GANの訓練ループでよく見るdetachは何をしている？ - Qiita

optimizerD = torch.optim.AdamW(D.parameters()) optimizerG = torch.optim.AdamW(G.parameters()) G_losses = [] D_losses = [] for epoch in tqdm(range(num_epochs)): for data, class_label in tqdm(train_loader, leave=False): real_data = data.to(device) y_real = torch.full((batch_size,1), real_label, dtype=torch.float32, device=device) y_fake = torch.full((batch_size,1), fake_label, dtype=torch.float32, d