297 lines
11 KiB
Python
297 lines
11 KiB
Python
# coding=utf-8
|
|
# Copyright 2021 The HuggingFace Inc. team. All rights reserved.
|
|
#
|
|
# 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.
|
|
|
|
|
|
from __future__ import annotations
|
|
|
|
import unittest
|
|
|
|
from transformers import DebertaConfig, is_tf_available
|
|
from transformers.testing_utils import require_tf, slow
|
|
|
|
from ...test_configuration_common import ConfigTester
|
|
from ...test_modeling_tf_common import TFModelTesterMixin, ids_tensor, random_attention_mask
|
|
from ...test_pipeline_mixin import PipelineTesterMixin
|
|
|
|
|
|
if is_tf_available():
|
|
import tensorflow as tf
|
|
|
|
from transformers import (
|
|
TFDebertaForMaskedLM,
|
|
TFDebertaForQuestionAnswering,
|
|
TFDebertaForSequenceClassification,
|
|
TFDebertaForTokenClassification,
|
|
TFDebertaModel,
|
|
)
|
|
|
|
|
|
class TFDebertaModelTester:
|
|
def __init__(
|
|
self,
|
|
parent,
|
|
batch_size=13,
|
|
seq_length=7,
|
|
is_training=True,
|
|
use_input_mask=True,
|
|
use_token_type_ids=True,
|
|
use_labels=True,
|
|
vocab_size=99,
|
|
hidden_size=32,
|
|
num_hidden_layers=2,
|
|
num_attention_heads=4,
|
|
intermediate_size=37,
|
|
hidden_act="gelu",
|
|
hidden_dropout_prob=0.1,
|
|
attention_probs_dropout_prob=0.1,
|
|
max_position_embeddings=512,
|
|
type_vocab_size=16,
|
|
type_sequence_label_size=2,
|
|
initializer_range=0.02,
|
|
num_labels=3,
|
|
num_choices=4,
|
|
scope=None,
|
|
):
|
|
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 = 2
|
|
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.relative_attention = False
|
|
self.max_relative_positions = -1
|
|
self.position_biased_input = True
|
|
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 = random_attention_mask([self.batch_size, self.seq_length])
|
|
|
|
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)
|
|
|
|
config = DebertaConfig(
|
|
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,
|
|
relative_attention=self.relative_attention,
|
|
max_relative_positions=self.max_relative_positions,
|
|
position_biased_input=self.position_biased_input,
|
|
initializer_range=self.initializer_range,
|
|
return_dict=True,
|
|
)
|
|
|
|
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
def create_and_check_model(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = TFDebertaModel(config=config)
|
|
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
|
|
|
|
inputs = [input_ids, input_mask]
|
|
result = model(inputs)
|
|
|
|
result = model(input_ids)
|
|
|
|
self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
|
|
|
|
def create_and_check_for_masked_lm(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = TFDebertaForMaskedLM(config=config)
|
|
inputs = {
|
|
"input_ids": input_ids,
|
|
"attention_mask": input_mask,
|
|
"token_type_ids": token_type_ids,
|
|
}
|
|
result = model(inputs)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
|
|
|
|
def create_and_check_for_sequence_classification(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
config.num_labels = self.num_labels
|
|
model = TFDebertaForSequenceClassification(config=config)
|
|
inputs = {
|
|
"input_ids": input_ids,
|
|
"attention_mask": input_mask,
|
|
"token_type_ids": token_type_ids,
|
|
}
|
|
|
|
result = model(inputs)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
|
|
|
|
def create_and_check_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 = TFDebertaForTokenClassification(config=config)
|
|
inputs = {
|
|
"input_ids": input_ids,
|
|
"attention_mask": input_mask,
|
|
"token_type_ids": token_type_ids,
|
|
}
|
|
result = model(inputs)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
|
|
|
|
def create_and_check_for_question_answering(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = TFDebertaForQuestionAnswering(config=config)
|
|
inputs = {
|
|
"input_ids": input_ids,
|
|
"attention_mask": input_mask,
|
|
"token_type_ids": token_type_ids,
|
|
}
|
|
|
|
result = model(inputs)
|
|
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))
|
|
|
|
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
|
|
|
|
|
|
@require_tf
|
|
class TFDebertaModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
|
|
all_model_classes = (
|
|
(
|
|
TFDebertaModel,
|
|
TFDebertaForMaskedLM,
|
|
TFDebertaForQuestionAnswering,
|
|
TFDebertaForSequenceClassification,
|
|
TFDebertaForTokenClassification,
|
|
)
|
|
if is_tf_available()
|
|
else ()
|
|
)
|
|
pipeline_model_mapping = (
|
|
{
|
|
"feature-extraction": TFDebertaModel,
|
|
"fill-mask": TFDebertaForMaskedLM,
|
|
"question-answering": TFDebertaForQuestionAnswering,
|
|
"text-classification": TFDebertaForSequenceClassification,
|
|
"token-classification": TFDebertaForTokenClassification,
|
|
"zero-shot": TFDebertaForSequenceClassification,
|
|
}
|
|
if is_tf_available()
|
|
else {}
|
|
)
|
|
|
|
test_head_masking = False
|
|
test_onnx = False
|
|
|
|
def setUp(self):
|
|
self.model_tester = TFDebertaModelTester(self)
|
|
self.config_tester = ConfigTester(self, config_class=DebertaConfig, hidden_size=37)
|
|
|
|
def test_config(self):
|
|
self.config_tester.run_common_tests()
|
|
|
|
def test_model(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
self.model_tester.create_and_check_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_for_masked_lm(*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_for_question_answering(*config_and_inputs)
|
|
|
|
def test_for_sequence_classification(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
self.model_tester.create_and_check_for_sequence_classification(*config_and_inputs)
|
|
|
|
def test_for_token_classification(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
self.model_tester.create_and_check_for_token_classification(*config_and_inputs)
|
|
|
|
@slow
|
|
def test_model_from_pretrained(self):
|
|
model = TFDebertaModel.from_pretrained("kamalkraj/deberta-base")
|
|
self.assertIsNotNone(model)
|
|
|
|
|
|
@require_tf
|
|
class TFDeBERTaModelIntegrationTest(unittest.TestCase):
|
|
@unittest.skip(reason="Model not available yet")
|
|
def test_inference_masked_lm(self):
|
|
pass
|
|
|
|
@slow
|
|
def test_inference_no_head(self):
|
|
model = TFDebertaModel.from_pretrained("kamalkraj/deberta-base")
|
|
input_ids = tf.constant([[0, 31414, 232, 328, 740, 1140, 12695, 69, 46078, 1588, 2]])
|
|
attention_mask = tf.constant([[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])
|
|
output = model(input_ids, attention_mask=attention_mask)[0]
|
|
|
|
expected_slice = tf.constant(
|
|
[
|
|
[
|
|
[-0.59855896, -0.80552566, -0.8462135],
|
|
[1.4484025, -0.93483794, -0.80593085],
|
|
[0.3122741, 0.00316059, -1.4131377],
|
|
]
|
|
]
|
|
)
|
|
tf.debugging.assert_near(output[:, 1:4, 1:4], expected_slice, atol=1e-4)
|