Comment by numbernine

Can someone explain me why one would need a container orchestration system to train ml models? Why wouldn't one just stick with containers?