-
Notifications
You must be signed in to change notification settings - Fork 0
/
new_main.py
223 lines (195 loc) · 8.57 KB
/
new_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
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
import torch
import torch.nn as nn
from torch.nn import init
from torch.autograd import Variable
import pickle
import numpy as np
import time
import random
from collections import defaultdict
import torch.nn.functional as F
import torch.utils.data
from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error
from math import sqrt
import datetime
import argparse
import os
import Parser
import Bulidata
import time
from tqdm import tqdm
from getData import get_last_data
from model import HetAD
from model import cross_entropy_loss
from data_input import get_train_data
if __name__ == '__main__':
args = Parser.Define_Params()
embed_dim = args.embed_dim
uu_neigh, ui_neigh, iu_neigh, ii_neigh, _, _ = get_last_data()
# User_num = len(ui_neigh)
# Item_num = len(iu_neigh)
# start_node_embedding = args.start_node_embedding
# u_self_embed = np.zeros((User_num + 1, embed_dim))
# i_self_embed = np.zeros((Item_num + 1, embed_dim))
# with open(start_node_embedding, 'r') as file:
# lines = file.readlines()
# for line in lines:
# line = line.strip('\n')
# line = line.split(' ')
# embed = line[1].split(', ')
# embeds = np.asarray(embed, dtype='float32')
# if line[0][0] == 'u':
# u_self_embed[int(line[0][1:])] = embeds
# elif line[0][0] == 'i':
# i_self_embed[int(line[0][1:])] = embeds
#
# # 这个是本身的权重，然后它的属性例如他评论的项目，的权重相加取均值把然后在聚合
#
# print(u_self_embed)
model = HetAD(embed_dim, uu_neigh, ui_neigh, iu_neigh, ii_neigh)
parameters = filter(lambda p: p.requires_grad, model.parameters())
# optimizer = torch.optim.RMSprop(model.parameters(), lr=args.lr, alpha=0.9)
optim = torch.optim.Adam(parameters, lr=args.lr, weight_decay=0)
model.init_weights()
# triple_list = get_train_data()
# uu_train_set = torch.utils.data.TensorDataset(torch.LongTensor(triple_list[0]))
# uu_train_loader = torch.utils.data.DataLoader(uu_train_set, batch_size=args.batch_size, shuffle=True)
# for type_data in uu_train_loader:
#
# c_id_batch = [int(x[0]) for x in type_data[0]]
# pos_id_batch = [int(x[1]) for x in type_data[0]]
# neg_id_batch = [int(x[2]) for x in type_data[0]]
# #
# # model.u_content_agg(c_id_batch)
# # model.i_content_agg(c_id_batch)
# print(c_id_batch)
# embed = model.types_agg(c_id_batch, 1)
# print(embed.shape)
# for single_embed in embed:
# print(single_embed)
# print(single_embed.detach().numpy().tolist())
# exit(0)
#
# exit(0)
# c_out = torch.zeros([3, 500, embed_dim])
# a_out = []
# a_out.append([0] * 500)
# a_out.append([0] * 230)
# a_out.append([0] * 210)
# print(c_out)
# for i in range(len(a_out)):
# for j in range(len(a_out[i])):
# a_out[i][j] = np.zeros(128)
# a_out = torch.tensor(a_out)
# print(a_out)
# exit(0)
# x = 0
# y = 0
# z = 0
# z1 = 0
# for i in range(100):
# triple_list = get_train_data()
# x = x + len(triple_list[0])
# y = y + len(triple_list[1])
# z = z + len(triple_list[2])
# z1 = z1 + len(triple_list[3])
# print(len(triple_list[0]))
# print(len(triple_list[1]))
# print(len(triple_list[2]))
# print(len(triple_list[3]))
# print(x/100, y/100, z/100, z1/100)
# exit(0)
start_i = 1
RESUME = True
if RESUME:
path_checkpoint = "./models/checkpoint/amazon_ckpt_best_newChangeModeladdII40.pth" # 断点路径
checkpoint = torch.load(path_checkpoint) # 加载断点
model.load_state_dict(checkpoint['net']) # 加载模型可学习参数
optim.load_state_dict(checkpoint['optimizer']) # 加载优化器参数
