sd-webui-controlnet
sd-webui-controlnet copied to clipboard
RuntimeError when using depth_leres on Mac
The normal depth preprocessor works fine, so does the "depth extension" for Automatic1111.
depth_leres throughs a runtime error:
RuntimeError: Input type (MPSFloatType) and weight type (torch.FloatTensor) should be the same
Complete error here: PasteBin
My Setup:
python: 3.10.8 • torch: 1.13.1 • xformers: N/A • gradio: 3.16.2 • commit: [0cc0ee1b]
has same issue
Hi @krummrey do you still have this error? I'm getting it too - pretty sure I've used depth leres successfully in the past on my M1 with no error.
I've gone through all preprocessors and found these as not working:
- depth-leres - RuntimeError: Input type (MPSFloatType) and weight type (torch.FloatTensor) should be the same
- scribble-pidinet - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, mps:0 and cpu!
- softedge-pidinet - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, mps:0 and cpu!
- softedge-pidisafe - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, mps:0 and cpu!
- t2ia_sketch_pidi - RuntimeError: Expected all tensors to be on the same device, but found at least two devices, mps:0 and cpu!
- t2ia_style_clipvision - blank image
Closed without any information?
has same issue
Closed without any information?
Please open this issue again.
any news?
Just want to add that it's 2024 and the bug still persists. CN v1.1.440 on Mac M1 Pro and can't use depth_leres or leres++
Just want to add that it's 2024 and the bug still persists. CN v1.1.440 on Mac M1 Pro and can't use depth_leres or leres++
With all respect, they do not care. And I hope they will continue with this path. I believe latest events and concerns changed Apple's attitude and effort focus, we might have something good this year like what they did, D3DMetal.
I really get bored too. Do not get me wrong. Always, almost always people focus on Nvidia and left the nice devices we have back. Some people even don't know they have a dedicated NN unit in these devices, called ANE. With a little help of Apple, we can get into the right path with these devices.