Code terminates without training while using accelerate

[muellerzr]

@jaydeepb

So would it automatically run the model on multiple GPUs?

Correct! There’s nothing special needed to do for DDP etc since accelerate is fully integrated (code-wise)