wehub-resource-sync
410 lines
17 KiB
Python
410 lines
17 KiB
Python
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
|
|
# Copyright 2019 Hugging Face inc.
|
|
#
|
|
# 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 os
|
|
import unittest
|
|
|
|
from paddlenlp.transformers.albert.tokenizer import (
|
|
AlbertChineseTokenizer,
|
|
AlbertEnglishTokenizer,
|
|
)
|
|
from paddlenlp.transformers.bert.tokenizer import BasicTokenizer, WordpieceTokenizer
|
|
|
|
from ...testing_utils import get_tests_dir, slow
|
|
from ..test_tokenizer_common import TokenizerTesterMixin, filter_non_english
|
|
|
|
SAMPLE_VOCAB = get_tests_dir("fixtures/spiece.model")
|
|
|
|
|
|
class AlbertEnglishTokenizerTest(TokenizerTesterMixin, unittest.TestCase):
|
|
tokenizer_class = AlbertEnglishTokenizer
|
|
from_pretrained_vocab_key = "sentencepiece_model_file"
|
|
test_sentencepiece = True
|
|
test_sentencepiece_ignore_case = True
|
|
|
|
def setUp(self):
|
|
super().setUp()
|
|
|
|
# We have a SentencePiece fixture for testing
|
|
tokenizer = AlbertEnglishTokenizer(SAMPLE_VOCAB)
|
|
tokenizer.save_pretrained(self.tmpdirname)
|
|
|
|
def get_input_output_texts(self, tokenizer):
|
|
input_text = "this is a test"
|
|
output_text = "this is a test"
|
|
return input_text, output_text
|
|
|
|
def test_convert_token_and_id(self):
|
|
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
|
|
token = "<pad>"
|
|
token_id = 0
|
|
|
|
self.assertEqual(self.get_tokenizer()._convert_token_to_id(token), token_id)
|
|
self.assertEqual(self.get_tokenizer()._convert_id_to_token(token_id), token)
|
|
|
|
def test_get_vocab(self):
|
|
vocab_keys = list(self.get_tokenizer().get_vocab().keys())
|
|
self.assertEqual(vocab_keys[0], "<pad>")
|
|
self.assertEqual(vocab_keys[1], "<unk>")
|
|
self.assertEqual(vocab_keys[-1], "▁eloquent")
|
|
self.assertEqual(len(vocab_keys), 30_000)
|
|
|
|
def test_vocab_size(self):
|
|
self.assertEqual(self.get_tokenizer().vocab_size, 30_000)
|
|
|
|
def test_full_tokenizer(self):
|
|
tokenizer = AlbertEnglishTokenizer(SAMPLE_VOCAB, keep_accents=True)
|
|
|
|
tokens = tokenizer.tokenize("This is a test")
|
|
self.assertListEqual(tokens, ["▁this", "▁is", "▁a", "▁test"])
|
|
|
|
self.assertListEqual(tokenizer.convert_tokens_to_ids(tokens), [48, 25, 21, 1289])
|
|
|
|
tokens = tokenizer.tokenize("I was born in 92000, and this is falsé.")
|
|
self.assertListEqual(
|
|
tokens, ["▁i", "▁was", "▁born", "▁in", "▁9", "2000", ",", "▁and", "▁this", "▁is", "▁fal", "s", "é", "."]
