Looks pretty cool. I haven't yet gotten into training any models at scale, but this seems to reduce the cognitive load associated with MLOps. Will start hacki

Looks pretty cool. I haven't yet gotten into training any models at scale, but this seems to reduce the cognitive load associated with MLOps. Will start hacking on it this weekend. The upsides to this are pretty compelling. Any downsides vs traditional MLOps?