701JAN 3110:29
articleHuggingFace Bloglast yr.
Mini-R1: Reproduce Deepseek R1 „aha moment“ a RL tutorial
Replaying the DeepSeek-R1 'aha moment', this post uses Group Relative Policy Optimization (GRPO) and the Countdown Game to train an open model via RL. It details a distributed setup with DeepSpeed and vLLM on 4× NVIDIA H100 GPUs and explains how GRPO replaces a value function with group-based baselines. The aim is self-verification and search abilities learned with minimal human data, illustrating a concrete RL workflow for LLMs.