|
|
)
|
|
ids = tokenizer.convert_tokens_to_ids(tokens)
|
|
self.assertListEqual(ids, [31, 23, 386, 19, 561, 3050, 15, 17, 48, 25, 8256, 18, 1, 9])
|
|
|
|
back_tokens = tokenizer.convert_ids_to_tokens(ids)
|
|
self.assertListEqual(
|
|
back_tokens,
|
|
["▁i", "▁was", "▁born", "▁in", "▁9", "2000", ",", "▁and", "▁this", "▁is", "▁fal", "s", "<unk>", "."],
|
|
)
|
|
|
|
def test_sequence_builders(self):
|
|
tokenizer = AlbertEnglishTokenizer(SAMPLE_VOCAB)
|
|
|
|
text = tokenizer.encode("sequence builders")["input_ids"]
|
|
text_2 = tokenizer.encode("multi-sequence build")["input_ids"]
|
|
|
|
encoded_sentence = tokenizer.build_inputs_with_special_tokens(text)
|
|
encoded_pair = tokenizer.build_inputs_with_special_tokens(text, text_2)
|
|
|
|
assert encoded_sentence == [tokenizer.cls_token_id] + text + [tokenizer.sep_token_id]
|
|
assert encoded_pair == [tokenizer.cls_token_id] + text + [tokenizer.sep_token_id] + text_2 + [
|
|
tokenizer.sep_token_id
|
|
]
|
|
|
|
@slow
|
|
def test_tokenizer_integration(self):
|
|
# fmt: off
|
|
expected_encoding = {
|
|
'attention_mask': [
|
|
[
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
|
|
],
|
|
[
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
|
|
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
|
],
|
|
[
|
|
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
|
]],
|
|
'input_ids': [
|
|
[
|
|
2, 21970, 13, 5, 6092, 167, 28, 7103, 2153, 673, 8, 7028, 12051,
|
|
18, 17, 7103, 2153, 673, 8, 3515, 18684, 8, 4461, 6, 1927, 297,
|
|
8, 12060, 2607, 18, 13, 5, 4461, 15, 10538, 38, 8, 135, 15, 822,
|
|
58, 15, 993, 10363, 15, 1460, 8005, 4461, 15, 993, 255, 2328, 9,
|
|
9, 9, 6, 26, 1112, 816, 3260, 13, 5, 103, 2377, 6, 17, 1112,
|
|
816, 2782, 13, 5, 103, 10641, 6, 29, 84, 2512, 2430, 782, 18684,
|
|
2761, 19, 808, 2430, 2556, 17, 855, 1480, 9477, 4091, 128,
|
|
11712, 15, 7103, 2153, 673, 17, 24883, 9990, 9, 3
|
|
],
|
|
[
|
|
2, 11502, 25, 1006, 20, 782, 8, 11809, 855, 1732,
|
|
19393, 18667, 37, 367, 21018, 69, 1854, 34, 11860,
|
|
19124, 27, 156, 225, 17, 193, 4141, 19, 65, 9124,
|
|
9, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0
|
|
],
|
|
[
|
|
2, 14, 2231, 886, 2385, 17659, 84, 14, 16792,
|
|
1952, 9, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0
|
|
]],
|
|
'token_type_ids': [
|
|
[
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
|
],
|
|
[
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
|
],
|
|
[
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
|
|
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
|
|
]]
|
|
} # noqa: E501
|
|
# fmt: on
|
|
|
|
self.tokenizer_integration_test_util(
|
|
expected_encoding=expected_encoding,
|
|
model_name="albert-base-v2",
|
|
)
|
|
|
|
|
|
class AlbertChineseTokenizerTest(TokenizerTesterMixin, unittest.TestCase):
|
|
|
|
tokenizer_class = AlbertChineseTokenizer
|
|
space_between_special_tokens = True
|
|
from_pretrained_filter = filter_non_english
|
|
test_seq2seq = True
|
|
|
|
def setUp(self):
|
|
super().setUp()
|
|
|
|
vocab_tokens = [
|
|
"[UNK]",
|
|
"[CLS]",
|
|
"[SEP]",
|
|
"[PAD]",
|
|
"[MASK]",
|
|
"want",
|
|
"##want",
|
|
"##ed",
|
|
"wa",
|
|
"un",
|
|
"runn",
|
|
"##ing",
|
|
",",
|
|
"low",
|
|
"lowest",
|
|
]
|
|
|
|
self.vocab_file = os.path.join(self.tmpdirname, AlbertChineseTokenizer.resource_files_names["vocab_file"])
|
|
with open(self.vocab_file, "w", encoding="utf-8") as vocab_writer:
|
|
vocab_writer.write("".join([x + "\n" for x in vocab_tokens]))
|
|
|
|
def get_input_output_texts(self, tokenizer):
|
|
input_text = "UNwant\u00E9d,running"
|
|
output_text = "unwanted, running"
|
|
return input_text, output_text
|
|
|
|
def test_full_tokenizer(self):
|
|
tokenizer = self.tokenizer_class(self.vocab_file)
|
|
|
|
tokens = tokenizer.tokenize("UNwant\u00E9d,running")
|
|
self.assertListEqual(tokens, ["un", "##want", "##ed", ",", "runn", "##ing"])
|
|
self.assertListEqual(tokenizer.convert_tokens_to_ids(tokens), [9, 6, 7, 12, 10, 11])
|
|
|
|
def test_chinese(self):
|
|
tokenizer = BasicTokenizer()
|
|
|
|
self.assertListEqual(tokenizer.tokenize("ah\u535A\u63A8zz"), ["ah", "\u535A", "\u63A8", "zz"])
|
|
|
|
def test_basic_tokenizer_lower(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=True)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? "), ["hello", "!", "how", "are", "you", "?"]
