Open dannyvfilms opened 1 year ago
Yup, this is expected. Apple changed Metal a lot; Torch isn't updated for these changes, etc.
I assume at this stage it wouldn't be possible to have any kind of expected timeframe, right?
Right, and it'd be a pytorch issue; most likely not an InvokeAI one.
I don't mean this to be in poor taste, but if anyone else finds themselves on the Sonoma beta and don't want to downgrade (or can't) but now you're stuck, however Diffusion Bee 2 structured their build still works on Sonoma.
I'll be returning as soon as possible to Invoke though.
This issue initially appears to be resolved with a combination of 3.0.2post1 and 14.0 Beta (23A5328b). Leaving open for now for further testing.
How did you eventually solve this ? 🥺 Here I have the same problem .
@JiayuMao02 are you using an updated version of InvokeAI?
Is there an existing issue for this?
OS
macOS
GPU
cpu
VRAM
16
What version did you experience this issue on?
3.0.1
What happened?
I should have seen this coming, but for an unrelated reason I upgraded MacOS to Sonoma 14.0 Beta (23A5328b) and later that day attempted to use InvokeAI. Now there are repeated ANE (Apple Neural Engine) errors during inference when invoking the model. I am running the model on Apple Silicon hardware (M1 Pro, 16GB).
Screenshots
Final output image:
Additional context
Contact Details
No response