Kubetorch makes RL easy, part 1: Launch code sandboxes (extremely easily!) as part of #RL training done with #TRL by @huggingface. KT makes it extremely simple to setup any agent for execution on separate compute (& separate images) for your RL loop.
📦Kubetorch🔥: Fast, debuggable, fault-tolerant ML on shared compute.
github.com/run-house/kube…

