Adding CTRL (squashed commit)
adding conversion script adding first draft of modeling & tokenization adding placeholder for test files bunch of changes registering the tokenizer/model/etc tests change link; something is very VERY wrong here weird end-of-word thingy going on i think the tokenization works now ; wrote the unit tests overall structure works;load w next the monster is alive! works after some cleanup as well adding emacs autosave to gitignore currently only supporting the 48 layer one; seems to infer fine on my macbook cleanup fixing some documentation fixing some documentation tests passing? now works on CUDA also adding greedy? adding greedy sampling works well
This commit is contained in:
keskarnitish
213
transformers/tests/modeling_ctrl_test.py
Normal file
213
transformers/tests/modeling_ctrl_test.py
Normal file
@@ -0,0 +1,213 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2018 Salesforce and HuggingFace Inc. team.
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
from __future__ import absolute_import
|
||||
from __future__ import division
|
||||
from __future__ import print_function
|
||||
|
||||
import unittest
|
||||
import pytest
|
||||
import shutil
|
||||
import pdb
|
||||
|
||||
from transformers import is_torch_available
|
||||
|
||||
if is_torch_available():
|
||||
from transformers import (CTRLConfig, CTRLModel, CTRL_PRETRAINED_MODEL_ARCHIVE_MAP,
|
||||
CTRLLMHeadModel)
|
||||
else:
|
||||
pytestmark = pytest.mark.skip("Require Torch")
|
||||
|
||||
from .modeling_common_test import (CommonTestCases, ids_tensor)
|
||||
from .configuration_common_test import ConfigTester
|
||||
|
||||
|
||||
class CTRLModelTest(CommonTestCases.CommonModelTester):
|
||||
|
||||
all_model_classes = (CTRLModel, CTRLLMHeadModel) if is_torch_available() else ()
|
||||
test_pruning = False
|
||||
test_torchscript = False
|
||||
test_resize_embeddings = False
|
||||
test_head_masking = False
|
||||
|
||||
class CTRLModelTester(object):
|
||||
|
||||
def __init__(self,
|
||||
parent,
|
||||
batch_size=13,
|
||||
seq_length=7,
|
||||
is_training=True,
|
||||
use_token_type_ids=True,
|
||||
use_input_mask=True,
|
||||
use_labels=True,
|
||||
use_mc_token_ids=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_input_mask = use_input_mask
|
||||
self.use_labels = use_labels
|
||||
self.use_mc_token_ids = use_mc_token_ids
|
||||
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)
|
||||
|
||||
input_mask = None
|
||||
if self.use_input_mask:
|
||||
input_mask = ids_tensor([self.batch_size, self.seq_length], vocab_size=2)
|
||||
|
||||
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)
|
||||
|
||||
mc_token_ids = None
|
||||
if self.use_mc_token_ids:
|
||||
mc_token_ids = ids_tensor([self.batch_size, self.num_choices], self.seq_length)
|
||||
|
||||
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 = CTRLConfig(
|
||||
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, input_mask, head_mask, token_type_ids, mc_token_ids, sequence_labels, token_labels, choice_labels
|
||||
|
||||
def check_loss_output(self, result):
|
||||
self.parent.assertListEqual(
|
||||
list(result["loss"].size()),
|
||||
[])
|
||||
|
||||
def create_and_check_ctrl_model(self, config, input_ids, input_mask, head_mask, token_type_ids, *args):
|
||||
model = CTRLModel(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, input_mask, head_mask, token_type_ids, *args):
|
||||
model = CTRLLMHeadModel(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 prepare_config_and_inputs_for_common(self):
|
||||
config_and_inputs = self.prepare_config_and_inputs()
|
||||
|
||||
(config, input_ids, input_mask, head_mask, token_type_ids,
|
||||
mc_token_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 = CTRLModelTest.CTRLModelTester(self)
|
||||
self.config_tester = ConfigTester(self, config_class=CTRLConfig, n_embd=37)
|
||||
|
||||
def test_config(self):
|
||||
self.config_tester.run_common_tests()
|
||||
|
||||
def test_ctrl_model(self):
|
||||
config_and_inputs = self.model_tester.prepare_config_and_inputs()
|
||||
self.model_tester.create_and_check_ctrl_model(*config_and_inputs)
|
||||
|
||||
def test_ctrl_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)
|
||||
|
||||
@pytest.mark.slow
|
||||
def test_model_from_pretrained(self):
|
||||
cache_dir = "/tmp/transformers_test/"
|
||||
for model_name in list(CTRL_PRETRAINED_MODEL_ARCHIVE_MAP.keys())[:1]:
|
||||
model = CTRLModel.from_pretrained(model_name, cache_dir=cache_dir)
|
||||
shutil.rmtree(cache_dir)
|
||||
self.assertIsNotNone(model)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
69
transformers/tests/tokenization_ctrl_test.py
Normal file
69
transformers/tests/tokenization_ctrl_test.py
Normal file
@@ -0,0 +1,69 @@
|
||||
# coding=utf-8
|
||||
# Copyright 2018 Salesforce and HuggingFace Inc. team.
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
from __future__ import absolute_import, division, print_function, unicode_literals
|
||||
|
||||
import os
|
||||
import unittest
|
||||
import json
|
||||
from io import open
|
||||
|
||||
from transformers.tokenization_ctrl import CTRLTokenizer, VOCAB_FILES_NAMES
|
||||
|
||||
from .tokenization_tests_commons import CommonTestCases
|
||||
|
||||
class CTRLTokenizationTest(CommonTestCases.CommonTokenizerTester):
|
||||
|
||||
tokenizer_class = CTRLTokenizer
|
||||
|
||||
def setUp(self):
|
||||
super(CTRLTokenizationTest, self).setUp()
|
||||
|
||||
# Adapted from Sennrich et al. 2015 and https://github.com/rsennrich/subword-nmt
|
||||
vocab = ['adapt', 're@@', 'a@@', 'apt', 'c@@', 't', '<unk>']
|
||||
vocab_tokens = dict(zip(vocab, range(len(vocab))))
|
||||
merges = ["#version: 0.2", 'a p', 'ap t</w>', 'r e', 'a d', 'ad apt</w>', '']
|
||||
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", encoding="utf-8") as fp:
|
||||
fp.write(json.dumps(vocab_tokens) + "\n")
|
||||
with open(self.merges_file, "w", encoding="utf-8") as fp:
|
||||
fp.write("\n".join(merges))
|
||||
|
||||
def get_tokenizer(self, **kwargs):
|
||||
kwargs.update(self.special_tokens_map)
|
||||
return CTRLTokenizer.from_pretrained(self.tmpdirname, **kwargs)
|
||||
|
||||
def get_input_output_texts(self):
|
||||
input_text = u"adapt react readapt apt"
|
||||
output_text = u"adapt react readapt apt"
|
||||
return input_text, output_text
|
||||
|
||||
def test_full_tokenizer(self):
|
||||
tokenizer = CTRLTokenizer(self.vocab_file, self.merges_file, **self.special_tokens_map)
|
||||
text = "adapt react readapt apt"
|
||||
bpe_tokens = 'adapt re@@ a@@ c@@ t re@@ adapt apt'.split()
|
||||
tokens = tokenizer.tokenize(text, add_prefix_space=True)
|
||||
self.assertListEqual(tokens, bpe_tokens)
|
||||
|
||||
input_tokens = tokens + [tokenizer.unk_token]
|
||||
|
||||
input_bpe_tokens = [0, 1, 2, 4, 5, 1, 0, 3, 6]
|
||||
self.assertListEqual(
|
||||
tokenizer.convert_tokens_to_ids(input_tokens), input_bpe_tokens)
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
unittest.main()
|
||||
Reference in New Issue
Block a user