Skip to content

GitLab

Unverified Commit fcad8018 authored by Philip May's avatar Philip May Committed by GitHub
Browse files

Add regression tests for slow sentencepiece tokenizers. (#11737) 

* add test_vocab_size for sentencepiece tok.

* add test_get_vocab for sentencepiece tok.

* add test_convert_token_and_id for sentencepiece tok.

* add test_tokenize_and_convert_tokens_to_string for all tok.

* improve test_tokenize_and_convert_tokens_to_string for sp. tok.

* add common tokenizer integration tests
- for albert
- for barthez

* add tokenizer integration tests to bert gen.

* add most tokenizer integration tests

* fix camembert tokenizer integration test

* add tokenizer integration test to marian

* add tokenizer integration test to reformer

* add typing and doc to tokenizer_integration_test_util

* fix tokenizer integration test of reformer

* improve test_sentencepiece_tokenize_and_convert_tokens_to_string

* empty commit to trigger CI

* fix tokenizer integration test of reformer

* remove code not needed anymore

* empty commit to trigger CI

* empty commit to trigger CI
parent c3d958b2
Hide whitespace changes
Inline Side-by-side
Showing with 626 additions and 113 deletions
tests/test_tokenization_albert.py
View file @ fcad8018
......@@ -47,6 +47,25 @@ class AlbertTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
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_rust_and_python_full_tokenizers(self):
if not self.test_rust_tokenizer:
return
......@@ -106,47 +125,12 @@ class AlbertTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
@slow
def test_tokenizer_integration(self):
tokenizer_classes = [self.tokenizer_class]
if self.test_rust_tokenizer:
tokenizer_classes.append(self.rust_tokenizer_class)
for tokenizer_class in tokenizer_classes:
tokenizer = tokenizer_class.from_pretrained("albert-base-v2")
sequences = [
"ALBERT: A Lite BERT for Self-supervised Learning of Language Representations",
"ALBERT incorporates two parameter reduction techniques",
"The first one is a factorized embedding parameterization. By decomposing the large vocabulary embedding matrix into two small matrices, we separate the size of the hidden layers from the size of vocabulary embedding.", # noqa: E231
]
encoding = tokenizer(sequences, padding=True)
decoded_sequences = [tokenizer.decode(seq, skip_special_tokens=True) for seq in encoding["input_ids"]]
# fmt: off
expected_encoding = {
'input_ids': [
[2, 2953, 45, 21, 13, 10601, 11502, 26, 1119, 8, 8542, 3762, 69, 2477, 16, 816, 18667, 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], # noqa: E231
[2, 2953, 13760, 81, 18906, 5895, 4212, 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], # noqa: E231
[2, 14, 64, 53, 25, 21, 3932, 1333, 11911, 69, 3258, 18906, 1829, 9, 34, 121, 960, 14717, 14, 370, 18630, 11911, 69, 3258, 8187, 77, 81, 284, 24849, 15, 95, 1725, 14, 1072, 16, 14, 3689, 9124, 37, 14, 1072, 16, 18630, 11911, 69, 3258, 9, 3]], # noqa: E231
'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], # noqa: E231
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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: E231
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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: E231
'attention_mask': [
[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], # noqa: E231
[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], # noqa: E231
[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] # noqa: E231
]
}
expected_decoded_sequence = [
"albert: a lite bert for self-supervised learning of language representations",
'albert incorporates two parameter reduction techniques',
'the first one is a factorized embedding parameterization. by decomposing the large vocabulary embedding matrix into two small matrices, we separate the size of the hidden layers from the size of vocabulary embedding.' # noqa: E231
]
# fmt: on
self.assertDictEqual(encoding.data, expected_encoding)
for expected, decoded in zip(expected_decoded_sequence, decoded_sequences):
self.assertEqual(expected, decoded)
# 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",
revision="6b6560eaf5ff2e250b00c50f380c5389a9c2d82e",
)
tests/test_tokenization_barthez.py
View file @ fcad8018
......@@ -39,6 +39,25 @@ class BarthezTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer.save_pretrained(self.tmpdirname, legacy_format=False)
self.tokenizer = tokenizer
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<pad>"
token_id = 1
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], "<s>")
self.assertEqual(vocab_keys[1], "<pad>")
self.assertEqual(vocab_keys[-1], "<mask>")
self.assertEqual(len(vocab_keys), 101_122)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 101_122)
@require_torch
def test_prepare_batch(self):
src_text = ["A long paragraph for summarization.", "Another paragraph for summarization."]
......@@ -75,3 +94,25 @@ class BarthezTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
ids = tokenizer.encode(sequence)
rust_ids = rust_tokenizer.encode(sequence)
self.assertListEqual(ids, rust_ids)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[0, 490, 14328, 4507, 354, 47, 43669, 95, 25, 78117, 20215, 19779, 190, 22, 400, 4, 35343, 80310, 603, 86, 24937, 105, 33438, 94762, 196, 39642, 7, 15, 15933, 173, 2, 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, 10534, 87, 25, 66, 3358, 196, 55289, 8, 82961, 81, 2204, 75203, 7, 15, 763, 12956, 216, 178, 14328, 9595, 1377, 69693, 7, 448, 71021, 196, 18106, 1437, 13974, 108, 9083, 4, 49315, 7, 39, 86, 1326, 2793, 46333, 4, 448, 196, 74588, 7, 49315, 7, 39, 21, 822, 38470, 74, 21, 66723, 62480, 8, 22050, 5, 2]], '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, 0, 0, 0, 0, 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, 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]]} # noqa: E501
