SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Better booleans handling in the TF models (#8777)

Browse Source 
* Apply on BERT and ALBERT

* Update TF Bart

* Add input processing to TF BART

* Add input processing for TF CTRL

* Add input processing to TF Distilbert

* Add input processing to TF DPR

* Add input processing to TF Electra

* Add deprecated arguments

* Add input processing to TF XLM

* Add input processing to TF Funnel

* Add input processing to TF GPT2

* Add input processing to TF Longformer

* Add input processing to TF Lxmert

* Apply style

* Add input processing to TF Mobilebert

* Add input processing to TF GPT

* Add input processing to TF Roberta

* Add input processing to TF T5

* Add input processing to TF TransfoXL

* Apply style

* Rebase on master

* Bug fix

* Retry to bugfix

* Retry bug fix

* Fix wrong model name

* Try another fix

* Fix BART

* Fix input precessing

* Apply style

* Put the deprecated warnings in the input processing function

* Remove the unused imports

* Raise an error when len(kwargs)>0

* test ModelOutput instead of TFBaseModelOutput

* Bug fix

* Address Patrick's comments

* Address Patrick's comments

* Address Sylvain's comments

* Add boolean processing for the inputs

* Apply style

* Missing optional

* Fix missing some input proc

* Update the template

* Fix missing inputs

* Missing input

* Fix args parameter

* Trigger CI

* Trigger CI

* Trigger CI

* Address Patrick's and Sylvain's comments

* Replace warn by warning

* Trigger CI

* Fix XLNET

* Fix detection
This commit is contained in:
Julien Plu
2020-12-04 15:08:29 +01:00
committed by GitHub
parent 4c3d98dddc
commit dcd3046f98
21 changed files with 597 additions and 643 deletions
Download Patch File Download Diff File Expand all files Collapse all files

