/
run_triplet_classification.py
236 lines (222 loc) · 10.5 KB
/
run_triplet_classification.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
import logging
import os
import time
import numpy as np
import torch
from transformers import AutoConfig, AutoTokenizer, HfArgumentParser, TrainingArguments, set_seed
from transformers import BertForSequenceClassification, BertForNextSentencePrediction
from model.trainer import KGCTrainer
from model.data_processor import DictDataset, KGProcessor
from model.data_collator import PoolingCollator, PromptCollator, TempCollator
from model.utils import DataArguments, ModelArguments
from model.bert_template_model import PTuneNSP
import pickle
logger = logging.getLogger(__name__)
def simple_accuracy(preds, labels):
return (preds == labels).mean()
def compute_metrics(task_name, preds, labels):
assert len(preds) == len(labels)
if task_name == 'kg':
return {'acc': simple_accuracy(preds, labels)}
else:
raise KeyError(task_name)
def main():
parser = HfArgumentParser((ModelArguments, DataArguments, TrainingArguments))
(model_args, data_args, training_args) = parser.parse_args_into_dataclasses()
print(training_args.learning_rate)
print(model_args.top_layer_nums)
if (
os.path.exists(training_args.output_dir)
and os.listdir(training_args.output_dir)
and training_args.do_train
and (not training_args.overwrite_output_dir)
):
raise ValueError(
f'Output directory ({training_args.output_dir}) already exists and is not empty. Use --overwrite_output_dir to overcome.')
logging.basicConfig(
format='%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt='%m/%d/%Y %H:%M:%S',
level=logging.INFO if training_args.local_rank in [-1, 0] else logging.WARN
)
assert (model_args.word_embedding_type in [None, "linear", "double-linear", "mlp"])
assert (model_args.top_additional_layer_type in [None, "linear", "double-linear", "mlp", "adapter-module"])
logger.warning(
'Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s',
training_args.local_rank,
training_args.device,
training_args.n_gpu,
bool(training_args.local_rank != -1),
training_args.fp16
)
logger.info('Training/evaluation parameters %s', training_args)
set_seed(training_args.seed)
if training_args.do_train and not model_args.load_checkpoint:
use_cache = True
logger.info("Using cached pretrained model.")
elif model_args.checkpoint_dir is None:
use_cache = True
logger.info("Zero shot prediction.")
else:
use_cache = False
logger.info("In prediction setting or using checkpoint.")
model_checkpoint = model_args.checkpoint_dir
if model_args.model_checkpoint_num:
print("Using checkpoint num: %d." % model_args.model_checkpoint_num)
model_checkpoint += "/checkpoint-%d" % model_args.model_checkpoint_num
logger.info("Using checkpoint from dir %s" % model_checkpoint)
tokenizer = AutoTokenizer.from_pretrained(
model_args.model_name_or_path if use_cache else model_args.checkpoint_dir,
cache_dir=model_args.model_cache_dir
)
is_world_process_zero = training_args.local_rank == -1 or torch.distributed.get_rank() == 0
processor = KGProcessor(data_args, tokenizer, is_world_process_zero)
(train_data, dev_data, test_data) = processor.get_dataset(training_args)
config = AutoConfig.from_pretrained(
model_args.config_name if model_args.config_name else model_args.model_name_or_path,
cache_dir=model_args.model_cache_dir
)
if not hasattr(config, 'real_vocab_size'):
config.real_vocab_size = config.vocab_size
if model_args.pos_weight is not None:
model_args.pos_weight = torch.tensor([model_args.pos_weight]).to(training_args.device)
if model_args.model_type == "template":
if use_cache:
tokenizer.add_special_tokens({'additional_special_tokens': [data_args.pseudo_token]})
pseudo_token_id = tokenizer.convert_tokens_to_ids(data_args.pseudo_token)
pad_token_id, unk_token_id = tokenizer.pad_token_id, tokenizer.unk_token_id
template = [data_args.begin_temp, data_args.mid_temp, data_args.end_temp]
print("="*10 + "using prompt model" + "="*10)
if model_args.use_NSP:
model = PTuneNSP.from_pretrained(
model_args.model_name_or_path if use_cache else model_checkpoint,
template=template,
pseudo_token_id=pseudo_token_id,
pad_token_id=pad_token_id,
unk_token_id=unk_token_id,
use_mlm_finetune=model_args.use_mlm_finetune,
use_head_finetune=model_args.use_head_finetune,
use_mlpencoder=model_args.use_mlpencoder,
word_embedding_type=model_args.word_embedding_type,
word_embedding_hidden_size=model_args.word_embedding_hidden_size,
word_embedding_dropout=model_args.word_embedding_dropout,
word_embedding_layernorm=model_args.word_embedding_layernorm,
top_additional_layer_type=model_args.top_additional_layer_type,
top_additional_layer_hidden_size=model_args.top_additional_layer_hidden_size,
top_use_dropout=model_args.top_use_dropout,
dropout_ratio=model_args.dropout_ratio,
top_use_layernorm=model_args.top_use_layernorm,
top_layer_nums=model_args.top_layer_nums,
adapter_type=model_args.adapter_type,
adapter_size=model_args.adapter_size,
from_tf=bool('.ckpt' in model_args.model_name_or_path),
config=config,
cache_dir=model_args.model_cache_dir
)
if not model_args.not_print_model:
print(model)
with open("exp-result-record.txt", "a") as f:
print(model, file=f)
else:
raise NotImplementedError()
data_collator = TempCollator(tokenizer, pseudo_token_id=pseudo_token_id,
prompt_temp=template, nsp=model_args.use_NSP)
elif model_args.model_type == "raw_bert":
print("="*10 + "using bert model" + "="*10)
if model_args.use_NSP:
model = BertForNextSentencePrediction.from_pretrained(
model_args.model_name_or_path,
from_tf=bool('.ckpt' in model_args.model_name_or_path),
config=config,
cache_dir=model_args.model_cache_dir
)
else:
model = BertForSequenceClassification.from_pretrained(
model_args.model_name_or_path,
from_tf=bool('.ckpt' in model_args.model_name_or_path),
config=config,
cache_dir=model_args.model_cache_dir
)
data_collator = PromptCollator(tokenizer, nsp=model_args.use_NSP)
else:
raise NotImplementedError()
trainer = KGCTrainer(
model=model,
args=training_args,
data_collator=data_collator,
train_dataset=train_data,
eval_dataset=dev_data,
prediction_loss_only=True
)
if data_args.group_shuffle:
print('using group shuffle')
trainer.use_group_shuffle(data_args.num_neg)
if training_args.do_train:
model_path = (
model_args.model_name_or_path
if model_args.model_name_or_path is not None and os.path.isdir(model_args.model_name_or_path)
else None
)
trainer.train(model_path=model_path)
trainer.save_model()
if trainer.is_world_master():
tokenizer.save_pretrained(training_args.output_dir)
if training_args.do_predict:
label_map = {'-1': 0, '1': 1}
#trainer.model.set_predict_mode()
trainer.prediction_loss_only = False
trainer.data_collator.set_predict_mode()
(dev_triples, dev_labels) = processor.get_dev_triples(return_label=True)
dev_labels = np.array([label_map[l] for l in dev_labels], dtype=int)
(_, tmp_features) = processor._create_examples_and_features(dev_triples)
all_input_ids = torch.tensor([f.input_ids for f in tmp_features], dtype=torch.long)
all_pos_indicator = torch.tensor([f.pos_indicator for f in tmp_features], dtype=torch.long)
eval_data = DictDataset(input_ids=all_input_ids, pos_indicator=all_pos_indicator)
trainer.data_collator.predict_mask_part = 0
preds = trainer.predict(eval_data).predictions
preds = torch.tensor(preds)
shape = preds.shape
if len(shape) > 1 and shape[1] > 1:
preds = torch.nn.functional.softmax(preds)[:, 0].numpy()
mean_dev = np.mean(preds)
print('mean_dev: ', mean_dev)
if len(shape) > 1 and shape[1] > 1:
a, b = 0, 1
else:
a, b = -5, 5
max_acc = 0
for i in range(1000):
m = (b - a) / 1000 * i + a
tmp_preds = preds - m
acc = np.mean((tmp_preds > 0).astype(int) == dev_labels)
if acc > max_acc:
max_acc = acc
max_m = m
print('max acc: ', max_acc)
print('max m: ', max_m)
mean_dev = max_m
# mean_dev = 0.5
(test_triples, test_labels) = processor.get_test_triples(return_label=True)
test_labels = np.array([label_map[l] for l in test_labels], dtype=int)
(_, tmp_features) = processor._create_examples_and_features(test_triples)
all_input_ids = torch.tensor([f.input_ids for f in tmp_features], dtype=torch.long)
all_pos_indicator = torch.tensor([f.pos_indicator for f in tmp_features], dtype=torch.long)
eval_data = DictDataset(input_ids=all_input_ids, pos_indicator=all_pos_indicator)
preds = trainer.predict(eval_data).predictions
preds = torch.tensor(preds)
shape = preds.shape
if len(shape) > 1 and shape[1] > 1:
preds = torch.nn.functional.softmax(preds)[:, 0].numpy()
with open("case_study/%s-test_info.pkl" % model_args.checkpoint_dir.strip().split("/")[-1], "wb") as f:
pickle.dump([mean_dev, preds], f)
preds = preds - mean_dev
acc = np.mean((preds > 0).astype(int) == test_labels)
print('test acc: ', acc)
with open("exp-result-record.txt", "a") as f:
f.writelines([
"Experiment: %s\n" % data_args.exp_info,
"\tDev max acc: %.06f\n" % max_acc,
"\tTest acc: %.06f\n" % acc
])
print(training_args.output_dir)
if __name__ == '__main__':
main()