Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "1e4cf8bb446fcbfcd428b15377bd12e254ad8baa"
Unverified Commit dad5ca83 authored by Joao Gante's avatar Joao Gante Committed by GitHub
Browse files

TF: Finalize `unpack_inputs`-related changes (#16499) 

* Add unpack_inputs to remaining models

* removed kwargs to `call()` in TF models

* fix TF T5 tests
parent be9474bd
Show whitespace changes
Inline Side-by-side
Showing with 25 additions and 136 deletions
src/transformers/modeling_tf_utils.py
View file @ dad5ca83
......@@ -312,10 +312,12 @@ def booleans_processing(config, **kwargs):
final_booleans = {}
if tf.executing_eagerly():
# Pure conv models (such as ConvNext) do not have `output_attentions`
final_booleans["output_attentions"] = kwargs.get("output_attentions", None)
if final_booleans["output_attentions"] is None:
final_booleans["output_attentions"] = config.output_attentions
# Pure conv models (such as ConvNext) do not have `output_attentions`. If the signature has
# `output_attentions`, it will be present here in `kwargs`, even if unset (in that case, as `None`)
if "output_attentions" in kwargs:
final_booleans["output_attentions"] = (
kwargs["output_attentions"] if kwargs["output_attentions"] is not None else config.output_attentions
)
final_booleans["output_hidden_states"] = (
kwargs["output_hidden_states"]
if kwargs["output_hidden_states"] is not None
......@@ -330,6 +332,9 @@ def booleans_processing(config, **kwargs):
kwargs["use_cache"] if kwargs["use_cache"] is not None else getattr(config, "use_cache", None)
)
else:
# Pure conv models (such as ConvNext) do not have `output_attentions`. If the signature has
# `output_attentions`, it will be present here in `kwargs`, even if unset (in that case, as `None`)
if "output_attentions" in kwargs:
final_booleans["output_attentions"] = config.output_attentions
final_booleans["output_hidden_states"] = config.output_hidden_states
......@@ -403,7 +408,7 @@ def input_processing(func, config, input_ids, **kwargs):
Two lists, one for the missing layers, and another one for the unexpected layers.
"""
signature = dict(inspect.signature(func).parameters)
signature.pop("kwargs", None)
has_kwargs = bool(signature.pop("kwargs", None))
signature.pop("self", None)
parameter_names = list(signature.keys())
output = {}
......@@ -433,11 +438,13 @@ def input_processing(func, config, input_ids, **kwargs):
elif "past_key_values" in kwargs["kwargs_call"] and "past" in parameter_names:
kwargs["past"] = kwargs["kwargs_call"].pop("past_key_values")
if has_kwargs:
output["kwargs"] = kwargs.pop("kwargs_call", {})
else:
if len(kwargs["kwargs_call"]) > 0:
raise ValueError(
f"The following keyword arguments are not supported by this model: {list(kwargs['kwargs_call'].keys())}."
)
kwargs.pop("kwargs_call")
for k, v in kwargs.items():
......
src/transformers/models/albert/modeling_tf_albert.py
View file @ dad5ca83
......@@ -551,7 +551,6 @@ class TFAlbertMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
if input_ids is not None and inputs_embeds is not None:
......@@ -785,7 +784,6 @@ class TFAlbertModel(TFAlbertPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
outputs = self.albert(
input_ids=input_ids,
......@@ -854,7 +852,6 @@ class TFAlbertForPreTraining(TFAlbertPreTrainedModel, TFAlbertPreTrainingLoss):
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
sentence_order_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFAlbertForPreTrainingOutput, Tuple[tf.Tensor]]:
r"""
Return:
......@@ -976,7 +973,6 @@ class TFAlbertForMaskedLM(TFAlbertPreTrainedModel, TFMaskedLanguageModelingLoss)
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1064,7 +1060,6 @@ class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel, TFSequenceClass
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1158,7 +1153,6 @@ class TFAlbertForTokenClassification(TFAlbertPreTrainedModel, TFTokenClassificat
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1244,7 +1238,6 @@ class TFAlbertForQuestionAnswering(TFAlbertPreTrainedModel, TFQuestionAnsweringL
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1355,7 +1348,6 @@ class TFAlbertForMultipleChoice(TFAlbertPreTrainedModel, TFMultipleChoiceLoss):
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/bart/modeling_tf_bart.py
View file @ dad5ca83
......@@ -679,7 +679,6 @@ class TFBartEncoder(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
"""
Args:
......@@ -834,7 +833,6 @@ class TFBartDecoder(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
Args:
......@@ -1273,7 +1271,6 @@ class TFBartForConditionalGeneration(TFBartPretrainedModel, TFCausalLanguageMode
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSeq2SeqLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......
