Last released Aug 14, 2026
An easy-to-use, fast toolkit to scale up RL post-training on a single node.