Skip to content

GitLab

"git@developer.sourcefind.cn:chenpangpang/transformers.git" did not exist on "482c9178d3271bfde17bc0404bf668e330577423"
Unverified Commit 3b3619a3 authored by Patrick von Platen's avatar Patrick von Platen Committed by GitHub
Browse files

[All models] fix docs after adding output attentions to all forward functions (#4909) 

* fix doc

* add format file

* add output attentions to all docs

* add also for bart

* fix naming

* re-add doc to config
parent ac99217e
Hide whitespace changes
Inline Side-by-side
Showing with 108 additions and 68 deletions
src/transformers/configuration_utils.py
View file @ 3b3619a3
...@@ -44,8 +44,10 @@ class PretrainedConfig(object): ...@@ -44,8 +44,10 @@ class PretrainedConfig(object):
Name of the task used to fine-tune the model. This can be used when converting from an original (TensorFlow or PyTorch) checkpoint. Name of the task used to fine-tune the model. This can be used when converting from an original (TensorFlow or PyTorch) checkpoint.
num_labels (:obj:`int`, `optional`, defaults to `2`): num_labels (:obj:`int`, `optional`, defaults to `2`):
Number of classes to use when the model is a classification model (sequences/tokens) Number of classes to use when the model is a classification model (sequences/tokens)
output_hidden_states (:obj:`string`, `optional`, defaults to :obj:`False`): output_hidden_states (:obj:`bool`, `optional`, defaults to :obj:`False`):
Should the model returns all hidden-states. Should the model returns all hidden-states.
output_attentions (:obj:`bool`, `optional`, defaults to :obj:`False`):
Should the model returns all attentions.
torchscript (:obj:`bool`, `optional`, defaults to :obj:`False`): torchscript (:obj:`bool`, `optional`, defaults to :obj:`False`):
Is the model used with Torchscript (for PyTorch models). Is the model used with Torchscript (for PyTorch models).
""" """
......
src/transformers/modeling_albert.py
View file @ 3b3619a3
...@@ -423,6 +423,8 @@ ALBERT_INPUTS_DOCSTRING = r""" ...@@ -423,6 +423,8 @@ ALBERT_INPUTS_DOCSTRING = r"""
Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -507,7 +509,7 @@ class AlbertModel(AlbertPreTrainedModel): ...@@ -507,7 +509,7 @@ class AlbertModel(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -631,7 +633,7 @@ class AlbertForPreTraining(AlbertPreTrainedModel): ...@@ -631,7 +633,7 @@ class AlbertForPreTraining(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -778,7 +780,7 @@ class AlbertForMaskedLM(AlbertPreTrainedModel): ...@@ -778,7 +780,7 @@ class AlbertForMaskedLM(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -873,7 +875,7 @@ class AlbertForSequenceClassification(AlbertPreTrainedModel): ...@@ -873,7 +875,7 @@ class AlbertForSequenceClassification(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -968,7 +970,7 @@ class AlbertForTokenClassification(AlbertPreTrainedModel): ...@@ -968,7 +970,7 @@ class AlbertForTokenClassification(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1074,7 +1076,7 @@ class AlbertForQuestionAnswering(AlbertPreTrainedModel): ...@@ -1074,7 +1076,7 @@ class AlbertForQuestionAnswering(AlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_bart.py
View file @ 3b3619a3
...@@ -89,6 +89,8 @@ BART_INPUTS_DOCSTRING = r""" ...@@ -89,6 +89,8 @@ BART_INPUTS_DOCSTRING = r"""
Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
If you want to change padding behavior, you should read :func:`~transformers.modeling_bart._prepare_decoder_inputs` and modify. If you want to change padding behavior, you should read :func:`~transformers.modeling_bart._prepare_decoder_inputs` and modify.