src/transformers/models/bert/modeling_tf_bert.py
View file @ dad5ca83
......@@ -737,7 +737,6 @@ class TFBertMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:
if not self.config.is_decoder:
......@@ -1067,7 +1066,6 @@ class TFBertModel(TFBertPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPoolingAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
encoder_hidden_states (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
......@@ -1174,7 +1172,6 @@ class TFBertForPreTraining(TFBertPreTrainedModel, TFBertPreTrainingLoss):
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
next_sentence_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBertForPreTrainingOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1302,7 +1299,6 @@ class TFBertForMaskedLM(TFBertPreTrainedModel, TFMaskedLanguageModelingLoss):
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1520,7 +1516,6 @@ class TFBertForNextSentencePrediction(TFBertPreTrainedModel, TFNextSentencePredi
return_dict: Optional[bool] = None,
next_sentence_label: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFNextSentencePredictorOutput, Tuple[tf.Tensor]]:
r"""
Return:
......@@ -1628,7 +1623,6 @@ class TFBertForSequenceClassification(TFBertPreTrainedModel, TFSequenceClassific
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......@@ -1723,7 +1717,6 @@ class TFBertForMultipleChoice(TFBertPreTrainedModel, TFMultipleChoiceLoss):
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......@@ -1857,7 +1850,6 @@ class TFBertForTokenClassification(TFBertPreTrainedModel, TFTokenClassificationL
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1949,7 +1941,6 @@ class TFBertForQuestionAnswering(TFBertPreTrainedModel, TFQuestionAnsweringLoss)
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/blenderbot/modeling_tf_blenderbot.py
View file @ dad5ca83
......@@ -662,7 +662,6 @@ class TFBlenderbotEncoder(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
"""
Args:
......@@ -823,7 +822,6 @@ class TFBlenderbotDecoder(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
r"""
Args:
......@@ -1276,7 +1274,6 @@ class TFBlenderbotForConditionalGeneration(TFBlenderbotPreTrainedModel, TFCausal
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFSeq2SeqLMOutput]:
r"""
labels (`tf.tensor` of shape `(batch_size, sequence_length)`, *optional*):
......
src/transformers/models/blenderbot_small/modeling_tf_blenderbot_small.py
View file @ dad5ca83
......@@ -667,7 +667,6 @@ class TFBlenderbotSmallEncoder(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
"""
Args:
......@@ -827,7 +826,6 @@ class TFBlenderbotSmallDecoder(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
r"""
Args:
......@@ -1253,7 +1251,6 @@ class TFBlenderbotSmallForConditionalGeneration(TFBlenderbotSmallPreTrainedModel
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFSeq2SeqLMOutput]:
r"""
labels (`tf.tensor` of shape `(batch_size, sequence_length)`, *optional*):
......
