--size parameters must be a power of 2

[hackerfriendly]

First, thank you for your effort! I've enjoyed experimenting with this code and I've learned a lot from your contributions.

Sorry I'm not much help here, but I noticed that specifying an x or y that isn't a power of 2 raises a RuntimeError. Not sure if that's intended behavior at this stage:

RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 24 but got size 25 for tensor number 1 in the list.

Full traceback follows.

$ ./clip_sample.py --size 640 400 "something"
Traceback (most recent call last):
File "./clip_sample.py", line 203, in
main()
File "./clip_sample.py", line 197, in main
run_all(args.n, args.batch_size)
File "./clip_sample.py", line 192, in run_all
outs = run(x[i:i+cur_batch_size], steps, clip_embed[i:i+cur_batch_size])
File "./clip_sample.py", line 180, in run
return sampling.cond_sample(model, x, steps, args.eta, extra_args, cond_fn_)
File "/work/env/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/sampling.py", line 62, in cond_sample
v = model(x, ts * steps[i], **extra_args)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 246, in forward
out = self.net(torch.cat([input, timestep_embed], dim=1))
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 63, in forward
return torch.cat([self.main(input), self.skip(input)], dim=1)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 63, in forward
return torch.cat([self.main(input), self.skip(input)], dim=1)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 63, in forward
return torch.cat([self.main(input), self.skip(input)], dim=1)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 63, in forward
return torch.cat([self.main(input), self.skip(input)], dim=1)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/container.py", line 141, in forward
input = module(input)
File "/work/env/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/work/v-diffusion-pytorch/diffusion/models/cc12m_1.py", line 63, in forward
return torch.cat([self.main(input), self.skip(input)], dim=1)
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 24 but got size 25 for tensor number 1 in the list.

[crowsonkb]

At some point I am going to work out what the size parameters have to be divisible by and add it to the model class for each model, because I am going to have to do this for upscaler models because they have no set native resolution...

[lopho]

From experimentation I determined the sides have to be divisible by 64.