it is barely relevant to big players, but is extremely valuable for small players: distributing your workflow to multiple gpus manually is not that simple thing

it is barely relevant to big players, but is extremely valuable for small players: distributing your workflow to multiple gpus manually is not that simple thing to do and there are a lot of much more interesting/important problem to solve than shoving your model to a gpu.