See diagram 1 in the paper for more info on the default strategy See diagram 1 in the paper for more info on the default strategy
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -928,7 +930,7 @@ class BartForConditionalGeneration(PretrainedBartModel): ...@@ -928,7 +930,7 @@ class BartForConditionalGeneration(PretrainedBartModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1078,7 +1080,7 @@ class BartForSequenceClassification(PretrainedBartModel): ...@@ -1078,7 +1080,7 @@ class BartForSequenceClassification(PretrainedBartModel):
Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape :obj:`(batch_size, num_heads, sequence_length, sequence_length)`. Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
Attentions weights after the attention softmax, used to compute the weighted average in the Attentions weights after the attention softmax, used to compute the weighted average in the
self-attention self-attention
......
src/transformers/modeling_bert.py
View file @ 3b3619a3
...@@ -594,6 +594,8 @@ BERT_INPUTS_DOCSTRING = r""" ...@@ -594,6 +594,8 @@ BERT_INPUTS_DOCSTRING = r"""
is used in the cross-attention if the model is configured as a decoder. is used in the cross-attention if the model is configured as a decoder.
Mask values selected in ``[0, 1]``: Mask values selected in ``[0, 1]``:
``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens. ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -674,7 +676,7 @@ class BertModel(BertPreTrainedModel): ...@@ -674,7 +676,7 @@ class BertModel(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -814,7 +816,7 @@ class BertForPreTraining(BertPreTrainedModel): ...@@ -814,7 +816,7 @@ class BertForPreTraining(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -928,7 +930,7 @@ class BertForMaskedLM(BertPreTrainedModel): ...@@ -928,7 +930,7 @@ class BertForMaskedLM(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1060,7 +1062,7 @@ class BertForNextSentencePrediction(BertPreTrainedModel): ...@@ -1060,7 +1062,7 @@ class BertForNextSentencePrediction(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1152,7 +1154,7 @@ class BertForSequenceClassification(BertPreTrainedModel): ...@@ -1152,7 +1154,7 @@ class BertForSequenceClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1251,7 +1253,7 @@ class BertForMultipleChoice(BertPreTrainedModel): ...@@ -1251,7 +1253,7 @@ class BertForMultipleChoice(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1354,7 +1356,7 @@ class BertForTokenClassification(BertPreTrainedModel): ...@@ -1354,7 +1356,7 @@ class BertForTokenClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1461,7 +1463,7 @@ class BertForQuestionAnswering(BertPreTrainedModel): ...@@ -1461,7 +1463,7 @@ class BertForQuestionAnswering(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_camembert.py
View file @ 3b3619a3
...@@ -49,6 +49,8 @@ CAMEMBERT_START_DOCSTRING = r""" ...@@ -49,6 +49,8 @@ CAMEMBERT_START_DOCSTRING = r"""
model. Initializing with a config file does not load the weights associated with the model, only the model. Initializing with a config file does not load the weights associated with the model, only the
configuration. configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights. Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
......
src/transformers/modeling_ctrl.py
View file @ 3b3619a3
...@@ -266,6 +266,8 @@ CTRL_INPUTS_DOCSTRING = r""" ...@@ -266,6 +266,8 @@ CTRL_INPUTS_DOCSTRING = r"""
use_cache (:obj:`bool`): use_cache (:obj:`bool`):
If `use_cache` is True, `past` key value states are returned and If `use_cache` is True, `past` key value states are returned and