|
|
)
|
|
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
|
|
|
def test_basic_tokenizer_lower_strip_accents_false(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=True, strip_accents=False)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hällo", "!", "how", "are", "you", "?"]
|
|
)
|
|
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["h\u00E9llo"])
|
|
|
|
def test_basic_tokenizer_lower_strip_accents_true(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=True)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hallo", "!", "how", "are", "you", "?"]
|
|
)
|
|
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
|
|
|
def test_basic_tokenizer_lower_strip_accents_default(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=True)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["hallo", "!", "how", "are", "you", "?"]
|
|
)
|
|
self.assertListEqual(tokenizer.tokenize("H\u00E9llo"), ["hello"])
|
|
|
|
def test_basic_tokenizer_no_lower(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=False)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? "), ["HeLLo", "!", "how", "Are", "yoU", "?"]
|
|
)
|
|
|
|
def test_basic_tokenizer_no_lower_strip_accents_false(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=False, strip_accents=False)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["HäLLo", "!", "how", "Are", "yoU", "?"]
|
|
)
|
|
|
|
def test_basic_tokenizer_no_lower_strip_accents_true(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=False, strip_accents=True)
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHäLLo!how \n Are yoU? "), ["HaLLo", "!", "how", "Are", "yoU", "?"]
|
|
)
|
|
|
|
def test_basic_tokenizer_respects_never_split_tokens(self):
|
|
tokenizer = BasicTokenizer(do_lower_case=False, never_split=["[UNK]"])
|
|
|
|
self.assertListEqual(
|
|
tokenizer.tokenize(" \tHeLLo!how \n Are yoU? [UNK]"), ["HeLLo", "!", "how", "Are", "yoU", "?", "[UNK]"]
|
|
)
|
|
|
|
def test_wordpiece_tokenizer(self):
|
|
vocab_tokens = ["[UNK]", "[CLS]", "[SEP]", "want", "##want", "##ed", "wa", "un", "runn", "##ing"]
|
|
|
|
vocab = {}
|
|
for (i, token) in enumerate(vocab_tokens):
|
|
vocab[token] = i
|
|
tokenizer = WordpieceTokenizer(vocab=vocab, unk_token="[UNK]")
|
|
|
|
self.assertListEqual(tokenizer.tokenize(""), [])
|
|
|
|
self.assertListEqual(tokenizer.tokenize("unwanted running"), ["un", "##want", "##ed", "runn", "##ing"])
|
|
|
|
self.assertListEqual(tokenizer.tokenize("unwantedX running"), ["[UNK]", "runn", "##ing"])
|
|
|
|
def test_clean_text(self):
|
|
tokenizer = self.get_tokenizer()
|
|
|
|
# Example taken from the issue https://github.com/huggingface/tokenizers/issues/340
|
|
self.assertListEqual([tokenizer.tokenize(t) for t in ["Test", "\xad", "test"]], [["[UNK]"], [], ["[UNK]"]])
|
|
|
|
@slow
|
|
def test_sequence_builders(self):
|
|
tokenizer = self.tokenizer_class.from_pretrained("albert-chinese-base")
|
|
|
|
text = tokenizer.encode("sequence builders", return_token_type_ids=None, add_special_tokens=False)["input_ids"]
|
|
text_2 = tokenizer.encode("multi-sequence build", return_token_type_ids=None, add_special_tokens=False)[
|
|
"input_ids"
|
|
]
|
|
|
|
encoded_sentence = tokenizer.build_inputs_with_special_tokens(text)
|
|
encoded_pair = tokenizer.build_inputs_with_special_tokens(text, text_2)
|
|
|
|
assert encoded_sentence == [101] + text + [102]
|
|
assert encoded_pair == [101] + text + [102] + text_2 + [102]
|
|
|
|
def test_offsets_with_special_characters(self):
|
|
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
|
|
with self.subTest(f"{tokenizer.__class__.__name__} ({pretrained_name})"):
|
|
tokenizer = self.tokenizer_class.from_pretrained(pretrained_name, **kwargs)
|
|
|
|
sentence = f"A, naïve {tokenizer.mask_token} AllenNLP sentence."
