it essentially just copies a chunk of the model to each one, works well for situations where each machine has limited vram
you run that on the remote nodes
it essentially just copies a chunk of the model to each one, works well for situations where each machine has limited vram