Merge pull request #1074 from huggingface/improved_testing
Shortcut to special tokens' ids - fix GPT2 & RoBERTa tokenizers - improved testing for GPT/GPT-2
This commit is contained in:
Thomas Wolf
@@ -214,6 +214,9 @@ class CommonTestCases:
|
||||
|
||||
config, inputs_dict = self.model_tester.prepare_config_and_inputs_for_common()
|
||||
|
||||
if "head_mask" in inputs_dict:
|
||||
del inputs_dict["head_mask"]
|
||||
|
||||
for model_class in self.all_model_classes:
|
||||
config.output_attentions = True
|
||||
config.output_hidden_states = False
|
||||
|
||||
@@ -18,31 +18,196 @@ from __future__ import print_function
|
||||
|
||||
import unittest
|
||||
import pytest
|
||||
import shutil
|
||||
|
||||
|
||||
from pytorch_transformers import (GPT2Config, GPT2Model,
|
||||
from pytorch_transformers import (GPT2Config, GPT2Model, GPT2_PRETRAINED_MODEL_ARCHIVE_MAP,
|
||||
GPT2LMHeadModel, GPT2DoubleHeadsModel)
|
||||
|
||||
from .modeling_common_test import CommonTestCases, ConfigTester
|
||||
from .modeling_common_test import CommonTestCases, ConfigTester, ids_tensor
|
||||
|
||||
class GPT2ModelTest(unittest.TestCase):
|
||||
|
||||
class GPT2ModelTest(CommonTestCases.CommonModelTester):
|
||||
|
||||
all_model_classes = (GPT2Model, GPT2LMHeadModel, GPT2DoubleHeadsModel)
|
||||
|
||||
class GPT2ModelTester(object):
|
||||
|
||||
def __init__(self,
|
||||
parent,
|
||||
batch_size=13,
|
||||
seq_length=7,
|
||||
is_training=True,
|
||||
use_token_type_ids=True,
|
||||
use_labels=True,
|
||||
vocab_size=99,
|
||||
hidden_size=32,
|
||||
num_hidden_layers=5,
|
||||
num_attention_heads=4,
|
||||
intermediate_size=37,
|
||||
hidden_act="gelu",
|
||||
hidden_dropout_prob=0.1,
|
||||
attention_probs_dropout_prob=0.1,
|
||||
max_position_embeddings=512,
|
||||
type_vocab_size=16,
|
||||
type_sequence_label_size=2,
|
||||
initializer_range=0.02,
|
||||
num_labels=3,
|
||||
num_choices=4,
|
||||
scope=None,
|
||||
):
|
||||
self.parent = parent
|
||||
self.batch_size = batch_size
|
||||
self.seq_length = seq_length
|
||||
self.is_training = is_training
|
||||
self.use_token_type_ids = use_token_type_ids
|
||||
self.use_labels = use_labels
|
||||
self.vocab_size = vocab_size
|
||||
self.hidden_size = hidden_size
|
||||
self.num_hidden_layers = num_hidden_layers
|
||||
self.num_attention_heads = num_attention_heads
|
||||
self.intermediate_size = intermediate_size
|
||||
self.hidden_act = hidden_act
|
||||
self.hidden_dropout_prob = hidden_dropout_prob
|
||||
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
||||
self.max_position_embeddings = max_position_embeddings
|
||||
self.type_vocab_size = type_vocab_size
|
||||
self.type_sequence_label_size = type_sequence_label_size
|
||||
self.initializer_range = initializer_range
|
||||
self.num_labels = num_labels
|
||||
self.num_choices = num_choices
|
||||
self.scope = scope
|
||||
|
||||
def prepare_config_and_inputs(self):
|
||||
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
|
||||
|
||||
token_type_ids = None
|
||||
if self.use_token_type_ids:
|
||||
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
|
||||
|
||||
sequence_labels = None
|
||||
token_labels = None
|
||||
choice_labels = None
|
||||
if self.use_labels:
|
||||
sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
|
||||
token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
|
||||
choice_labels = ids_tensor([self.batch_size], self.num_choices)
|
||||
|
||||
config = GPT2Config(
|
||||
vocab_size_or_config_json_file=self.vocab_size,
|
||||
n_embd=self.hidden_size,
|
||||
n_layer=self.num_hidden_layers,
