2019-09-21 05:14:51 +08:00
|
|
|
# coding=utf-8
|
2020-12-08 07:36:34 +08:00
|
|
|
# Copyright 2020 The HuggingFace Team. All rights reserved.
|
2019-09-21 05:14:51 +08:00
|
|
|
#
|
|
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
|
|
# you may not use this file except in compliance with the License.
|
|
|
|
# You may obtain a copy of the License at
|
|
|
|
#
|
|
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
#
|
|
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
# See the License for the specific language governing permissions and
|
|
|
|
# limitations under the License.
|
2019-12-22 23:20:32 +08:00
|
|
|
|
2019-09-21 05:14:51 +08:00
|
|
|
|
|
|
|
import unittest
|
|
|
|
|
2019-12-21 22:57:32 +08:00
|
|
|
from transformers import RobertaConfig, is_tf_available
|
2020-10-19 02:51:24 +08:00
|
|
|
from transformers.testing_utils import require_sentencepiece, require_tf, require_tokenizers, slow
|
2019-12-21 22:57:32 +08:00
|
|
|
|
2019-12-22 20:44:13 +08:00
|
|
|
from .test_configuration_common import ConfigTester
|
2019-12-22 21:57:20 +08:00
|
|
|
from .test_modeling_tf_common import TFModelTesterMixin, ids_tensor
|
2019-09-21 05:14:51 +08:00
|
|
|
|
|
|
|
|
|
|
|
if is_tf_available():
|
|
|
|
import numpy
|
2020-08-24 23:03:01 +08:00
|
|
|
import tensorflow as tf
|
|
|
|
|
2020-11-17 10:43:42 +08:00
|
|
|
from transformers.models.roberta.modeling_tf_roberta import (
|
2020-08-24 23:03:01 +08:00
|
|
|
TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST,
|
2019-12-21 22:46:46 +08:00
|
|
|
TFRobertaForMaskedLM,
|
2020-08-24 23:03:01 +08:00
|
|
|
TFRobertaForMultipleChoice,
|
|
|
|
TFRobertaForQuestionAnswering,
|
2019-12-21 22:46:46 +08:00
|
|
|
TFRobertaForSequenceClassification,
|
|
|
|
TFRobertaForTokenClassification,
|
2020-08-24 23:03:01 +08:00
|
|
|
TFRobertaModel,
|
2019-12-21 22:46:46 +08:00
|
|
|
)
|
2019-09-21 05:14:51 +08:00
|
|
|
|
|
|
|
|
2020-06-16 20:03:43 +08:00
|
|
|
class TFRobertaModelTester:
|
|
|
|
def __init__(
|
2020-08-26 23:20:22 +08:00
|
|
|
self,
|
|
|
|
parent,
|
2020-06-16 20:03:43 +08:00
|
|
|
):
|
|
|
|
self.parent = parent
|
|
|
|
self.batch_size = 13
|
|
|
|
self.seq_length = 7
|
|
|
|
self.is_training = True
|
|
|
|
self.use_input_mask = True
|
|
|
|
self.use_token_type_ids = True
|
|
|
|
self.use_labels = True
|
|
|
|
self.vocab_size = 99
|
|
|
|
self.hidden_size = 32
|
|
|
|
self.num_hidden_layers = 5
|
|
|
|
self.num_attention_heads = 4
|
|
|
|
self.intermediate_size = 37
|
|
|
|
self.hidden_act = "gelu"
|
|
|
|
self.hidden_dropout_prob = 0.1
|
|
|
|
self.attention_probs_dropout_prob = 0.1
|
|
|
|
self.max_position_embeddings = 512
|
|
|
|
self.type_vocab_size = 16
|
|
|
|
self.type_sequence_label_size = 2
|
|
|
|
self.initializer_range = 0.02
|
|
|
|
self.num_labels = 3
|
|
|
|
self.num_choices = 4
|
|
|
|
self.scope = None
|
|
|
|
|
|
|
|
def prepare_config_and_inputs(self):
|
|
|
|
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
|
|
|
|
|
|
|
|
input_mask = None
|
|
|
|
if self.use_input_mask:
|
|
|
|
input_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2)
|
|
|
|
|
|
|
|
token_type_ids = None
|
|
|
|
if self.use_token_type_ids:
|
|
|
|
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
|
|
|
|
|
|
|
|
sequence_labels = None
|
|
|
|
token_labels = None
|
|
|
|
choice_labels = None
|
|
|
|
if self.use_labels:
|
|
|
|
sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
|
|
|
|
token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
|
|
|
|
choice_labels = ids_tensor([self.batch_size], self.num_choices)
