-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtraining_vqgan.py
156 lines (133 loc) · 8.28 KB
/
training_vqgan.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
144
145
146
147
148
149
150
151
152
153
154
155
156
import os
import argparse
import matplotlib.pyplot as plt
from tqdm import tqdm
import numpy as np
import torch
import torch.nn.functional as F
from torchvision import utils as vutils
from discriminator import Discriminator
from lpips import LPIPS
from vqgan import VQGAN
from utils import load_data, weights_init
import matplotlib.pyplot as plt
from torch import autocast
from torch.cuda.amp import GradScaler
class TrainVQGAN:
def __init__(self, args):
self.vqgan = VQGAN(args).to(device=args.device)
self.vqgan.load_state_dict(
torch.load(os.path.join("/media/userdisk1/code/VQGAN-pytorch/checkpoints", "vqgan_epoch_308.pt")))
self.discriminator = Discriminator(args).to(device=args.device)
self.discriminator.load_state_dict(
torch.load(os.path.join("/media/userdisk1/code/VQGAN-pytorch/checkpoints", "discriminator_epoch_308.pt")))
# self.discriminator.apply(weights_init)
self.perceptual_loss = LPIPS().eval().to(device=args.device)
self.opt_vq, self.opt_disc = self.configure_optimizers(args)
# self.usage_codebook = np.zeros(shape=(args.num_codebook_vectors))
self.prepare_training()
self.train(args)
def configure_optimizers(self, args):
lr = args.learning_rate
opt_vq = torch.optim.Adam(
list(self.vqgan.encoder.parameters()) +
list(self.vqgan.decoder.parameters()) +
list(self.vqgan.codebook.parameters()) +
list(self.vqgan.quant_conv.parameters()) +
list(self.vqgan.post_quant_conv.parameters()),
lr=lr, eps=1e-08, betas=(args.beta1, args.beta2)
)
opt_disc = torch.optim.Adam(self.discriminator.parameters(),
lr=lr, eps=1e-08, betas=(args.beta1, args.beta2))
return opt_vq, opt_disc
@staticmethod
def prepare_training():
os.makedirs("results", exist_ok=True)
os.makedirs("checkpoints", exist_ok=True)
def train(self, args):
train_dataset = load_data(args)
steps_per_epoch = len(train_dataset)
scaler = GradScaler()
for epoch in range(309, args.epochs):
usage_codebook = np.zeros(shape=(args.num_codebook_vectors))
with tqdm(range(len(train_dataset))) as pbar:
for i, imgs in zip(pbar, train_dataset):
self.discriminator.zero_grad()
self.opt_disc.zero_grad()
with autocast(device_type='cuda', dtype=torch.float16):
imgs = imgs.to(device=args.device)
with torch.no_grad():
decoded_images, _, q_loss = self.vqgan(imgs)
disc_real = self.discriminator(imgs)
disc_fake = self.discriminator(decoded_images.detach())
d_loss_real = torch.mean(F.relu(1. - disc_real))
d_loss_fake = torch.mean(F.relu(1. + disc_fake))
disc_factor = self.vqgan.adopt_weight(args.disc_factor, epoch * steps_per_epoch + i,
threshold=args.disc_start)
gan_loss = disc_factor * 0.5 * (d_loss_real + d_loss_fake)
scaler.scale(gan_loss).backward()
scaler.step(self.opt_disc)
scaler.update()
# gan_loss.backward()
# print(self.vqgan.decoder.conv_out.weight.grad.max(), self.discriminator.model[-1].weight.grad.max())
# self.opt_disc.step()
self.vqgan.zero_grad()
self.opt_vq.zero_grad()
with autocast(device_type='cuda', dtype=torch.float16):
# imgs = imgs.to(device=args.device)
decoded_images, min_encoding_indices, q_loss = self.vqgan(imgs)
perceptual_loss = self.perceptual_loss(imgs.contiguous(), decoded_images.contiguous())