|
|
tokens = tokenizer.encode(
|
|
sentence,
|
|
return_attention_mask=False,
|
|
return_token_type_ids=False,
|
|
return_offsets_mapping=True,
|
|
add_special_tokens=True,
|
|
)
|
|
|
|
do_lower_case = tokenizer.do_lower_case if hasattr(tokenizer, "do_lower_case") else False
|
|
expected_results = (
|
|
[
|
|
((0, 0), tokenizer.cls_token),
|
|
((0, 1), "A"),
|
|
((1, 2), ","),
|
|
((3, 5), "na"),
|
|
((5, 6), "##ï"),
|
|
((6, 8), "##ve"),
|
|
((9, 15), tokenizer.mask_token),
|
|
((16, 21), "Allen"),
|
|
((21, 23), "##NL"),
|
|
((23, 24), "##P"),
|
|
((25, 33), "sentence"),
|
|
((33, 34), "."),
|
|
((0, 0), tokenizer.sep_token),
|
|
]
|
|
if not do_lower_case
|
|
else [
|
|
((0, 0), tokenizer.cls_token),
|
|
((0, 1), "a"),
|
|
((1, 2), ","),
|
|
((3, 8), "naive"),
|
|
((9, 15), tokenizer.mask_token),
|
|
((16, 21), "allen"),
|
|
((21, 23), "##nl"),
|
|
((23, 24), "##p"),
|
|
((25, 33), "sentence"),
|
|
((33, 34), "."),
|
|
((0, 0), tokenizer.sep_token),
|
|
]
|
|
)
|
|
|
|
self.assertEqual(
|
|
[e[1] for e in expected_results], tokenizer.convert_ids_to_tokens(tokens["input_ids"])
|
|
)
|
|
self.assertEqual([e[0] for e in expected_results], tokens["offset_mapping"])
|
|
|
|
def test_change_tokenize_chinese_chars(self):
|
|
list_of_commun_chinese_char = ["的", "人", "有"]
|
|
text_with_chinese_char = "".join(list_of_commun_chinese_char)
|
|
for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
|
|
with self.subTest(f"{tokenizer.__class__.__name__} ({pretrained_name})"):
|
|
|
|
kwargs["tokenize_chinese_chars"] = True
|
|
tokenizer = self.tokenizer_class.from_pretrained(pretrained_name, **kwargs)
|
|
|
|
ids_without_spe_char_p = tokenizer.encode(
|
|
text_with_chinese_char, return_token_type_ids=None, add_special_tokens=False
|
|
)["input_ids"]
|
|
|
|
tokens_without_spe_char_p = tokenizer.convert_ids_to_tokens(ids_without_spe_char_p)
|
|
|
|
# it is expected that each Chinese character is not preceded by "##"
|
|
self.assertListEqual(tokens_without_spe_char_p, list_of_commun_chinese_char)
|