Error(s) in loading state_dict for Generator:

[askasjoe]

seems no 'g_ema' layer in model.py? any suggestions? thanks~

[askasjoe]

[r1cebank]

Sorry for the late reply, are you training or trying to run trained checkpoints? Can you share more info?

[askasjoe]

Sorry for the late reply, are you training or trying to run trained checkpoints? Can you share more info?

yea~ i was doing the inference python generate.py --size 512 --ckpt checkpoint/790000.pt using the the ckpt file checkpoints_790000_v1.zip from here https://github.com/diva-eng/stylegan-waifu-generator/releases/ and met the RuntimeError mentioned above, thanks~

[r1cebank]

could you try again with the load_state_dict line replaced with g_ema.load_state_dict(checkpoint["g_ema"], strict=False). sorry I dont have the machine I did the training and getting this to run on my current windows system is a pain 😢

[askasjoe]

this way worked, thank you very much~ even though the color style looks a bit weird

[r1cebank]

Glad to hear it worked, the color problem is probably due to the g_ema layer issue😓, if I found a way to make it work I’ll let you know.

[askasjoe]

Glad to hear it worked, the color problem is probably due to the g_ema layer issuesweat, if I found a way to make it work I’ll let you know.

got it~ thank you very much 😊