I was trying to get something similar setup for Q-Diffusion, but the PyTorch cuda dependencies seem to be making this trickier. I did a manual fix to use "mps" torch_device, but I see calls to .cuda() throughput the code. I tried replacing those calls, but couldn't quite get it working. I imagine I need a different environment.yaml file for Mac but preliminary trial and errors aren't quite cutting it.
Has anyone managed to get this to work with Apple Mx silicon?
I have got Stable Diffusion working with Apple M3 Silicon as per the instructions from diffusers for MPS https://huggingface.co/docs/diffusers/optimization/mps.
I was trying to get something similar setup for Q-Diffusion, but the PyTorch cuda dependencies seem to be making this trickier. I did a manual fix to use "mps" torch_device, but I see calls to .cuda() throughput the code. I tried replacing those calls, but couldn't quite get it working. I imagine I need a different environment.yaml file for Mac but preliminary trial and errors aren't quite cutting it.
Has anyone managed to get this to work with Apple Mx silicon?