/
generate_t5_pegasus.py
260 lines (223 loc) · 8.93 KB
/
generate_t5_pegasus.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
#!/usr/bin/env python
# coding=utf-8
import argparse
from functools import partial
from tqdm import tqdm
import jieba
import torch
import logging
import sacrebleu
import numpy as np
from accelerate import Accelerator
from datasets import load_dataset
from torch.utils.data.dataloader import DataLoader
from transformers import (
DataCollatorForSeq2Seq,
set_seed,
BertTokenizer,
T5ForConditionalGeneration
)
def postprocess_text(preds, labels):
preds = [pred.replace(" ", "") for pred in preds]
labels = [label.replace(" ", "") for label in labels]
return preds, labels
def gen_args():
parser = argparse.ArgumentParser(description="generation for t5 pegasus")
parser.add_argument(
"--model_name", type=str, default='fnlp/bart-base-chinese',
)
parser.add_argument(
"--validation_file", type=str, default='data/eval.json',
help="A csv or a json file containing the validation data."
)
parser.add_argument(
"--ignore_pad_token_for_loss",
type=bool,
default=True,
help="Whether to ignore the tokens corresponding to " "padded labels in the loss computation or not.",
)
parser.add_argument(
"--max_source_length",
type=int,
default=1024,
help="The maximum total input sequence length after "
"tokenization.Sequences longer than this will be truncated, sequences shorter will be padded.",
)
parser.add_argument(
"--preprocessing_num_workers",
type=int,
default=None,
help="The number of processes to use for the preprocessing.",
)
parser.add_argument(
"--overwrite_cache", type=bool, default=False, help="Overwrite the cached training and evaluation sets"
)
parser.add_argument(
"--max_target_length",
type=int,
default=1024,
help="The maximum total sequence length for target text after "
"tokenization. Sequences longer than this will be truncated, sequences shorter will be padded."
"during ``evaluate`` and ``predict``.",
)
parser.add_argument(
"--max_length",
type=int,
default=1024,
help=(
"The maximum total input sequence length after tokenization. Sequences longer than this will be truncated,"
" sequences shorter will be padded if `--pad_to_max_lengh` is passed."
),
)
parser.add_argument(
"--per_device_eval_batch_size",
type=int,
default=8,
help="Batch size (per device) for the evaluation dataloader.",
)
parser.add_argument(
"--seed",
type=int,
default=815,
)
parser.add_argument("--output_path", type=str, default='steps', help="Where to store the final model.")
parser.add_argument(
"--log_file", type=str, default=None, help="log file."
)
parser.add_argument(
"--num_beams",
type=int,
default=5,
help="Number of beams to use for evaluation. This argument will be "
"passed to ``model.generate``, which is used during ``evaluate`` and ``predict``.",
)
args = parser.parse_args()
# Sanity checks
if args.validation_file is None:
raise ValueError("Need either a dataset name or a validation file.")
else:
if args.validation_file is not None:
extension = args.validation_file.split(".")[-1]
assert extension in ["csv", "json"], "`validation_file` should be a csv or a json file."
return args
class T5PegasusTokenizer(BertTokenizer):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.pre_tokenizer = partial(jieba.cut, HMM=False)
def _tokenize(self, text, *arg, **kwargs):
split_tokens = []
for text in self.pre_tokenizer(text):
if text in self.vocab:
split_tokens.append(text)
else:
split_tokens.extend(super()._tokenize(text))
return split_tokens
def write_gen_file(predict_sys, out_file):
with open(out_file, 'a+') as f:
for line in predict_sys:
f.write(line + '\n')
print('write complete!')
class MODEL:
def __init__(self, checkpoint):
self.model = T5ForConditionalGeneration.from_pretrained(checkpoint)
self.tokenizer = T5PegasusTokenizer.from_pretrained(checkpoint)
def predict(self, args):
logger = logging.getLogger(__name__)
logging.basicConfig(
filename=args.log_file,
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO,
)
accelerator = Accelerator()
set_seed(args.seed)
data_files = {'validation': args.validation_file}
extension = args.validation_file.split(".")[-1]
raw_datasets = load_dataset(extension, data_files=data_files)
column_names = raw_datasets['validation'].column_names
text_column, summary_column = column_names[0], column_names[1]
# Temporarily set max_target_length for training
padding = True
def preprocess_function(examples):
inputs = examples[text_column]
targets = examples[summary_column]
inputs = [inp for inp in inputs]
model_inputs = self.tokenizer(
inputs,
return_token_type_ids=False,
max_length=args.max_source_length,
padding=padding,
truncation=True
)
with self.tokenizer.as_target_tokenizer():
labels = self.tokenizer(
targets,
return_token_type_ids=False,
max_length=args.max_target_length,
padding=padding,
truncation=True
)
if padding == "max_length" and args.ignore_pad_token_for_loss:
labels["input_ids"] = [
[(l if l != self.tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
processed_datasets = raw_datasets.map(
preprocess_function, batched=True, remove_columns=column_names,
load_from_cache_file=False
)
eval_dataset = processed_datasets["validation"]
label_pad_token_id = -100 if args.ignore_pad_token_for_loss else self.tokenizer.pad_token_id
data_collator = DataCollatorForSeq2Seq(
self.tokenizer,
model=self.model,
label_pad_token_id=label_pad_token_id,
pad_to_multiple_of=8 if accelerator.use_fp16 else None,
)
eval_dataloader = DataLoader(eval_dataset, collate_fn=data_collator, batch_size=args.per_device_eval_batch_size)
model, eval_dataloader = accelerator.prepare(
self.model, eval_dataloader
)
model.eval()
gen_kwargs = {
"max_length": args.max_target_length,
"num_beams": args.num_beams,
}
logger.info("***** start evaluating *****")
predict_sys = []
predict_ref = []
for step, batch in enumerate(tqdm(eval_dataloader)):
with torch.no_grad():
generated_tokens = accelerator.unwrap_model(model).generate(
batch["input_ids"],
attention_mask=batch["attention_mask"],
**gen_kwargs,
)
generated_tokens = accelerator.pad_across_processes(
generated_tokens, dim=1, pad_index=self.tokenizer.pad_token_id
)
labels = accelerator.pad_across_processes(
batch["labels"], dim=1,
pad_index=self.tokenizer.pad_token_id
)
generated_tokens = accelerator.gather(generated_tokens).cpu().numpy()
labels = accelerator.gather(labels).cpu().numpy()
if args.ignore_pad_token_for_loss:
# Replace -100 in the labels as we can't decode them.
labels = np.where(labels != -100, labels, self.tokenizer.pad_token_id)
decoded_preds = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
decoded_labels = self.tokenizer.batch_decode(labels, skip_special_tokens=True)
decoded_preds, decoded_labels = postprocess_text(decoded_preds, decoded_labels)
predict_sys.extend(decoded_preds)
predict_ref.extend(decoded_labels)
# predict_ref = [predict_ref]
bleu_score = sacrebleu.corpus_bleu(predict_sys, [predict_ref], tokenize='zh').score
logging.info(f" current score = {bleu_score}")
write_gen_file(predict_sys, args.output_path)
def main():
args = gen_args()
model = MODEL(args.model_name)
model.predict(args)
if __name__ == '__main__':
main()