-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
235 lines (208 loc) · 10.4 KB
/
train.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
import sys
import argparse
from tqdm import tqdm
from loguru import logger
import numpy as np
from scipy.stats import spearmanr
import torch
import torch.nn.functional as F
from torch.utils.data import Dataset, DataLoader
from dataloader import TrainDataset, TestDataset, DevDataset, load_sts_data, load_sts_data_unsup, random_swap_word,random_delete_word, load_data_sup, TrainDataset_sup
from model import SimcseModel_dropout, simcse_unsup_loss, SimcseModel_sup, simcse_sup_loss
from transformers import BertModel, BertConfig, BertTokenizer
import matplotlib.pyplot as plt
# from transformers import AutoTokenizer, AutoModelForMaskedLM
# tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
#
# model = AutoModelForMaskedLM.from_pretrained("bert-base-uncased")
def show(file1,file2):
input_1 = torch.load(file1)
input_2 = torch.load(file2)
corrcoef_update1 = [0.0]
corrcoef_update2 = [0.0]
for line in input_1:
corrcoef_update1.append(line)
for line in input_2:
corrcoef_update2.append(line)
plt.ion()
for i in range(1, len(corrcoef_update2)):
ix = corrcoef_update1[:i]
iy = corrcoef_update2[:i]
plt.cla()
plt.plot(ix)
plt.plot(iy)
# plt.title("loss")
# plt.plot(ix, iy)
# plt.xlabel("epoch")
# plt.ylabel("acc")
plt.pause(0.1)
plt.ioff()
plt.show()
# plt.figure('frame time')
# plt.subplot(211)
# plt.plot(corrcoef_update1, '.r', )
# plt.grid(True)
# plt.subplot(212)
# plt.plot(corrcoef_update1)
# plt.plot(corrcoef_update2)
# plt.grid(True)
# plt.show()
def train_unsup(model, train_dl, dev_dl, optimizer, device, save_path1,save_path2):
"""模型训练函数"""
model.train()
best = 0
loss_list = []
for batch_idx, source in enumerate(tqdm(train_dl), start=1):
# 维度转换 [batch, 2, seq_len] -> [batch * 2, sql_len]
real_batch_num = source.get('input_ids').shape[0]
input_ids = source.get('input_ids').view(real_batch_num * 2, -1).to(device)
attention_mask = source.get('attention_mask').view(real_batch_num * 2, -1).to(device)
token_type_ids = source.get('token_type_ids').view(real_batch_num * 2, -1).to(device)
out = model(input_ids, attention_mask, token_type_ids)
loss = simcse_unsup_loss(out, device)
optimizer.zero_grad()
loss.backward()
optimizer.step()
if batch_idx % 5 == 0:
logger.info(f'loss: {loss.item():.4f}')
corrcoef = eval(model, dev_dl)
model.train()
if best < corrcoef:
best = corrcoef
torch.save(model.state_dict(), save_path1)
torch.save(loss_list, save_path2)
logger.info(f"in batch: {batch_idx} save model,higher corrcoef: {best:.4f} ")
loss_list.append(best)
def train_sup(model, train_dl, dev_dl, optimizer, save_path, batch_size):
"""模型训练函数
"""
model.train()
best = 0
early_stop_batch = 0
for batch_idx, source in enumerate(tqdm(train_dl), start=1):
# 维度转换 [batch, 3, seq_len] -> [batch * 3, sql_len]
real_batch_num = source.get('input_ids').shape[0]
input_ids = source.get('input_ids').view(real_batch_num * 3, -1).to(DEVICE)
attention_mask = source.get('attention_mask').view(real_batch_num * 3, -1).to(DEVICE)
token_type_ids = source.get('token_type_ids').view(real_batch_num * 3, -1).to(DEVICE)
# 训练
out = model(input_ids, attention_mask, token_type_ids)
loss = simcse_sup_loss(out)
optimizer.zero_grad()
loss.backward()
optimizer.step()
# 评估
if batch_idx % 10 == 0:
logger.info(f'loss: {loss.item():.4f}')
corrcoef = eval(model, dev_dl)
model.train()
if best < corrcoef:
early_stop_batch = 0
best = corrcoef
torch.save(model.state_dict(), save_path)
logger.info(f"higher corrcoef: {best:.4f} in batch: {batch_idx}, save model")
continue
early_stop_batch += 1
if early_stop_batch == 10:
logger.info(f"corrcoef doesn't improve for {early_stop_batch} batch, early stop!")
logger.info(f"train use sample number: {(batch_idx - 10) * batch_size}")
return
DEVICE = torch.device('cuda:3' if torch.cuda.is_available() else 'cpu')
def eval(model, dataloader) -> float:
"""模型评估函数
批量预测, batch结果拼接, 一次性求spearman相关度
"""
model.eval()
sim_tensor = torch.tensor([], device=DEVICE)
label_array = np.array([])
with torch.no_grad():
for source, target, label in dataloader:
# source [batch, 1, seq_len] -> [batch, seq_len]
source_input_ids = source.get('input_ids').squeeze(1).to(DEVICE)
source_attention_mask = source.get('attention_mask').squeeze(1).to(DEVICE)
source_token_type_ids = source.get('token_type_ids').squeeze(1).to(DEVICE)
source_pred = model(source_input_ids, source_attention_mask, source_token_type_ids)
# target [batch, 1, seq_len] -> [batch, seq_len]
target_input_ids = target.get('input_ids').squeeze(1).to(DEVICE)
target_attention_mask = target.get('attention_mask').squeeze(1).to(DEVICE)
target_token_type_ids = target.get('token_type_ids').squeeze(1).to(DEVICE)
target_pred = model(target_input_ids, target_attention_mask, target_token_type_ids)
# concat
sim = F.cosine_similarity(source_pred, target_pred, dim=-1)
sim_tensor = torch.cat((sim_tensor, sim), dim=0)
label_array = np.append(label_array, np.array(label))
# corrcoef
return spearmanr(label_array, sim_tensor.cpu().numpy()).correlation
def main(args):
args.device = torch.device("cuda:3" if torch.cuda.is_available() else "cpu")
train_path_sp = args.data_path + "cnsd-sts-train.txt"
train_path_unsp = args.data_path + "cnsd-sts-train_unsup.txt"
dev_path_sp = args.data_path + "cnsd-sts-dev.txt"
test_path_sp = args.data_path + "cnsd-sts-test.txt"
dev_data_source = load_sts_data(dev_path_sp)
test_data_source = load_sts_data(test_path_sp)
tokenizer = BertTokenizer.from_pretrained(args.pretrain_model_path)
if args.un_supervise:
train_data_source = load_sts_data_unsup(train_path_unsp)
sentence = [data[0] for data in train_data_source]
if args.addreverse:
train_sents = random_swap_word(sentence, 0.1)
else:
if args.addremove:
train_sents = random_delete_word(sentence, 0.1)
else:
train_sents = [data[0] for data in train_data_source]
train_dataset = TrainDataset(train_sents, tokenizer, max_len=args.max_length)
else:
train_data_source = load_data_sup(train_path_sp)
# train_sents = [data[0] for data in train_data_source] + [data[1] for data in train_data_source]
train_dataset = TrainDataset_sup(train_data_source)
train_dataloader = DataLoader(train_dataset, batch_size=args.batch_size, shuffle=True, num_workers=12)
test_dataset = TestDataset(test_data_source, tokenizer, max_len=args.max_length)
test_dataloader = DataLoader(test_dataset, batch_size=args.batch_size, shuffle=True, num_workers=12)
dev_dataset = DevDataset(dev_data_source, tokenizer, max_len=args.max_length)
dev_dataloader = DataLoader(dev_dataset, batch_size=args.batch_size, shuffle=True, num_workers=12)
assert args.pooler in ['cls', "pooler", "last-avg", "first-last-avg"]
if args.un_supervise:
model_drop = SimcseModel_dropout(pretrained_model=args.pretrain_model_path, pooling=args.pooler,
dropout=args.dropout).to(args.device)
optimizer = torch.optim.AdamW(model_drop.parameters(), lr=args.lr)
train_unsup(model_drop, train_dataloader, test_dataloader, optimizer, args.device, args.save_path1, args.save_path4)
model_drop.load_state_dict(torch.load(args.save_path1))
dev_corrcoef = eval(model_drop, dev_dataloader)
test_corrcoef = eval(model_drop, test_dataloader)
logger.info(f'dev_corrcoef: {dev_corrcoef:.4f}')
logger.info(f'test_corrcoef: {test_corrcoef:.4f}')
else:
model_sup = SimcseModel_sup(pretrained_model=args.pretrain_model_path, pooling=args.pooler).to(args.device)
optimizer = torch.optim.AdamW(model_sup.parameters(), lr=args.lr)
train_sup(model_sup, train_dataloader, test_dataloader, optimizer, args.save_path, batch_size=args.batch_size)
model_sup.load_state_dict(torch.load(args.save_path))
dev_corrcoef = eval(model_sup, dev_dataloader)
test_corrcoef = eval(model_sup, test_dataloader)
logger.info(f'dev_corrcoef: {dev_corrcoef:.4f}')
logger.info(f'test_corrcoef: {test_corrcoef:.4f}')
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--device", type=str, default='gpu', help="gpu or cpu")
parser.add_argument("--save_path1", type=str, default='../model_save/simcse_unsup.pt')
parser.add_argument("--save_path2", type=str, default='../model_save/train.pth')
parser.add_argument("--save_path3", type=str, default='../model_save/train_none.pth')
parser.add_argument("--save_path4", type=str, default='../model_save/train_test.pth')
parser.add_argument("--un_supervise", type=bool, default=True)
parser.add_argument("--addreverse", type=bool, default=False)
parser.add_argument("--addremove", type=bool, default=False)
parser.add_argument("--lr", type=float, default=1e-5)
parser.add_argument("--dropout", type=float, default=0.15)
parser.add_argument("--batch_size", type=float, default=64)
parser.add_argument("--max_length", type=int, default=64, help="max length of input sentences")
parser.add_argument("--data_path", type=str, default="../data/STS-B/")
parser.add_argument("--pretrain_model_path", type=str,
default="bert-base-chinese")
parser.add_argument("--pooler", type=str, choices=['cls', "pooler", "last-avg", "first-last-avg"],
default='cls', help='which pooler to use')
args = parser.parse_args()
# logger.add("../log/train.log")
# logger.info(args)
main(args)
# show(args.save_path2,args.save_path3)