src/transformers/models/clip/modeling_tf_clip.py
View file @ dad5ca83
......@@ -504,7 +504,6 @@ class TFCLIPTextTransformer(tf.keras.layers.Layer):
output_hidden_states: bool,
return_dict: bool,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
input_shape = shape_list(input_ids)
......@@ -593,7 +592,6 @@ class TFCLIPTextMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
if input_ids is None:
raise ValueError("You have to specify input_ids")
......@@ -632,7 +630,6 @@ class TFCLIPVisionTransformer(tf.keras.layers.Layer):
output_hidden_states: bool,
return_dict: bool,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
embedding_output = self.embeddings(pixel_values=pixel_values)
......@@ -683,7 +680,6 @@ class TFCLIPVisionMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
if pixel_values is None:
......@@ -762,7 +758,6 @@ class TFCLIPMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
if input_ids is None:
......@@ -796,7 +791,6 @@ class TFCLIPMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
if pixel_values is None:
raise ValueError("You have to specify pixel_values")
......@@ -826,7 +820,6 @@ class TFCLIPMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFCLIPOutput, Tuple[tf.Tensor]]:
if input_ids is None:
......@@ -1058,7 +1051,6 @@ class TFCLIPTextModel(TFCLIPPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
r"""
Returns:
......@@ -1153,7 +1145,6 @@ class TFCLIPVisionModel(TFCLIPPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
r"""
Returns:
......@@ -1258,7 +1249,6 @@ class TFCLIPModel(TFCLIPPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
r"""
Returns:
......@@ -1297,7 +1287,6 @@ class TFCLIPModel(TFCLIPPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> tf.Tensor:
r"""
Returns:
......@@ -1345,7 +1334,6 @@ class TFCLIPModel(TFCLIPPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFCLIPOutput, Tuple[tf.Tensor]]:
r"""
Returns:
......
src/transformers/models/convbert/modeling_tf_convbert.py
View file @ dad5ca83
......@@ -581,7 +581,6 @@ class TFConvBertMainLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
......@@ -751,7 +750,6 @@ class TFConvBertModel(TFConvBertPreTrainedModel):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
outputs = self.convbert(
input_ids=input_ids,
......@@ -870,7 +868,6 @@ class TFConvBertForMaskedLM(TFConvBertPreTrainedModel, TFMaskedLanguageModelingL
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFMaskedLMOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -979,7 +976,6 @@ class TFConvBertForSequenceClassification(TFConvBertPreTrainedModel, TFSequenceC
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFSequenceClassifierOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1073,7 +1069,6 @@ class TFConvBertForMultipleChoice(TFConvBertPreTrainedModel, TFMultipleChoiceLos
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFMultipleChoiceModelOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1188,7 +1183,6 @@ class TFConvBertForTokenClassification(TFConvBertPreTrainedModel, TFTokenClassif
return_dict: Optional[bool] = None,
labels: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFTokenClassifierOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1268,7 +1262,6 @@ class TFConvBertForQuestionAnswering(TFConvBertPreTrainedModel, TFQuestionAnswer
start_positions: Optional[tf.Tensor] = None,
end_positions: Optional[tf.Tensor] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFQuestionAnsweringModelOutput]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/convnext/modeling_tf_convnext.py
View file @ dad5ca83
......@@ -293,7 +293,6 @@ class TFConvNextMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
output_hidden_states = (
output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
......@@ -439,7 +438,6 @@ class TFConvNextModel(TFConvNextPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor]]:
r"""
Returns:
......@@ -518,7 +516,6 @@ class TFConvNextForImageClassification(TFConvNextPreTrainedModel, TFSequenceClas
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/ctrl/modeling_tf_ctrl.py
View file @ dad5ca83
......@@ -268,7 +268,6 @@ class TFCTRLMainLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
# If using past key value states, only the last tokens
......@@ -541,7 +540,6 @@ class TFCTRLModel(TFCTRLPreTrainedModel):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
outputs = self.transformer(
input_ids=input_ids,
......@@ -653,7 +651,6 @@ class TFCTRLLMHeadModel(TFCTRLPreTrainedModel, TFCausalLanguageModelingLoss):
return_dict=None,
labels=None,
training=False,
**kwargs,
):
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -765,7 +762,6 @@ class TFCTRLForSequenceClassification(TFCTRLPreTrainedModel, TFSequenceClassific
return_dict=None,
labels=None,
training=False,
**kwargs,
):
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......
src/transformers/models/deberta/modeling_tf_deberta.py
View file @ dad5ca83
......@@ -928,7 +928,6 @@ class TFDebertaMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
if input_ids is not None and inputs_embeds is not None:
......@@ -1096,7 +1095,6 @@ class TFDebertaModel(TFDebertaPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
outputs = self.deberta(
input_ids=input_ids,
......@@ -1156,7 +1154,6 @@ class TFDebertaForMaskedLM(TFDebertaPreTrainedModel, TFMaskedLanguageModelingLos
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1242,7 +1239,6 @@ class TFDebertaForSequenceClassification(TFDebertaPreTrainedModel, TFSequenceCla
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......@@ -1325,7 +1321,6 @@ class TFDebertaForTokenClassification(TFDebertaPreTrainedModel, TFTokenClassific
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1404,7 +1399,6 @@ class TFDebertaForQuestionAnswering(TFDebertaPreTrainedModel, TFQuestionAnswerin
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/deberta_v2/modeling_tf_deberta_v2.py
View file @ dad5ca83
......@@ -1028,7 +1028,6 @@ class TFDebertaV2MainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
if input_ids is not None and inputs_embeds is not None:
......@@ -1198,7 +1197,6 @@ class TFDebertaV2Model(TFDebertaV2PreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
outputs = self.deberta(
input_ids=input_ids,
......@@ -1259,7 +1257,6 @@ class TFDebertaV2ForMaskedLM(TFDebertaV2PreTrainedModel, TFMaskedLanguageModelin
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1346,7 +1343,6 @@ class TFDebertaV2ForSequenceClassification(TFDebertaV2PreTrainedModel, TFSequenc
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......@@ -1430,7 +1426,6 @@ class TFDebertaV2ForTokenClassification(TFDebertaV2PreTrainedModel, TFTokenClass
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` or `np.ndarray` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1510,7 +1505,6 @@ class TFDebertaV2ForQuestionAnswering(TFDebertaV2PreTrainedModel, TFQuestionAnsw
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` or `np.ndarray` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/distilbert/modeling_tf_distilbert.py
View file @ dad5ca83
......@@ -372,7 +372,6 @@ class TFDistilBertMainLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
......@@ -543,7 +542,6 @@ class TFDistilBertModel(TFDistilBertPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutput, Tuple[tf.Tensor]]:
outputs = self.distilbert(
input_ids=input_ids,
......@@ -647,7 +645,6 @@ class TFDistilBertForMaskedLM(TFDistilBertPreTrainedModel, TFMaskedLanguageModel
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -735,7 +732,6 @@ class TFDistilBertForSequenceClassification(TFDistilBertPreTrainedModel, TFSeque
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -817,7 +813,6 @@ class TFDistilBertForTokenClassification(TFDistilBertPreTrainedModel, TFTokenCla
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -911,7 +906,6 @@ class TFDistilBertForMultipleChoice(TFDistilBertPreTrainedModel, TFMultipleChoic
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1021,7 +1015,6 @@ class TFDistilBertForQuestionAnswering(TFDistilBertPreTrainedModel, TFQuestionAn
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/dpr/modeling_tf_dpr.py
View file @ dad5ca83
......@@ -174,7 +174,6 @@ class TFDPREncoderLayer(tf.keras.layers.Layer):
output_hidden_states: bool = None,
return_dict: bool = None,
