Cannot convert a MPS Tensor to float64 dtype as the MPS framework doesn't support float64. Please use float32 instead.
#2
by
brenzel
- opened
Get following error on my MacBook Pro:
Cannot convert a MPS Tensor to float64 dtype as the MPS framework doesn't support float64. Please use float32 instead.
What am I doing wrong?
I don't think this is GGUF specific, but my guess is that it's failing on this line:
scale = torch.arange(0, dim, 2, dtype=torch.float64, device=pos.device) / dim
You could try replace it with this and see if it fixes it:
scale = torch.arange(0, dim, 2, dtype=(torch.float32 if is_mps else torch.float64), device=pos.device) / dim
(Or just wait for it to be fixed in base ComfyUI)
It's working. Thank you.
brenzel
changed discussion status to
closed