1 GPU when multiple are available about finetuner HOT 2 OPEN

Hi @jose,

I think it's because there's a bug in the accelerate package that doesn't detect multiple GPUs on your computer. I checked the documentation and it says that the code accelerator = Accelerator() will automatically detect all available GPUs, but that's only one for you. So I don't think it's a problem with my code, maybe you can use the torch.nn.parallel.DistributedDataParallel to implement parallel training.

from finetuner.

Thanks. Unfortunately, I'm not that familiar with torch to be able to make a change like that. I will keep using your base code even if in my case only detects 1 GPU. It takes longer to run but at least it works. :-)

from finetuner.