can be used to speed up decoding (see `past`). Defaults to `True`. can be used to speed up decoding (see `past`). Defaults to `True`.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -332,7 +334,7 @@ class CTRLModel(CTRLPreTrainedModel): ...@@ -332,7 +334,7 @@ class CTRLModel(CTRLPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -521,7 +523,7 @@ class CTRLLMHeadModel(CTRLPreTrainedModel): ...@@ -521,7 +523,7 @@ class CTRLLMHeadModel(CTRLPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_distilbert.py
View file @ 3b3619a3
...@@ -377,6 +377,8 @@ DISTILBERT_INPUTS_DOCSTRING = r""" ...@@ -377,6 +377,8 @@ DISTILBERT_INPUTS_DOCSTRING = r"""
Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -421,7 +423,7 @@ class DistilBertModel(DistilBertPreTrainedModel): ...@@ -421,7 +423,7 @@ class DistilBertModel(DistilBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -523,7 +525,7 @@ class DistilBertForMaskedLM(DistilBertPreTrainedModel): ...@@ -523,7 +525,7 @@ class DistilBertForMaskedLM(DistilBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -616,7 +618,7 @@ class DistilBertForSequenceClassification(DistilBertPreTrainedModel): ...@@ -616,7 +618,7 @@ class DistilBertForSequenceClassification(DistilBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -713,7 +715,7 @@ class DistilBertForQuestionAnswering(DistilBertPreTrainedModel): ...@@ -713,7 +715,7 @@ class DistilBertForQuestionAnswering(DistilBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -812,7 +814,7 @@ class DistilBertForTokenClassification(DistilBertPreTrainedModel): ...@@ -812,7 +814,7 @@ class DistilBertForTokenClassification(DistilBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_electra.py
View file @ 3b3619a3
...@@ -220,6 +220,8 @@ ELECTRA_INPUTS_DOCSTRING = r""" ...@@ -220,6 +220,8 @@ ELECTRA_INPUTS_DOCSTRING = r"""
is used in the cross-attention if the model is configured as a decoder. is used in the cross-attention if the model is configured as a decoder.
Mask values selected in ``[0, 1]``: Mask values selected in ``[0, 1]``:
``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens. ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -281,7 +283,7 @@ class ElectraModel(ElectraPreTrainedModel): ...@@ -281,7 +283,7 @@ class ElectraModel(ElectraPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -404,7 +406,7 @@ class ElectraForSequenceClassification(ElectraPreTrainedModel): ...@@ -404,7 +406,7 @@ class ElectraForSequenceClassification(ElectraPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -494,7 +496,7 @@ class ElectraForPreTraining(ElectraPreTrainedModel): ...@@ -494,7 +496,7 @@ class ElectraForPreTraining(ElectraPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -597,7 +599,7 @@ class ElectraForMaskedLM(ElectraPreTrainedModel): ...@@ -597,7 +599,7 @@ class ElectraForMaskedLM(ElectraPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -691,7 +693,7 @@ class ElectraForTokenClassification(ElectraPreTrainedModel): ...@@ -691,7 +693,7 @@ class ElectraForTokenClassification(ElectraPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_flaubert.py
View file @ 3b3619a3
...@@ -100,6 +100,8 @@ FLAUBERT_INPUTS_DOCSTRING = r""" ...@@ -100,6 +100,8 @@ FLAUBERT_INPUTS_DOCSTRING = r"""
Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -140,7 +142,7 @@ class FlaubertModel(XLMModel): ...@@ -140,7 +142,7 @@ class FlaubertModel(XLMModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_gpt2.py
View file @ 3b3619a3
...@@ -335,6 +335,8 @@ GPT2_INPUTS_DOCSTRING = r""" ...@@ -335,6 +335,8 @@ GPT2_INPUTS_DOCSTRING = r"""
If `past` is used, optionally only the last `inputs_embeds` have to be input (see `past`). If `past` is used, optionally only the last `inputs_embeds` have to be input (see `past`).