# fmt: on
# moussaKam/mbarthez is a french model. So we also use french texts.
sequences = [
"Le transformeur est un modèle d'apprentissage profond introduit en 2017, "
"utilisé principalement dans le domaine du traitement automatique des langues (TAL).",
"À l'instar des réseaux de neurones récurrents (RNN), les transformeurs sont conçus "
"pour gérer des données séquentielles, telles que le langage naturel, pour des tâches "
"telles que la traduction et la synthèse de texte.",
]
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="moussaKam/mbarthez",
revision="c2e4ecbca5e3cd2c37fe1ac285ca4fbdf1366fb6",
sequences=sequences,
)
tests/test_tokenization_bert_generation.py
View file @ fcad8018
......@@ -40,6 +40,25 @@ class BertGenerationTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = BertGenerationTokenizer(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
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], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "<pad>")
self.assertEqual(len(vocab_keys), 1_002)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_000)
def test_full_tokenizer(self):
tokenizer = BertGenerationTokenizer(SAMPLE_VOCAB, keep_accents=True)
......@@ -209,3 +228,15 @@ class BertGenerationTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
with torch.no_grad():
model(**encoded_sequence)
model(**batch_encoded_sequence)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[39286, 458, 36335, 2001, 456, 13073, 13266, 455, 113, 7746, 1741, 11157, 391, 13073, 13266, 455, 113, 3967, 35412, 113, 4936, 109, 3870, 2377, 113, 30084, 45720, 458, 134, 17496, 112, 503, 11672, 113, 118, 112, 5665, 13347, 38687, 112, 1496, 31389, 112, 3268, 47264, 134, 962, 112, 16377, 8035, 23130, 430, 12169, 15518, 28592, 458, 146, 41697, 109, 391, 12169, 15518, 16689, 458, 146, 41358, 109, 452, 726, 4034, 111, 763, 35412, 5082, 388, 1903, 111, 9051, 391, 2870, 48918, 1900, 1123, 550, 998, 112, 9586, 15985, 455, 391, 410, 22955, 37636, 114], [448, 17496, 419, 3663, 385, 763, 113, 27533, 2870, 3283, 13043, 1639, 24713, 523, 656, 24013, 18550, 2521, 517, 27014, 21244, 420, 1212, 1465, 391, 927, 4833, 388, 578, 11786, 114, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [484, 2169, 7687, 21932, 18146, 726, 363, 17032, 3391, 114, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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="google/bert_for_seq_generation_L-24_bbc_encoder",
revision="c817d1fd1be2ffa69431227a1fe320544943d4db",
)
tests/test_tokenization_big_bird.py
View file @ fcad8018
......@@ -43,6 +43,25 @@ class BigBirdTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = self.tokenizer_class(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
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], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "[MASK]")
self.assertEqual(len(vocab_keys), 1_004)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_000)
def test_rust_and_python_full_tokenizers(self):
if not self.test_rust_tokenizer:
return
......@@ -202,3 +221,15 @@ class BigBirdTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
decoded_text = tokenizer.decode(tokenizer("Paris is the [MASK].").input_ids)
self.assertTrue(decoded_text == "[CLS] Paris is the [MASK].[SEP]")
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[65, 39286, 458, 36335, 2001, 456, 13073, 13266, 455, 113, 7746, 1741, 11157, 391, 13073, 13266, 455, 113, 3967, 35412, 113, 4936, 109, 3870, 2377, 113, 30084, 45720, 458, 134, 17496, 112, 503, 11672, 113, 118, 112, 5665, 13347, 38687, 112, 1496, 31389, 112, 3268, 47264, 134, 962, 112, 16377, 8035, 23130, 430, 12169, 15518, 28592, 458, 146, 41697, 109, 391, 12169, 15518, 16689, 458, 146, 41358, 109, 452, 726, 4034, 111, 763, 35412, 5082, 388, 1903, 111, 9051, 391, 2870, 48918, 1900, 1123, 550, 998, 112, 9586, 15985, 455, 391, 410, 22955, 37636, 114, 66], [65, 448, 17496, 419, 3663, 385, 763, 113, 27533, 2870, 3283, 13043, 1639, 24713, 523, 656, 24013, 18550, 2521, 517, 27014, 21244, 420, 1212, 1465, 391, 927, 4833, 388, 578, 11786, 114, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [65, 484, 2169, 7687, 21932, 18146, 726, 363, 17032, 3391, 114, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="google/bigbird-roberta-base",
revision="215c99f1600e06f83acce68422f2035b2b5c3510",
)
tests/test_tokenization_camembert.py
View file @ fcad8018
......@@ -18,7 +18,7 @@ import unittest
from transformers import CamembertTokenizer, CamembertTokenizerFast
from transformers.file_utils import is_torch_available
from transformers.testing_utils import require_sentencepiece, require_tokenizers
from transformers.testing_utils import require_sentencepiece, require_tokenizers, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -45,6 +45,25 @@ class CamembertTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = CamembertTokenizer(SAMPLE_VOCAB)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<pad>"
token_id = 1
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], "<s>NOTUSED")
self.assertEqual(vocab_keys[1], "<pad>")
self.assertEqual(vocab_keys[-1], "<mask>")
self.assertEqual(len(vocab_keys), 1_004)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_005)
def test_rust_and_python_bpe_tokenizers(self):
tokenizer = CamembertTokenizer(SAMPLE_BPE_VOCAB)
tokenizer.save_pretrained(self.tmpdirname)
......@@ -88,3 +107,25 @@ class CamembertTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
ids = tokenizer.encode(sequence)
rust_ids = rust_tokenizer.encode(sequence)
