Start training on CPU if couldnt auto detect GPU/s

[biomedsciai]

Start training on CPU if couldnt auto detect GPU/s

[mosheraboh]

Thanks @smartdanny !
You can modify the code to something like this:
num_allocated = GPU.choose_and_enable_multiple_gpus(NUM_GPUS, force_gpus=force_gpus, use_cpu_if_fail=True)
if num_allocated == 0:
TRAIN_COMMON_PARAMS["trainer.accelerator"] = "cpu"

BTW, we test fusemedml only on linux servers. Let me know if you have any IOS-related issues.

[shatz01]

right, using use_cpu_if_fail fixed :) thanks