use_cache (:obj:`bool`): use_cache (:obj:`bool`):
If `use_cache` is True, `past` key value states are returned and can be used to speed up decoding (see `past`). Defaults to `True`. If `use_cache` is True, `past` key value states are returned and can be used to speed up decoding (see `past`). Defaults to `True`.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -395,7 +397,7 @@ class GPT2Model(GPT2PreTrainedModel): ...@@ -395,7 +397,7 @@ class GPT2Model(GPT2PreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -582,7 +584,7 @@ class GPT2LMHeadModel(GPT2PreTrainedModel): ...@@ -582,7 +584,7 @@ class GPT2LMHeadModel(GPT2PreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -703,7 +705,7 @@ class GPT2DoubleHeadsModel(GPT2PreTrainedModel): ...@@ -703,7 +705,7 @@ class GPT2DoubleHeadsModel(GPT2PreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_longformer.py
View file @ 3b3619a3
...@@ -487,6 +487,8 @@ LONGFORMER_INPUTS_DOCSTRING = r""" ...@@ -487,6 +487,8 @@ LONGFORMER_INPUTS_DOCSTRING = r"""
Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -597,7 +599,7 @@ class LongformerModel(RobertaModel): ...@@ -597,7 +599,7 @@ class LongformerModel(RobertaModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -726,7 +728,7 @@ class LongformerForMaskedLM(BertPreTrainedModel): ...@@ -726,7 +728,7 @@ class LongformerForMaskedLM(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -825,7 +827,7 @@ class LongformerForSequenceClassification(BertPreTrainedModel): ...@@ -825,7 +827,7 @@ class LongformerForSequenceClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -948,7 +950,7 @@ class LongformerForQuestionAnswering(BertPreTrainedModel): ...@@ -948,7 +950,7 @@ class LongformerForQuestionAnswering(BertPreTrainedModel):
Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (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 Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
...@@ -1069,7 +1071,7 @@ class LongformerForTokenClassification(BertPreTrainedModel): ...@@ -1069,7 +1071,7 @@ class LongformerForTokenClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1172,7 +1174,7 @@ class LongformerForMultipleChoice(BertPreTrainedModel): ...@@ -1172,7 +1174,7 @@ class LongformerForMultipleChoice(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_mmbt.py
View file @ 3b3619a3
...@@ -141,6 +141,8 @@ MMBT_INPUTS_DOCSTRING = r""" Inputs: ...@@ -141,6 +141,8 @@ MMBT_INPUTS_DOCSTRING = r""" Inputs:
is used in the cross-attention if the model is configured as a decoder. is used in the cross-attention if the model is configured as a decoder.
Mask values selected in ``[0, 1]``: Mask values selected in ``[0, 1]``:
``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens. ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
......
src/transformers/modeling_openai.py
View file @ 3b3619a3
...@@ -324,6 +324,8 @@ OPENAI_GPT_INPUTS_DOCSTRING = r""" ...@@ -324,6 +324,8 @@ OPENAI_GPT_INPUTS_DOCSTRING = r"""
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to :obj:`False`): output_attentions (:obj:`bool`, `optional`, defaults to :obj:`False`):
Should the model returns attentions weights. Should the model returns attentions weights.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -377,7 +379,7 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel): ...@@ -377,7 +379,7 @@ class OpenAIGPTModel(OpenAIGPTPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -521,7 +523,7 @@ class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel): ...@@ -521,7 +523,7 @@ class OpenAIGPTLMHeadModel(OpenAIGPTPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -638,7 +640,7 @@ class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel): ...@@ -638,7 +640,7 @@ class OpenAIGPTDoubleHeadsModel(OpenAIGPTPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_reformer.py
View file @ 3b3619a3
...@@ -1505,6 +1505,8 @@ REFORMER_INPUTS_DOCSTRING = r""" ...@@ -1505,6 +1505,8 @@ REFORMER_INPUTS_DOCSTRING = r"""
bucketing. Setting `num_hashes` overwrites the default `num_hashes` defined bucketing. Setting `num_hashes` overwrites the default `num_hashes` defined
in `config.num_hashes`. in `config.num_hashes`.
For more information, see `num_hashes` in :class:`transformers.ReformerConfig`. For more information, see `num_hashes` in :class:`transformers.ReformerConfig`.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -1561,7 +1563,7 @@ class ReformerModel(ReformerPreTrainedModel): ...@@ -1561,7 +1563,7 @@ class ReformerModel(ReformerPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. Hidden-states of the model at the output of each layer plus the initial embedding outputs.
all_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True``): all_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1763,7 +1765,7 @@ class ReformerModelWithLMHead(ReformerPreTrainedModel): ...@@ -1763,7 +1765,7 @@ class ReformerModelWithLMHead(ReformerPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. Hidden-states of the model at the output of each layer plus the initial embedding outputs.