self.assertListEqual(ids, rust_ids)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[5, 54, 7196, 297, 30, 23, 776, 18, 11, 3215, 3705, 8252, 22, 3164, 1181, 2116, 29, 16, 813, 25, 791, 3314, 20, 3446, 38, 27575, 120, 6, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], [5, 468, 17, 11, 9088, 20, 1517, 8, 22804, 18818, 10, 38, 629, 607, 607, 142, 19, 7196, 867, 56, 10326, 24, 2267, 20, 416, 5072, 15612, 233, 734, 7, 2399, 27, 16, 3015, 1649, 7, 24, 20, 4338, 2399, 27, 13, 3400, 14, 13, 6189, 8, 930, 9, 6]], '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, 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, 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]]} # noqa: E501
# fmt: on
# camembert is a french model. So we also use french texts.
sequences = [
"Le transformeur est un modèle d'apprentissage profond introduit en 2017, "
"utilisé principalement dans le domaine du traitement automatique des langues (TAL).",
"À l'instar des réseaux de neurones récurrents (RNN), les transformeurs sont conçus "
"pour gérer des données séquentielles, telles que le langage naturel, pour des tâches "
"telles que la traduction et la synthèse de texte.",
]
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="camembert-base",
revision="3a0641d9a1aeb7e848a74299e7e4c4bca216b4cf",
sequences=sequences,
)
tests/test_tokenization_common.py
View file @ fcad8018
......@@ -24,7 +24,7 @@ import tempfile
import unittest
from collections import OrderedDict
from itertools import takewhile
from typing import TYPE_CHECKING, Dict, List, Tuple, Union
from typing import TYPE_CHECKING, Any, Dict, List, Tuple, Union
from huggingface_hub import HfApi
from requests.exceptions import HTTPError
......@@ -175,13 +175,74 @@ class TokenizerTesterMixin:
def get_rust_tokenizer(self, **kwargs) -> PreTrainedTokenizerFast:
return self.rust_tokenizer_class.from_pretrained(self.tmpdirname, **kwargs)
# def get_input_output_texts(self) -> Tuple[str, str]:
# """Feel free to overwrite"""
# # TODO: @property
# return (
# "This is a test",
# "This is a test",
# )
def tokenizer_integration_test_util(
self,
expected_encoding: Dict,
model_name: str,
revision: str = None,
sequences: List[str] = None,
decode_kwargs: Dict[str, Any] = None,
padding: bool = True,
):
"""
Util for integration test.
Text is tokenized and then reverted back to text. Both results are then checked.
Args:
expected_encoding:
The expected result of the tokenizer output.
model_name:
The model name of the tokenizer to load and use.
revision:
The full git revision number of the model. This is to pin the
tokenizer config and to avoid that tests start to fail if the
config gets changed upstream.
sequences:
Can overwrite the texts that are used to check the tokenizer.
This is useful if the tokenizer supports non english languages
like france.
decode_kwargs:
Additional args for the ``decode`` function which reverts the
tokenized text back to a string.
padding:
Activates and controls padding of the tokenizer.
"""
decode_kwargs = {} if decode_kwargs is None else decode_kwargs
if sequences is None:
sequences = [
"Transformers (formerly known as pytorch-transformers and pytorch-pretrained-bert) provides "
"general-purpose architectures (BERT, GPT-2, RoBERTa, XLM, DistilBert, XLNet...) for Natural "
"Language Understanding (NLU) and Natural Language Generation (NLG) with over 32+ pretrained "
"models in 100+ languages and deep interoperability between Jax, PyTorch and TensorFlow.",
"BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly "
"conditioning on both left and right context in all layers.",
"The quick brown fox jumps over the lazy dog.",
]
tokenizer_classes = [self.tokenizer_class]
if self.test_rust_tokenizer:
tokenizer_classes.append(self.rust_tokenizer_class)
for tokenizer_class in tokenizer_classes:
tokenizer = tokenizer_class.from_pretrained(
model_name,
revision=revision, # to pin the tokenizer version
)
encoding = tokenizer(sequences, padding=padding)
decoded_sequences = [
tokenizer.decode(seq, skip_special_tokens=True, **decode_kwargs) for seq in encoding["input_ids"]
]
encoding_data = encoding.data
self.assertDictEqual(encoding_data, expected_encoding)
for expected, decoded in zip(sequences, decoded_sequences):
if self.test_sentencepiece_ignore_case:
expected = expected.lower()
self.assertEqual(expected, decoded)
def assert_padded_input_match(self, input_r: list, input_p: list, max_length: int, pad_token_id: int):
# Ensure we match max_length
......@@ -224,6 +285,30 @@ class TokenizerTesterMixin:
for i in range(len(batch_encode_plus_sequences["input_ids"]))
]
# TODO: this test could be extended to all tokenizers - not just the sentencepiece
def test_sentencepiece_tokenize_and_convert_tokens_to_string(self):
"""Test ``_tokenize`` and ``convert_tokens_to_string``."""
if not self.test_sentencepiece:
return
tokenizer = self.get_tokenizer()
text = "This is text to test the tokenizer."
if self.test_sentencepiece_ignore_case:
text = text.lower()
tokens = tokenizer.tokenize(text)
self.assertTrue(len(tokens) > 0)
# check if converting back to original text works
reverse_text = tokenizer.convert_tokens_to_string(tokens)
if self.test_sentencepiece_ignore_case:
reverse_text = reverse_text.lower()
self.assertEqual(reverse_text, text)
def test_subword_regularization_tokenizer(self) -> None:
if not self.test_sentencepiece:
return
......
tests/test_tokenization_deberta_v2.py
View file @ fcad8018
......@@ -17,7 +17,7 @@ import os
import unittest
from transformers import DebertaV2Tokenizer
from transformers.testing_utils import require_sentencepiece, require_tokenizers
from transformers.testing_utils import require_sentencepiece, require_tokenizers, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -47,6 +47,25 @@ class DebertaV2TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
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], "[PAD]")
self.assertEqual(len(vocab_keys), 30_001)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 30_000)