|
|
|
|
|
|
|
|
config = RobertaConfig(
|
|
|
|
vocab_size=self.vocab_size,
|
|
|
|
hidden_size=self.hidden_size,
|
|
|
|
num_hidden_layers=self.num_hidden_layers,
|
|
|
|
num_attention_heads=self.num_attention_heads,
|
|
|
|
intermediate_size=self.intermediate_size,
|
|
|
|
hidden_act=self.hidden_act,
|
|
|
|
hidden_dropout_prob=self.hidden_dropout_prob,
|
|
|
|
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
|
|
|
|
max_position_embeddings=self.max_position_embeddings,
|
|
|
|
type_vocab_size=self.type_vocab_size,
|
|
|
|
initializer_range=self.initializer_range,
|
|
|
|
)
|
|
|
|
|
|
|
|
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
|
|
|
|
def create_and_check_roberta_model(
|
|
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
):
|
|
|
|
model = TFRobertaModel(config=config)
|
|
|
|
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(inputs)
|
2020-06-16 20:03:43 +08:00
|
|
|
|
|
|
|
inputs = [input_ids, input_mask]
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(inputs)
|
2020-06-16 20:03:43 +08:00
|
|
|
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(input_ids)
|
2020-06-16 20:03:43 +08:00
|
|
|
|
2020-08-13 16:29:06 +08:00
|
|
|
self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
|
2020-06-16 20:03:43 +08:00
|
|
|
|
|
|
|
def create_and_check_roberta_for_masked_lm(
|
|
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
):
|
|
|
|
model = TFRobertaForMaskedLM(config=config)
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model([input_ids, input_mask, token_type_ids])
|
2020-08-13 16:29:06 +08:00
|
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
|
2020-06-16 20:03:43 +08:00
|
|
|
|
|
|
|
def create_and_check_roberta_for_token_classification(
|
|
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
):
|
|
|
|
config.num_labels = self.num_labels
|
|
|
|
model = TFRobertaForTokenClassification(config=config)
|
|
|
|
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(inputs)
|
2020-08-13 16:29:06 +08:00
|
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
|
2020-06-16 20:03:43 +08:00
|
|
|
|
|
|
|
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}
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(inputs)
|
2020-08-13 16:29:06 +08:00
|
|
|
self.parent.assertEqual(result.start_logits.shape, (self.batch_size, self.seq_length))
|
|
|
|
self.parent.assertEqual(result.end_logits.shape, (self.batch_size, self.seq_length))
|
2020-06-16 20:03:43 +08:00
|
|
|
|
2020-06-24 23:37:20 +08:00
|
|
|
def create_and_check_roberta_for_multiple_choice(
|
|
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
):
|
|
|
|
config.num_choices = self.num_choices
|
|
|
|
model = TFRobertaForMultipleChoice(config=config)
|
|
|
|
multiple_choice_inputs_ids = tf.tile(tf.expand_dims(input_ids, 1), (1, self.num_choices, 1))
|
|
|
|
multiple_choice_input_mask = tf.tile(tf.expand_dims(input_mask, 1), (1, self.num_choices, 1))
|
|
|
|
multiple_choice_token_type_ids = tf.tile(tf.expand_dims(token_type_ids, 1), (1, self.num_choices, 1))
|
|
|
|
inputs = {
|
|
|
|
"input_ids": multiple_choice_inputs_ids,
|
|
|
|
"attention_mask": multiple_choice_input_mask,
|
|
|
|
"token_type_ids": multiple_choice_token_type_ids,
|
|
|
|
}
|
2020-08-05 23:34:39 +08:00
|
|
|
result = model(inputs)
|
2020-08-13 16:29:06 +08:00
|
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
|
2020-06-24 23:37:20 +08:00
|
|
|
|
2020-06-16 20:03:43 +08:00
|
|
|
def prepare_config_and_inputs_for_common(self):
|
|
|
|
config_and_inputs = self.prepare_config_and_inputs()
|
|
|
|
(
|
|
|
|
config,
|
|
|
|
input_ids,
|
|
|
|
token_type_ids,
|
|
|
|
input_mask,
|
|
|
|
sequence_labels,
|
|
|
|
token_labels,
|
|
|
|
choice_labels,
|
|
|
|
) = config_and_inputs
|
|
|
|
inputs_dict = {"input_ids": input_ids, "token_type_ids": token_type_ids, "attention_mask": input_mask}
|
|
|
|
return config, inputs_dict
|
|
|
|
|
|
|
|
|
2019-12-07 02:57:38 +08:00
|
|
|
@require_tf
|
2019-12-22 21:57:20 +08:00
|
|
|
class TFRobertaModelTest(TFModelTesterMixin, unittest.TestCase):
|
2019-09-21 05:14:51 +08:00
|
|
|
|
2019-12-21 22:46:46 +08:00
|
|
|
all_model_classes = (
|
2020-04-17 22:45:30 +08:00
|
|
|
(
|
|
|
|
TFRobertaModel,
|
|
|
|
TFRobertaForMaskedLM,
|
|
|
|
TFRobertaForSequenceClassification,
|
|
|
|
TFRobertaForTokenClassification,
|
|
|
|
TFRobertaForQuestionAnswering,
|
|
|
|
)
|
|
|
|
if is_tf_available()
|
|
|
|
else ()
|
2019-12-21 22:46:46 +08:00
|
|
|
)
|
2021-01-26 16:50:00 +08:00
|
|
|
test_head_masking = False
|
2021-02-15 20:55:10 +08:00
|
|
|
test_onnx = False
|
2019-09-21 05:14:51 +08:00
|
|
|
|
|
|
|
def setUp(self):
|
2020-06-16 20:03:43 +08:00
|
|
|
self.model_tester = TFRobertaModelTester(self)
|
2019-09-21 05:14:51 +08:00
|
|
|
self.config_tester = ConfigTester(self, config_class=RobertaConfig, hidden_size=37)
|
|
|
|
|
|
|
|
def test_config(self):
|
|
|
|
self.config_tester.run_common_tests()
|
|
|
|
|
|
|
|
def test_roberta_model(self):
|
|
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
|
|
self.model_tester.create_and_check_roberta_model(*config_and_inputs)
|
|
|
|
|
|
|
|
def test_for_masked_lm(self):
|
|
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
|
|
self.model_tester.create_and_check_roberta_for_masked_lm(*config_and_inputs)
|
|
|
|
|
2020-02-04 07:39:54 +08:00
|
|
|
def test_for_token_classification(self):
|
|
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
|
|
self.model_tester.create_and_check_roberta_for_token_classification(*config_and_inputs)
|
|
|
|
|
2020-04-17 22:45:30 +08:00
|
|
|
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)
|
2020-06-24 23:37:20 +08:00
|
|
|
|
|
|
|
def test_for_multiple_choice(self):
|
|
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
|
|
self.model_tester.create_and_check_roberta_for_multiple_choice(*config_and_inputs)
|
2020-04-17 22:45:30 +08:00
|
|
|
|
2019-12-07 02:57:38 +08:00
|
|
|
@slow
|
2019-09-21 05:14:51 +08:00
|
|
|
def test_model_from_pretrained(self):
|
2020-06-02 21:39:33 +08:00
|
|
|
for model_name in TF_ROBERTA_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
|
2020-04-24 05:12:59 +08:00
|
|
|
model = TFRobertaModel.from_pretrained(model_name)
|
2019-09-21 05:14:51 +08:00
|
|
|
self.assertIsNotNone(model)
|
|
|
|
|
|
|
|
|
2020-06-03 00:59:00 +08:00
|
|
|
@require_tf
|
2020-10-19 02:51:24 +08:00
|
|
|
@require_sentencepiece
|
|
|
|
@require_tokenizers
|
2019-09-21 05:14:51 +08:00
|
|
|
class TFRobertaModelIntegrationTest(unittest.TestCase):
|
2019-12-07 02:57:38 +08:00
|
|
|
@slow
|
2019-09-21 05:14:51 +08:00
|
|
|
def test_inference_masked_lm(self):
|
2019-12-21 22:46:46 +08:00
|
|
|
model = TFRobertaForMaskedLM.from_pretrained("roberta-base")
|
2019-12-07 02:57:38 +08:00
|
|
|
|
2019-12-21 22:46:46 +08:00
|
|
|
input_ids = tf.constant([[0, 31414, 232, 328, 740, 1140, 12695, 69, 46078, 1588, 2]])
|
2019-09-21 05:14:51 +08:00
|
|
|
output = model(input_ids)[0]
|
|
|
|
expected_shape = [1, 11, 50265]
|
2019-12-21 22:46:46 +08:00
|
|
|
self.assertEqual(list(output.numpy().shape), expected_shape)
