/
train.py
executable file
·124 lines (102 loc) · 4.58 KB
/
train.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
import os
import torch
import argparse
import timm
import numpy as np
import utils
import time
import torch.nn.functional as F
import warnings
warnings.filterwarnings(action='ignore')
def train():
parser = argparse.ArgumentParser()
parser.add_argument('--net','-n', default = 'vit_tiny_patch16_224', type=str)
parser.add_argument('--data', '-d', type=str)
parser.add_argument('--gpu', '-g', default = '0', type=str)
parser.add_argument('--save_path', '-s', type=str)
parser.add_argument('--jigsawnum', '-j', default = 3, type=int)
args = parser.parse_args()
config = utils.read_conf('conf/'+args.data+'.json')
device = 'cuda:'+args.gpu
model_name = args.net
dataset_path = config['id_dataset']
save_path = config['save_path'] + args.save_path
num_classes = int(config['num_classes'])
class_range = list(range(0, num_classes))
jigsaw_num = args.jigsawnum
batch_size = int(int(config['batch_size'])) #/2)
max_epoch = int(config['epoch'])
lrde = [10]
print(model_name, dataset_path.split('/')[-2], batch_size, class_range, jigsaw_num)
if not os.path.exists(config['save_path']):
os.mkdir(config['save_path'])
if not os.path.exists(save_path):
os.mkdir(save_path)
else:
raise ValueError('save_path already exists')
if 'cifar' in args.data:
train_loader, valid_loader = utils.get_cifar_jigsaw(args.data, dataset_path, batch_size, jigsaw_num = jigsaw_num)
elif 'tinyimagenet' in args.data:
train_loader, valid_loader = utils.get_tinyimagenet_jigsaw(dataset_path, batch_size, size=224)
elif 'mnist' in args.data:
train_loader, valid_loader = utils.get_mnist_jigsaw(dataset_path, batch_size, jigsaw_num = jigsaw_num)
elif 'imagenet' in args.data:
train_loader, valid_loader = utils.get_imagenet_jigsaw(dataset_path, batch_size)
print(args.net)
model = timm.create_model(args.net, pretrained=True, num_classes=num_classes)
model.to(device)
criterion = torch.nn.CrossEntropyLoss()
model.eval()
print(utils.validation_accuracy(model, valid_loader, device))
optimizer = torch.optim.SGD(model.parameters(), lr = 0.003, momentum=0.9, weight_decay = 5e-4)
if 'mnist' in args.data:
optimizer = torch.optim.SGD(model.parameters(), lr = 0.0003, momentum=0.9, weight_decay = 1e-4)
if 'imagenet' in args.data:
optimizer = torch.optim.SGD(model.parameters(), lr = 0.003, momentum=0.9, weight_decay = 0)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, lrde)
saver = timm.utils.CheckpointSaver(model, optimizer, checkpoint_dir= save_path, max_history = 2)
print(train_loader.dataset[0][0].shape)
start_time = time.time()
for epoch in range(15):
## training
model.train()
total_loss = 0
total = 0
correct = 0
for batch_idx, (inputs, inputs_jigsaw, targets) in enumerate(train_loader):
inputs, inputs_jigsaw, targets = inputs.to(device), inputs_jigsaw.to(device), targets.to(device)
optimizer.zero_grad()
input_ = torch.cat((inputs, inputs_jigsaw), 0)
# compute output
output_ = model(input_)
outputs = output_[:len(targets)]
outputs_corr = output_[len(targets):]
loss_in = criterion(outputs, targets)
loss_out = torch.norm(F.relu(outputs_corr), dim=1).mean()
loss = loss_in + loss_out
loss.backward()
optimizer.step()
total_loss += loss
total += targets.size(0)
_, predicted = outputs[:len(targets)].max(1)
correct += predicted.eq(targets).sum().item()
print('\r', batch_idx, len(train_loader), 'Loss: %.3f | Acc: %.3f%% (%d/%d)'
% (total_loss/(batch_idx+1), 100.*correct/total, correct, total), end = '')
train_accuracy = correct/total
train_avg_loss = total_loss/len(train_loader)
print()
## validation
model.eval()
total_loss = 0
total = 0
correct = 0
valid_accuracy = utils.validation_accuracy(model, valid_loader, device)
scheduler.step()
saver.save_checkpoint(epoch, metric = valid_accuracy)
print('EPOCH {:4}, TRAIN [loss - {:.4f}, acc - {:.4f}], VALID [acc - {:.4f}]\n'.format(epoch, train_avg_loss, train_accuracy, valid_accuracy))
print(scheduler.get_last_lr())
end_time = time.time()
elapsed_time = end_time - start_time
print(elapsed_time)
if __name__ =='__main__':
train()