/
train.py
219 lines (185 loc) · 8.66 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
#!usr/bin/env python
#-*- coding:utf-8 -*-
import os
import tqdm
import pandas as pd
import torch
import torch.nn as nn
from torch.optim import Adam
from transformers import BertTokenizer, BertModel, BertConfig
from optim_schedule import ScheduledOptim
from torch.utils.data import Dataset
from torch.utils.data import DataLoader
from model import SoftMaskedBert
from sklearn.model_selection import KFold
MAX_INPUT_LEN = 512
os.environ["CUDA_VISIBLE_DEVICES"] = "1"
class SoftMaskedBertTrainer():
def __init__(self, bert, tokenizer, device, hidden=256, layer_n=1, lr=2e-5, gama=0.8, betas=(0.9, 0.999), weight_decay=0.01, warmup_steps=10000):
self.device = device
self.tokenizer = tokenizer
self.model = SoftMaskedBert(bert, self.tokenizer, hidden, layer_n, self.device).to(self.device)
# if torch.cuda.device_count() > 1:
# print("Using %d GPUS for train" % torch.cuda.device_count())
# self.model = nn.DataParallel(self.model, device_ids=[0,1,2])
optim = Adam(self.model.parameters(), lr=lr, betas=betas, weight_decay=weight_decay)
self.optim_schedule = ScheduledOptim(optim, hidden, n_warmup_steps=warmup_steps)
self.criterion_c = nn.NLLLoss()
self.criterion_d = nn.BCELoss()
self.gama = gama
self.log_freq = 100
def train(self, train_data, epoch):
self.model.train()
return self.iteration(epoch, train_data)
def evaluate(self, val_data, epoch):
self.model.eval()
return self.iteration(epoch, val_data, train=False)
def inference(self, data_loader):
self.model.eval()
out_put = []
data_loader = tqdm.tqdm(enumerate(data_loader),
desc="%s" % 'Inference:',
total=len(data_loader),
bar_format="{l_bar}{r_bar}")
for i, data in data_loader:
# 0. batch_data will be sent into the device(GPU or cpu)
data = {key: value.to(self.device) for key, value in data.items()}
out, prob = self.model(data["input_ids"], data["input_mask"], data["segment_ids"]) #prob [batch_size, seq_len, 1]
out_put.extend(out.argmax(dim=-1).cpu().numpy().tolist())
return [''.join(self.tokenizer.convert_ids_to_tokens(x)) for x in out_put]
def save(self, file_path):
torch.save(self.model.cpu(), file_path)
self.model.to(self.device)
print('Model save {}'.format(file_path))
def load(self, file_path):
if not os.path.exists(file_path):
return
self.model = torch.load(file_path)
self.model.to(self.device)
def iteration(self, epoch, data_loader, train=True):
str_code = "train" if train else "val"
# Setting the tqdm progress bar
data_loader = tqdm.tqdm(enumerate(data_loader),
desc="EP_%s:%d" % (str_code, epoch),
total=len(data_loader),
bar_format="{l_bar}{r_bar}")
avg_loss = 0.0
# total_correct = 0
total_element = 0
c_correct = 0
d_correct = 0
for i, data in data_loader:
# 0. batch_data will be sent into the device(GPU or cpu)
data = {key: value.to(self.device) for key, value in data.items()}
out, prob = self.model(data["input_ids"], data["input_mask"], data["segment_ids"]) #prob [batch_size, seq_len, 1]
prob = prob.reshape(-1, prob.shape[1])
loss_d = self.criterion_d(prob, data['label'].float())
loss_c = self.criterion_c(out.transpose(1, 2), data["output_ids"])
loss = self.gama * loss_c + (1-self.gama) * loss_d
if train:
self.optim_schedule.zero_grad()
loss.backward(retain_graph=True)
self.optim_schedule.step_and_update_lr()
# correct = out.argmax(dim=-1).eq(data["output_ids"]).sum().item()
out = out.argmax(dim=-1)
c_correct += sum([out[i].equal(data['output_ids'][i]) for i in range(len(out))])
prob = torch.round(prob).long()
d_correct += sum([prob[i].equal(data['label'][i]) for i in range(len(prob))])