training: bool = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPooling, Tuple[tf.Tensor, ...]]:
outputs = self.bert_model(
input_ids=input_ids,
......@@ -235,7 +234,6 @@ class TFDPRSpanPredictorLayer(tf.keras.layers.Layer):
output_hidden_states: bool = False,
return_dict: bool = False,
training: bool = False,
**kwargs,
) -> Union[TFDPRReaderOutput, Tuple[tf.Tensor, ...]]:
# notations: N - number of questions in a batch, M - number of passages per questions, L - sequence length
n_passages, sequence_length = shape_list(input_ids) if input_ids is not None else shape_list(inputs_embeds)[:2]
......@@ -294,7 +292,6 @@ class TFDPRSpanPredictor(TFPreTrainedModel):
output_hidden_states: bool = False,
return_dict: bool = False,
training: bool = False,
**kwargs,
) -> Union[TFDPRReaderOutput, Tuple[tf.Tensor, ...]]:
outputs = self.encoder(
input_ids=input_ids,
......@@ -328,7 +325,6 @@ class TFDPREncoder(TFPreTrainedModel):
output_hidden_states: bool = False,
return_dict: bool = False,
training: bool = False,
**kwargs,
) -> Union[TFDPRReaderOutput, Tuple[tf.Tensor, ...]]:
outputs = self.encoder(
input_ids=input_ids,
......@@ -560,7 +556,6 @@ class TFDPRContextEncoder(TFDPRPretrainedContextEncoder):
output_hidden_states=None,
return_dict=None,
training: bool = False,
**kwargs,
) -> Union[TFDPRContextEncoderOutput, Tuple[tf.Tensor, ...]]:
r"""
Return:
......@@ -648,7 +643,6 @@ class TFDPRQuestionEncoder(TFDPRPretrainedQuestionEncoder):
output_hidden_states=None,
return_dict=None,
training: bool = False,
**kwargs,
) -> Union[TFDPRQuestionEncoderOutput, Tuple[tf.Tensor, ...]]:
r"""
Return:
......@@ -734,7 +728,6 @@ class TFDPRReader(TFDPRPretrainedReader):
output_hidden_states: bool = None,
return_dict=None,
training: bool = False,
**kwargs,
) -> Union[TFDPRReaderOutput, Tuple[tf.Tensor, ...]]:
r"""
Return:
......
src/transformers/models/electra/modeling_tf_electra.py
View file @ dad5ca83
......@@ -719,7 +719,6 @@ class TFElectraMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
if not self.config.is_decoder:
use_cache = False
......@@ -953,7 +952,6 @@ class TFElectraModel(TFElectraPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
encoder_hidden_states (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
......@@ -1043,7 +1041,6 @@ class TFElectraForPreTraining(TFElectraPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFElectraForPreTrainingOutput, Tuple[tf.Tensor]]:
r"""
Returns:
......@@ -1180,7 +1177,6 @@ class TFElectraForMaskedLM(TFElectraPreTrainedModel, TFMaskedLanguageModelingLos
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMaskedLMOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1290,7 +1286,6 @@ class TFElectraForSequenceClassification(TFElectraPreTrainedModel, TFSequenceCla
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1383,7 +1378,6 @@ class TFElectraForMultipleChoice(TFElectraPreTrainedModel, TFMultipleChoiceLoss)
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFMultipleChoiceModelOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1501,7 +1495,6 @@ class TFElectraForTokenClassification(TFElectraPreTrainedModel, TFTokenClassific
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFTokenClassifierOutput, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1583,7 +1576,6 @@ class TFElectraForQuestionAnswering(TFElectraPreTrainedModel, TFQuestionAnswerin
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFQuestionAnsweringModelOutput, Tuple[tf.Tensor]]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/encoder_decoder/modeling_tf_encoder_decoder.py
