I managed to let Automatic1111 version of SD run on CPU but painfully slow.My mac did pass the mps support verification in Python environment.But I can’t find clear instructions on how to let SD run on GPUs.
I found this comment from @fillipux is very inspiring.
He managed to solve it by…
go through every file and changetorch.device("cpu"/"cuda")to torch.device("mps")in register_buffer() in ddim.py, change to attr = attr.to(torch.device("mps"), torch.float32)in layer_norm() in functional.py (part of pytorch I guess), change to return torch.layer_norm(input.contiguous(), ...in terminal write export PYTORCH_ENABLE_MPS_FALLBACK=1dozens of other things I have forgotten about...
Are there any easier solutions?
System: MacbookPro Intel Core i7 Radeon Pro 555 Ventura