Open Moulick opened 1 month ago
same here
Are both models: GFPGANv1.4.pth and inswapper_128_fp16.onnx compatible with the CoreML architecture ? Because I tried to force the core.py file to utilize the CoreMLExecutionProvider instead of the CPUExecutionProvider, but it always reverts to CPU only. I also tried to make ONNX being more verbose-like, but I got no extra verbose output out of it. Any other idea ?
@julywint007, out of curiosity, why did you thumbs-down @Moulick? Do you think there's something obvious we're overlooking?
Also same issue on M3 Pro
close
pip uninstall onnxruntime onnxruntime-silicon pip install onnxruntime-silicon==1.13.1 python run.py --execution-provider coreml
I have an M3 Pro, but after following the below, it still uses the CPU and not the GPU
pip uninstall onnxruntime onnxruntime-silicon pip install onnxruntime-silicon==1.13.1 python run.py --execution-provider coreml
with
--execution-provider coreml
, it enables both the CPU and GPU, is there a way to force disable the CPU provider?
@julywint007 Huh?
seems it's impossible for live
cam to use GPU on M2/M3 chips? Is there anyone able to do that?
@julywint007 are you able to use apple GPU?
Just wonder if CoreMLExecutionProvider has really been tested on apple silicon, it seems GPU properties are not working at all.
@yiwenzhouzoe @julywint007 Better alternative thats compatible with Apple Silicon and CoreML utilizing full its potential is FaceFusion and it works on same principles https://github.com/facefusion/facefusion
@yiwenzhouzoe @julywint007 Better alternative thats compatible with Apple Silicon and CoreML utilizing full its potential is FaceFusion and it works on same principles https://github.com/facefusion/facefusion
Does FaceFusion allow for live swapping?
Yep
Yep
Awesome! Thank you.
I have an M3 Pro, but after following the below, it still uses the CPU and not the GPU
with
--execution-provider coreml
, it enables both the CPU and GPU, is there a way to force disable the CPU provider?