-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_positive_only.py
377 lines (313 loc) · 16.6 KB
/
train_positive_only.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
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
import os
import sys
import numpy as np
from tqdm import tqdm
import pickle
import argparse
import torch
import torch.nn as nn
from transformers import TrainingArguments
from trainer import modifiedSeq2SeqTrainer
from trie import Trie
from utils import (reform_input,
get_config,
load_dictionary,
load_trie,
load_cui_label,
read_file_bylines,
convert_sets_to_lists
)
import copy
import json
# import wandb
import ast
def train(config):
config.max_steps = config.max_steps // config.gradient_accumulate
config.save_steps = config.max_steps
training_args = TrainingArguments(
output_dir=config.model_save_path,
num_train_epochs=config.num_train_epochs, # total number of training epochs
per_device_train_batch_size=config.per_device_train_batch_size, # batch size per device during training
per_device_eval_batch_size=config.per_device_eval_batch_size, # batch size for evaluation
warmup_steps=config.warmup_steps, # number of warmup steps for learning rate scheduler
weight_decay=config.weight_decay, # strength of weight decay
logging_dir=config.logging_path, # directory for storing logs
logging_steps=config.logging_steps,
save_steps=config.save_steps,
evaluation_strategy=config.evaluation_strategy,
learning_rate=config.init_lr,
label_smoothing_factor=config.label_smoothing_factor,
max_grad_norm=config.max_grad_norm,
max_steps=config.max_steps,
lr_scheduler_type=config.lr_scheduler_type,
seed=config.seed,
gradient_accumulation_steps=config.gradient_accumulate,
)
if config.t5:
from models import T5EntityPromptModel
from transformers import T5Tokenizer, T5Config
from datagen import prepare_trainer_dataset
t5conf = T5Config.from_pretrained('./t5-large')
t5conf.dropout_rate = config.dropout
tokenizer = T5Tokenizer.from_pretrained('./t5-large')
model = T5EntityPromptModel.from_pretrained(config.model_load_path,
config = t5conf,
finetune = config.finetune,
n_tokens = (config.prompt_tokens_enc, config.prompt_tokens_dec),
load_prompt = config.load_prompt,
soft_prompt_path = config.model_load_path,
initialize_from_vocab = config.init_from_vocab,
)
else:
from models import BartEntityPromptModel
from transformers import BartTokenizer, BartConfig
from datagen import prepare_trainer_dataset as prepare_trainer_dataset
bartconf = BartConfig.from_pretrained(config.model_load_path)
bartconf.max_position_embeddings = config.max_position_embeddings
bartconf.attention_dropout = config.attention_dropout
bartconf.dropout = config.dropout
tokenizer = BartTokenizer.from_pretrained(config.model_token_path,
max_length=1024,
)
model = BartEntityPromptModel.from_pretrained(config.model_load_path,
config = bartconf,
finetune = config.finetune,
n_tokens = (config.prompt_tokens_enc, config.prompt_tokens_dec),
load_prompt = config.load_prompt,
soft_prompt_path = config.model_load_path,
no_finetune_decoder = config.no_finetune_decoder,
)
train_dataset, _, _ = prepare_trainer_dataset(tokenizer,
config.dataset_path,
prefix_mention_is = config.prefix_mention_is,
evaluate = config.evaluation,
dataset=config.dataset
)
if config.unlikelihood_loss:
print('loading trie......')
with open(config.trie_path, "rb") as f:
trie = Trie.load_from_dict(pickle.load(f))
print('trie loaded.......')
trainer = modifiedSeq2SeqTrainer(
model=model, # the instantiated Transformers model to be trained
args=training_args, # training arguments, defined above
train_dataset=train_dataset,
fairseq_loss=config.fairseq_loss,
enc_num = config.prompt_tokens_enc,
dec_num = config.prompt_tokens_dec,
prefix_allowed_tokens_fn = lambda batch_id, sent: trie.get(sent.tolist()),
rdrop = config.rdrop,
)
else:
trainer = modifiedSeq2SeqTrainer(
model=model, # the instantiated Transformers model to be trained
args=training_args, # training arguments, defined above
train_dataset=train_dataset,
fairseq_loss=config.fairseq_loss,
enc_num = config.prompt_tokens_enc,
dec_num = config.prompt_tokens_dec,
rdrop = config.rdrop,
)
trainer.train()
trainer.save_model(config.model_save_path)
def evalu(config):
from fairseq_beam import SequenceGenerator, PrefixConstrainedBeamSearch, PrefixConstrainedBeamSearchWithSampling
if config.t5:
from models import T5EntityPromptModel
from transformers import T5Tokenizer, T5Config
from datagen import prepare_trainer_dataset_t5 as prepare_trainer_dataset
t5conf = T5Config.from_pretrained('./t5-large')
t5conf.dropout_rate = config.dropout
tokenizer = T5Tokenizer.from_pretrained('./t5-large')
model = T5EntityPromptModel.from_pretrained(config.model_load_path,
config = t5conf,
n_tokens = (config.prompt_tokens_enc, config.prompt_tokens_dec),
load_prompt = True,
soft_prompt_path = config.model_load_path
)
else:
from models import BartEntityPromptModel
from transformers import BartTokenizer, BartConfig
from datagen import prepare_trainer_dataset
tokenizer = BartTokenizer.from_pretrained(config.model_token_path)
bartconf = BartConfig.from_pretrained(config.model_load_path)
bartconf.max_position_embeddings = config.max_position_embeddings
bartconf.attention_dropout = config.attention_dropout
bartconf.dropout = config.dropout
bartconf.max_length = config.max_length
model = BartEntityPromptModel.from_pretrained(config.model_load_path,
config = bartconf,
n_tokens = (config.prompt_tokens_enc, config.prompt_tokens_dec),
load_prompt = True,
soft_prompt_path=config.model_load_path
)
model = model.cuda().to(model.device)
train_dataset, dev_dataset, test_dataset = prepare_trainer_dataset(tokenizer,
config.dataset_path,
prefix_mention_is = config.prefix_mention_is,
evaluate = config.evaluation,
)
if config.testset:
print('eval on test set...')
eval_dataset = test_dataset
else:
print('eval on develop set...')
eval_dataset = dev_dataset
cui2str, str2cui = load_dictionary(config)
cui_labels = load_cui_label(config)
trie = load_trie(config)
# if config.rerank:
# print('loading retrieved names......')
# with open(config.retrieved_path, 'r') as f:
# retrieved_names = [line.split('\t')[0].split(' ') for line in f.readlines()]
# print('retrieved names loaded.')
# for i, l in tqdm(enumerate(retrieved_names)):
# for cui in list(l):
# if cui in cui2str:
# continue
# else:
# retrieved_names[i].remove(cui)
# print('loading tokenized names......')
# with open(config.dataset_path+'/tokenized.json', 'r') as f:
# tokenized_names = json.load(f)
# print('tokenized names loaded.')
# if config.gold_sty:
# print('loading tokenized names......')
# with open(config.dataset_path+'/tokenized.json', 'r') as f:
# tokenized_names = json.load(f)
# print('tokenized names loaded.')
# print('loading sty to cui dict.....')
# with open(config.dataset_path+'/sty2cui.json', 'r') as f:
# sty2cuis = json.load(f)
# with open(config.dataset_path+'/sty.json', 'r') as f:
# cuis2sty = json.load(f)
# print('sty to cui dict loaded.')
# trie_dict = {}
# for sty in sty2cuis:
# names = []
# for cui in tqdm(sty2cuis[sty]):
# names += tokenized_names[cui]
# trie_dict[sty] = Trie(names)
if config.wandb:
wandb.init(project=f'{config.model_name}_finetuning')
if config.beam_threshold == 0:
print('without using beam threshold')
beam_strategy = PrefixConstrainedBeamSearch(
tgt_dict=None,
prefix_allowed_tokens_fn=lambda batch_id, sent: trie.get(sent.tolist())
)
else:
beam_strategy = PrefixConstrainedBeamSearchWithSampling(
tgt_dict=None,
prefix_allowed_tokens_fn=lambda batch_id, sent: trie.get(sent.tolist()),
logit_thresholding=config.beam_threshold,
)
fairseq_generator = SequenceGenerator(
models = model,
tgt_dict = None,
beam_size=config.num_beams,
max_len_a=0,
max_len_b=config.max_length,
min_len=config.min_length,
eos=model.config.eos_token_id,
search_strategy=beam_strategy,
##### all hyperparams below are set to default
normalize_scores=True,
len_penalty=config.length_penalty,
unk_penalty=0.0,
temperature=0.7,
match_source_len=False,
no_repeat_ngram_size=0,
symbols_to_strip_from_output=None,
lm_model=None,
lm_weight=1.0,
)
results = list()
cui_results = list()
results_score = list()
given_mentions = list()
corrects = list()
input_ids = []
decoder_input_ids = []
attention_mask = []
scores = {f'count_top{k+1}': 0 for k in range(10)}
for i in tqdm(range(0, len(eval_dataset))):
input_ids.append(eval_dataset[i]['input_ids'])
attention_mask.append(eval_dataset[i]['attention_mask'])
decoder_input_ids.append(eval_dataset[i]['decoder_input_ids_test'])
if i % config.per_device_eval_batch_size == 0:
input_ids, attention_mask = reform_input(torch.stack(input_ids), attention_mask=torch.stack(attention_mask), ending_token=model.config.eos_token_id)
sample = {'net_input':{'input_ids':input_ids, 'attention_mask':attention_mask}}
result_tokens, posi_scores = fairseq_generator.forward(
sample=sample,
prefix_mention_is = config.prefix_mention_is,
prefix_tokens=decoder_input_ids[0].unsqueeze(0).cuda() if config.prefix_mention_is else None,
)
for ba, beam_sent in enumerate(result_tokens):
result = []
cui_result = []
for be, sent in enumerate(beam_sent):
if config.prefix_mention_is:
result.append(tokenizer.decode(sent[len(decoder_input_ids[0]):], skip_special_tokens=True))
else:
result.append(tokenizer.decode(sent, skip_special_tokens=True))
for r in result:
if r.strip(' ') in str2cui:
cui_result.append(str2cui[r.strip(' ')])
else:
cui_result.append(r)
given_mention = tokenizer.decode(decoder_input_ids[0])[:-3].strip()
given_mentions.append(given_mention)
cui_results.append(cui_result)
results.append(result)
results_score.append(posi_scores)
for k in range(10):
accumulated_results = set(cui for cuis in cui_result[:k+1] for cui in cuis)
if cui_labels[i].intersection(accumulated_results):
scores[f'count_top{k+1}'] += 1
if k == 0:
corrects.append('correct')
else:
if k == 0:
corrects.append('wrong')
input_ids = []
decoder_input_ids = []
attention_mask = []
print('=============Top1 Precision :\t',round(scores['count_top1']/(i+1)*100, 3))
print('=============Top2 Precision :\t',round(scores['count_top2']/(i+1)*100, 3))
print('=============Top3 Precision :\t',round(scores['count_top3']/(i+1)*100, 3))
print('=============Top4 Precision :\t',round(scores['count_top4']/(i+1)*100, 3))
print('=============Top5 Precision :\t',round(scores['count_top5']/(i+1)*100, 3))
with open(f'./logs/{config.model_name}.txt', 'a+') as f:
f.write(config.model_load_path+'\n')
f.write(f'=============Top1 Precision :\t{str(round(scores["count_top1"]/(i+1)*100, 3))}\n')
f.write(f'=============Top2 Precision :\t{str(round(scores["count_top2"]/(i+1)*100, 3))}\n')
f.write(f'=============Top3 Precision :\t{str(round(scores["count_top3"]/(i+1)*100, 3))}\n')
f.write(f'=============Top4 Precision :\t{str(round(scores["count_top4"]/(i+1)*100, 3))}\n')
f.write(f'=============Top5 Precision :\t{str(round(scores["count_top5"]/(i+1)*100, 3))}\n\n')
zipped_list = [
{
'correctness': convert_sets_to_lists(correct),
'given_mention': convert_sets_to_lists(given_mention),
'result': convert_sets_to_lists(result),
'cui_label': convert_sets_to_lists(cui_label),
'cui_result': convert_sets_to_lists(cui_result)
}
for correct, result, given_mention, cui_label, cui_result in zip(corrects, results, given_mentions, cui_labels, cui_results)
]
result_score = {f'count_top{k+1}': round(scores[f'count_top{k+1}'] / (i + 1) * 100, 3) for k in range(len(scores))}
zipped_list.insert(0, result_score)
os.makedirs(config.model_load_path, exist_ok=True)
if config.testset:
with open(os.path.join(config.model_load_path, 'results_test_pos.json'), 'w') as f:
json.dump(zipped_list, f, indent=2)
else:
with open(os.path.join(config.model_load_path, 'results_dev_pos.json'), 'w') as f:
json.dump(zipped_list, f, indent=2)
return scores
if __name__ == '__main__':
config = get_config()
if config.evaluation:
evalu(config)
else:
train(config)