-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathmain.py
More file actions
172 lines (146 loc) · 6.8 KB
/
main.py
File metadata and controls
172 lines (146 loc) · 6.8 KB
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
import datetime
import torch
from torch.utils.data import dataloader
import pickle
import numpy as np
from params import args
from preprocess import csr_to_tensor
from dataset import RecDataset_beh, RecDataset
from train_and_test import Trainer
from torch.utils.tensorboard import SummaryWriter
class Model:
def __init__(self):
self.train_data_path = args.dataset_path + 'trn_'
self.test_data_path = args.dataset_path + 'tst_int'
self.target_behavior = args.target_behavior
self.behaviors_data_csr = {}
self.behavior_mats = {}
self.behavior_mats_t = {}
self.behaviors = ['click', 'fav', 'cart', 'buy']
self.user_num = -1
self.item_num = -1
now_time = datetime.datetime.now()
self.time = datetime.datetime.strftime(now_time, '%Y_%m_%d__%H_%M_%S')
self.epoch = 0
self.train_loss = []
self.hr_history = []
self.ndcg_history = []
self.best_HR = 0
self.best_NDCG = 0
self.best_epoch = 0
self.cnt = 0
# Load data
for i in range(len(self.behaviors)):
behavior = self.behaviors[i]
with open(self.train_data_path + behavior, 'rb') as fs:
data = pickle.load(fs)
self.behaviors_data_csr[i] = data
if data.get_shape()[0] > self.user_num:
self.user_num = data.get_shape()[0] # 17435
if data.get_shape()[1] > self.item_num:
self.item_num = data.get_shape()[1] # 35920
if behavior == args.target_behavior:
self.train_mat_csr = data
self.train_label_csr = 1 * (self.train_mat_csr != 0) # Change timestamp to 1
self.labelP = np.squeeze(np.array(
np.sum(self.train_label_csr, axis=0))) # [17435], total interacted user_num for each item
# Change data to tensor
for i in range(len(self.behaviors)):
csr_data = (self.behaviors_data_csr[i] != 0) * 1
self.behavior_mats[i] = csr_to_tensor(csr_data)
self.behavior_mats_t[i] = csr_to_tensor(csr_data.T)
# Build train and test dataloader
target_users, target_items = self.train_label_csr.nonzero()
target_user_item = np.hstack((target_users.reshape(-1, 1), target_items.reshape(-1, 1)))
self.train_dataset = RecDataset_beh(self.behaviors, self.behaviors_data_csr, target_user_item, self.item_num,
self.target_behavior, mode='train')
self.train_loader = dataloader.DataLoader(self.train_dataset, batch_size=args.batch_size, shuffle=True,
num_workers=0, pin_memory=True)
with open(self.test_data_path, 'rb') as fs:
data = pickle.load(fs)
test_user = np.array([idx for idx, i in enumerate(data) if i is not None])
test_item = np.array([i for idx, i in enumerate(data) if i is not None])
test_target_user_item = np.hstack((test_user.reshape(-1, 1), test_item.reshape(-1, 1)))
test_dataset = RecDataset(test_target_user_item)
self.test_loader = dataloader.DataLoader(test_dataset, batch_size=args.batch_size, shuffle=False, num_workers=0,
pin_memory=True)
def save_history(self):
history = dict()
history['loss'] = self.train_loss
history['HR'] = self.hr_history
history['NDCG'] = self.ndcg_history
model_name = self.time
with open(r'./History/' + model_name + '.his', 'wb') as fs:
pickle.dump(history, fs)
def save_model(self):
history = dict()
history['loss'] = self.train_loss
history['HR'] = self.hr_history
history['NDCG'] = self.ndcg_history
model_name = self.time
savePath = r'./Model/' + model_name + r'.pth'
params = {
'epoch': self.epoch,
'model': self.best_model,
'history': history,
'user_embed': self.user_embed,
'user_embeds': self.user_embeds,
'item_embed': self.item_embed,
'item_embeds': self.item_embeds,
}
torch.save(params, savePath)
def run(self):
# torch.autograd.set_detect_anomaly(True)
trainer = Trainer(self.train_loader, self.behaviors, self.user_num, self.item_num, self.behavior_mats,
self.behavior_mats_t, self.behaviors_data_csr, self.train_mat_csr)
# trainer.test_epoch(self.test_loader)
# trainer.train_epoch()
for i in range(args.epoch_num+1):
self.epoch = i+1
gnn, epoch_loss, user_embed, item_embed, user_embeds, item_embeds = trainer.train_epoch()
self.train_loss.append(epoch_loss)
print(f"epoch {self.epoch}, epoch loss {epoch_loss}")
HR, NDCG = trainer.test_epoch(self.test_loader)
self.hr_history.append(HR)
self.ndcg_history.append(NDCG)
self.save_metric()
if HR > self.best_HR:
self.cnt = 0
self.best_HR = HR
self.best_epoch = self.epoch
self.user_embed = user_embed
self.item_embed = item_embed
self.user_embeds = user_embeds
self.item_embeds = item_embeds
self.best_model = gnn
self.save_history()
self.save_model()
if NDCG > self.best_NDCG:
self.cnt = 0
self.best_NDCG = NDCG
self.best_epoch = self.epoch
self.user_embed = user_embed
self.item_embed = item_embed
self.user_embeds = user_embeds
self.item_embeds = item_embeds
self.best_model = gnn
self.save_history()
self.save_model()
if HR < self.best_HR and NDCG < self.best_NDCG:
self.cnt += 1
if self.cnt == args.patience:
print(f"Early stop at {self.best_epoch} : best HR: {self.best_HR}, best_NDCG: {self.best_NDCG} \n")
self.save_history()
self.save_model()
break
def save_metric(self):
path = './runs/' + self.time + '/' + str(self.epoch) + '/'
writer = SummaryWriter(path)
for i in range(self.epoch):
writer.add_scalar('Loss', self.train_loss[i], i)
writer.add_scalar('NDCG', self.ndcg_history[i], i)
writer.add_scalar('HR', self.hr_history[i], i)
if __name__ == '__main__':
model = Model()
model.run()
# model.save_metric()