Executing the accelerate script within a child process

🤗Accelerate
1

Hi, :slight_smile:

I’d like to train a large model on 8 gpus. But the training function is called by another python code. Is there a function to launch the distributed training from within another process?

Any advice is welcome
Thanks!