Is there a way to run this model on Apple M1 that doesn't have CUDA support?

[gurpreetsinghmukker]

I get the following error when I try to run the SixDRepNet() model on an image.

File "/Users/gurpreetmukker/Desktop/face_detection/face_detection/lib/python3.11/site-packages/torch/cuda/__init__.py", line 239, in _lazy_init raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

Thanks

[thohemp]

The error occurs in the face detection package, that is used to generate the face crops for 6DRepNet.
Modifying the detector or changing it should solve the issue. 6DRepNet itself can run on CPU, even though it's not recommended.

[lheiler]

For me it worked to set line 17 in site-packages/sixdrepnet/regressor.py from
def __init__(self, gpu_id : int=0, dict_path: str=''): to def __init__(self, gpu_id : int=-1, dict_path: str=''):.
I tested the same picture in the Colab and on my M1 Max and I got the following results for pitch, yaw, roll:
M1 Max = [-3.0097966] [-14.455278] [1.3092302]
Google Colab: [-3.0097966] [-14.455285] [1.3092275]

For me that is close enough. :)