This repository was archived by the owner on Dec 16, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 2.2k
/
Copy pathpretrained_transformer_mismatched_embedder_test.py
273 lines (214 loc) · 10.3 KB
/
pretrained_transformer_mismatched_embedder_test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
import pytest
import torch
from allennlp.common import Params, cached_transformers
from allennlp.common.checks import ConfigurationError
from allennlp.data import Token, Vocabulary
from allennlp.data.batch import Batch
from allennlp.data.fields import TextField
from allennlp.data.instance import Instance
from allennlp.data.token_indexers import PretrainedTransformerMismatchedIndexer
from allennlp.modules.text_field_embedders import BasicTextFieldEmbedder
from allennlp.modules.token_embedders import PretrainedTransformerMismatchedEmbedder
from allennlp.common.testing import AllenNlpTestCase
class TestPretrainedTransformerMismatchedEmbedder(AllenNlpTestCase):
def teardown_method(self):
super().teardown_method()
cached_transformers._clear_caches()
@pytest.mark.parametrize("train_parameters", [True, False])
def test_end_to_end(self, train_parameters: bool):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased")
sentence1 = ["A", ",", "AllenNLP", "sentence", "."]
sentence2 = ["AllenNLP", "is", "great"]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
params = Params(
{
"token_embedders": {
"bert": {
"type": "pretrained_transformer_mismatched",
"model_name": "bert-base-uncased",
"train_parameters": train_parameters,
}
}
}
)
token_embedder = BasicTextFieldEmbedder.from_params(vocab=vocab, params=params)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
assert tokens["bert"]["offsets"].tolist() == [
[[1, 1], [2, 2], [3, 5], [6, 6], [7, 7]],
[[1, 3], [4, 4], [5, 5], [0, 0], [0, 0]],
]
# Attention mask
bert_vectors = token_embedder(tokens)
assert bert_vectors.size() == (2, max(len(sentence1), len(sentence2)), 768)
assert not torch.isnan(bert_vectors).any()
assert bert_vectors.requires_grad == train_parameters
def test_long_sequence_splitting_end_to_end(self):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased", max_length=4)
sentence1 = ["A", ",", "AllenNLP", "sentence", "."]
sentence2 = ["AllenNLP", "is", "great"]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
params = Params(
{
"token_embedders": {
"bert": {
"type": "pretrained_transformer_mismatched",
"model_name": "bert-base-uncased",
"max_length": 4,
}
}
}
)
token_embedder = BasicTextFieldEmbedder.from_params(vocab=vocab, params=params)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
assert tokens["bert"]["mask"].tolist() == [
[True, True, True, True, True],
[True, True, True, False, False],
]
assert tokens["bert"]["offsets"].tolist() == [
[[1, 1], [2, 2], [3, 5], [6, 6], [7, 7]],
[[1, 3], [4, 4], [5, 5], [0, 0], [0, 0]],
]
bert_vectors = token_embedder(tokens)
assert bert_vectors.size() == (2, max(len(sentence1), len(sentence2)), 768)
assert not torch.isnan(bert_vectors).any()
def test_token_without_wordpieces(self):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased")
sentence1 = ["A", "", "AllenNLP", "sentence", "."]
sentence2 = ["AllenNLP", "", "great"]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
params = Params(
{
"token_embedders": {
"bert": {
"type": "pretrained_transformer_mismatched",
"model_name": "bert-base-uncased",
}
}
}
)
token_embedder = BasicTextFieldEmbedder.from_params(vocab=vocab, params=params)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
assert tokens["bert"]["offsets"].tolist() == [
[[1, 1], [-1, -1], [2, 4], [5, 5], [6, 6]],
[[1, 3], [-1, -1], [4, 4], [0, 0], [0, 0]],
]
bert_vectors = token_embedder(tokens)
assert bert_vectors.size() == (2, max(len(sentence1), len(sentence2)), 768)
assert not torch.isnan(bert_vectors).any()
assert all(bert_vectors[0, 1] == 0)
assert all(bert_vectors[1, 1] == 0)
def test_exotic_tokens_no_nan_grads(self):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased")
sentence1 = ["A", "", "AllenNLP", "sentence", "."]
sentence2 = ["A", "\uf732\uf730\uf730\uf733", "AllenNLP", "sentence", "."]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
token_embedder = BasicTextFieldEmbedder(
{"bert": PretrainedTransformerMismatchedEmbedder("bert-base-uncased")}
)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
bert_vectors = token_embedder(tokens)
test_loss = bert_vectors.mean()
test_loss.backward()
for name, param in token_embedder.named_parameters():
grad = param.grad
assert (grad is None) or (not torch.any(torch.isnan(grad)).item())
@pytest.mark.parametrize("sub_token_mode", ("first", "avg"))
def test_end_to_end_for_first_sub_token_embedding(self, sub_token_mode: str):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased")
sentence1 = ["A", ",", "AllenNLP", "sentence", "."]
sentence2 = ["AllenNLP", "is", "open", "source", "NLP", "library"]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
params = Params(
{
"token_embedders": {
"bert": {
"type": "pretrained_transformer_mismatched",
"model_name": "bert-base-uncased",
"sub_token_mode": sub_token_mode,
}
}
}
)
token_embedder = BasicTextFieldEmbedder.from_params(vocab=vocab, params=params)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
assert tokens["bert"]["mask"].tolist() == [
[True, True, True, True, True, False],
[True, True, True, True, True, True],
]
assert tokens["bert"]["offsets"].tolist() == [
[[1, 1], [2, 2], [3, 5], [6, 6], [7, 7], [0, 0]],
[[1, 3], [4, 4], [5, 5], [6, 6], [7, 8], [9, 9]],
]
# Attention mask
bert_vectors = token_embedder(tokens)
assert bert_vectors.size() == (2, max(len(sentence1), len(sentence2)), 768)
assert not torch.isnan(bert_vectors).any()
@pytest.mark.parametrize("sub_token_mode", ["last"])
def test_throws_error_on_incorrect_sub_token_mode(self, sub_token_mode: str):
token_indexer = PretrainedTransformerMismatchedIndexer("bert-base-uncased")
sentence1 = ["A", ",", "AllenNLP", "sentence", "."]
sentence2 = ["AllenNLP", "is", "open", "source", "NLP", "library"]
tokens1 = [Token(word) for word in sentence1]
tokens2 = [Token(word) for word in sentence2]
vocab = Vocabulary()
params = Params(
{
"token_embedders": {
"bert": {
"type": "pretrained_transformer_mismatched",
"model_name": "bert-base-uncased",
"sub_token_mode": sub_token_mode,
}
}
}
)
token_embedder = BasicTextFieldEmbedder.from_params(vocab=vocab, params=params)
instance1 = Instance({"tokens": TextField(tokens1, {"bert": token_indexer})})
instance2 = Instance({"tokens": TextField(tokens2, {"bert": token_indexer})})
batch = Batch([instance1, instance2])
batch.index_instances(vocab)
padding_lengths = batch.get_padding_lengths()
tensor_dict = batch.as_tensor_dict(padding_lengths)
tokens = tensor_dict["tokens"]
with pytest.raises(ConfigurationError):
token_embedder(tokens)