-
Notifications
You must be signed in to change notification settings - Fork 6
/
model.py
134 lines (98 loc) · 4.84 KB
/
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
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
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
from gensim.models.keyedvectors import KeyedVectors
#import word2vec
#from gensim.models import word2vec
torch.manual_seed(1)
class CNN_Text(nn.Module):
def __init__(self, args, text_field):
super(CNN_Text, self).__init__()
self.args = args
#new_model = word2vec.Word2Vec.load(args.word2vec_model)
V = args.embed_num # 词汇个数，61382，统计得到
D = args.embed_dim # 词向量维度，128，人为设定
#V = text_field.vocab.vectors.shape[0] # 85362
#D = text_field.vocab.vectors.shape[1] # 128
#V = 34265
#D = 128
#print(V, D, type(V), type(D))
C = args.class_num # 分类个数
Ci = 1
Co = args.kernel_num
Ks = args.kernel_sizes
if args.use_word2vec: # 使用预训练模型
self.embed = nn.Embedding(V, D).from_pretrained(text_field.vocab.vectors)
else: # 随机初始化
self.embed = nn.Embedding(V, D)
#self.embed = nn.Embedding(V, D).from_pretrained(text_field.vocab.vectors) # 使用预训练模型
#self.embed.weight.data = torch.Tensor(new_model.vectors).clone()
#self.embed.weight.data = torch.Tensor(wv_model.wv.vectors).clone()
# self.convs1 = [nn.Conv2d(Ci, Co, (K, D)) for K in Ks]
self.convs1 = nn.ModuleList([nn.Conv2d(Ci, Co, (K, D)) for K in Ks])
'''
self.conv13 = nn.Conv2d(Ci, Co, (3, D))
self.conv14 = nn.Conv2d(Ci, Co, (4, D))
self.conv15 = nn.Conv2d(Ci, Co, (5, D))
'''
self.dropout = nn.Dropout(args.dropout)
self.fc1 = nn.Linear(len(Ks)*Co, C)
def conv_and_pool(self, x, conv):
x = F.relu(conv(x)).squeeze(3) # (N, Co, W) # 去掉第四维
x = F.max_pool1d(x, x.size(2)).squeeze(2)
return x
def forward(self, x):
#print(self.embed.weight.sum())
#i = torch.tensor(1)
#i = i.cuda()
#print(dir(self.embed))
#import os
#os._exit(1)
#print(x.shape[0], x.shape[1])
x = self.embed(x).view(x.shape[0], x.shape[1], -1) # (N, W, D) N应该为batch，W为词的个数，D为词向量维度
#print(x.shape) # [64, 44, 128]
if self.args.static: # 静态了，就不允许反向传播时修改embed了
x = Variable(x) # 如果运行了这一行，embed不会被反向传播改变
x = x.unsqueeze(1) # (N, Ci, W, D) # Ci是输入通道个数，这里为1
x = [F.relu(conv(x)).squeeze(3) for conv in self.convs1] # [(N, Co, W), ...]*len(Ks)
x = [F.max_pool1d(i, i.size(2)).squeeze(2) for i in x] # [(N, Co), ...]*len(Ks)
x = torch.cat(x, 1)
'''
x1 = self.conv_and_pool(x,self.conv13) #(N,Co)
x2 = self.conv_and_pool(x,self.conv14) #(N,Co)
x3 = self.conv_and_pool(x,self.conv15) #(N,Co)
x = torch.cat((x1, x2, x3), 1) # (N,len(Ks)*Co)
'''
x = self.dropout(x) # (N, len(Ks)*Co)
logit = self.fc1(x) # (N, C)
return logit
class LSTM_Text(nn.Module):
def __init__(self, args, text_field):
super(LSTM_Text, self).__init__()
self.args = args
#V = args.embed_num # 词汇个数，61382，统计得到
#D = args.embed_dim # 词向量维度，128，人为设定
V = text_field.vocab.vectors.shape[0] # 85362
D = text_field.vocab.vectors.shape[1] # 128
C = args.class_num # 分类个数
Ci = 1
Co = args.kernel_num
Ks = args.kernel_sizes
self.embed = nn.Embedding(V, D) # 随机初始化
self.embed = nn.Embedding(V, D).from_pretrained(text_field.vocab.vectors) # 使用预训练模型
# LSTM
# 如果num_layers = 1, 则dropout = 0。只有num_layers > 1, dropout > 0
self.lstm = nn.LSTM( input_size=D, hidden_size=args.hidden_size, num_layers=args.lstm_num, dropout=0,
batch_first=True, bidirectional=True)
self.bn2 = nn.BatchNorm1d(args.hidden_size*2)
self.fc = nn.Linear(args.hidden_size*2, C)
def forward(self, x):
x = self.embed(x).view(x.shape[0], x.shape[1], -1) # (N, W, D) N应该为batch，W为词的个数，D为词向量维度
#print(x.shape) # [64, 44, 128]
if self.args.static: # 静态了，就不允许反向传播时修改embed了
x = Variable(x) # 如果运行了这一行，embed不会被方向传播改变
#x = x.unsqueeze(1) # (N, Ci, W, D) # Ci是输入通道个数，这里为1
output, ht = self.lstm(x, None) # (N, C)
print(output.shape, ht.shape)
return output