# initializate optimizer
opt_enc = optim.Adam(enc.parameters(), lr=LEARNING_RATE, betas=(0.0, 0.9))
padding_epoch = len(str(enc_epochs))
padding_i = len(str(len(loader)))
writer_enc = SummaryWriter(f"logs_kernel_3/enc")
for epoch in range(enc_epochs):
# Target labels not needed! <3 unsupervised
for batch_idx, (real, _) in enumerate(loader):
real = real.to(device)
cur_batch_size = real.shape[0]
opt_enc.zero_grad()
z = enc(real)
fake = gen(z)
real_features = critic.forward_features(real)
fake_features = critic.forward_features(fake)
#izif architecture
loss_imgs = criterion(fake, real)
loss_features = criterion(fake_features, real_features)
enc_loss = loss_imgs + kappa * loss_features
enc_loss.backward()
opt_enc.step()
# if i % CRITIC_ITERATIONS == 0:
# e_losses.append(e_loss)
# enc.eval()
writer_enc.add_scalar('enc_loss', enc_loss.item(), epoch)
print(f"[Epoch {epoch:{padding_epoch}}/{enc_epochs}] "
f"[Batch {batch_idx:{padding_i}}/{len(loader)}] "