all_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True``): all_attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_roberta.py
View file @ 3b3619a3
...@@ -130,6 +130,8 @@ ROBERTA_INPUTS_DOCSTRING = r""" ...@@ -130,6 +130,8 @@ ROBERTA_INPUTS_DOCSTRING = r"""
Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
This is useful if you want more control over how to convert `input_ids` indices into associated vectors This is useful if you want more control over how to convert `input_ids` indices into associated vectors
than the model's internal embedding lookup matrix. than the model's internal embedding lookup matrix.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -208,7 +210,7 @@ class RobertaForMaskedLM(BertPreTrainedModel): ...@@ -208,7 +210,7 @@ class RobertaForMaskedLM(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -328,7 +330,7 @@ class RobertaForSequenceClassification(BertPreTrainedModel): ...@@ -328,7 +330,7 @@ class RobertaForSequenceClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -423,7 +425,7 @@ class RobertaForMultipleChoice(BertPreTrainedModel): ...@@ -423,7 +425,7 @@ class RobertaForMultipleChoice(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -521,7 +523,7 @@ class RobertaForTokenClassification(BertPreTrainedModel): ...@@ -521,7 +523,7 @@ class RobertaForTokenClassification(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -649,7 +651,7 @@ class RobertaForQuestionAnswering(BertPreTrainedModel): ...@@ -649,7 +651,7 @@ class RobertaForQuestionAnswering(BertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
......
src/transformers/modeling_t5.py
View file @ 3b3619a3
...@@ -841,6 +841,8 @@ T5_INPUTS_DOCSTRING = r""" ...@@ -841,6 +841,8 @@ T5_INPUTS_DOCSTRING = r"""
Mask to nullify selected heads of the self-attention modules. Mask to nullify selected heads of the self-attention modules.
Mask values selected in ``[0, 1]``: Mask values selected in ``[0, 1]``:
``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**. ``1`` indicates the head is **not masked**, ``0`` indicates the head is **masked**.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -914,7 +916,7 @@ class T5Model(T5PreTrainedModel): ...@@ -914,7 +916,7 @@ class T5Model(T5PreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape
:obj:`(batch_size, num_heads, sequence_length, sequence_length)`. :obj:`(batch_size, num_heads, sequence_length, sequence_length)`.
...@@ -1050,7 +1052,7 @@ class T5ForConditionalGeneration(T5PreTrainedModel): ...@@ -1050,7 +1052,7 @@ class T5ForConditionalGeneration(T5PreTrainedModel):
Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer)
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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``): attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or ``config.output_attentions=True``):
Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape Tuple of :obj:`torch.FloatTensor` (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. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention.
......
src/transformers/modeling_tf_albert.py
View file @ 3b3619a3
...@@ -688,6 +688,8 @@ ALBERT_INPUTS_DOCSTRING = r""" ...@@ -688,6 +688,8 @@ ALBERT_INPUTS_DOCSTRING = r"""
training (:obj:`boolean`, `optional`, defaults to :obj:`False`): training (:obj:`boolean`, `optional`, defaults to :obj:`False`):
Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them
(if set to :obj:`False`) for evaluation. (if set to :obj:`False`) for evaluation.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -719,7 +721,7 @@ class TFAlbertModel(TFAlbertPreTrainedModel): ...@@ -719,7 +721,7 @@ class TFAlbertModel(TFAlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -771,7 +773,7 @@ class TFAlbertForPreTraining(TFAlbertPreTrainedModel): ...@@ -771,7 +773,7 @@ class TFAlbertForPreTraining(TFAlbertPreTrainedModel):
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)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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.
