-
Notifications
You must be signed in to change notification settings - Fork 5
/
train.py
143 lines (108 loc) · 4.37 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
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
import os
import math
import argparse
import torch
import torch.nn as nn
import torch.optim as optim
import torch.backends.cudnn as cudnn
from PIL import Image
from torch.autograd import Variable
from torchvision.utils import save_image
from torchvision import datasets
from torch.utils.data import DataLoader
from torchvision import utils
from data.dataloader_canny import GetData
from loss.InpaintingLoss import InpaintingLossWithGAN
from models.LBAMModel import LBAMModel, VGG16FeatureExtractor
from MECNet.models import EdgeModel
from MECNet.config import Config
torch.set_num_threads(6)
parser = argparse.ArgumentParser()
parser.add_argument('--numOfWorkers', type=int, default=4,
help='workers for dataloader')
parser.add_argument('--modelsSavePath', type=str, default='',
help='path for saving models')
parser.add_argument('--logPath', type=str,
default='')
parser.add_argument('--batchSize', type=int, default=16)
parser.add_argument('--loadSize', type=int, default=256,
help='image loading size')
parser.add_argument('--cropSize', type=int, default=256,
help='image training size')
parser.add_argument('--dataRoot', type=str,
default='')
parser.add_argument('--maskRoot', type=str,
default='')
parser.add_argument('--pretrained',type=str, default='', help='pretrained models for finetuning')
parser.add_argument('--train_epochs', type=int, default=500, help='training epochs')
args = parser.parse_args()
cuda = torch.cuda.is_available()
if cuda:
print('Cuda is available!')
cudnn.enable = True
cudnn.benchmark = True
batchSize = args.batchSize
loadSize = (args.loadSize, args.loadSize)
cropSize = (args.cropSize, args.cropSize)
if not os.path.exists(args.modelsSavePath):
os.makedirs(args.modelsSavePath)
config = Config("config.yml")
edge_model = EdgeModel(config).to(config.DEVICE)
edge_model.load()
edge_model.cuda()
edge_model = nn.DataParallel(edge_model, device_ids=[0,1,2,3])
dataRoot = args.dataRoot
maskRoot = args.maskRoot
imgData = GetData(dataRoot, maskRoot, loadSize, cropSize)
data_loader = DataLoader(imgData, batch_size=batchSize,
shuffle=True, num_workers=args.numOfWorkers, drop_last=False, pin_memory=True)
num_epochs = args.train_epochs
netG = LBAMModel(5, 3)
if args.pretrained != '':
netG.load_state_dict(torch.load(args.pretrained))
numOfGPUs = torch.cuda.device_count()
if cuda:
netG = netG.cuda()
if numOfGPUs > 1:
netG = nn.DataParallel(netG, device_ids=range(numOfGPUs))
count = 1
G_optimizer = optim.Adam(netG.parameters(), lr=0.000025, betas=(0.5, 0.9))
criterion = InpaintingLossWithGAN(args.logPath, VGG16FeatureExtractor(), lr=0.00001, betasInit=(0.0, 0.9), Lamda=10.0)
if cuda:
criterion = criterion.cuda()
if numOfGPUs > 1:
criterion = nn.DataParallel(criterion, device_ids=range(numOfGPUs))
print('OK!')
for i in range(1, num_epochs + 1):
netG.train()
for inputImgs, GT, masks,img_gray, edge, masks_over in (data_loader):
if cuda:
inputImgs = inputImgs.cuda()
GT = GT.cuda()
masks = masks.cuda()
edge = edge.cuda()
masks_over = masks_over.cuda()
netG.zero_grad()
outputs = edge_model(img_gray, edge, masks_over)
outputs_merged = (outputs * masks_over) + (edge * (1 - masks_over))
inputImgs = torch.cat((inputImgs, outputs_merged), 1)
# print(inputImgs2.shape)
fake_images = netG(inputImgs, masks,outputs_merged)
G_loss = criterion(inputImgs[:, 0:3, :, :], masks, fake_images, GT, count, i)
G_loss = G_loss.sum()
G_optimizer.zero_grad()
G_loss.backward()
G_optimizer.step()
with open('/home/wangdongsheng/LBAM_version6/loss2.txt', 'a') as file:
file.write('Generator Loss of epoch{} is {}\n'.format(i, G_loss.item()))
count += 1
""" if (count % 4000 == 0):
torch.save(netG.module.state_dict(), args.modelsSavePath +
'/Places_{}.pth'.format(i)) """
if ( i % 10 == 0):
if numOfGPUs > 1 :
torch.save(netG.module.state_dict(), args.modelsSavePath +
'/LBAM_{}.pth'.format(i%50))
else:
torch.save(netG.state_dict(), args.modelsSavePath +
'/LBAM_{}.pth'.format(i%50))