start_i = checkpoint['i'] # 设置开始的epoch
for i in range(start_i, args.iter_n):
triple_list = get_train_data()
while len(triple_list[0]) <= 2450 or len(triple_list[1]) <= 2320 or len(triple_list[2]) <= 2070 or len(triple_list[3]) <= 2210:
triple_list = get_train_data()
# 各种数据的类型数目不一样所以不能用trainloader去统一，最后可以（统一一下数据数量）看一下效果
uu_train_set = torch.utils.data.TensorDataset(torch.LongTensor(triple_list[0]))
ui_train_set = torch.utils.data.TensorDataset(torch.LongTensor(triple_list[1]))
iu_train_set = torch.utils.data.TensorDataset(torch.LongTensor(triple_list[2]))
ii_train_set = torch.utils.data.TensorDataset(torch.LongTensor(triple_list[3]))
uu_train_loader = torch.utils.data.DataLoader(uu_train_set, batch_size= 245, shuffle=True)
ui_train_loader = torch.utils.data.DataLoader(ui_train_set, batch_size= 232, shuffle=True)
iu_train_loader = torch.utils.data.DataLoader(iu_train_set, batch_size= 207, shuffle=True)
ii_train_loader = torch.utils.data.DataLoader(ii_train_set, batch_size= 221, shuffle=True)
uu_out = torch.zeros([len(triple_list), 245, embed_dim])
ui_out = torch.zeros([len(triple_list), 232, embed_dim])
iu_out = torch.zeros([len(triple_list), 207, embed_dim])
ii_out = torch.zeros([len(triple_list), 221, embed_dim])
epochs = 0
all_loss = 0
for uu_data, ui_data, iu_data, ii_data in zip(uu_train_loader, ui_train_loader, iu_train_loader, ii_train_loader):
if len(uu_data[0]) == 245 and len(ui_data[0]) == 232 and len(iu_data[0]) == 207 and len(ii_data[0]) == 221:
epochs = epochs + 1
c_pre, p_pre, n_pre = model(uu_data, 1, i)
#
uu_out[0] = c_pre
uu_out[1] = p_pre
uu_out[2] = n_pre
c_pre, p_pre, n_pre = model(ui_data, 2, i)
ui_out[0] = c_pre
ui_out[1] = p_pre
ui_out[2] = n_pre
c_pre, p_pre, n_pre = model(iu_data, 3, i)
iu_out[0] = c_pre
iu_out[1] = p_pre
iu_out[2] = n_pre
c_pre, p_pre, n_pre = model(ii_data, 4, i)
ii_out[0] = c_pre
ii_out[1] = p_pre
ii_out[2] = n_pre
loss = cross_entropy_loss(uu_out, ui_out, iu_out, ii_out, embed_dim)
all_loss = all_loss + loss
optim.zero_grad()
loss.backward(retain_graph=True)
optim.step()
# print('after %dth iter, %d epochs, the loss is %.4f' %(i, epochs, loss))
if i % 10 == 0:
all_loss = all_loss/epochs
print('after %dth iter, the whole loss is %.4f'%(i, all_loss))
if i % 20 == 0:
checkpoint = {
"net": model.state_dict(),
'optimizer': optim.state_dict(),
"i": i
}
if not os.path.isdir("./models/checkpoint"):
os.makedirs("./models/checkpoint")
torch.save(checkpoint, './models/checkpoint/amazon_ckpt_best_newChangeModeladdII%s.pth' % (str(i)))
u_all_batch = []
for i in range(1, args.u_l+1):
u_all_batch.append(i)
all_u_node = model.types_agg(u_all_batch, 1)
i_all_batch = []
for j in range(1, args.i_l+1):
i_all_batch.append(j)
all_i_node = model.types_agg(i_all_batch, 2)
f = open('./edges/new_node_embedding5.txt', 'r+')
f.truncate()
f.close()
# 然后再在这里面写
with open('./edges/new_node_embedding5.txt', 'a') as file:
# for i in range(len(self.u_self_embed)):
# file.write('u' + str(i) + ' ' + str(self.u_self_embed[i])[1: -1] + '\n')
# for j in range(len(self.i_self_embed)):
# file.write('i' + str(j) + ' ' + str(self.i_self_embed[j])[1: -1] + '\n')
u_num = 1
i_num = 1
for u_node in all_u_node:
node_embedding = u_node.detach().numpy().tolist()
file.write('u' + str(u_num) + ' ' + str(node_embedding)[1: -1] + '\n')
u_num = u_num + 1
for i_node in all_i_node:
node_embedding = i_node.detach().numpy().tolist()
file.write('i' + str(i_num) + ' ' + str(node_embedding)[1: -1] + '\n')
i_num = i_num + 1
print('更新成功')