...@@ -831,7 +833,7 @@ class TFAlbertForMaskedLM(TFAlbertPreTrainedModel): ...@@ -831,7 +833,7 @@ class TFAlbertForMaskedLM(TFAlbertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -905,7 +907,7 @@ class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClass ...@@ -905,7 +907,7 @@ class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClass
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -993,7 +995,7 @@ class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificat ...@@ -993,7 +995,7 @@ class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificat
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -1089,7 +1091,7 @@ class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringL ...@@ -1089,7 +1091,7 @@ class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringL
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -1196,7 +1198,7 @@ class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss): ...@@ -1196,7 +1198,7 @@ class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
......
src/transformers/modeling_tf_bert.py
View file @ 3b3619a3
...@@ -682,6 +682,8 @@ BERT_INPUTS_DOCSTRING = r""" ...@@ -682,6 +682,8 @@ BERT_INPUTS_DOCSTRING = r"""
training (:obj:`boolean`, `optional`, defaults to :obj:`False`): training (:obj:`boolean`, `optional`, defaults to :obj:`False`):
Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them
(if set to :obj:`False`) for evaluation. (if set to :obj:`False`) for evaluation.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -713,7 +715,7 @@ class TFBertModel(TFBertPreTrainedModel): ...@@ -713,7 +715,7 @@ class TFBertModel(TFBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`.
...@@ -765,7 +767,7 @@ class TFBertForPreTraining(TFBertPreTrainedModel): ...@@ -765,7 +767,7 @@ class TFBertForPreTraining(TFBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -819,7 +821,7 @@ class TFBertForMaskedLM(TFBertPreTrainedModel): ...@@ -819,7 +821,7 @@ class TFBertForMaskedLM(TFBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -869,7 +871,7 @@ class TFBertForNextSentencePrediction(TFBertPreTrainedModel): ...@@ -869,7 +871,7 @@ class TFBertForNextSentencePrediction(TFBertPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -945,7 +947,7 @@ class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassific ...@@ -945,7 +947,7 @@ class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassific
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -1044,7 +1046,7 @@ class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss): ...@@ -1044,7 +1046,7 @@ class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -1169,7 +1171,7 @@ class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationL ...@@ -1169,7 +1171,7 @@ class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationL
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
...@@ -1266,7 +1268,7 @@ class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss) ...@@ -1266,7 +1268,7 @@ class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss)
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`:
......
src/transformers/modeling_tf_camembert.py
View file @ 3b3619a3
...@@ -62,6 +62,8 @@ CAMEMBERT_START_DOCSTRING = r""" ...@@ -62,6 +62,8 @@ CAMEMBERT_START_DOCSTRING = r"""
config (:class:`~transformers.CamembertConfig`): Model configuration class with all the parameters of the config (:class:`~transformers.CamembertConfig`): Model configuration class with all the parameters of the
model. Initializing with a config file does not load the weights associated with the model, only the configuration. model. Initializing with a config file does not load the weights associated with the model, only the configuration.
Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights. Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
......
src/transformers/modeling_tf_ctrl.py
View file @ 3b3619a3
...@@ -464,6 +464,8 @@ CTRL_INPUTS_DOCSTRING = r""" ...@@ -464,6 +464,8 @@ CTRL_INPUTS_DOCSTRING = r"""
training (:obj:`boolean`, `optional`, defaults to :obj:`False`): training (:obj:`boolean`, `optional`, defaults to :obj:`False`):
Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them Whether to activate dropout modules (if set to :obj:`True`) during training or to de-activate them
(if set to :obj:`False`) for evaluation. (if set to :obj:`False`) for evaluation.
output_attentions (:obj:`bool`, `optional`, defaults to `:obj:`None`):
If set to ``True``, the attentions tensors of all attention layers are returned. See ``attentions`` under returned tensors for more detail.
""" """
...@@ -492,7 +494,7 @@ class TFCTRLModel(TFCTRLPreTrainedModel): ...@@ -492,7 +494,7 @@ class TFCTRLModel(TFCTRLPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
Tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`.
...@@ -572,7 +574,7 @@ class TFCTRLLMHeadModel(TFCTRLPreTrainedModel): ...@@ -572,7 +574,7 @@ class TFCTRLLMHeadModel(TFCTRLPreTrainedModel):
of shape :obj:`(batch_size, sequence_length, hidden_size)`. 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. 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 ``config.output_attentions=True``):
Tuple of :obj:`tf.Tensor` (one for each layer) of shape 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)`.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment