Yolox continous issues without cuda

[Raj2032]

I want to train model with cpu only

I am having errors like this:

lib/python3.14/site-packages/torch/cuda/__init__.py", line 417, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")

AssertionError: Torch not compiled with CUDA enabled

when running:

CUDA_VISIBLE_DEVICES="" python tools/train.py -f ../yolox_custom.py -b 

I want to train model with only cpu, is there a way I can do so?

[RinMar]

All Statements are cuda hardcoded. By surrounding all of these Statements with a check if cuda is available this could be solved

[Raj2032]

Which file and which lines/statements should I change?

[RinMar]

Sadly its multiple lines in multiple files. I did it locally and you basically need to look for Statements involving cude and add a check if cuda is available and otherwise solve the instructions on cpu

[Raj2032]

@RinMar By any chance is it possible to use AMD graphics card instead (could be via ROCm)?

[RinMar]

I mean the code has hardcoded cuda Statements. In Theorie it could probably work but you would probably need to adjust many files