5fa1a76
1
You can resume training with the sharded state dicts with the [~accelerate.Accelerator.load_state]` method.
