main_vdsr.py
5.88 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
import argparse, os
import torch
import random
import torch.backends.cudnn as cudnn
import torch.nn as nn
import torch.optim as optim
from torch.autograd import Variable
from torch.utils.data import DataLoader
from vdsr import Net
from dataset import DatasetFromHdf5
## Custom
from data import FeatureDataset
# Training settings
parser = argparse.ArgumentParser(description="PyTorch VDSR")
parser.add_argument("--batchSize", type=int, default=128, help="Training batch size")
parser.add_argument("--nEpochs", type=int, default=50, help="Number of epochs to train for")
parser.add_argument("--lr", type=float, default=0.1, help="Learning Rate. Default=0.1")
parser.add_argument("--step", type=int, default=10, help="Sets the learning rate to the initial LR decayed by momentum every n epochs, Default: n=10")
parser.add_argument("--cuda", action="store_true", help="Use cuda?")
parser.add_argument("--resume", default="", type=str, help="Path to checkpoint (default: none)")
parser.add_argument("--start-epoch", default=1, type=int, help="Manual epoch number (useful on restarts)")
parser.add_argument("--clip", type=float, default=0.4, help="Clipping Gradients. Default=0.4")
# 1->3 custom
parser.add_argument("--threads", type=int, default=3, help="Number of threads for data loader to use, Default: 1")
parser.add_argument("--momentum", default=0.9, type=float, help="Momentum, Default: 0.9")
parser.add_argument("--weight-decay", "--wd", default=1e-4, type=float, help="Weight decay, Default: 1e-4")
parser.add_argument('--pretrained', default='', type=str, help='path to pretrained model (default: none)')
parser.add_argument("--gpus", default="0", type=str, help="gpu ids (default: 0)")
## custom
parser.add_argument("--dataPath", type=str)
parser.add_argument("--featureType", type=str, default="p2")
parser.add_argument("--scaleFactor",type=int, default=4)
# parser.add_argument("--trainingData", type=DataLoader)
def main():
global opt, model
opt = parser.parse_args()
print(opt)
cuda = opt.cuda
if cuda:
print("=> use gpu id: '{}'".format(opt.gpus))
os.environ["CUDA_VISIBLE_DEVICES"] = opt.gpus
if not torch.cuda.is_available():
raise Exception("No GPU found or Wrong gpu id, please run without --cuda")
opt.seed = random.randint(1, 10000)
print("Random Seed: ", opt.seed)
torch.manual_seed(opt.seed)
if cuda:
torch.cuda.manual_seed(opt.seed)
cudnn.benchmark = True
print("===> Loading datasets")
if os.path.isfile('dataloader/training_data_loader.pth'):
training_data_loader = torch.load('dataloader/training_data_loader.pth')
else:
train_set = FeatureDataset(opt.dataPath,opt.featureType,opt.scaleFactor,False)
train_size = 100 #우선은 100개만
test_size = len(train_set) - train_size
train_dataset, test_dataset = torch.utils.data.random_split(train_set, [train_size, test_size])
training_data_loader = DataLoader(dataset=train_dataset, num_workers=3, batch_size=8, shuffle=True, pin_memory=False)
torch.save(training_data_loader, 'dataloader/training_data_loader.pth'.format(DataLoader))
print("===> Building model")
model = Net(opt.scaleFactor)
criterion = nn.MSELoss(size_average=False)
print("===> Setting GPU")
if cuda:
model = model.cuda()
criterion = criterion.cuda()
# optionally resume from a checkpoint
if opt.resume:
if os.path.isfile(opt.resume):
print("=> loading checkpoint '{}'".format(opt.resume))
checkpoint = torch.load(opt.resume)
opt.start_epoch = checkpoint["epoch"] + 1
model.load_state_dict(checkpoint["model"].state_dict())
else:
print("=> no checkpoint found at '{}'".format(opt.resume))
# optionally copy weights from a checkpoint
if opt.pretrained:
if os.path.isfile(opt.pretrained):
print("=> loading model '{}'".format(opt.pretrained))
weights = torch.load(opt.pretrained)
model.load_state_dict(weights['model'].state_dict())
else:
print("=> no model found at '{}'".format(opt.pretrained))
print("===> Setting Optimizer")
optimizer = optim.SGD(model.parameters(), lr=opt.lr, momentum=opt.momentum, weight_decay=opt.weight_decay)
print("===> Training")
for epoch in range(opt.start_epoch, opt.nEpochs + 1):
train(training_data_loader, optimizer, model, criterion, epoch)
save_checkpoint(model, epoch)
def adjust_learning_rate(optimizer, epoch):
"""Sets the learning rate to the initial LR decayed by 10 every 10 epochs"""
lr = opt.lr * (0.1 ** (epoch // opt.step))
return lr
def train(training_data_loader, optimizer, model, criterion, epoch):
lr = adjust_learning_rate(optimizer, epoch-1)
for param_group in optimizer.param_groups:
param_group["lr"] = lr
print("Epoch = {}, lr = {}".format(epoch, optimizer.param_groups[0]["lr"]))
model.train()
for iteration, batch in enumerate(training_data_loader, 1):
input, target = Variable(batch[0]), Variable(batch[1], requires_grad=False)
if opt.cuda:
input = input.cuda()
target = target.cuda()
loss = criterion(model(input), target)
optimizer.zero_grad()
loss.backward()
nn.utils.clip_grad_norm(model.parameters(),opt.clip)
optimizer.step()
if iteration%10 == 0:
# loss.data[0] --> loss.data
print("===> Epoch[{}]({}/{}): Loss: {:.10f}".format(epoch, iteration, len(training_data_loader), loss.data))
def save_checkpoint(model, epoch):
model_out_path = "checkpoint/" + "model_epoch_{}.pth".format(epoch)
state = {"epoch": epoch ,"model": model}
if not os.path.exists("checkpoint/"):
os.makedirs("checkpoint/")
torch.save(state, model_out_path)
print("Checkpoint saved to {}".format(model_out_path))
if __name__ == "__main__":
main()