def test_rust_and_python_full_tokenizers(self):
if not self.test_rust_tokenizer:
return
......@@ -108,57 +127,14 @@ class DebertaV2TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
encoded_pair,
)
@slow
def test_tokenizer_integration(self):
tokenizer_classes = [self.tokenizer_class]
if self.test_rust_tokenizer:
tokenizer_classes.append(self.rust_tokenizer_class)
for tokenizer_class in tokenizer_classes:
tokenizer = tokenizer_class.from_pretrained("microsoft/deberta-xlarge-v2")
sequences = [
[
"DeBERTa: Decoding-enhanced BERT with Disentangled Attention",
"DeBERTa: Decoding-enhanced BERT with Disentangled Attention",
],
[
"Recent progress in pre-trained neural language models has significantly improved the performance of many natural language processing (NLP) tasks.",
"DeBERTa: Decoding-enhanced BERT with Disentangled Attention",
],
[
"In this paper we propose a new model architecture DeBERTa",
"DeBERTa: Decoding-enhanced BERT with Disentangled Attention",
],
]
encoding = tokenizer(sequences, padding=True)
decoded_sequences = [tokenizer.decode(seq, skip_special_tokens=True) for seq in encoding["input_ids"]]
# fmt: off
expected_encoding = {
'input_ids': [
[1, 1804, 69418, 191, 43, 117056, 18, 44596, 448, 37132, 19, 8655, 10625, 69860, 21149, 2, 1804, 69418, 191, 43, 117056, 18, 44596, 448, 37132, 19, 8655, 10625, 69860, 21149, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1, 9755, 1944, 11, 1053, 18, 16899, 12730, 1072, 1506, 45, 2497, 2510, 5, 610, 9, 127, 699, 1072, 2101, 36, 99388, 53, 2930, 4, 2, 1804, 69418, 191, 43, 117056, 18, 44596, 448, 37132, 19, 8655, 10625, 69860, 21149, 2],
[1, 84, 32, 778, 42, 9441, 10, 94, 735, 3372, 1804, 69418, 191, 2, 1804, 69418, 191, 43, 117056, 18, 44596, 448, 37132, 19, 8655, 10625, 69860, 21149, 2, 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, 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, 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, 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]],
'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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[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]
]
}
expected_decoded_sequences = [
'DeBERTa: Decoding-enhanced BERT with Disentangled Attention DeBERTa: Decoding-enhanced BERT with Disentangled Attention',
'Recent progress in pre-trained neural language models has significantly improved the performance of many natural language processing (NLP) tasks. DeBERTa: Decoding-enhanced BERT with Disentangled Attention',
'In this paper we propose a new model architecture DeBERTa DeBERTa: Decoding-enhanced BERT with Disentangled Attention'
]
# fmt: on
self.assertDictEqual(encoding.data, expected_encoding)
for expected, decoded in zip(expected_decoded_sequences, decoded_sequences):
self.assertEqual(expected, decoded)
# fmt: off
expected_encoding = {'input_ids': [[1, 32732, 36, 19390, 486, 27, 35052, 81436, 18, 60685, 1225, 7, 35052, 81436, 18, 9367, 16899, 18, 15937, 53, 594, 773, 18, 16287, 30465, 36, 69418, 6, 107805, 36979, 10993, 69418, 191, 6, 12692, 829, 6, 8655, 16555, 92459, 6, 12692, 9431, 20850, 14, 4184, 6369, 9875, 36, 1323, 23941, 53, 7, 4184, 6369, 11005, 36, 20582, 1186, 53, 19, 105, 3049, 1896, 1053, 16899, 1506, 11, 37978, 4243, 7, 1237, 31869, 200, 42754, 6, 19645, 45050, 3425, 7, 107535, 4, 2], [1, 448, 37132, 13, 667, 8, 1053, 18, 23611, 1237, 72356, 12820, 34, 104134, 1209, 35, 13313, 6627, 21, 202, 347, 7, 164, 2399, 11, 46, 4485, 4, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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, 23, 1232, 2864, 15785, 14951, 105, 5, 8581, 1250, 4, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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]], '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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="microsoft/deberta-v2-xlarge",
revision="ad6e42c1532ddf3a15c39246b63f5559d558b670",
)
tests/test_tokenization_m2m_100.py
View file @ fcad8018
......@@ -20,7 +20,7 @@ from shutil import copyfile
from transformers import M2M100Tokenizer, is_torch_available
from transformers.file_utils import is_sentencepiece_available
from transformers.testing_utils import nested_simplify, require_sentencepiece, require_tokenizers, require_torch
from transformers.testing_utils import nested_simplify, require_sentencepiece, require_tokenizers, require_torch, slow
if is_sentencepiece_available():
......@@ -69,6 +69,25 @@ class M2M100TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
"This is a test",
)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "</s>"
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], "</s>")
self.assertEqual(vocab_keys[1], "<unk>")
self.assertEqual(vocab_keys[-1], "<s>")
self.assertEqual(len(vocab_keys), 10)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 117)
@unittest.skip("Skip this test while all models are still to be uploaded.")
def test_pretrained_model_lists(self):
pass
......@@ -90,6 +109,18 @@ class M2M100TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
text = tokenizer.convert_tokens_to_string(tokens)
self.assertEqual(text, "This is a test")
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[128022, 110108, 397, 11, 38272, 2247, 124811, 285, 18105, 1586, 207, 7, 39534, 4428, 397, 1019, 18105, 1586, 207, 7, 41337, 16786, 241, 7, 20214, 17, 125690, 10398, 7, 44378, 58069, 68342, 7798, 7343, 11, 299, 33310, 4, 158, 37350, 94077, 4569, 299, 33310, 90, 4, 52840, 290, 4, 31270, 112, 299, 682, 4, 52840, 39953, 14079, 193, 52519, 90894, 17894, 120697, 11, 40445, 551, 17, 1019, 52519, 90894, 17756, 963, 11, 40445, 480, 17, 9792, 1120, 5173, 1393, 6240, 16786, 241, 120996, 28, 1245, 1393, 118240, 11123, 1019, 93612, 2691, 10618, 98058, 120409, 1928, 279, 4, 40683, 367, 178, 207, 1019, 103, 103121, 506, 65296, 5, 2], [128022, 21217, 367, 117, 125450, 128, 719, 7, 7308, 40, 93612, 12669, 1116, 16704, 71, 17785, 3699, 15592, 35, 144, 9584, 241, 11943, 713, 950, 799, 2247, 88427, 150, 149, 118813, 120706, 1019, 106906, 81518, 28, 1224, 22799, 397, 5, 2, 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], [128022, 1658, 123311, 5155, 5578, 4722, 279, 14947, 2366, 1120, 1197, 14, 1348, 9232, 5, 2, 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]], '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, 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], [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, 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="facebook/m2m100_418M",
revision="c168bae485c864188cf9aa0e4108b0b6934dc91e",
)
@require_torch
@require_sentencepiece
......
tests/test_tokenization_marian.py
View file @ fcad8018
......@@ -21,7 +21,7 @@ from shutil import copyfile
from transformers import BatchEncoding, MarianTokenizer
from transformers.file_utils import is_sentencepiece_available, is_tf_available, is_torch_available
from transformers.testing_utils import require_sentencepiece
from transformers.testing_utils import require_sentencepiece, slow
if is_sentencepiece_available():
......@@ -74,6 +74,25 @@ class MarianTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
"This is a test",
)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "</s>"
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], "</s>")
self.assertEqual(vocab_keys[1], "<unk>")
self.assertEqual(vocab_keys[-1], "<pad>")
self.assertEqual(len(vocab_keys), 9)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 9)
def test_tokenizer_equivalence_en_de(self):
en_de_tokenizer = MarianTokenizer.from_pretrained(f"{ORG_NAME}opus-mt-en-de")
batch = en_de_tokenizer(["I am a small frog"], return_tensors=None)
......@@ -101,3 +120,16 @@ class MarianTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
batch_smaller = tok(["I am a tiny frog", "I am a small frog"], padding=True, return_tensors=FRAMEWORK)
self.assertIsInstance(batch_smaller, BatchEncoding)
self.assertEqual(batch_smaller.input_ids.shape, (2, 10))
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[43495, 462, 20, 42164, 1369, 52, 464, 132, 1703, 492, 13, 7491, 38999, 6, 8, 464, 132, 1703, 492, 13, 4669, 37867, 13, 7525, 27, 1593, 988, 13, 33972, 7029, 6, 20, 8251, 383, 2, 270, 5866, 3788, 2, 2353, 8251, 12338, 2, 13958, 387, 2, 3629, 6953, 188, 2900, 2, 13958, 8011, 11501, 23, 8460, 4073, 34009, 20, 435, 11439, 27, 8, 8460, 4073, 6004, 20, 9988, 375, 27, 33, 266, 1945, 1076, 1350, 37867, 3288, 5, 577, 1076, 4374, 8, 5082, 5, 26453, 257, 556, 403, 2, 242, 132, 383, 316, 492, 8, 10767, 6, 316, 304, 4239, 3, 0], [148, 15722, 19, 1839, 12, 1350, 13, 22327, 5082, 5418, 47567, 35938, 59, 318, 19552, 108, 2183, 54, 14976, 4835, 32, 547, 1114, 8, 315, 2417, 5, 92, 19088, 3, 0, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100], [36, 6395, 12570, 39147, 11597, 6, 266, 4, 45405, 7296, 3, 0, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100, 58100]], '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, 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], [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, 0]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="Helsinki-NLP/opus-mt-en-de",
revision="1a8c2263da11e68e50938f97e10cd57820bd504c",
decode_kwargs={"use_source_tokenizer": True},
)
tests/test_tokenization_mbart50.py
View file @ fcad8018
......@@ -17,7 +17,7 @@ import tempfile
import unittest
from transformers import SPIECE_UNDERLINE, BatchEncoding, MBart50Tokenizer, MBart50TokenizerFast, is_torch_available
from transformers.testing_utils import nested_simplify, require_sentencepiece, require_tokenizers, require_torch
from transformers.testing_utils import nested_simplify, require_sentencepiece, require_tokenizers, require_torch, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -47,6 +47,25 @@ class MBartTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = MBart50Tokenizer(SAMPLE_VOCAB, src_lang="en_XX", tgt_lang="ro_RO", keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
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], "<s>")
self.assertEqual(vocab_keys[1], "<pad>")
self.assertEqual(vocab_keys[-1], "<mask>")
self.assertEqual(len(vocab_keys), 1_054)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_054)
def test_full_tokenizer(self):
tokenizer = MBart50Tokenizer(SAMPLE_VOCAB, src_lang="en_XX", tgt_lang="ro_RO", keep_accents=True)
......@@ -82,6 +101,18 @@ class MBartTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
# fmt: on
)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[250004, 11062, 82772, 7, 15, 82772, 538, 51529, 237, 17198, 1290, 206, 9, 215175, 1314, 136, 17198, 1290, 206, 9, 56359, 42, 122009, 9, 16466, 16, 87344, 4537, 9, 4717, 78381, 6, 159958, 7, 15, 24480, 618, 4, 527, 22693, 5428, 4, 2777, 24480, 9874, 4, 43523, 594, 4, 803, 18392, 33189, 18, 4, 43523, 24447, 12399, 100, 24955, 83658, 9626, 144057, 15, 839, 22335, 16, 136, 24955, 83658, 83479, 15, 39102, 724, 16, 678, 645, 2789, 1328, 4589, 42, 122009, 115774, 23, 805, 1328, 46876, 7, 136, 53894, 1940, 42227, 41159, 17721, 823, 425, 4, 27512, 98722, 206, 136, 5531, 4970, 919, 17336, 5, 2], [250004, 20080, 618, 83, 82775, 47, 479, 9, 1517, 73, 53894, 333, 80581, 110117, 18811, 5256, 1295, 51, 152526, 297, 7986, 390, 124416, 538, 35431, 214, 98, 15044, 25737, 136, 7108, 43701, 23, 756, 135355, 7, 5, 2, 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], [250004, 581, 63773, 119455, 6, 147797, 88203, 7, 645, 70, 21, 3285, 10269, 5, 2, 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]], '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, 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], [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, 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="facebook/mbart-large-50",
revision="d3913889c59cd5c9e456b269c376325eabad57e2",
)
@require_torch
@require_sentencepiece
......
tests/test_tokenization_pegasus.py
View file @ fcad8018
......@@ -16,7 +16,7 @@ import unittest
from transformers import PegasusTokenizer, PegasusTokenizerFast
from transformers.file_utils import cached_property
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, require_torch
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, require_torch, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -50,6 +50,25 @@ class PegasusTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
def get_input_output_texts(self, tokenizer):
return ("This is a test", "This is a test")
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "</s>"
token_id = 1
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], "</s>")
self.assertEqual(vocab_keys[-1], "v")
self.assertEqual(len(vocab_keys), 1_103)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_103)
def test_mask_tokens_rust_pegasus(self):
rust_tokenizer = self.rust_tokenizer_class.from_pretrained(self.tmpdirname)
py_tokenizer = self.tokenizer_class.from_pretrained(self.tmpdirname)
......@@ -97,6 +116,18 @@ class PegasusTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
assert targets["input_ids"].shape == (2, 5)
assert len(batch) == 2 # input_ids, attention_mask.
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[38979, 143, 18485, 606, 130, 26669, 87686, 121, 54189, 1129, 111, 26669, 87686, 121, 9114, 14787, 121, 13249, 158, 592, 956, 121, 14621, 31576, 143, 62613, 108, 9688, 930, 43430, 11562, 62613, 304, 108, 11443, 897, 108, 9314, 17415, 63399, 108, 11443, 7614, 18316, 118, 4284, 7148, 12430, 143, 1400, 25703, 158, 111, 4284, 7148, 11772, 143, 21297, 1064, 158, 122, 204, 3506, 1754, 1133, 14787, 1581, 115, 33224, 4482, 111, 1355, 110, 29173, 317, 50833, 108, 20147, 94665, 111, 77198, 107, 1], [110, 62613, 117, 638, 112, 1133, 121, 20098, 1355, 79050, 13872, 135, 1596, 53541, 1352, 141, 13039, 5542, 124, 302, 518, 111, 268, 2956, 115, 149, 4427, 107, 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], [139, 1235, 2799, 18289, 17780, 204, 109, 9474, 1296, 107, 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]], '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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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="google/bigbird-pegasus-large-arxiv",
revision="ba85d0851d708441f91440d509690f1ab6353415",
)
@require_sentencepiece
@require_tokenizers
......
tests/test_tokenization_reformer.py
View file @ fcad8018
......@@ -41,6 +41,25 @@ class ReformerTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = ReformerTokenizer(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
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], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "j")
self.assertEqual(len(vocab_keys), 1_000)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_000)
def test_rust_and_python_full_tokenizers(self):
if not self.test_rust_tokenizer:
return
......@@ -328,3 +347,25 @@ class ReformerTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
with torch.no_grad():
model(**encoded_sequence)
model(**batch_encoded_sequence)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[108, 265, 24, 111, 4, 258, 156, 7, 51, 279, 58, 7, 76, 25, 69, 278], [140, 243, 264, 134, 17, 267, 77, 263, 22, 262, 297, 258, 304, 177, 279, 266, 14, 89, 13, 35, 261, 299, 272, 137, 275, 278]], '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]]} # noqa: E501
# fmt: on
# This tokenizer does not know some characters like ")".
# That is the reason why we use very simple texts here.
# Also see https://github.com/huggingface/transformers/pull/11737#issuecomment-850769064
sequences = [
"This is a very simple sentence.",
"The quick brown fox jumps over the lazy dog.",
]
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="google/reformer-crime-and-punishment",
revision="0e6c3decb8211d49bf881013425dc8b0448b3f5a",
padding=False,
sequences=sequences,
)
tests/test_tokenization_speech_to_text.py
View file @ fcad8018
......@@ -20,7 +20,7 @@ from shutil import copyfile
from transformers import SPIECE_UNDERLINE, is_sentencepiece_available
from transformers.models.speech_to_text import Speech2TextTokenizer
from transformers.models.speech_to_text.tokenization_speech_to_text import VOCAB_FILES_NAMES, save_json
from transformers.testing_utils import require_sentencepiece, require_tokenizers
from transformers.testing_utils import require_sentencepiece, require_tokenizers, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -60,6 +60,25 @@ class SpeechToTextTokenizerTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = Speech2TextTokenizer.from_pretrained(self.tmpdirname)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<pad>"
token_id = 1
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], "<s>")
self.assertEqual(vocab_keys[1], "<pad>")
self.assertEqual(vocab_keys[-1], "j")
self.assertEqual(len(vocab_keys), 1_001)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_001)
def test_full_tokenizer(self):
tokenizer = Speech2TextTokenizer.from_pretrained(self.tmpdirname)
......@@ -89,6 +108,18 @@ class SpeechToTextTokenizerTest(TokenizerTesterMixin, unittest.TestCase):
# fmt: on
)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[3791, 797, 31, 11, 64, 797, 31, 2429, 433, 12, 1176, 12, 20, 786, 915, 142, 2413, 240, 37, 3238, 797, 31, 11, 35, 93, 915, 142, 2413, 240, 37, 5540, 567, 1276, 93, 37, 610, 40, 62, 455, 657, 1042, 123, 780, 177, 37, 309, 241, 1298, 514, 20, 292, 2737, 114, 2469, 241, 85, 64, 302, 548, 528, 423, 4, 509, 406, 423, 37, 601, 4, 777, 302, 548, 528, 423, 284, 4, 3388, 511, 459, 4, 3555, 40, 321, 302, 705, 4, 3388, 511, 583, 326, 5, 5, 5, 62, 3310, 560, 177, 2680, 217, 1508, 32, 31, 853, 418, 64, 583, 511, 1605, 62, 35, 93, 560, 177, 2680, 217, 1508, 1521, 64, 583, 511, 519, 62, 20, 1515, 764, 20, 149, 261, 5625, 7972, 20, 5540, 567, 1276, 93, 3925, 1675, 11, 15, 802, 7972, 576, 217, 1508, 11, 35, 93, 1253, 2441, 15, 289, 652, 31, 416, 321, 3842, 115, 40, 911, 8, 476, 619, 4, 380, 142, 423, 335, 240, 35, 93, 264, 8, 11, 335, 569, 420, 163, 5, 2], [260, 548, 528, 423, 20, 451, 20, 2681, 1153, 3434, 20, 5540, 37, 567, 126, 1253, 2441, 3376, 449, 210, 431, 1563, 177, 767, 5540, 11, 1203, 472, 11, 2953, 685, 285, 364, 706, 1153, 20, 6799, 20, 2869, 20, 4464, 126, 40, 2429, 20, 1040, 866, 2664, 418, 20, 318, 20, 1726, 186, 20, 265, 522, 35, 93, 2191, 4634, 20, 1040, 12, 6799, 15, 228, 2356, 142, 31, 11, 5, 2, 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], [2575, 2666, 684, 1582, 1176, 12, 627, 149, 619, 20, 4902, 563, 11, 20, 149, 261, 3420, 2356, 174, 142, 4714, 131, 5, 2, 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, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], '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, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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="facebook/s2t-small-mustc-en-de-st",
revision="a14f04cf0776c02f62a8cb800cf7909e15ea23ad",
)
@require_sentencepiece
class SpeechToTextTokenizerMultilinguialTest(unittest.TestCase):
......@@ -108,6 +139,9 @@ class SpeechToTextTokenizerMultilinguialTest(unittest.TestCase):
self.assertEqual(self.tokenizer.lang_code_to_id["it"], 9)
self.assertEqual(self.tokenizer.lang_code_to_id["de"], 11)
def test_vocab_size(self):
self.assertEqual(self.tokenizer.vocab_size, 10_000)
def test_tokenizer_decode_ignores_language_codes(self):
self.assertIn(ES_CODE, self.tokenizer.all_special_ids)
generated_ids = [ES_CODE, 4, 1601, 47, 7647, 2]
......
tests/test_tokenization_t5.py
View file @ fcad8018
......@@ -17,7 +17,7 @@ import unittest
from transformers import SPIECE_UNDERLINE, AddedToken, BatchEncoding, T5Tokenizer, T5TokenizerFast
from transformers.file_utils import cached_property, is_tf_available, is_torch_available
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers
from transformers.testing_utils import get_tests_dir, require_sentencepiece, require_tokenizers, slow
from .test_tokenization_common import TokenizerTesterMixin
......@@ -48,6 +48,25 @@ class T5TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = T5Tokenizer(SAMPLE_VOCAB)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
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], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "<pad>")
self.assertEqual(len(vocab_keys), 1_101)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_100)
def test_full_tokenizer(self):
tokenizer = T5Tokenizer(SAMPLE_VOCAB)
......@@ -274,3 +293,15 @@ class T5TokenizationTest(TokenizerTesterMixin, unittest.TestCase):
self.assertTrue(special_token_id in p_output)
self.assertTrue(special_token_id in r_output)
self.assertTrue(special_token_id in cr_output)
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[31220, 7, 41, 14034, 801, 38, 3, 102, 63, 17, 127, 524, 18, 7031, 2032, 277, 11, 3, 102, 63, 17, 127, 524, 18, 2026, 17, 10761, 18, 7041, 61, 795, 879, 18, 19681, 4648, 7, 41, 12920, 382, 6, 350, 6383, 4949, 6, 2158, 12920, 382, 9, 6, 3, 4, 11160, 6, 2043, 17153, 279, 49, 17, 6, 3, 4, 434, 9688, 11439, 21, 6869, 10509, 17725, 41, 567, 9138, 61, 11, 6869, 10509, 11946, 41, 18207, 517, 61, 28, 147, 3538, 1220, 7140, 10761, 2250, 16, 910, 1220, 8024, 11, 1659, 1413, 32, 883, 2020, 344, 2215, 226, 6, 12901, 382, 127, 524, 11, 4738, 7, 127, 15390, 5, 1], [272, 24203, 19, 876, 12, 554, 18, 9719, 1659, 2647, 26352, 6497, 7, 45, 73, 9339, 400, 26, 1499, 57, 22801, 10760, 30, 321, 646, 11, 269, 2625, 16, 66, 7500, 5, 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], [37, 1704, 4216, 3, 20400, 4418, 7, 147, 8, 19743, 1782, 5, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '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, 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, 0, 0, 0, 0, 0, 0, 0, 0, 0], [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, 0, 0, 0, 0, 0, 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="t5-base",
revision="5a7ff2d8f5117c194c7e32ec1ccbf04642cca99b",
)
tests/test_tokenization_xlm_prophetnet.py
View file @ fcad8018
......@@ -40,6 +40,25 @@ class XLMProphetNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = XLMProphetNetTokenizer(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
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], "[CLS]")
self.assertEqual(vocab_keys[-1], "j")
self.assertEqual(len(vocab_keys), 1_012)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_012)
def test_full_tokenizer(self):
tokenizer = XLMProphetNetTokenizer(SAMPLE_VOCAB, keep_accents=True)
......@@ -124,3 +143,15 @@ class XLMProphetNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
symbols = "Hello World!"
original_tokenizer_encodings = [35389, 6672, 49, 2]
self.assertListEqual(original_tokenizer_encodings, self.big_tokenizer.encode(symbols))
