A generate.py bug for loading trained model

[YYZ1021]

Hi,

When I generate images by using my model, I found it will output nothing.

I have debugded the code and I found that you didn't pass the option 'network' into the 'main' function.

pfgmpp/generate.py

Line 285 in acad8da

def main(ckpt, end_ckpt, outdir, subdirs, seeds, class_idx, max_batch_size, save_images, pfgmpp, aug_dim, edm, use_pickle, device=torch.device('cuda'), **sampler_kwargs):

And the 'main' function will search the model checkpoint in the 'outdir' path.

pfgmpp/generate.py

Line 298 in acad8da

stats = glob.glob(os.path.join(outdir, "training-state-*.pkl"))

Now I put the model checkpoint into the 'outdir' folder, and it works.

However, it may be confused. Please check it, thank you.

[Newbeeer]

Thanks for spotting this. I will add this feature tomorrow.