View file @ dad5ca83
......@@ -23,7 +23,7 @@ import tensorflow as tf
from ...configuration_utils import PretrainedConfig
from ...modeling_tf_outputs import TFBaseModelOutput, TFSeq2SeqLMOutput
from ...modeling_tf_utils import TFCausalLanguageModelingLoss, TFPreTrainedModel, get_initializer, input_processing
from ...modeling_tf_utils import TFCausalLanguageModelingLoss, TFPreTrainedModel, get_initializer, unpack_inputs
from ...tf_utils import shape_list
from ...utils import (
DUMMY_INPUTS,
......@@ -491,6 +491,7 @@ class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss):
config = EncoderDecoderConfig.from_encoder_decoder_configs(encoder.config, decoder.config, **kwargs)
return cls(encoder=encoder, decoder=decoder, config=config)
@unpack_inputs
@add_start_docstrings_to_model_forward(ENCODER_DECODER_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
@replace_return_docstrings(output_type=TFSeq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
def call(
......@@ -559,9 +560,7 @@ class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss):
if encoder_outputs is None:
encoder_processing_inputs = {
"func": self.encoder.call,
"config": self.encoder.config,
encoder_inputs = {
"input_ids": input_ids,
"attention_mask": attention_mask,
"inputs_embeds": inputs_embeds,
......@@ -569,14 +568,10 @@ class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss):
"output_hidden_states": output_hidden_states,
"return_dict": return_dict,
"training": training,
"kwargs_call": {},
}
# Add arguments to encoder from `kwargs_encoder`
for k, v in kwargs_encoder.items():
encoder_processing_inputs[k] = v
encoder_inputs = input_processing(**encoder_processing_inputs)
encoder_inputs.update(kwargs_encoder)
# Handle the case where the inputs are passed as a single dict which contains `labels`.
# The `labels` shouldn't be passed to `self.encoder` below, because it is a based model without this
......@@ -607,9 +602,7 @@ class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss):
labels, self.config.pad_token_id, self.config.decoder_start_token_id
)
decoder_processing_inputs = {
"func": self.decoder.call,
"config": self.decoder.config,
decoder_inputs = {
"input_ids": decoder_input_ids,
"attention_mask": decoder_attention_mask,
"encoder_hidden_states": encoder_hidden_states,
......@@ -621,14 +614,11 @@ class TFEncoderDecoderModel(TFPreTrainedModel, TFCausalLanguageModelingLoss):
"past_key_values": past_key_values,
"return_dict": return_dict,
"training": training,
"kwargs_call": {},
}
# Add arguments to decoder from `kwargs_decoder`
for k, v in kwargs_decoder.items():
decoder_processing_inputs[k] = v
decoder_inputs.update(kwargs_decoder)
decoder_inputs = input_processing(**decoder_processing_inputs)
decoder_outputs = self.decoder(**decoder_inputs)
logits = decoder_outputs[0]
......
src/transformers/models/flaubert/modeling_tf_flaubert.py
View file @ dad5ca83
......@@ -258,7 +258,6 @@ class TFFlaubertModel(TFFlaubertPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFBaseModelOutput]:
outputs = self.transformer(
input_ids=input_ids,
......@@ -490,7 +489,6 @@ class TFFlaubertMainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFBaseModelOutput]:
# removed: src_enc=None, src_len=None
......@@ -808,7 +806,6 @@ class TFFlaubertWithLMHeadModel(TFFlaubertPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[Tuple, TFFlaubertWithLMHeadModelOutput]:
transformer_outputs = self.transformer(
......
src/transformers/models/funnel/modeling_tf_funnel.py
View file @ dad5ca83
......@@ -761,7 +761,6 @@ class TFFunnelBaseLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
if input_ids is not None and inputs_embeds is not None:
......@@ -835,7 +834,6 @@ class TFFunnelMainLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
......@@ -1117,7 +1115,6 @@ class TFFunnelBaseModel(TFFunnelPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFBaseModelOutput]:
return self.funnel(
input_ids=input_ids,
......@@ -1165,7 +1162,6 @@ class TFFunnelModel(TFFunnelPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFBaseModelOutput]:
return self.funnel(
......@@ -1293,7 +1289,6 @@ class TFFunnelForMaskedLM(TFFunnelPreTrainedModel, TFMaskedLanguageModelingLoss)
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFMaskedLMOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1369,7 +1364,6 @@ class TFFunnelForSequenceClassification(TFFunnelPreTrainedModel, TFSequenceClass
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFSequenceClassifierOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1455,7 +1449,6 @@ class TFFunnelForMultipleChoice(TFFunnelPreTrainedModel, TFMultipleChoiceLoss):
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFMultipleChoiceModelOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1566,7 +1559,6 @@ class TFFunnelForTokenClassification(TFFunnelPreTrainedModel, TFTokenClassificat
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFTokenClassifierOutput]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -1645,7 +1637,6 @@ class TFFunnelForQuestionAnswering(TFFunnelPreTrainedModel, TFQuestionAnsweringL
start_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
end_positions: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: bool = False,
**kwargs,
) -> Union[Tuple[tf.Tensor], TFQuestionAnsweringModelOutput]:
r"""
start_positions (`tf.Tensor` of shape `(batch_size,)`, *optional*):
......
src/transformers/models/gpt2/modeling_tf_gpt2.py
View file @ dad5ca83
......@@ -367,7 +367,6 @@ class TFGPT2MainLayer(tf.keras.layers.Layer):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
if input_ids is not None and inputs_embeds is not None:
......@@ -730,7 +729,6 @@ class TFGPT2Model(TFGPT2PreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFBaseModelOutputWithPastAndCrossAttentions, Tuple[tf.Tensor]]:
r"""
encoder_hidden_states (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
......@@ -920,7 +918,6 @@ class TFGPT2LMHeadModel(TFGPT2PreTrainedModel, TFCausalLanguageModelingLoss):
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFCausalLMOutputWithCrossAttentions, Tuple[tf.Tensor]]:
r"""
encoder_hidden_states (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
......@@ -1038,7 +1035,6 @@ class TFGPT2DoubleHeadsModel(TFGPT2PreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFGPT2DoubleHeadsModelOutput, Tuple[tf.Tensor]]:
r"""
mc_token_ids (`tf.Tensor` or `Numpy array` of shape `(batch_size, num_choices)`, *optional*, default to index of the last token of the input):
......@@ -1195,7 +1191,6 @@ class TFGPT2ForSequenceClassification(TFGPT2PreTrainedModel, TFSequenceClassific
return_dict: Optional[bool] = None,
labels: Optional[Union[np.ndarray, tf.Tensor]] = None,
training: Optional[bool] = False,
**kwargs,
) -> Union[TFSequenceClassifierOutputWithPast, Tuple[tf.Tensor]]:
r"""
labels (`tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......
src/transformers/models/gptj/modeling_tf_gptj.py
View file @ dad5ca83
......@@ -390,7 +390,6 @@ class TFGPTJMainLayer(tf.keras.layers.Layer):
output_hidden_states=None,
return_dict=None,
training=False,
**kwargs,
):
if input_ids is not None and inputs_embeds is not None:
......@@ -672,7 +671,6 @@ class TFGPTJModel(TFGPTJPreTrainedModel):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
):
r"""
use_cache (`bool`, *optional*, defaults to `True`):
......@@ -781,7 +779,6 @@ class TFGPTJForCausalLM(TFGPTJPreTrainedModel, TFCausalLanguageModelingLoss):
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
):
r"""
labels (`np.ndarray` or `tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
......@@ -886,7 +883,6 @@ class TFGPTJForSequenceClassification(TFGPTJPreTrainedModel, TFSequenceClassific
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
):
r"""
labels (`np.ndarray` or `tf.Tensor` of shape `(batch_size,)`, *optional*):
......@@ -1011,7 +1007,6 @@ class TFGPTJForQuestionAnswering(TFGPTJPreTrainedModel, TFQuestionAnsweringLoss)
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
training: Optional[bool] = False,
**kwargs,
):
r"""
start_positions (`np.ndarray` or `tf.Tensor` of shape `(batch_size,)`, *optional*):
......
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