Maxwell J Jacobson

I was able to solve this by changing line 171 from: first_pass_img = torch.from_numpy(first_pass_img).unsqueeze(0).transpose(1,3).transpose(2,3).float().to(gpu_id) to first_pass_img = torch.from_numpy(first_pass_img).unsqueeze(0).transpose(1,3).transpose(2,3).contiguous().float().to(gpu_id) just making that img var contiguous.

According to this: https://github.com/kjsman/stable-diffusion-pytorch/issues/7 The weights are from a Stable Diffusion pretrained model, so there likely isn't training code.

Hey Ivangvozdanovic! You could define one like this: `class ProgIdent(ProgInertBlock): def __init__(self): super().__init__() def runBlock(self, x): return x def runActivation(self, x): return x` this would define a new inert block...