avg_loss += loss.item()
# total_correct += c_correct
# # total_element += data["label"].nelement()
total_element += len(data)
post_fix = {
"epoch": epoch,
"iter": i,
"avg_loss": avg_loss / (i + 1),
"d_acc": d_correct / total_element,
"c_acc": c_correct / total_element
}
if i % self.log_freq == 0:
data_loader.write(str(post_fix))
print("EP%d_%s, avg_loss=" % (epoch, str_code), avg_loss / len(data_loader), "d_acc=",
d_correct / total_element, "c_acc", c_correct / total_element)
return avg_loss / len(data_loader)
class BertDataset(Dataset):
def __init__(self, tokenizer, dataset, max_len=512, pad_first=True, mode='train'):
self.tokenizer = tokenizer
self.dataset = dataset
self.max_len = max_len
self.data_size = len(dataset)
self.pad_first = pad_first
self.mode = mode
def __len__(self):
return self.data_size
def __getitem__(self, item):
item = self.dataset.iloc[item]
input_ids = item['random_text']
input_ids = ['[CLS]'] + list(input_ids)[:min(len(input_ids), self.max_len - 2)] + ['[SEP]']
# convert to bert ids
input_ids = self.tokenizer.convert_tokens_to_ids(input_ids)
input_mask = [1] * len(input_ids)
segment_ids = [0] * len(input_ids)
pad_len = self.max_len - len(input_ids)
if self.pad_first:
input_ids = [0] * pad_len + input_ids
input_mask = [0] * pad_len + input_mask
segment_ids = [0] * pad_len + segment_ids
else:
input_ids = input_ids + [0] * pad_len
input_mask = input_mask + [0] * pad_len
segment_ids = segment_ids + [0] * pad_len
output = {
'input_ids': input_ids,
'input_mask': input_mask,
'segment_ids': segment_ids,
}
if self.mode == 'train':
output_ids = item['origin_text']
label = item['label']
label = [int(x) for x in label if x != ' ']
output_ids = ['[CLS]'] + list(output_ids)[:min(len(output_ids), self.max_len - 2)] + ['[SEP]']
label = [0] + label[:min(len(label), self.max_len - 2)] + [0]
output_ids = self.tokenizer.convert_tokens_to_ids(output_ids)
pad_label_len = self.max_len - len(label)
if self.pad_first:
output_ids = [0] * pad_len + output_ids
label = [0] * pad_label_len + label
else:
output_ids = output_ids + [0] * pad_len
label = label + [0] * pad_label_len
output = {
'input_ids': input_ids,
'input_mask': input_mask,
'segment_ids': segment_ids,
'output_ids': output_ids,
'label': label
}
return {key: torch.tensor(value) for key, value in output.items()}
if __name__ == '__main__':
dataset = pd.read_csv('data/processed_data/all_same_765376/train.csv')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
config = BertConfig.from_pretrained('data/chinese_wwm_pytorch/bert_config.json')
tokenizer = BertTokenizer.from_pretrained('data/chinese_wwm_pytorch/vocab.txt')
kf = KFold(n_splits=5, shuffle=True)
for k, (train_index, val_index) in enumerate(kf.split(range(len(dataset)))):
print('Start train {} ford'.format(k))
bert = BertModel.from_pretrained('data/chinese_wwm_pytorch/pytorch_model.bin', config=config)
train = dataset.iloc[train_index]
val = dataset.iloc[val_index]
train = BertDataset(tokenizer, train, max_len=152)
train = DataLoader(train, batch_size=8, num_workers=2)
val = BertDataset(tokenizer, val, max_len=152)
val = DataLoader(val, batch_size=8, num_workers=2)
trainer = SoftMaskedBertTrainer(bert, tokenizer, device)
best_loss = 100000
for e in range(100):
trainer.train(train, e)
val_loss = trainer.evaluate(val, e)
if best_loss > val_loss:
best_loss = val_loss
trainer.save('best_model_{}ford.pt'.format(k))
print('Best val loss {}'.format(best_loss))
trainer.load('best_model_{}ford.pt'.format(k))
for i in trainer.inference(val):
print(i)
print('\n')