plt.figure(figsize=(10, 10))
plt.axis("off")
plt.imshow(
np.transpose(
vutils.make_grid(gen_data, nrow=10, padding=2, normalize=True),
(1, 2, 0)))
plt.savefig("Epoch_%d {}".format(params['dataset']) % (epoch + 1))
plt.close('all')
# Save network weights.
if (epoch + 1) % params['save_epoch'] == 0:
torch.save(
{
'netG': netG.state_dict(),
'discriminator': discriminator.state_dict(),
'netD': netD.state_dict(),
'netQ': netQ.state_dict(),
'optimD': optimD.state_dict(),
'optimG': optimG.state_dict(),
'params': params
}, 'checkpoint/model_epoch_%d_{}'.format(params['dataset']) %
(epoch + 1))
training_time = time.time() - start_time
print("-" * 50)
print('Training finished!\nTotal Time for Training: %.2fm' %
(training_time / 60))
print("-" * 50)
# Generate image to check performance of trained generator.
with torch.no_grad():
# Generate image to check performance of generator.
if (epoch + 1) == 1 or (epoch + 1) == params['num_epochs']/2:
with torch.no_grad():
gen_data = netG(fixed_noise).detach().cpu()
plt.figure(figsize=(10, 10))
plt.axis("off")
plt.imshow(np.transpose(vutils.make_grid(gen_data, nrow=10, padding=2, normalize=True), (1,2,0)))
plt.savefig("Epoch_%d {}".format(params['dataset']) %(epoch+1))
plt.close('all')
# Save network weights.
if (epoch+1) % params['save_epoch'] == 0:
torch.save({
'netG' : netG.state_dict(),
'discriminator' : discriminator.state_dict(),
'netD' : netD.state_dict(),
'netQ' : netQ.state_dict(),
'optimD' : optimD.state_dict(),
'optimG' : optimG.state_dict(),
'params' : params
}, 'checkpoint/model_epoch_%d_{}'.format(params['dataset']) %(epoch+1))
training_time = time.time() - start_time
print("-"*50)
print('Training finished!\nTotal Time for Training: %.2fm' %(training_time / 60))
print("-"*50)
# Save model weights
torch.save(netG.state_dict(), 'NetG_Weights.pth')
torch.save(discriminator.state_dict(), 'Discriminator_Weights.pth')
torch.save(netD.state_dict(), 'NetD_Weights.pth')