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[11073, 82783, 18, 26, 82783, 549, 51540, 248, 17209, 1301, 217, 20, 215186, 1325, 147, 17209, 1301, 217, 20, 56370, 53, 122020, 20, 16477, 27, 87355, 4548, 20, 4728, 78392, 17, 159969, 18, 26, 24491, 629, 15, 538, 22704, 5439, 15, 2788, 24491, 9885, 15, 43534, 605, 15, 814, 18403, 33200, 29, 15, 43534, 24458, 12410, 111, 24966, 83669, 9637, 144068, 26, 850, 22346, 27, 147, 24966, 83669, 83490, 26, 39113, 735, 27, 689, 656, 2800, 1339, 4600, 53, 122020, 115785, 34, 816, 1339, 46887, 18, 147, 53905, 1951, 42238, 41170, 17732, 834, 436, 15, 27523, 98733, 217, 147, 5542, 4981, 930, 17347, 16, 2], [20091, 629, 94, 82786, 58, 490, 20, 1528, 84, 53905, 344, 80592, 110128, 18822, 5267, 1306, 62, 152537, 308, 7997, 401, 124427, 549, 35442, 225, 109, 15055, 25748, 147, 7119, 43712, 34, 767, 135366, 18, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [592, 63784, 119466, 17, 147808, 88214, 18, 656, 81, 32, 3296, 10280, 16, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], '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, 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], [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, 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="microsoft/xprophetnet-large-wiki100-cased",
revision="1acad1643ddd54a44df6a1b797ada8373685d90e",
)
tests/test_tokenization_xlm_roberta.py
View file @ fcad8018
......@@ -42,6 +42,25 @@ class XLMRobertaTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer = XLMRobertaTokenizer(SAMPLE_VOCAB, keep_accents=True)
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<pad>"
token_id = 1
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], "<s>")
self.assertEqual(vocab_keys[1], "<pad>")
self.assertEqual(vocab_keys[-1], "<mask>")
self.assertEqual(len(vocab_keys), 1_002)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_002)
def test_full_tokenizer(self):
tokenizer = XLMRobertaTokenizer(SAMPLE_VOCAB, keep_accents=True)
......@@ -229,3 +248,15 @@ class XLMRobertaTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
# xlmr.encode(symbols)
self.assertListEqual(original_tokenizer_encodings, self.big_tokenizer.encode(symbols))
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[0, 11062, 82772, 7, 15, 82772, 538, 51529, 237, 17198, 1290, 206, 9, 215175, 1314, 136, 17198, 1290, 206, 9, 56359, 42, 122009, 9, 16466, 16, 87344, 4537, 9, 4717, 78381, 6, 159958, 7, 15, 24480, 618, 4, 527, 22693, 5428, 4, 2777, 24480, 9874, 4, 43523, 594, 4, 803, 18392, 33189, 18, 4, 43523, 24447, 12399, 100, 24955, 83658, 9626, 144057, 15, 839, 22335, 16, 136, 24955, 83658, 83479, 15, 39102, 724, 16, 678, 645, 2789, 1328, 4589, 42, 122009, 115774, 23, 805, 1328, 46876, 7, 136, 53894, 1940, 42227, 41159, 17721, 823, 425, 4, 27512, 98722, 206, 136, 5531, 4970, 919, 17336, 5, 2], [0, 20080, 618, 83, 82775, 47, 479, 9, 1517, 73, 53894, 333, 80581, 110117, 18811, 5256, 1295, 51, 152526, 297, 7986, 390, 124416, 538, 35431, 214, 98, 15044, 25737, 136, 7108, 43701, 23, 756, 135355, 7, 5, 2, 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, 581, 63773, 119455, 6, 147797, 88203, 7, 645, 70, 21, 3285, 10269, 5, 2, 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]], '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, 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], [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, 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="xlm-roberta-base",
revision="d9d8a8ea5eb94b1c6654ae9249df7793cd2933d3",
)
tests/test_tokenization_xlnet.py
View file @ fcad8018
......@@ -42,6 +42,25 @@ class XLNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
tokenizer.sanitize_special_tokens()
tokenizer.save_pretrained(self.tmpdirname)
def test_convert_token_and_id(self):
"""Test ``_convert_token_to_id`` and ``_convert_id_to_token``."""
token = "<s>"
token_id = 1
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], "<unk>")
self.assertEqual(vocab_keys[1], "<s>")
self.assertEqual(vocab_keys[-1], "<eod>")
self.assertEqual(len(vocab_keys), 1_006)
def test_vocab_size(self):
self.assertEqual(self.get_tokenizer().vocab_size, 1_000)
def test_full_tokenizer(self):
tokenizer = XLNetTokenizer(SAMPLE_VOCAB, keep_accents=True)
......@@ -180,3 +199,15 @@ class XLNetTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
assert encoded_sentence == text + [4, 3]
assert encoded_pair == text + [4] + text_2 + [4, 3]
@slow
def test_tokenizer_integration(self):
# fmt: off
expected_encoding = {'input_ids': [[17, 21442, 270, 17, 10, 14645, 318, 34, 17, 4546, 3145, 787, 13, 7752, 22018, 23, 21, 17, 4546, 3145, 787, 13, 3352, 14431, 13, 5500, 11, 1176, 580, 13, 16819, 4797, 23, 17, 10, 17135, 658, 19, 457, 7932, 13, 184, 19, 3154, 17135, 6468, 19, 1404, 12269, 19, 4229, 5356, 16264, 46, 19, 17, 20545, 10395, 9, 9, 9, 11, 28, 6421, 9531, 20729, 17, 10, 353, 17022, 11, 21, 6421, 9531, 16949, 17, 10, 11509, 753, 11, 33, 95, 2421, 7385, 956, 14431, 2626, 25, 842, 7385, 4836, 21, 1429, 2272, 9855, 3120, 161, 24738, 19, 13203, 658, 218, 787, 21, 430, 18482, 847, 2637, 9, 4, 3], [5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 322, 22178, 27, 1064, 22, 956, 13, 11101, 1429, 5854, 24313, 18953, 40, 422, 24366, 68, 1758, 37, 10483, 14257, 31, 207, 263, 21, 203, 3773, 25, 71, 9735, 9, 4, 3], [5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 32, 2049, 3442, 17, 13894, 3380, 23, 95, 18, 17634, 2288, 9, 4, 3]], '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, 2], [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 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, 2], [3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]], '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], [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 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, 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, 0, 0, 0, 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, 1, 1]]} # noqa: E501
# fmt: on
self.tokenizer_integration_test_util(
expected_encoding=expected_encoding,
model_name="xlnet-base-cased",
revision="c841166438c31ec7ca9a106dee7bb312b73ae511",
)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment