GAN：数据生成的把戏师 - ToB企服应用市场:ToB评测及商务社交产业平台

import torch
import torch.nn as nn
import torch.optim as optim
from torchvision.utils import save_image
# 定义生成器
class Generator(nn.Module):
def __init__(self, ngpu):
super(Generator, self).__init__()
self.ngpu = ngpu
self.main = nn.Sequential(
# 输入是Z，大小为 (nz, 1, 1)
nn.ConvTranspose2d(nz, ngf * 8, 4, 1, 0, bias=False),
nn.BatchNorm2d(ngf * 8),
nn.ReLU(True),
# 状态大小: (ngf*8) x 4 x 4
nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf * 4),
nn.ReLU(True),
# 状态大小: (ngf*4) x 8 x 8
nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf * 2),
nn.ReLU(True),
# 状态大小: (ngf*2) x 16 x 16
nn.ConvTranspose2d(ngf * 2, ngf, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf),
nn.ReLU(True),
# 状态大小: (ngf) x 32 x 32
nn.ConvTranspose2d(ngf, nc, 4, 2, 1, bias=False),
nn.Tanh()
# 输出大小: (nc) x 64 x 64
)
def forward(self, input):
return self.main(input)
# 定义判别器
class Discriminator(nn.Module):
def __init__(self, ngpu):
super(Discriminator, self).__init__()
self.ngpu = ngpu
self.main = nn.Sequential(
# 输入大小: 3 x 64 x 64
nn.Conv2d(nc, ndf, 4, 2, 1, bias=False),
nn.LeakyReLU(0.2, inplace=True),
# 状态大小: (ndf) x 32 x 32
nn.Conv2d(ndf, ndf * 2, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 2),
nn.LeakyReLU(0.2, inplace=True),
# 状态大小: (ndf*2) x 16 x 16
nn.Conv2d(ndf * 2, ndf * 4, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 4),
nn.LeakyReLU(0.2, inplace=True),
# 状态大小: (ndf*4) x 8 x 8
nn.Conv2d(ndf * 4, ndf * 8, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 8),
nn.LeakyReLU(0.2, inplace=True),
# 状态大小: (ndf*8) x 4 x 4
nn.Conv2d(ndf * 8, 1, 4, 1, 0, bias=False),
nn.Sigmoid()
)
def forward(self, input):
return self.main(input).view(-1)
# 初始化网络
netG = Generator(ngpu).to(device)
netD = Discriminator(ngpu).to(device)
# 应用权重初始化
netG.apply(weights_init)
netD.apply(weights_init)
# 设置损失函数和优化器
criterion = nn.BCELoss()
optimizerD = optim.Adam(netD.parameters(), lr=0.0002, betas=(0.5, 0.999))
optimizerG = optim.Adam(netG.parameters(), lr=0.0002, betas=(0.5, 0.999))
# 训练GAN
for epoch in range(num_epochs):
for i, data in enumerate(dataloader, 0):
# 创建标签
real = torch.ones(batch_size, 1, device=device)
fake = torch.zeros(batch_size, 1, device=device)
# 获取真实图像
real_imgs = data[0].to(device)
# 训练判别器
netD.zero_grad()
output = netD(real_imgs).view(-1)
errD_real = criterion(output, real)
errD_real.backward()
D_x = output.mean().item()
# 生成假图像并训练判别器
noise = torch.randn(batch_size, nz, 1, 1, device=device)
fake_imgs = netG(noise)
output = netD(fake_imgs.detach()).view(-1)
errD_fake = criterion(output, fake)
errD_fake.backward()
D_G_z1 = output.mean().item()
optimizerD.step()
# 训练生成器
netG.zero_grad()
output = netD(fake_imgs).view(-1)
errG = criterion(output, real)
errG.backward()
D_G_z2 = output.mean().item()
optimizerG.step()
# 打印训练进度
if i % 50 == 0:
print('[%d/%d][%d/%d] Loss_D: %.4f Loss_G: %.4f D(x): %.4f D(G(z)): %.4f / %.4f'
% (epoch, num_epochs, i, len(dataloader), errD_real.item() + errD_fake.item(), errG.item(), D_x, D_G_z1, D_G_z2))
# 保存生成的图像
if epoch % 100 == 0:
with torch.no_grad():
fake_imgs = netG(fixed_noise).detach().cpu()
img_list.append(make_grid(fake_imgs, padding=2, normalize=True))
save_image(fake_imgs, f'gan/fake_samples_epoch_{epoch}.png', normalize=True)
# 保存训练好的模型
torch.save(netG.state_dict(), 'gan/netG.pth')
torch.save(netD.state_dict(), 'gan/netD.pth')

复制代码