511 lines
20 KiB
Python
511 lines
20 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.
|
|
"""Testing suite for the PyTorch RemBERT model."""
|
|
|
|
import unittest
|
|
|
|
from transformers import is_torch_available
|
|
from transformers.testing_utils import require_torch, slow, torch_device
|
|
|
|
from ...test_configuration_common import ConfigTester
|
|
from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor, random_attention_mask
|
|
from ...test_pipeline_mixin import PipelineTesterMixin
|
|
|
|
|
|
if is_torch_available():
|
|
import torch
|
|
|
|
from transformers import (
|
|
RemBertConfig,
|
|
RemBertForCausalLM,
|
|
RemBertForMaskedLM,
|
|
RemBertForMultipleChoice,
|
|
RemBertForQuestionAnswering,
|
|
RemBertForSequenceClassification,
|
|
RemBertForTokenClassification,
|
|
RemBertModel,
|
|
)
|
|
|
|
|
|
class RemBertModelTester:
|
|
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,
|
|
input_embedding_size=18,
|
|
output_embedding_size=43,
|
|
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 = batch_size
|
|
self.seq_length = seq_length
|
|
self.is_training = is_training
|
|
self.use_input_mask = use_input_mask
|
|
self.use_token_type_ids = use_token_type_ids
|
|
self.use_labels = use_labels
|
|
self.vocab_size = vocab_size
|
|
self.hidden_size = hidden_size
|
|
self.input_embedding_size = input_embedding_size
|
|
self.output_embedding_size = output_embedding_size
|
|
self.num_hidden_layers = num_hidden_layers
|
|
self.num_attention_heads = num_attention_heads
|
|
self.intermediate_size = intermediate_size
|
|
self.hidden_act = hidden_act
|
|
self.hidden_dropout_prob = hidden_dropout_prob
|
|
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
|
self.max_position_embeddings = max_position_embeddings
|
|
self.type_vocab_size = type_vocab_size
|
|
self.type_sequence_label_size = type_sequence_label_size
|
|
self.initializer_range = initializer_range
|
|
self.num_labels = num_labels
|
|
self.num_choices = num_choices
|
|
self.scope = scope
|
|
|
|
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)
|
|
choice_labels = ids_tensor([self.batch_size], self.num_choices)
|
|
|
|
config = RemBertConfig(
|
|
vocab_size=self.vocab_size,
|
|
hidden_size=self.hidden_size,
|
|
input_embedding_size=self.input_embedding_size,
|
|
output_embedding_size=self.output_embedding_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,
|
|
is_decoder=False,
|
|
initializer_range=self.initializer_range,
|
|
)
|
|
|
|
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
|
|
def prepare_config_and_inputs_for_decoder(self):
|
|
(
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
) = self.prepare_config_and_inputs()
|
|
|
|
config.is_decoder = True
|
|
encoder_hidden_states = floats_tensor([self.batch_size, self.seq_length, self.hidden_size])
|
|
encoder_attention_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2)
|
|
|
|
return (
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
)
|
|
|
|
def create_and_check_model(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = RemBertModel(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids)
|
|
result = model(input_ids, token_type_ids=token_type_ids)
|
|
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_model_as_decoder(
|
|
self,
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
):
|
|
config.add_cross_attention = True
|
|
model = RemBertModel(config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(
|
|
input_ids,
|
|
attention_mask=input_mask,
|
|
token_type_ids=token_type_ids,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
)
|
|
result = model(
|
|
input_ids,
|
|
attention_mask=input_mask,
|
|
token_type_ids=token_type_ids,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
)
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids)
|
|
self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
|
|
|
|
def create_and_check_for_causal_lm(
|
|
self,
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
):
|
|
model = RemBertForCausalLM(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
|
|
|
|
def create_and_check_for_masked_lm(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = RemBertForMaskedLM(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
|
|
|
|
def create_and_check_decoder_model_past_large_inputs(
|
|
self,
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
):
|
|
config.is_decoder = True
|
|
config.add_cross_attention = True
|
|
model = RemBertForCausalLM(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
|
|
# first forward pass
|
|
outputs = model(
|
|
input_ids,
|
|
attention_mask=input_mask,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
use_cache=True,
|
|
)
|
|
past_key_values = outputs.past_key_values
|
|
|
|
# create hypothetical multiple next token and extent to next_input_ids
|
|
next_tokens = ids_tensor((self.batch_size, 3), config.vocab_size)
|
|
next_mask = ids_tensor((self.batch_size, 3), vocab_size=2)
|
|
|
|
# append to next input_ids and
|
|
next_input_ids = torch.cat([input_ids, next_tokens], dim=-1)
|
|
next_attention_mask = torch.cat([input_mask, next_mask], dim=-1)
|
|
|
|
output_from_no_past = model(
|
|
next_input_ids,
|
|
attention_mask=next_attention_mask,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
output_hidden_states=True,
|
|
)["hidden_states"][0]
|
|
output_from_past = model(
|
|
next_tokens,
|
|
attention_mask=next_attention_mask,
|
|
encoder_hidden_states=encoder_hidden_states,
|
|
encoder_attention_mask=encoder_attention_mask,
|
|
past_key_values=past_key_values,
|
|
output_hidden_states=True,
|
|
)["hidden_states"][0]
|
|
|
|
# select random slice
|
|
random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item()
|
|
output_from_no_past_slice = output_from_no_past[:, -3:, random_slice_idx].detach()
|
|
output_from_past_slice = output_from_past[:, :, random_slice_idx].detach()
|
|
|
|
self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1])
|
|
|
|
# test that outputs are equal for slice
|
|
self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3))
|
|
|
|
def create_and_check_for_question_answering(
|
|
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
|
|
):
|
|
model = RemBertForQuestionAnswering(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(
|
|
input_ids,
|
|
attention_mask=input_mask,
|
|
token_type_ids=token_type_ids,
|
|
start_positions=sequence_labels,
|
|
end_positions=sequence_labels,
|
|
)
|
|
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 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 = RemBertForSequenceClassification(config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels)
|
|
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 = RemBertForTokenClassification(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
|
|
|
|
def create_and_check_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 = RemBertForMultipleChoice(config=config)
|
|
model.to(torch_device)
|
|
model.eval()
|
|
multiple_choice_inputs_ids = input_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
|
|
multiple_choice_token_type_ids = token_type_ids.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
|
|
multiple_choice_input_mask = input_mask.unsqueeze(1).expand(-1, self.num_choices, -1).contiguous()
|
|
result = model(
|
|
multiple_choice_inputs_ids,
|
|
attention_mask=multiple_choice_input_mask,
|
|
token_type_ids=multiple_choice_token_type_ids,
|
|
labels=choice_labels,
|
|
)
|
|
self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
|
|
|
|
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_torch
|
|
class RemBertModelTest(ModelTesterMixin, PipelineTesterMixin, unittest.TestCase):
|
|
all_model_classes = (
|
|
(
|
|
RemBertModel,
|
|
RemBertForMaskedLM,
|
|
RemBertForCausalLM,
|
|
RemBertForMultipleChoice,
|
|
RemBertForQuestionAnswering,
|
|
RemBertForSequenceClassification,
|
|
