-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
173 lines (143 loc) · 7.05 KB
/
main.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
from transformers import AutoConfig,AutoTokenizer
import torch
import torch.nn.functional as F
import os
from tqdm import tqdm
from model import P_Tuningv2_Model,compute_loss
from data_process import load_data
from sklearn.metrics import accuracy_score
def seed_everything(seed=3427):
# random.seed(seed)
os.environ['PYTHONHASHSEED'] = str(seed)
# np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.deterministic = True
def init_tokenizer(path):
# [X]
tokenizer = AutoTokenizer.from_pretrained(path)
tokenizer.add_tokens(['[X]'])
return tokenizer
auto_config = AutoConfig.from_pretrained("roberta_data")
class p_tuningv2_config():
# ptuning parameter
prefix_projection = True
pre_seq_len = 128
prefix_hidden_size = 768
# roberta parameter
hidden_size = auto_config.hidden_size
num_hidden_layers = auto_config.num_hidden_layers
num_attention_heads = auto_config.num_attention_heads
hidden_dropout_prob = auto_config.hidden_dropout_prob
def train():
best_acc = 0
for e in range(epoch):
print(f"当前是第{e}轮")
model.train()
pbar = tqdm(train_data_loader)
for data in pbar:
optimzer.zero_grad()
input_ids_prompt_sen_1 = data['input_ids_prompt_sen_1'].to(device)
attention_masks_prompt_sen_1 = data['attention_masks_prompt_sen_1'].to(device)
input_ids_template_sen_1 = data['input_ids_template_sen_1'].to(device)
attention_masks_template_sen_1 = data['attention_masks_template_sen_1'].to(device)
input_ids_prompt_sen_2 = data['input_ids_prompt_sen_2'].to(device)
attention_masks_prompt_sen_2 = data['attention_masks_prompt_sen_2'].to(device)
input_ids_template_sen_2 = data['input_ids_template_sen_2'].to(device)
attention_masks_template_sen_2 = data['attention_masks_template_sen_2'].to(device)
# h_i-h_i^
s1 = model.forward(input_ids_prompt_sen_1, attention_masks_prompt_sen_1, input_ids_template_sen_1,
attention_masks_template_sen_1)
# h_i'-h_i'^
s2 = model.forward(input_ids_prompt_sen_2, attention_masks_prompt_sen_2, input_ids_template_sen_2,
attention_masks_template_sen_2)
loss = compute_loss(s1,s2)
loss.backward()
optimzer.step()
pbar.update()
pbar.set_description(f"当前loss={loss.item():.6f}")
pred = []
label = []
model.eval()
for data in tqdm(dev_data_loader):
input_ids_prompt_sen_1 = data["input_ids_prompt_sen_1"].to(device)
attention_masks_prompt_sen_1 = data["attention_masks_prompt_sen_1"].to(device)
input_ids_template_sen_1 = data["input_ids_template_sen_1"].to(device)
attention_masks_template_sen_1 = data["attention_masks_template_sen_1"].to(device)
input_ids_prompt_sen_2 = data["input_ids_prompt_sen_2"].to(device)
attention_masks_prompt_sen_2 = data["attention_masks_prompt_sen_2"].to(device)
input_ids_template_sen_2 = data["input_ids_template_sen_2"].to(device)
attention_masks_template_sen_2 = data["attention_masks_template_sen_2"].to(device)
labels = data["label"].to(device)
# ,sim,sentence_embedding
# 前向传播
with torch.no_grad():
# 前向传播
sentence_embedding_a = model.forward(input_ids_prompt_sen_1, attention_masks_prompt_sen_1,
input_ids_template_sen_1, attention_masks_template_sen_1)
sentence_embedding_b = model.forward(input_ids_prompt_sen_2, attention_masks_prompt_sen_2,
input_ids_template_sen_2, attention_masks_template_sen_2)
similarity = F.cosine_similarity(sentence_embedding_a, sentence_embedding_b, dim=-1)
# 获取预测值
similarity = similarity.detach().cpu().numpy()
pred.extend(similarity)
# 获取标签
label.extend(labels.cpu().numpy())
# 计算验证集准确率
pred = torch.tensor(pred)
pred = (pred >= 0.7).long().detach().cpu().numpy()
acc = accuracy_score(pred, label)
print('dev acc:', acc)
print()
# 如果当前准确率大于最佳准确率，则保存模型参数
if acc > best_acc:
torch.save(model.state_dict(), './model_weight/pt2_prompt.bin')
best_acc = acc
def infer(thres_hold=0.5):
# t1->emb->cos
print(f"prompt bert 开始推理")
model.load_state_dict(torch.load('./model_weight/pt2_prompt.bin' ,map_location=device))
model.eval()
pre = []
label = []
for data in dev_data_loader:
input_ids_prompt_sen_1 = data["input_ids_prompt_sen_1"].to(device)
attention_masks_prompt_sen_1 = data["attention_masks_prompt_sen_1"].to(device)
input_ids_template_sen_1 = data["input_ids_template_sen_1"].to(device)
attention_masks_template_sen_1 = data["attention_masks_template_sen_1"].to(device)
input_ids_prompt_sen_2 = data["input_ids_prompt_sen_2"].to(device)
attention_masks_prompt_sen_2 = data["attention_masks_prompt_sen_2"].to(device)
input_ids_template_sen_2 = data["input_ids_template_sen_2"].to(device)
attention_masks_template_sen_2 = data["attention_masks_template_sen_2"].to(device)
labels = data["label"].to(device)
with torch.no_grad():
# 前向传播
sentence_embedding_a = model.forward(input_ids_prompt_sen_1, attention_masks_prompt_sen_1
,input_ids_template_sen_1 ,attention_masks_template_sen_1)
sentence_embedding_b = model.forward(input_ids_prompt_sen_2, attention_masks_prompt_sen_2
,input_ids_template_sen_2 ,attention_masks_template_sen_2)
similarity = F.cosine_similarity(sentence_embedding_a ,sentence_embedding_b ,dim=-1)
similarity = similarity.detach().cpu().numpy()
pre.append(similarity)
label.append(labels.cpu().numpy())
pre = torch.tensor(pre)
pre = (pre >= thres_hold).long().detach().cpu().numpy()
acc = accuracy_score(pre, label)
print('dev acc:', acc)
print()
if __name__ == '__main__':
seed_everything()
pretrained_path = "roberta_data"
tokenizer = init_tokenizer(pretrained_path)
ptv2_cfg = p_tuningv2_config()
train_data_loader, dev_data_loader = load_data(tokenizer, 2)
device = 'cuda' if torch.cuda.is_available() else 'cpu'
model = P_Tuningv2_Model(ptv2_cfg,tokenizer).to(device)
print(model)
# train args
epoch = 10
lr = 3e-5
optimzer = torch.optim.AdamW(model.parameters(), lr=lr)
train()
infer()
pass