-
Notifications
You must be signed in to change notification settings - Fork 0
/
run.py
118 lines (89 loc) · 4.11 KB
/
run.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
import time
import os
import torch
from torch import optim
from src.app.g1 import Generator as Generator_g1
from src.app.g2 import Generator as Generator_g2
from src.app.discriminator import Discriminator
from src.app.training import train_model
from src.settings import settings
from src.utils import utils
def load_checkpoint(path, generator1, generator2, discriminator, optim_g1, optim_g2, optim_d):
if not os.path.exists(path):
print("No checkpoint found.")
return 1, [], [], []
checkpoint = torch.load(path)
generator1.load_state_dict(checkpoint['generator1_state_dict'])
generator2.load_state_dict(checkpoint['generator2_state_dict'])
discriminator.load_state_dict(checkpoint['discriminator_state_dict'])
optim_g1.load_state_dict(checkpoint['optimizer_g1_state_dict'])
optim_g2.load_state_dict(checkpoint['optimizer_g2_state_dict'])
optim_d.load_state_dict(checkpoint['optimizer_d_state_dict'])
epoch = checkpoint['epoch'] + 1
losses_g1 = checkpoint['losses_g1']
losses_g2 = checkpoint['losses_g2']
losses_d = checkpoint['losses_d']
print(f'Checkpoint loaded, starting from epoch {epoch}')
return epoch, losses_g1, losses_g2, losses_d
def setup_training(params):
utils.check_if_gpu_available()
utils.check_if_set_seed(params["seed"])
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
generator1 = Generator_g1(
params["z_dim"], params["channels_img"], params["features_g"], img_size=params['image_size']
).to(device)
generator1.apply(utils.weights_init)
generator2 = Generator_g2(
params["z_dim"], params["channels_img"], params["features_g"], img_size=params['image_size']
).to(device)
generator2.apply(utils.weights_init)
discriminator = Discriminator(
params["channels_img"], params["features_d"], params["alpha"], img_size=params['image_size']
).to(device)
discriminator.apply(utils.weights_init)
optim_g1 = optim.Adam(
generator1.parameters(), lr=params["lr_g"],
betas=(params['g_beta_min'], params['g_beta_max'])
)
optim_g2 = optim.Adam(
generator2.parameters(), lr=params["lr_g"],
betas=(params['g_beta_min'], params['g_beta_max'])
)
optim_d = optim.Adam(
discriminator.parameters(), lr=params["lr_d"],
betas=(params['d_beta_min'], params['d_beta_max'])
)
return generator1, generator2, discriminator, optim_g1, optim_g2, optim_d, device
def setup_directories_and_checkpoint(path_data, params, generator1, generator2, discriminator, optim_g1, optim_g2, optim_d):
training_version = utils.create_next_version_directory(
path_data, params['continue_last_training']
)
data_dir = os.path.join(path_data, training_version)
print('Training version:', training_version)
last_epoch, losses_g1, losses_g2, losses_d = load_checkpoint(
os.path.join(data_dir, 'weights', 'checkpoint.pth'),
generator1, generator2, discriminator, optim_g1, optim_g2, optim_d
)
return data_dir, last_epoch, losses_g1, losses_g2, losses_d
def main(params, path_data, path_dataset):
time_start = time.time()
utils.print_datetime()
generator1, generator2, discriminator, optim_g1, optim_g2, optim_d, device = setup_training(
params
)
data_loader = utils.dataloader(path_dataset, params["image_size"], params["batch_size"])
data_dir, last_epoch, losses_g1, losses_g2, losses_d = setup_directories_and_checkpoint(
path_data, params, generator1, generator2, discriminator, optim_g1, optim_g2, optim_d
)
training_config = utils.create_train_config(
params, generator1, generator2, discriminator, optim_g1, optim_g2, optim_d,
data_loader, last_epoch, losses_g1, losses_g2, losses_d, data_dir, device
)
train_model(**training_config)
time_end = time.time()
time_total = (time_end - time_start) / 60
print(f"The code took {round(time_total, 1)} minutes to execute.")
utils.print_datetime()
if __name__ == '__main__':
PARAMS = utils.get_params(settings.PATH_PARAMS)
main(PARAMS, settings.PATH_DATA, settings.PATH_DATASET)