del last loss

flux/sampling_lore.py

@@ -348,6 +348,7 @@ def denoise_with_noise_optim(
                 trainable_noise.grad *= source_mask[0]
             optimizer.step()
             print(f"Time {t_curr:.4f} Step {j+1}/{training_epochs}, Loss: {total_loss.item():.6f}")
+            del total_loss,attn_maps
             torch.cuda.empty_cache()
             torch.cuda.synchronize()