|
2019-09-21 05:14:51 +08:00
|
|
|
# compare the actual values for a slice.
|
|
|
|
expected_slice = tf.constant(
|
2020-03-04 04:31:31 +08:00
|
|
|
[[[33.8802, -4.3103, 22.7761], [4.6539, -2.8098, 13.6253], [1.8228, -3.6898, 8.8600]]]
|
2019-09-21 05:14:51 +08:00
|
|
|
)
|
2020-03-04 04:31:31 +08:00
|
|
|
self.assertTrue(numpy.allclose(output[:, :3, :3].numpy(), expected_slice.numpy(), atol=1e-4))
|
2019-09-21 05:14:51 +08:00
|
|
|
|
2019-12-07 02:57:38 +08:00
|
|
|
@slow
|
2019-09-21 05:14:51 +08:00
|
|
|
def test_inference_no_head(self):
|
2019-12-21 22:46:46 +08:00
|
|
|
model = TFRobertaModel.from_pretrained("roberta-base")
|
2019-12-07 02:57:38 +08:00
|
|
|
|
2019-12-21 22:46:46 +08:00
|
|
|
input_ids = tf.constant([[0, 31414, 232, 328, 740, 1140, 12695, 69, 46078, 1588, 2]])
|
2019-09-21 05:14:51 +08:00
|
|
|
output = model(input_ids)[0]
|
|
|
|
# compare the actual values for a slice.
|
|
|
|
expected_slice = tf.constant(
|
2020-03-04 04:31:31 +08:00
|
|
|
[[[-0.0231, 0.0782, 0.0074], [-0.1854, 0.0540, -0.0175], [0.0548, 0.0799, 0.1687]]]
|
2019-09-21 05:14:51 +08:00
|
|
|
)
|
2020-03-04 04:31:31 +08:00
|
|
|
self.assertTrue(numpy.allclose(output[:, :3, :3].numpy(), expected_slice.numpy(), atol=1e-4))
|
2019-09-21 05:14:51 +08:00
|
|
|
|
2019-12-07 02:57:38 +08:00
|
|
|
@slow
|
2019-09-21 05:14:51 +08:00
|
|
|
def test_inference_classification_head(self):
|
2019-12-21 22:46:46 +08:00
|
|
|
model = TFRobertaForSequenceClassification.from_pretrained("roberta-large-mnli")
|
2019-12-07 02:57:38 +08:00
|
|
|
|
2019-12-21 22:46:46 +08:00
|
|
|
input_ids = tf.constant([[0, 31414, 232, 328, 740, 1140, 12695, 69, 46078, 1588, 2]])
|
2019-09-21 05:14:51 +08:00
|
|
|
output = model(input_ids)[0]
|
|
|
|
expected_shape = [1, 3]
|
2019-12-21 22:46:46 +08:00
|
|
|
self.assertEqual(list(output.numpy().shape), expected_shape)
|
|
|
|
expected_tensor = tf.constant([[-0.9469, 0.3913, 0.5118]])
|
2020-03-04 04:31:31 +08:00
|
|
|
self.assertTrue(numpy.allclose(output.numpy(), expected_tensor.numpy(), atol=1e-4))
|