54
src/transformers/models/roberta/modeling_tf_roberta.py
View File

@@ -467,6 +467,7 @@ class TFRobertaMainLayer(tf.keras.layers.Layer):
def __init__(self, config, **kwargs):
super().__init__(**kwargs)
self.config = config
self.num_hidden_layers = config.num_hidden_layers
self.initializer_range = config.initializer_range
self.output_attentions = config.output_attentions
@@ -511,6 +512,7 @@ class TFRobertaMainLayer(tf.keras.layers.Layer):
):
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -523,13 +525,6 @@ class TFRobertaMainLayer(tf.keras.layers.Layer):
training=training,
kwargs_call=kwargs,
)
output_attentions = (
inputs["output_attentions"] if inputs["output_attentions"] is not None else self.output_attentions
)
output_hidden_states = (
inputs["output_hidden_states"] if inputs["output_hidden_states"] is not None else self.output_hidden_states
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.return_dict
if inputs["input_ids"] is not None and inputs["inputs_embeds"] is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
@@ -584,16 +579,16 @@ class TFRobertaMainLayer(tf.keras.layers.Layer):
embedding_output,
extended_attention_mask,
inputs["head_mask"],
output_attentions,
output_hidden_states,
return_dict,
inputs["output_attentions"],
inputs["output_hidden_states"],
inputs["return_dict"],
training=inputs["training"],
)
sequence_output = encoder_outputs[0]
pooled_output = self.pooler(sequence_output)
if not return_dict:
if not inputs["return_dict"]:
return (
sequence_output,
pooled_output,
@@ -739,6 +734,7 @@ class TFRobertaModel(TFRobertaPreTrainedModel):
):
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -844,6 +840,7 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLos
"""
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -857,7 +854,6 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLos
training=training,
kwargs_call=kwargs,
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.roberta.return_dict
outputs = self.roberta(
inputs["input_ids"],
attention_mask=inputs["attention_mask"],
@@ -867,7 +863,7 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLos
inputs_embeds=inputs["inputs_embeds"],
output_attentions=inputs["output_attentions"],
output_hidden_states=inputs["output_hidden_states"],
return_dict=return_dict,
return_dict=inputs["return_dict"],
training=inputs["training"],
)
@@ -876,7 +872,7 @@ class TFRobertaForMaskedLM(TFRobertaPreTrainedModel, TFMaskedLanguageModelingLos
loss = None if inputs["labels"] is None else self.compute_loss(inputs["labels"], prediction_scores)
if not return_dict:
if not inputs["return_dict"]:
output = (prediction_scores,) + outputs[2:]
return ((loss,) + output) if loss is not None else output
@@ -961,6 +957,7 @@ class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceCla
"""
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -974,7 +971,6 @@ class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceCla
training=training,
kwargs_call=kwargs,
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.roberta.return_dict
outputs = self.roberta(
inputs["input_ids"],
attention_mask=inputs["attention_mask"],
@@ -984,16 +980,16 @@ class TFRobertaForSequenceClassification(TFRobertaPreTrainedModel, TFSequenceCla
inputs_embeds=inputs["inputs_embeds"],
output_attentions=inputs["output_attentions"],
output_hidden_states=inputs["output_hidden_states"],
return_dict=return_dict,
return_dict=inputs["return_dict"],
training=inputs["training"],
)
sequence_output = outputs[0]
logits = self.classifier(sequence_output, training=training)
logits = self.classifier(sequence_output, training=inputs["training"])
loss = None if inputs["labels"] is None else self.compute_loss(inputs["labels"], logits)
if not return_dict:
if not inputs["return_dict"]:
output = (logits,) + outputs[2:]
return ((loss,) + output) if loss is not None else output
@@ -1062,6 +1058,7 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
"""
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -1075,7 +1072,6 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
training=training,
kwargs_call=kwargs,
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.roberta.return_dict
if inputs["input_ids"] is not None:
num_choices = shape_list(inputs["input_ids"])[1]
@@ -1103,7 +1099,7 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
inputs["inputs_embeds"],
inputs["output_attentions"],
inputs["output_hidden_states"],
return_dict=return_dict,
return_dict=inputs["return_dict"],
training=inputs["training"],
)
pooled_output = outputs[1]
@@ -1113,7 +1109,7 @@ class TFRobertaForMultipleChoice(TFRobertaPreTrainedModel, TFMultipleChoiceLoss)
loss = None if inputs["labels"] is None else self.compute_loss(inputs["labels"], reshaped_logits)
if not return_dict:
if not inputs["return_dict"]:
output = (reshaped_logits,) + outputs[2:]
return ((loss,) + output) if loss is not None else output
@@ -1175,6 +1171,7 @@ class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassific
"""
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -1188,7 +1185,6 @@ class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassific
training=training,
kwargs_call=kwargs,
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.roberta.return_dict
outputs = self.roberta(
inputs["input_ids"],
attention_mask=inputs["attention_mask"],
@@ -1198,18 +1194,17 @@ class TFRobertaForTokenClassification(TFRobertaPreTrainedModel, TFTokenClassific
inputs_embeds=inputs["inputs_embeds"],
output_attentions=inputs["output_attentions"],
output_hidden_states=inputs["output_hidden_states"],
return_dict=return_dict,
return_dict=inputs["return_dict"],
training=inputs["training"],
)
sequence_output = outputs[0]
sequence_output = self.dropout(sequence_output, training=training)
sequence_output = self.dropout(sequence_output, training=inputs["training"])
logits = self.classifier(sequence_output)
loss = None if inputs["labels"] is None else self.compute_loss(inputs["labels"], logits)
if not return_dict:
if not inputs["return_dict"]:
output = (logits,) + outputs[2:]
return ((loss,) + output) if loss is not None else output
@@ -1276,6 +1271,7 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
"""
inputs = input_processing(
func=self.call,
config=self.config,
input_ids=input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
@@ -1290,7 +1286,6 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
training=training,
kwargs_call=kwargs,
)
return_dict = inputs["return_dict"] if inputs["return_dict"] is not None else self.roberta.return_dict
outputs = self.roberta(
inputs["input_ids"],
attention_mask=inputs["attention_mask"],
@@ -1300,10 +1295,9 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
inputs_embeds=inputs["inputs_embeds"],
output_attentions=inputs["output_attentions"],
output_hidden_states=inputs["output_hidden_states"],
return_dict=return_dict,
return_dict=inputs["return_dict"],
training=inputs["training"],
)
sequence_output = outputs[0]
logits = self.qa_outputs(sequence_output)
@@ -1317,7 +1311,7 @@ class TFRobertaForQuestionAnswering(TFRobertaPreTrainedModel, TFQuestionAnswerin
labels["end_position"] = inputs["end_positions"]
loss = self.compute_loss(labels, (start_logits, end_logits))
if not return_dict:
if not inputs["return_dict"]:
output = (start_logits, end_logits) + outputs[2:]
return ((loss,) + output) if loss is not None else output