-
Notifications
You must be signed in to change notification settings - Fork 22
/
trainer.py
133 lines (102 loc) · 5.67 KB
/
trainer.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
import os
import time
from tqdm import tqdm, trange
import numpy as np
import torch
from utils.loader import load_seed, load_device, load_data, load_model_params, load_model_optimizer, \
load_ema, load_loss_fn, load_batch
from utils.logger import Logger, set_log, start_log, train_log
class Trainer(object):
def __init__(self, config):
super(Trainer, self).__init__()
self.config = config
self.log_folder_name, self.log_dir, self.ckpt_dir = set_log(self.config)
self.seed = load_seed(self.config.seed)
self.device = load_device()
self.train_loader, self.test_loader = load_data(self.config)
self.params_x, self.params_adj = load_model_params(self.config)
def train(self, ts):
self.config.exp_name = ts
self.ckpt = f'{ts}'
print('\033[91m' + f'{self.ckpt}' + '\033[0m')
# -------- Load models, optimizers, ema --------
self.model_x, self.optimizer_x, self.scheduler_x = load_model_optimizer(self.params_x, self.config.train,
self.device)
self.model_adj, self.optimizer_adj, self.scheduler_adj = load_model_optimizer(self.params_adj, self.config.train,
self.device)
self.ema_x = load_ema(self.model_x, decay=self.config.train.ema)
self.ema_adj = load_ema(self.model_adj, decay=self.config.train.ema)
logger = Logger(str(os.path.join(self.log_dir, f'{self.ckpt}.log')), mode='a')
logger.log(f'{self.ckpt}', verbose=False)
start_log(logger, self.config)
train_log(logger, self.config)
self.loss_fn = load_loss_fn(self.config)
# -------- Training --------
for epoch in trange(0, (self.config.train.num_epochs), desc = '[Epoch]', position = 1, leave=False):
self.train_x = []
self.train_adj = []
self.test_x = []
self.test_adj = []
t_start = time.time()
self.model_x.train()
self.model_adj.train()
for _, train_b in enumerate(self.train_loader):
self.optimizer_x.zero_grad()
self.optimizer_adj.zero_grad()
x, adj = load_batch(train_b, self.device)
loss_subject = (x, adj)
loss_x, loss_adj = self.loss_fn(self.model_x, self.model_adj, *loss_subject)
loss_x.backward()
loss_adj.backward()
torch.nn.utils.clip_grad_norm_(self.model_x.parameters(), self.config.train.grad_norm)
torch.nn.utils.clip_grad_norm_(self.model_adj.parameters(), self.config.train.grad_norm)
self.optimizer_x.step()
self.optimizer_adj.step()
# -------- EMA update --------
self.ema_x.update(self.model_x.parameters())
self.ema_adj.update(self.model_adj.parameters())
self.train_x.append(loss_x.item())
self.train_adj.append(loss_adj.item())
if self.config.train.lr_schedule:
self.scheduler_x.step()
self.scheduler_adj.step()
self.model_x.eval()
self.model_adj.eval()
for _, test_b in enumerate(self.test_loader):
x, adj = load_batch(test_b, self.device)
loss_subject = (x, adj)
with torch.no_grad():
self.ema_x.store(self.model_x.parameters())
self.ema_x.copy_to(self.model_x.parameters())
self.ema_adj.store(self.model_adj.parameters())
self.ema_adj.copy_to(self.model_adj.parameters())
loss_x, loss_adj = self.loss_fn(self.model_x, self.model_adj, *loss_subject)
self.test_x.append(loss_x.item())
self.test_adj.append(loss_adj.item())
self.ema_x.restore(self.model_x.parameters())
self.ema_adj.restore(self.model_adj.parameters())
mean_train_x = np.mean(self.train_x)
mean_train_adj = np.mean(self.train_adj)
mean_test_x = np.mean(self.test_x)
mean_test_adj = np.mean(self.test_adj)
# -------- Log losses --------
logger.log(f'{epoch+1:03d} | {time.time()-t_start:.2f}s | '
f'test x: {mean_test_x:.3e} | test adj: {mean_test_adj:.3e} | '
f'train x: {mean_train_x:.3e} | train adj: {mean_train_adj:.3e} | ', verbose=False)
# -------- Save checkpoints --------
if epoch % self.config.train.save_interval == self.config.train.save_interval-1:
save_name = f'_{epoch+1}' if epoch < self.config.train.num_epochs - 1 else ''
torch.save({
'model_config': self.config,
'params_x' : self.params_x,
'params_adj' : self.params_adj,
'x_state_dict': self.model_x.state_dict(),
'adj_state_dict': self.model_adj.state_dict(),
'ema_x': self.ema_x.state_dict(),
'ema_adj': self.ema_adj.state_dict()
}, f'./checkpoints/{self.config.data.data}/{self.ckpt + save_name}.pth')
if epoch % self.config.train.print_interval == self.config.train.print_interval-1:
tqdm.write(f'[EPOCH {epoch+1:04d}] test adj: {mean_test_adj:.3e} | train adj: {mean_train_adj:.3e} | '
f'test x: {mean_test_x:.3e} | train x: {mean_train_x:.3e}')
print(' ')
return self.ckpt