Face-Generator/train.py at main · MasonAndrewHarrison/Face-Generator · GitHub

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
import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import transforms
import matplotlib.pyplot as plt
from torchvision.datasets import ImageFolder
from torch.utils.data import DataLoader
from model import Critic, Generator, initialize_weights
from utils import gradient_penalty
import os


device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

z_dim = 100
image_dim = 64
batch_size = 64
num_epochs = 50
disc_features = 128
gen_features = 128
critic_iterations = 5
Lambda_GP = 10

critic = Critic(disc_features).to(device)
generator = Generator(z_dim, gen_features).to(device)

if os.path.exists("Generator.pth"):
    generator.load_state_dict(torch.load("Generator.pth", map_location=device))
    generator.to(device)

else:
    initialize_weights(generator)

if os.path.exists("Critic.pth"):
    critic.load_state_dict(torch.load("Critic.pth", map_location=device))
    critic.to(device)

else:
    initialize_weights(critic)

transform = transforms.Compose([
    transforms.Resize(64),
    transforms.CenterCrop(64),
    transforms.ToTensor(),
    transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])

dataset = ImageFolder(root='dataset/', transform=transform)
loader = DataLoader(dataset, batch_size=batch_size, shuffle=True)

opt_critic = optim.Adam(critic.parameters(), lr=1e-4, betas=(0.5, 0.999))
opt_gen = optim.Adam(generator.parameters(), lr=1e-4, betas=(0.5, 0.999))

fixed_noise = torch.randn(batch_size, z_dim, 1, 1).to(device)

critic.train()
generator.train()
sample_iter = 0

os.makedirs('samples', exist_ok=True)

for epoch in range(num_epochs):
    for i, (real, _) in enumerate(loader):

        real_image = real.to(device)
        current_batch_size,_,_,_ = real_image.shape

        for _ in range(critic_iterations):
            z_noise = torch.randn(current_batch_size, z_dim, 1, 1).to(device)
            fake_image = generator(z_noise)

            critic_real = critic(real_image).reshape(-1)
            critic_fake = critic(fake_image.detach()).reshape(-1)

            gp = gradient_penalty(critic, real_image, fake_image, device=device)
            loss_critic = (
                -(torch.mean(critic_real) - torch.mean(critic_fake))
                + (Lambda_GP * gp)
            )
            critic.zero_grad()
            loss_critic.backward()
            opt_critic.step()

        ### Train Generator: min -E[critic(gen_fake)] ###
        z_noise = torch.randn(current_batch_size, z_dim, 1, 1).to(device)
        fake_image = generator(z_noise)
        output = critic(fake_image).reshape(-1)
        loss_gen = -torch.mean(output)
        generator.zero_grad()
        loss_gen.backward()
        opt_gen.step()

        if i % 100 == 0:

            print("saved model for epoch :", epoch+1)
            torch.save(generator.state_dict(), "Generator.pth")
            torch.save(critic.state_dict(), "Critic.pth")

        if i % 1 == 0:
            print(f" Generator Loss: {loss_gen.item()}, Critic Loss: {loss_critic.item()}")

        if i % 200 == 0:
            generator.eval()
            with torch.no_grad():
                sample_iter += 1
                print(f"saved sample {sample_iter}")
                fake = generator(fixed_noise).detach().cpu()
                fake = fake[0, :, :, :].permute(1, 2, 0) * 0.5 + 0.5
                plt.imsave(f"samples/fake_images_epoch_{sample_iter}.png", fake.numpy())
                #plt.imshow(fake)
                #plt.show()
            generator.train()