Repository navigation
Expand file tree
/
Copy pathtest_tokenizer.py
More file actions
156 lines (138 loc) Β· 5.82 KB
/
Copy pathtest_tokenizer.py
File metadata and controls
156 lines (138 loc) Β· 5.82 KB
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
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
# Copyright 2020 The HuggingFace 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.
import json
import os
import unittest
from paddlenlp.transformers import BartTokenizer
from ..test_tokenizer_common import TokenizerTesterMixin, filter_roberta_detectors
VOCAB_FILES_NAMES = {
"vocab_file": "vocab.json",
"merges_file": "merges.txt",
}
class TestTokenizationBart(TokenizerTesterMixin, unittest.TestCase):
tokenizer_class = BartTokenizer
test_rust_tokenizer = False
test_offsets = False
from_pretrained_filter = filter_roberta_detectors
def setUp(self):
super().setUp()
vocab = [
"l",
"o",
"w",
"e",
"r",
"s",
"t",
"i",
"d",
"n",
"\u0120",
"\u0120l",
"\u0120n",
"\u0120lo",
"\u0120low",
"er",
"\u0120lowest",
"\u0120newer",
"\u0120wider",
"<unk>",
"<s>",
"</s>",
"<pad>",
"<mask>",
]
vocab_tokens = dict(zip(vocab, range(len(vocab))))
merges = ["#version: 0.2", "\u0120 l", "\u0120l o", "\u0120lo w", "e r", ""]
self.special_tokens_map = {
"bos_token": "<s>",
"eos_token": "</s>",
"cls_token": "<s>",
"sep_token": "</s>",
"unk_token": "<unk>",
"pad_token": "<pad>",
"mask_token": "<mask>",
}
self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"])
self.merges_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["merges_file"])
with open(self.vocab_file, "w", encoding="utf-8") as fp:
fp.write(json.dumps(vocab_tokens) + "\n")
with open(self.merges_file, "w", encoding="utf-8") as fp:
fp.write("\n".join(merges))
def get_tokenizer(self, **kwargs):
kwargs.update(self.special_tokens_map)
return self.tokenizer_class.from_pretrained(self.tmpdirname, **kwargs)
def get_input_output_texts(self, tokenizer):
return "lower newer", "lower newer"
def default_tokenizer(self):
return BartTokenizer.from_pretrained("bart-large")
def test_prepare_batch(self):
src_text = ["A long paragraph for summarization.", "Another paragraph for summarization."]
expected_src_tokens = [0, 250, 251, 17818, 13, 39186, 1938, 4, 2]
for tokenizer in [BartTokenizer.from_pretrained("bart-large")]:
batch = tokenizer(
text=src_text,
max_length=len(expected_src_tokens),
padding=True,
return_attention_mask=True,
return_tensors="pd",
)
self.assertEqual([2, 9], batch.input_ids.shape)
self.assertEqual([2, 9], batch.attention_mask.shape)
result = batch.input_ids.tolist()[0]
self.assertListEqual(expected_src_tokens, result)
# Test that special tokens are reset
def test_prepare_batch_empty_target_text(self):
src_text = ["A long paragraph for summarization.", "Another paragraph for summarization."]
for tokenizer in [BartTokenizer.from_pretrained("bart-large")]:
batch = tokenizer(text=src_text, padding=True, return_tensors="pd", return_attention_mask=True)
# check if input_ids are returned and no labels
self.assertIn("input_ids", batch)
self.assertIn("attention_mask", batch)
self.assertNotIn("labels", batch)
self.assertNotIn("decoder_attention_mask", batch)
def test_tokenizer_as_target_length(self):
tgt_text = [
"Summary of the text.",
"Another summary.",
]
for tokenizer in [BartTokenizer.from_pretrained("bart-large")]:
targets = tokenizer(text=tgt_text, max_length=32, padding="max_length", return_tensors="pd")
self.assertEqual(32, targets["input_ids"].shape[1])
def test_prepare_batch_not_longer_than_maxlen(self):
for tokenizer in [BartTokenizer.from_pretrained("bart-large", max_len=1024)]:
batch = tokenizer(
text=["I am a small frog" * 1024, "I am a small frog"],
padding=True,
truncation=True,
return_tensors="pd",
)
self.assertEqual(batch.input_ids.shape, [2, 1024])
def test_special_tokens(self):
src_text = ["A long paragraph for summarization."]
tgt_text = [
"Summary of the text.",
]
for tokenizer in [BartTokenizer.from_pretrained("bart-large")]:
inputs = tokenizer(text=src_text, return_tensors="pd")
targets = tokenizer(text=tgt_text, return_tensors="pd")
input_ids = inputs["input_ids"]
labels = targets["input_ids"]
self.assertTrue((input_ids[:, 0] == tokenizer.bos_token_id).all().item())
self.assertTrue((labels[:, 0] == tokenizer.bos_token_id).all().item())
self.assertTrue((input_ids[:, -1] == tokenizer.eos_token_id).all().item())
self.assertTrue((labels[:, -1] == tokenizer.eos_token_id).all().item())
def test_pretokenized_inputs(self):
pass