rec_loss = torch.abs(imgs.contiguous() - decoded_images.contiguous())
disc_fake = self.discriminator(decoded_images)
perceptual_rec_loss = args.perceptual_loss_factor * perceptual_loss + args.rec_loss_factor * rec_loss
perceptual_rec_loss = perceptual_rec_loss.mean()
g_loss = -torch.mean(disc_fake)
lbd = self.vqgan.calculate_lambda(perceptual_rec_loss, g_loss)
vq_loss = perceptual_rec_loss + q_loss + disc_factor * lbd * g_loss
# min_encoding_indices_int = min_encoding_indices.cpu()
# for j in range(len(min_encoding_indices_int)):
# usage_codebook[min_encoding_indices_int[j]] += 1
scaler.scale(vq_loss).backward()
scaler.step(self.opt_vq)
scaler.update()
if i % 1000 == 0:
with torch.no_grad():
real_fake_images = torch.cat((imgs.add(1).mul(0.5)[:4],
torch.clamp(decoded_images.add(1).mul(0.5), min=0.0, max=1)[
:4]))
vutils.save_image(real_fake_images, os.path.join("results", f"{epoch}_{i}.jpg"), nrow=4)
pbar.set_postfix(
VQ_Loss=np.round(vq_loss.cpu().detach().numpy().item(), 5),
GAN_Loss=np.round(gan_loss.cpu().detach().numpy().item(), 3),
lbd=lbd.data.cpu().numpy()
)
pbar.update(0)
torch.save(self.vqgan.state_dict(), os.path.join("checkpoints", f"vqgan_epoch_{epoch}.pt"))
torch.save(self.discriminator.state_dict(),
os.path.join("checkpoints", f"discriminator_epoch_{epoch}.pt"))
# plt.imshow(usage_codebook.reshape(32, 32))
# plt.savefig(os.path.join("checkpoints", f"codebook_epoch_{epoch}.png"))
# np.save(os.path.join("checkpoints", f"codebook_epoch_{epoch}.npy"), usage_codebook)
if __name__ == '__main__':
parser = argparse.ArgumentParser(description="VQGAN")
parser.add_argument('--latent-dim', type=int, default=256, help='Latent dimension n_z (default: 256)')
parser.add_argument('--image-size', type=int, default=256, help='Image height and width (default: 256)')
parser.add_argument('--num-codebook-vectors', type=int, default=1024,
help='Number of codebook vectors (default: 256)')
parser.add_argument('--beta', type=float, default=0.25, help='Commitment loss scalar (default: 0.25)')
parser.add_argument('--image-channels', type=int, default=3, help='Number of channels of images (default: 3)')
parser.add_argument('--dataset-path', type=str, default='/data', help='Path to data (default: /data)')
parser.add_argument('--device', type=str, default="cuda", help='Which device the training is on')
parser.add_argument('--batch-size', type=int, default=6, help='Input batch size for training (default: 6)')
parser.add_argument('--epochs', type=int, default=1000, help='Number of epochs to train (default: 50)')
parser.add_argument('--learning-rate', type=float, default=2.25e-05, help='Learning rate (default: 0.0002)')
parser.add_argument('--beta1', type=float, default=0.5, help='Adam beta param (default: 0.0)')
parser.add_argument('--beta2', type=float, default=0.9, help='Adam beta param (default: 0.999)')
parser.add_argument('--disc-start', type=int, default=10000, help='When to start the discriminator (default: 0)')
parser.add_argument('--disc-factor', type=float, default=0.2, help='')
parser.add_argument('--rec-loss-factor', type=float, default=1., help='Weighting factor for reconstruction loss.')
parser.add_argument('--perceptual-loss-factor', type=float, default=1.0,
help='Weighting factor for perceptual loss.')
args = parser.parse_args()
args.dataset_path = r"/media/userdisk1/code/jpg/"
train_vqgan = TrainVQGAN(args)