no checkpoint save about stylegan-nada HOT 4 CLOSED

Are you having this issue with the Collab, or with the offline training script?

If it's the Collab, you can add:

        if (args.save_interval is not None) and (i > 0) and (i % args.save_interval == 0):
            torch.save(
                {
                    "g_ema": net.generator_trainable.generator.state_dict(),
                    "g_optim": g_optim.state_dict(),
                },
                f"{ckpt_dir}/{str(i).zfill(6)}.pt",
            )

to the training loop.

I'll update the Collab to add model saving support.

from stylegan-nada.

Colab has been updated. Let me know if you're still having issues.

from stylegan-nada.

Thank you for your quick reply!
I have this issue with the offline training script, I have change offline script according colab, and the issue have been solved, thank you!

from stylegan-nada.

Happy to help. It doesn't look like the checkpoint code is missing from the train script, so I'm not sure what was causing your issue. Closing for now.

from stylegan-nada.