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

Question Answering support for Albert and Roberta in TF (#3812)

Browse Source 
* Add TFAlbertForQuestionAnswering

* Add TFRobertaForQuestionAnswering

* Update TFAutoModel with Roberta/Albert for QA

* Clean `super` TF Albert calls
This commit is contained in:
Pierric Cistac
2020-04-17 10:45:30 -04:00
committed by GitHub
parent f399c00610
commit 6d00033e97
6 changed files with 193 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
tests/test_modeling_tf_albert.py
View File

@@ -28,6 +28,7 @@ if is_tf_available():
TFAlbertModel,
TFAlbertForMaskedLM,
TFAlbertForSequenceClassification,
TFAlbertForQuestionAnswering,
TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP,
)
@@ -36,7 +37,9 @@ if is_tf_available():
class TFAlbertModelTest(TFModelTesterMixin, unittest.TestCase):
all_model_classes = (
(TFAlbertModel, TFAlbertForMaskedLM, TFAlbertForSequenceClassification) if is_tf_available() else ()
(TFAlbertModel, TFAlbertForMaskedLM, TFAlbertForSequenceClassification, TFAlbertForQuestionAnswering)
if is_tf_available()
else ()
)
class TFAlbertModelTester(object):
@@ -175,6 +178,19 @@ class TFAlbertModelTest(TFModelTesterMixin, unittest.TestCase):
}
self.parent.assertListEqual(list(result["logits"].shape), [self.batch_size, self.num_labels])
def create_and_check_albert_for_question_answering(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = TFAlbertForQuestionAnswering(config=config)
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
start_logits, end_logits = model(inputs)
result = {
"start_logits": start_logits.numpy(),
"end_logits": end_logits.numpy(),
}
self.parent.assertListEqual(list(result["start_logits"].shape), [self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["end_logits"].shape), [self.batch_size, self.seq_length])
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
(
@@ -208,6 +224,10 @@ class TFAlbertModelTest(TFModelTesterMixin, unittest.TestCase):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_albert_for_sequence_classification(*config_and_inputs)
def test_for_question_answering(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_albert_for_question_answering(*config_and_inputs)
@slow
def test_model_from_pretrained(self):
for model_name in list(TF_ALBERT_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:

28
tests/test_modeling_tf_roberta.py
View File

@@ -31,6 +31,7 @@ if is_tf_available():
TFRobertaForMaskedLM,
TFRobertaForSequenceClassification,
TFRobertaForTokenClassification,
TFRobertaForQuestionAnswering,
TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP,
)
@@ -39,7 +40,15 @@ if is_tf_available():
class TFRobertaModelTest(TFModelTesterMixin, unittest.TestCase):
all_model_classes = (
(TFRobertaModel, TFRobertaForMaskedLM, TFRobertaForSequenceClassification) if is_tf_available() else ()
(
TFRobertaModel,
TFRobertaForMaskedLM,
TFRobertaForSequenceClassification,
TFRobertaForTokenClassification,
TFRobertaForQuestionAnswering,
)
if is_tf_available()
else ()
)
class TFRobertaModelTester(object):
@@ -171,6 +180,19 @@ class TFRobertaModelTest(TFModelTesterMixin, unittest.TestCase):
list(result["logits"].shape), [self.batch_size, self.seq_length, self.num_labels]
)
def create_and_check_roberta_for_question_answering(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = TFRobertaForQuestionAnswering(config=config)
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
start_logits, end_logits = model(inputs)
result = {
"start_logits": start_logits.numpy(),
"end_logits": end_logits.numpy(),
}
self.parent.assertListEqual(list(result["start_logits"].shape), [self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["end_logits"].shape), [self.batch_size, self.seq_length])
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
(
@@ -204,6 +226,10 @@ class TFRobertaModelTest(TFModelTesterMixin, unittest.TestCase):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_roberta_for_token_classification(*config_and_inputs)
def test_for_question_answering(self):
config_and_inputs = self.model_tester.prepare_config_and_inputs()
self.model_tester.create_and_check_roberta_for_question_answering(*config_and_inputs)
@slow
def test_model_from_pretrained(self):
for model_name in list(TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]: