceLoss):
             `num_choices` is the size of the second dimension of the input tensors. (see `input_ids` above).
 
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1023,17 +1023,17 @@ class TFXLMForTokenClassification(TFXLMPreTrainedModel, TFTokenClassificationLos
         :obj:`tuple(tf.Tensor)` comprising various elements depending on the configuration (:class:`~transformers.BertConfig`) and inputs:
         scores (:obj:`Numpy array` or :obj:`tf.Tensor` of shape :obj:`(batch_size, sequence_length, config.num_labels)`):
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1125,13 +1125,13 @@ class TFXLMForQuestionAnsweringSimple(TFXLMPreTrainedModel, TFQuestionAnsweringL
         end_scores (:obj:`tf.Tensor` or :obj:`Numpy array` of shape :obj:`(batch_size, sequence_length,)`):
             Span-end scores (before SoftMax).
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.

src/transformers/modeling_tf_xlnet.py

@@ -927,13 +927,13 @@ class TFXLNetLMHeadModel(TFXLNetPreTrainedModel):
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -1019,13 +1019,13 @@ class TFXLNetForSequenceClassification(TFXLNetPreTrainedModel, TFSequenceClassif
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -1127,17 +1127,17 @@ class TFXLNetForMultipleChoice(TFXLNetPreTrainedModel, TFMultipleChoiceLoss):
             `num_choices` is the size of the second dimension of the input tensors. (see `input_ids` above).
 
             Classification scores (before SoftMax).
-        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when
-        ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
+        hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
             tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
             tuple of :obj:`tf.Tensor` (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.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
 
     Examples::
 
@@ -1274,13 +1274,13 @@ class TFXLNetForTokenClassification(TFXLNetPreTrainedModel, TFTokenClassificatio
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.
@@ -1384,13 +1384,13 @@ class TFXLNetForQuestionAnsweringSimple(TFXLNetPreTrainedModel, TFQuestionAnswer
             Can be used (see `past` input) to speed up sequential decoding. The token ids which have their past given to this model
             should not be passed as input ids as they have already been computed.
         hidden_states (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for the output of the embeddings + one for the output of each layer)
+            tuple of :obj:`tf.Tensor` (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(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
-            Tuple of :obj:`tf.Tensor` or :obj:`Numpy array` (one for each layer) of shape
-            :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
+        attentions (:obj:`tuple(tf.Tensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``):
+            tuple of :obj:`tf.Tensor` (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.

src/transformers/modeling_transfo_xl.py

@@ -772,7 +772,7 @@ class TransfoXLModel(TransfoXLPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -978,7 +978,7 @@ class TransfoXLLMHeadModel(TransfoXLPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 

src/transformers/modeling_xlm.py

@@ -419,7 +419,7 @@ class XLMModel(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -666,7 +666,7 @@ class XLMWithLMHeadModel(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -755,7 +755,7 @@ class XLMForSequenceClassification(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -861,7 +861,7 @@ class XLMForQuestionAnsweringSimple(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -999,7 +999,7 @@ class XLMForQuestionAnswering(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1095,7 +1095,7 @@ class XLMForTokenClassification(XLMPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 

src/transformers/modeling_xlnet.py

@@ -779,7 +779,7 @@ class XLNetModel(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1081,7 +1081,7 @@ class XLNetLMHeadModel(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1202,7 +1202,7 @@ class XLNetForSequenceClassification(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1310,7 +1310,7 @@ class XLNetForTokenClassification(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1425,7 +1425,7 @@ class XLNetForMultipleChoice(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1552,7 +1552,7 @@ class XLNetForQuestionAnsweringSimple(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.
 
@@ -1698,7 +1698,7 @@ class XLNetForQuestionAnswering(XLNetPreTrainedModel):
             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 ``output_attentions=True`` is passed or ``config.output_attentions=True``):
+        attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or 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)`.