RemBertForTokenClassification,
|
|
)
|
|
if is_torch_available()
|
|
else ()
|
|
)
|
|
all_generative_model_classes = (RemBertForCausalLM,) if is_torch_available() else ()
|
|
pipeline_model_mapping = (
|
|
{
|
|
"feature-extraction": RemBertModel,
|
|
"fill-mask": RemBertForMaskedLM,
|
|
"question-answering": RemBertForQuestionAnswering,
|
|
"text-classification": RemBertForSequenceClassification,
|
|
"text-generation": RemBertForCausalLM,
|
|
"token-classification": RemBertForTokenClassification,
|
|
"zero-shot": RemBertForSequenceClassification,
|
|
}
|
|
if is_torch_available()
|
|
else {}
|
|
)
|
|
|
|
def setUp(self):
|
|
self.model_tester = RemBertModelTester(self)
|
|
self.config_tester = ConfigTester(self, config_class=RemBertConfig, 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_model_various_embeddings(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
for type in ["absolute", "relative_key", "relative_key_query"]:
|
|
config_and_inputs[0].position_embedding_type = type
|
|
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_multiple_choice(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
|
self.model_tester.create_and_check_for_multiple_choice(*config_and_inputs)
|
|
|
|
def test_decoder_model_past_with_large_inputs(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs_for_decoder()
|
|
self.model_tester.create_and_check_decoder_model_past_large_inputs(*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)
|
|
|
|
def test_model_as_decoder(self):
|
|
config_and_inputs = self.model_tester.prepare_config_and_inputs_for_decoder()
|
|
self.model_tester.create_and_check_model_as_decoder(*config_and_inputs)
|
|
|
|
def test_model_as_decoder_with_default_input_mask(self):
|
|
# This regression test was failing with PyTorch < 1.3
|
|
(
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
) = self.model_tester.prepare_config_and_inputs_for_decoder()
|
|
|
|
input_mask = None
|
|
|
|
self.model_tester.create_and_check_model_as_decoder(
|
|
config,
|
|
input_ids,
|
|
token_type_ids,
|
|
input_mask,
|
|
sequence_labels,
|
|
token_labels,
|
|
choice_labels,
|
|
encoder_hidden_states,
|
|
encoder_attention_mask,
|
|
)
|
|
|
|
@slow
|
|
def test_model_from_pretrained(self):
|
|
model_name = "google/rembert"
|
|
model = RemBertModel.from_pretrained(model_name)
|
|
self.assertIsNotNone(model)
|
|
|
|
|
|
@require_torch
|
|
class RemBertModelIntegrationTest(unittest.TestCase):
|
|
@slow
|
|
def test_inference_model(self):
|
|
# Test exact values at the last hidden layer
|
|
model = RemBertModel.from_pretrained("google/rembert")
|
|
input_ids = torch.tensor([[312, 56498, 313, 2125, 313]])
|
|
segment_ids = torch.tensor([[0, 0, 0, 1, 1]])
|
|
with torch.no_grad():
|
|
output = model(input_ids, token_type_ids=segment_ids, output_hidden_states=True)
|
|
|
|
hidden_size = 1152
|
|
|
|
expected_shape = torch.Size((1, 5, hidden_size))
|
|
self.assertEqual(output["last_hidden_state"].shape, expected_shape)
|
|
|
|
expected_implementation = torch.tensor(
|
|
[
|
|
[
|
|
[0.0754, -0.2022, 0.1904],
|
|
[-0.3354, -0.3692, -0.4791],
|
|
[-0.2314, -0.6729, -0.0749],
|
|
[-0.0396, -0.3105, -0.4234],
|
|
[-0.1571, -0.0525, 0.5353],
|
|
]
|
|
]
|
|
)
|
|
|
|
# Running on the original tf implementation gives slightly different results here.
|
|
# Not clear why this variations is present
|
|
# TODO: Find reason for discrepancy
|
|
# expected_original_implementation = [[
|
|
# [0.07630594074726105, -0.20146065950393677, 0.19107051193714142],
|
|
# [-0.3405614495277405, -0.36971670389175415, -0.4808273911476135],
|
|
# [-0.22587086260318756, -0.6656315922737122, -0.07844287157058716],
|
|
# [-0.04145475849509239, -0.3077218234539032, -0.42316967248916626],
|
|
# [-0.15887849032878876, -0.054529931396245956, 0.5356100797653198]
|
|
# ]]
|
|
|
|
self.assertTrue(torch.allclose(output["last_hidden_state"][:, :, :3], expected_implementation, atol=1e-4))
|