GRPO alignment algorithm trainer based on OpenRLHF
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).