"ggml/src/ggml-musa/CMakeLists.txt" did not exist on "4cc1a6143387f41e2466536abcd6a2620b63a35b"
test_modeling_albert.py 13.5 KB
Newer Older
Lysandre's avatar
Lysandre committed
1
# coding=utf-8
Sylvain Gugger's avatar
Sylvain Gugger committed
2
# Copyright 2020 The HuggingFace Team. All rights reserved.
Lysandre's avatar
Lysandre committed
3
4
5
6
7
8
9
10
11
12
13
14
#
# 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.
Aymeric Augustin's avatar
Aymeric Augustin committed
15

Lysandre's avatar
Lysandre committed
16

17
18
import unittest

19
from transformers import AlbertConfig, is_torch_available
20
from transformers.models.auto import get_values
21
from transformers.testing_utils import require_torch, slow, torch_device
Lysandre's avatar
Lysandre committed
22

Yih-Dar's avatar
Yih-Dar committed
23
24
from ...test_configuration_common import ConfigTester
from ...test_modeling_common import ModelTesterMixin, ids_tensor, random_attention_mask
Lysandre's avatar
Lysandre committed
25

Aymeric Augustin's avatar
Aymeric Augustin committed
26

Lysandre's avatar
Lysandre committed
27
if is_torch_available():
28
29
    import torch

30
    from transformers import (
31
        MODEL_FOR_PRETRAINING_MAPPING,
32
        AlbertForMaskedLM,
33
        AlbertForMultipleChoice,
34
35
        AlbertForPreTraining,
        AlbertForQuestionAnswering,
36
        AlbertForSequenceClassification,
37
        AlbertForTokenClassification,
38
        AlbertModel,
39
    )
Sylvain Gugger's avatar
Sylvain Gugger committed
40
    from transformers.models.albert.modeling_albert import ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST
Lysandre's avatar
Lysandre committed
41
42


43
44
class AlbertModelTester:
    def __init__(
Lysandre's avatar
Lysandre committed
45
46
        self,
        parent,
Yih-Dar's avatar
Yih-Dar committed
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
        batch_size=13,
        seq_length=7,
        is_training=True,
        use_input_mask=True,
        use_token_type_ids=True,
        use_labels=True,
        vocab_size=99,
        embedding_size=16,
        hidden_size=36,
        num_hidden_layers=6,
        num_hidden_groups=6,
        num_attention_heads=6,
        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,
70
71
    ):
        self.parent = parent
Yih-Dar's avatar
Yih-Dar committed
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
        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.embedding_size = embedding_size
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_hidden_groups = num_hidden_groups
        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
95
96
97
98
99
100

    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:
101
            input_mask = random_attention_mask([self.batch_size, self.seq_length])
102
103
104
105
106
107
108
109
110
111
112
113
114

        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)

115
116
117
118
119
120
        config = self.get_config()

        return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels

    def get_config(self):
        return AlbertConfig(
121
122
123
124
125
126
127
128
129
130
131
132
133
134
            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,
            num_hidden_groups=self.num_hidden_groups,
        )

135
    def create_and_check_model(
136
137
138
139
140
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        model = AlbertModel(config=config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
141
142
143
        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)
Stas Bekman's avatar
Stas Bekman committed
144
145
        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
        self.parent.assertEqual(result.pooler_output.shape, (self.batch_size, self.hidden_size))
146

147
    def create_and_check_for_pretraining(
148
149
150
151
152
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        model = AlbertForPreTraining(config=config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
153
        result = model(
154
155
156
157
158
159
            input_ids,
            attention_mask=input_mask,
            token_type_ids=token_type_ids,
            labels=token_labels,
            sentence_order_label=sequence_labels,
        )
Stas Bekman's avatar
Stas Bekman committed
160
161
        self.parent.assertEqual(result.prediction_logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
        self.parent.assertEqual(result.sop_logits.shape, (self.batch_size, config.num_labels))
162

163
    def create_and_check_for_masked_lm(
164
165
166
167
168
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        model = AlbertForMaskedLM(config=config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
169
        result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
Stas Bekman's avatar
Stas Bekman committed
170
        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
171

172
    def create_and_check_for_question_answering(
173
174
175
176
177
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        model = AlbertForQuestionAnswering(config=config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
178
        result = model(
179
180
181
182
183
184
            input_ids,
            attention_mask=input_mask,
            token_type_ids=token_type_ids,
            start_positions=sequence_labels,
            end_positions=sequence_labels,
        )
Stas Bekman's avatar
Stas Bekman committed
185
186
        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))
187

188
    def create_and_check_for_sequence_classification(
189
190
191
192
193
194
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        config.num_labels = self.num_labels
        model = AlbertForSequenceClassification(config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
195
        result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels)
Stas Bekman's avatar
Stas Bekman committed
196
        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_labels))
197