|
||||
n_head=self.num_attention_heads,
|
||||
# intermediate_size=self.intermediate_size,
|
||||
# hidden_act=self.hidden_act,
|
||||
# hidden_dropout_prob=self.hidden_dropout_prob,
|
||||
# attention_probs_dropout_prob=self.attention_probs_dropout_prob,
|
||||
n_positions=self.max_position_embeddings,
|
||||
n_ctx=self.max_position_embeddings
|
||||
# type_vocab_size=self.type_vocab_size,
|
||||
# initializer_range=self.initializer_range
|
||||
)
|
||||
|
||||
head_mask = ids_tensor([self.num_hidden_layers, self.num_attention_heads], 2)
|
||||
|
||||
return config, input_ids, head_mask, token_type_ids, sequence_labels, token_labels, choice_labels
|
||||
|
||||
def check_loss_output(self, result):
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
|
||||
def create_and_check_gpt2_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = GPT2Model(config=config)
|
||||
model.eval()
|
||||
|
||||
model(input_ids, token_type_ids=token_type_ids, head_mask=head_mask)
|
||||
model(input_ids, token_type_ids=token_type_ids)
|
||||
sequence_output, presents = model(input_ids)
|
||||
|
||||
result = {
|
||||
"sequence_output": sequence_output,
|
||||
"presents": presents,
|
||||
}
|
||||
self.parent.assertListEqual(
|
||||
list(result["sequence_output"].size()),
|
||||
[self.batch_size, self.seq_length, self.hidden_size])
|
||||
self.parent.assertEqual(len(result["presents"]), config.n_layer)
|
||||
|
||||
def create_and_check_lm_head_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = GPT2LMHeadModel(config)
|
||||
model.eval()
|
||||
|
||||
loss, lm_logits, _ = model(input_ids, token_type_ids=token_type_ids, labels=input_ids)
|
||||
|
||||
result = {
|
||||
"loss": loss,
|
||||
"lm_logits": lm_logits
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
self.parent.assertListEqual(
|
||||
list(result["lm_logits"].size()),
|
||||
[self.batch_size, self.seq_length, self.vocab_size])
|
||||
|
||||
def create_and_check_double_lm_head_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = GPT2DoubleHeadsModel(config)
|
||||
model.eval()
|
||||
|
||||
loss, lm_logits, mc_logits, _ = model(input_ids, token_type_ids=token_type_ids, lm_labels=input_ids)
|
||||
|
||||
result = {
|
||||
"loss": loss,
|
||||
"lm_logits": lm_logits
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
self.parent.assertListEqual(
|
||||
list(result["lm_logits"].size()),
|
||||
[self.batch_size, self.seq_length, self.vocab_size])
|
||||
|
||||
def prepare_config_and_inputs_for_common(self):
|
||||
config_and_inputs = self.prepare_config_and_inputs()
|
||||
(config, input_ids, head_mask, token_type_ids, sequence_labels, token_labels, choice_labels) = config_and_inputs
|
||||
inputs_dict = {
|
||||
'input_ids': input_ids,
|
||||
'token_type_ids': token_type_ids,
|
||||
'head_mask': head_mask
|
||||
}
|
||||
|
||||
return config, inputs_dict
|
||||
|
||||
def setUp(self):
|
||||
self.model_tester = GPT2ModelTest.GPT2ModelTester(self)
|
||||
self.config_tester = ConfigTester(self, config_class=GPT2Config, n_embd=37)
|
||||
|
||||
def test_config(self):
|
||||
config_tester = ConfigTester(self, config_class=GPT2Config, n_embd=37)
|
||||
config_tester.run_common_tests()
|
||||
self.config_tester.run_common_tests()
|
||||
|
||||
def test_model(self):
|
||||
model_tester = CommonTestCases.GPTModelTester(self, config_class=GPT2Config, base_model_class=GPT2Model,
|
||||
lm_head_model_class=GPT2LMHeadModel,
|
||||
double_head_model_class=GPT2DoubleHeadsModel)
|
||||
model_tester.run_common_tests(test_presents=True)
|
||||
def test_gpt2_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_gpt2_model(*config_and_inputs)
|
||||
|
||||
def test_gpt2_lm_head_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_lm_head_model(*config_and_inputs)
|
||||
|
||||
def test_gpt2_double_lm_head_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_double_lm_head_model(*config_and_inputs)
|
||||
|
||||
@pytest.mark.slow
|
||||
def test_pretrained(self):
|
||||
model_tester = CommonTestCases.GPTModelTester(self, config_class=GPT2Config, base_model_class=GPT2Model,
|
||||
lm_head_model_class=GPT2LMHeadModel,
|
||||
double_head_model_class=GPT2DoubleHeadsModel)
|
||||
model_tester.run_slow_tests()
|
||||
def test_model_from_pretrained(self):
|
||||
cache_dir = "/tmp/pytorch_transformers_test/"
|
||||
for model_name in list(GPT2_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:
|
||||
model = GPT2Model.from_pretrained(model_name, cache_dir=cache_dir)
|
||||
shutil.rmtree(cache_dir)
|
||||
self.assertIsNotNone(model)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
@@ -18,31 +18,194 @@ from __future__ import print_function
|
||||
|
||||
import unittest
|
||||
import pytest
|
||||
import shutil
|
||||
|
||||
|
||||
from pytorch_transformers import (OpenAIGPTConfig, OpenAIGPTModel,
|
||||
OpenAIGPTLMHeadModel, OpenAIGPTDoubleHeadsModel)
|
||||
from pytorch_transformers import (OpenAIGPTConfig, OpenAIGPTModel, OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP,
|
||||
OpenAIGPTLMHeadModel, OpenAIGPTDoubleHeadsModel)
|
||||
|
||||
from .modeling_common_test import CommonTestCases, ConfigTester
|
||||
from .modeling_common_test import CommonTestCases, ConfigTester, ids_tensor
|
||||
|
||||
class OpenAIModelTest(unittest.TestCase):
|
||||
|
||||
class OpenAIGPTModelTest(CommonTestCases.CommonModelTester):
|
||||
|
||||
all_model_classes = (OpenAIGPTModel, OpenAIGPTLMHeadModel, OpenAIGPTDoubleHeadsModel)
|
||||
|
||||
class OpenAIGPTModelTester(object):
|
||||
|
||||
def __init__(self,
|
||||
parent,
|
||||
batch_size=13,
|
||||
seq_length=7,
|
||||
is_training=True,
|
||||
use_token_type_ids=True,
|
||||
use_labels=True,
|
||||
vocab_size=99,
|
||||
hidden_size=32,
|
||||
num_hidden_layers=5,
|
||||
num_attention_heads=4,
|
||||
intermediate_size=37,
|
||||
hidden_act="gelu",
|
||||
hidden_dropout_prob=0.1,
|
||||
attention_probs_dropout_prob=0.1,
|
||||
max_position_embeddings=512,
|
||||
type_vocab_size=16,
|
||||
type_sequence_label_size=2,
|
||||
initializer_range=0.02,
|
||||
num_labels=3,
|
||||
num_choices=4,
|
||||
scope=None,
|
||||
):
|
||||
self.parent = parent
|
||||
self.batch_size = batch_size
|
||||
self.seq_length = seq_length
|
||||
self.is_training = is_training
|
||||
self.use_token_type_ids = use_token_type_ids
|
||||
self.use_labels = use_labels
|
||||
self.vocab_size = vocab_size
|
||||
self.hidden_size = hidden_size
|
||||
self.num_hidden_layers = num_hidden_layers
|
||||
self.num_attention_heads = num_attention_heads
|
||||
self.intermediate_size = intermediate_size
|
||||
self.hidden_act = hidden_act
|
||||
self.hidden_dropout_prob = hidden_dropout_prob
|
||||
self.attention_probs_dropout_prob = attention_probs_dropout_prob
|
||||
self.max_position_embeddings = max_position_embeddings
|
||||
self.type_vocab_size = type_vocab_size
|
||||
self.type_sequence_label_size = type_sequence_label_size
|
||||
self.initializer_range = initializer_range
|
||||
self.num_labels = num_labels
|
||||
self.num_choices = num_choices
|
||||
self.scope = scope
|
||||
|
||||
def prepare_config_and_inputs(self):
|
||||
input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
|
||||
|
||||
token_type_ids = None
|
||||
if self.use_token_type_ids:
|
||||
token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
|
||||
|
||||
sequence_labels = None
|
||||
token_labels = None
|
||||
choice_labels = None
|
||||
if self.use_labels:
|
||||
sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
|
||||
token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
|
||||
choice_labels = ids_tensor([self.batch_size], self.num_choices)
|
||||
|
||||
config = OpenAIGPTConfig(
|
||||
vocab_size_or_config_json_file=self.vocab_size,
|
||||
n_embd=self.hidden_size,
|
||||
n_layer=self.num_hidden_layers,
|
||||
n_head=self.num_attention_heads,
|
||||
# intermediate_size=self.intermediate_size,
|
||||
# hidden_act=self.hidden_act,
|
||||
# hidden_dropout_prob=self.hidden_dropout_prob,
|
||||
# attention_probs_dropout_prob=self.attention_probs_dropout_prob,
|
||||
n_positions=self.max_position_embeddings,
|
||||
n_ctx=self.max_position_embeddings
|
||||
# type_vocab_size=self.type_vocab_size,
|
||||
# initializer_range=self.initializer_range
|
||||
)
|
||||
|
||||
head_mask = ids_tensor([self.num_hidden_layers, self.num_attention_heads], 2)
|
||||
|
||||
return config, input_ids, head_mask, token_type_ids, sequence_labels, token_labels, choice_labels
|
||||
|
||||
def check_loss_output(self, result):
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
|
||||
def create_and_check_openai_gpt_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = OpenAIGPTModel(config=config)
|
||||
model.eval()
|
||||
|
||||
model(input_ids, token_type_ids=token_type_ids, head_mask=head_mask)
|
||||
model(input_ids, token_type_ids=token_type_ids)
|
||||
(sequence_output,) = model(input_ids)
|
||||
|
||||
result = {
|
||||
"sequence_output": sequence_output
|
||||
}
|
||||
self.parent.assertListEqual(
|
||||
list(result["sequence_output"].size()),
|
||||
[self.batch_size, self.seq_length, self.hidden_size])
|
||||
|
||||
def create_and_check_lm_head_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = OpenAIGPTLMHeadModel(config)
|
||||
model.eval()
|
||||
|
||||
loss, lm_logits = model(input_ids, token_type_ids=token_type_ids, labels=input_ids)
|
||||
|
||||
result = {
|
||||
"loss": loss,
|
||||
"lm_logits": lm_logits
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
self.parent.assertListEqual(
|
||||
list(result["lm_logits"].size()),
|
||||
[self.batch_size, self.seq_length, self.vocab_size])
|
||||
|
||||
def create_and_check_double_lm_head_model(self, config, input_ids, head_mask, token_type_ids, *args):
|
||||
model = OpenAIGPTDoubleHeadsModel(config)
|
||||
model.eval()
|
||||
|
||||
loss, lm_logits, mc_logits = model(input_ids, token_type_ids=token_type_ids, lm_labels=input_ids)
|
||||
|
||||
result = {
|
||||
"loss": loss,
|
||||
"lm_logits": lm_logits
|
||||
}
|
||||
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
self.parent.assertListEqual(
|
||||
list(result["lm_logits"].size()),
|
||||
[self.batch_size, self.seq_length, self.vocab_size])
|
||||
|
||||
def prepare_config_and_inputs_for_common(self):
|
||||
config_and_inputs = self.prepare_config_and_inputs()
|
||||
(config, input_ids, head_mask, token_type_ids, sequence_labels, token_labels, choice_labels) = config_and_inputs
|
||||
inputs_dict = {
|
||||
'input_ids': input_ids,
|
||||
'token_type_ids': token_type_ids,
|
||||
'head_mask': head_mask
|
||||
}
|
||||
|
||||
return config, inputs_dict
|
||||
|
||||
def setUp(self):
|
||||
self.model_tester = OpenAIGPTModelTest.OpenAIGPTModelTester(self)
|
||||
self.config_tester = ConfigTester(self, config_class=OpenAIGPTConfig, n_embd=37)
|
||||
|
||||
def test_config(self):
|
||||
config_tester = ConfigTester(self, config_class=OpenAIGPTConfig, n_embd=37)
|
||||
config_tester.run_common_tests()
|
||||
self.config_tester.run_common_tests()
|
||||
|
||||
def test_model(self):
|
||||
model_tester = CommonTestCases.GPTModelTester(self, config_class=OpenAIGPTConfig, base_model_class=OpenAIGPTModel,
|
||||
lm_head_model_class=OpenAIGPTLMHeadModel,
|
||||
double_head_model_class=OpenAIGPTDoubleHeadsModel)
|
||||
model_tester.run_common_tests(test_presents=False)
|
||||
def test_openai_gpt_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_openai_gpt_model(*config_and_inputs)
|
||||
|
||||
def test_openai_gpt_lm_head_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_lm_head_model(*config_and_inputs)
|
||||
|
||||
def test_openai_gpt_double_lm_head_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_double_lm_head_model(*config_and_inputs)
|
||||
|
||||
@pytest.mark.slow
|
||||
def test_pretrained(self):
|
||||
model_tester = CommonTestCases.GPTModelTester(self, config_class=OpenAIGPTConfig, base_model_class=OpenAIGPTModel,
|
||||
lm_head_model_class=OpenAIGPTLMHeadModel,
|
||||
double_head_model_class=OpenAIGPTDoubleHeadsModel)
|
||||
model_tester.run_slow_tests()
|
||||
def test_model_from_pretrained(self):
|
||||
cache_dir = "/tmp/pytorch_transformers_test/"
|
||||
for model_name in list(OPENAI_GPT_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:
|
||||
model = OpenAIGPTModel.from_pretrained(model_name, cache_dir=cache_dir)
|
||||
shutil.rmtree(cache_dir)
|
||||
self.assertIsNotNone(model)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
|
||||
@@ -17,6 +17,7 @@ from __future__ import absolute_import, division, print_function, unicode_litera
|
||||
import os
|
||||
import unittest
|
||||
import json
|
||||
from io import open
|
||||
|
||||
from pytorch_transformers.tokenization_gpt2 import GPT2Tokenizer, VOCAB_FILES_NAMES
|
||||
|
||||
@@ -31,17 +32,18 @@ class GPT2TokenizationTest(CommonTestCases.CommonTokenizerTester):
|
||||
|
||||
# Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt
|
||||
vocab = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n",
|
||||
"lo", "low", "er",
|
||||
"low", "lowest", "newer", "wider", "<unk>"]
|
||||
"\u0120", "\u0120l", "\u0120n",
|
||||
"\u0120lo", "\u0120low", "er",
|
||||
"\u0120lowest", "\u0120newer", "\u0120wider", "<unk>"]
|
||||
vocab_tokens = dict(zip(vocab, range(len(vocab))))
|
||||
merges = ["#version: 0.2", "l o", "lo w", "e r", ""]
|
||||
merges = ["#version: 0.2", "\u0120 l", "\u0120l o", "\u0120lo w", "e r", ""]
|
||||
self.special_tokens_map = {"unk_token": "<unk>"}
|
||||
|
||||
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") as fp:
|
||||
fp.write(json.dumps(vocab_tokens))
|
||||
with open(self.merges_file, "w") as fp:
|
||||
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):
|
||||
@@ -50,18 +52,18 @@ class GPT2TokenizationTest(CommonTestCases.CommonTokenizerTester):
|
||||
|
||||
def get_input_output_texts(self):
|
||||
input_text = u"lower newer"
|
||||
output_text = u"lower<unk>newer"
|
||||
output_text = u" lower newer"
|
||||
return input_text, output_text
|
||||
|
||||
def test_full_tokenizer(self):
|
||||
tokenizer = GPT2Tokenizer(self.vocab_file, self.merges_file, **self.special_tokens_map)
|
||||
text = "lower"
|
||||
bpe_tokens = ["low", "er"]
|
||||
text = "lower newer"
|
||||
bpe_tokens = ["\u0120low", "er", "\u0120", "n", "e", "w", "er"]
|
||||
tokens = tokenizer.tokenize(text)
|
||||
self.assertListEqual(tokens, bpe_tokens)
|
||||
|
||||
input_tokens = tokens + [tokenizer.unk_token]
|
||||
input_bpe_tokens = [13, 12, 17]
|
||||
input_bpe_tokens = [14, 15, 10, 9, 3, 2, 15, 19]
|
||||
self.assertListEqual(
|
||||
tokenizer.convert_tokens_to_ids(input_tokens), input_bpe_tokens)
|
||||
|
||||
|
||||
@@ -17,6 +17,7 @@ from __future__ import absolute_import, division, print_function, unicode_litera
|
||||
import os
|
||||
import json
|
||||
import unittest
|
||||
from io import open
|
||||
|
||||
from pytorch_transformers.tokenization_roberta import RobertaTokenizer, VOCAB_FILES_NAMES
|
||||
from .tokenization_tests_commons import CommonTestCases
|
||||
@@ -30,17 +31,18 @@ class RobertaTokenizationTest(CommonTestCases.CommonTokenizerTester):
|
||||
|
||||
# Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt
|
||||
vocab = ["l", "o", "w", "e", "r", "s", "t", "i", "d", "n",
|
||||
"lo", "low", "er",
|
||||
"low", "lowest", "newer", "wider", "<unk>"]
|
||||
"\u0120", "\u0120l", "\u0120n",
|
||||
"\u0120lo", "\u0120low", "er",
|
||||
"\u0120lowest", "\u0120newer", "\u0120wider", "<unk>"]
|
||||
vocab_tokens = dict(zip(vocab, range(len(vocab))))
|
||||
merges = ["#version: 0.2", "l o", "lo w", "e r", ""]
|
||||
merges = ["#version: 0.2", "\u0120 l", "\u0120l o", "\u0120lo w", "e r", ""]
|
||||
self.special_tokens_map = {"unk_token": "<unk>"}
|
||||
|
||||
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") as fp:
|
||||
fp.write(json.dumps(vocab_tokens))
|
||||
with open(self.merges_file, "w") as fp:
|
||||
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):
|
||||
@@ -49,18 +51,18 @@ class RobertaTokenizationTest(CommonTestCases.CommonTokenizerTester):
|
||||
|
||||
def get_input_output_texts(self):
|
||||
input_text = u"lower newer"
|
||||
output_text = u"lower<unk>newer"
|
||||
output_text = u" lower newer"
|
||||
return input_text, output_text
|
||||
|
||||
def test_full_tokenizer(self):
|
||||
tokenizer = RobertaTokenizer(self.vocab_file, self.merges_file, **self.special_tokens_map)
|
||||
text = "lower"
|
||||
bpe_tokens = ["low", "er"]
|
||||
text = "lower newer"
|
||||
bpe_tokens = ["\u0120low", "er", "\u0120", "n", "e", "w", "er"]
|
||||
tokens = tokenizer.tokenize(text)
|
||||
self.assertListEqual(tokens, bpe_tokens)
|
||||
|
||||
input_tokens = tokens + [tokenizer.unk_token]
|
||||
input_bpe_tokens = [13, 12, 17]
|
||||
input_bpe_tokens = [14, 15, 10, 9, 3, 2, 15, 19]
|
||||
self.assertListEqual(
|
||||
tokenizer.convert_tokens_to_ids(input_tokens), input_bpe_tokens)
|
||||
|
||||
|
||||
@@ -120,7 +120,7 @@ class CommonTestCases:
|
||||
self.assertGreater(tokens[-2], tokenizer.vocab_size - 1)
|
||||
|
||||
new_toks_2 = {'eos_token': ">>>>|||<||<<|<<",
|
||||
'pad_token': "<<<<<|||>|>>>>|>"}
|
||||
'pad_token': "<<<<<|||>|>>>>|>"}
|
||||
added_toks_2 = tokenizer.add_special_tokens(new_toks_2)
|
||||
vocab_size_3 = tokenizer.vocab_size
|
||||
all_size_3 = len(tokenizer)
|
||||
@@ -137,8 +137,8 @@ class CommonTestCases:
|
||||
self.assertGreater(tokens[0], tokens[1])
|
||||
self.assertGreater(tokens[-2], tokenizer.vocab_size - 1)
|
||||
self.assertGreater(tokens[-2], tokens[-3])
|
||||
self.assertEqual(tokens[0], tokenizer.convert_tokens_to_ids(tokenizer.eos_token))
|
||||
self.assertEqual(tokens[-2], tokenizer.convert_tokens_to_ids(tokenizer.pad_token))
|
||||
self.assertEqual(tokens[0], tokenizer.eos_token_id)
|
||||
self.assertEqual(tokens[-2], tokenizer.pad_token_id)
|
||||
|
||||
|
||||
def test_required_methods_tokenizer(self):
|
||||
|
||||
Reference in New Issue
Block a user