-
Notifications
You must be signed in to change notification settings - Fork 0
/
Exp3_Model.py
49 lines (40 loc) · 1.71 KB
/
Exp3_Model.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
import torch
import torch.nn as nn
class TextCNN_Model(nn.Module):
def __init__(self, configs,pre_embedding):
super(TextCNN_Model, self).__init__()
vocab_size = configs.vocab_size
embedding_dimension = configs.embedding_dimension
label_num = configs.label_num
# 词嵌入和dropout
self.pre_embed = nn.Embedding.from_pretrained(pre_embedding,freeze=False)
self.sent_embed = nn.Embedding(vocab_size+2, embedding_dimension)
self.pos1_embed = nn.Embedding(153, 10)
self.pos2_embed = nn.Embedding(153, 10)
self.dropout = nn.Dropout(configs.dropout)
self.covns = nn.ModuleList([
nn.Sequential(
nn.Conv1d(in_channels=embedding_dimension+20,
out_channels=128,
kernel_size=k),
nn.Tanh(),
nn.MaxPool1d(kernel_size=98-k+1)) for k in range(2,6)
])
self.linear = nn.Linear(128*4, label_num)
def forward(self, x):
sent = x['text']
pos1 = x['pos1']
pos2 = x['pos2']
sent_embed = self.pre_embed(sent)
pos1_embed = self.pos1_embed(pos1)
pos2_embed = self.pos2_embed(pos2)
input = torch.cat([sent_embed,pos1_embed,pos2_embed],dim=2) #[64,100,120]
input = input.permute(0,2,1) #[64,120,100]
input = self.dropout(input)
input = [conv(input) for conv in self.covns] #4
output = torch.cat(input,dim=1) #[64,512,1]
output = self.dropout(output)
output = output.view(-1,output.size(1)) #[64,512]
output = self.dropout(output)
output = self.linear(output) #[64,44]
return output