198
    def create_and_check_for_token_classification(
199
200
201
202
203
204
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        config.num_labels = self.num_labels
        model = AlbertForTokenClassification(config=config)
        model.to(torch_device)
        model.eval()
Sylvain Gugger's avatar
Sylvain Gugger committed
205
        result = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
Stas Bekman's avatar
Stas Bekman committed
206
        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.num_labels))
207

208
    def create_and_check_for_multiple_choice(
209
210
211
212
213
214
215
216
217
        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
    ):
        config.num_choices = self.num_choices
        model = AlbertForMultipleChoice(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()
Sylvain Gugger's avatar
Sylvain Gugger committed
218
        result = model(
219
220
221
222
223
            multiple_choice_inputs_ids,
            attention_mask=multiple_choice_input_mask,
            token_type_ids=multiple_choice_token_type_ids,
            labels=choice_labels,
        )
Stas Bekman's avatar
Stas Bekman committed
224
        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.num_choices))
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240

    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


241
@require_torch
242
class AlbertModelTest(ModelTesterMixin, unittest.TestCase):
Lysandre's avatar
Lysandre committed
243

244
245
246
247
248
249
250
251
252
253
254
255
256
    all_model_classes = (
        (
            AlbertModel,
            AlbertForPreTraining,
            AlbertForMaskedLM,
            AlbertForMultipleChoice,
            AlbertForSequenceClassification,
            AlbertForTokenClassification,
            AlbertForQuestionAnswering,
        )
        if is_torch_available()
        else ()
    )
257
    fx_compatible = True
Lysandre's avatar
Lysandre committed
258

259
260
261
262
263
    # special case for ForPreTraining model
    def _prepare_for_class(self, inputs_dict, model_class, return_labels=False):
        inputs_dict = super()._prepare_for_class(inputs_dict, model_class, return_labels=return_labels)

        if return_labels:
264
            if model_class in get_values(MODEL_FOR_PRETRAINING_MAPPING):
265
266
267
268
269
270
271
272
                inputs_dict["labels"] = torch.zeros(
                    (self.model_tester.batch_size, self.model_tester.seq_length), dtype=torch.long, device=torch_device
                )
                inputs_dict["sentence_order_label"] = torch.zeros(
                    self.model_tester.batch_size, dtype=torch.long, device=torch_device
                )
        return inputs_dict

Lysandre's avatar
Lysandre committed
273
    def setUp(self):
274
        self.model_tester = AlbertModelTester(self)
Lysandre's avatar
Lysandre committed
275
276
277
278
279
        self.config_tester = ConfigTester(self, config_class=AlbertConfig, hidden_size=37)

    def test_config(self):
        self.config_tester.run_common_tests()

280
    def test_model(self):
Lysandre's avatar
Lysandre committed
281
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
282
        self.model_tester.create_and_check_model(*config_and_inputs)
Lysandre's avatar
Lysandre committed
283

284
285
    def test_for_pretraining(self):
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
286
        self.model_tester.create_and_check_for_pretraining(*config_and_inputs)
287

Lysandre's avatar
Lysandre committed
288
289
    def test_for_masked_lm(self):
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
290
        self.model_tester.create_and_check_for_masked_lm(*config_and_inputs)
Lysandre's avatar
Lysandre committed
291

292
293
    def test_for_multiple_choice(self):
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
294
        self.model_tester.create_and_check_for_multiple_choice(*config_and_inputs)
295

296
297
    def test_for_question_answering(self):
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
298
        self.model_tester.create_and_check_for_question_answering(*config_and_inputs)
299
300
301

    def test_for_sequence_classification(self):
        config_and_inputs = self.model_tester.prepare_config_and_inputs()
302
303
304
305
306
307
308
        self.model_tester.create_and_check_for_sequence_classification(*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)
309

310
    @slow
Lysandre's avatar
Lysandre committed
311
    def test_model_from_pretrained(self):
312
        for model_name in ALBERT_PRETRAINED_MODEL_ARCHIVE_LIST[:1]:
313
            model = AlbertModel.from_pretrained(model_name)
Lysandre's avatar
Lysandre committed
314
            self.assertIsNotNone(model)
315
316
317
318
319
320


@require_torch
class AlbertModelIntegrationTest(unittest.TestCase):
    @slow
    def test_inference_no_head_absolute_embedding(self):
321
        model = AlbertModel.from_pretrained("albert-base-v2")
322
        input_ids = torch.tensor([[0, 345, 232, 328, 740, 140, 1695, 69, 6078, 1588, 2]])
323
        attention_mask = torch.tensor([[0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]])
324
325
        with torch.no_grad():
            output = model(input_ids, attention_mask=attention_mask)[0]
326
        expected_shape = torch.Size((1, 11, 768))
327
328
        self.assertEqual(output.shape, expected_shape)
        expected_slice = torch.tensor(
329
            [[[-0.6513, 1.5035, -0.2766], [-0.6515, 1.5046, -0.2780], [-0.6512, 1.5049, -0.2784]]]
330
331
        )

332
        self.assertTrue(torch.allclose(output[:, 1:4, 1:4], expected_slice, atol=1e-4))