mlx-grpo
🧠 Train your own DeepSeek-R1 style reasoning model on Mac! First MLX implementation of GRPO - the breakthrough technique behind R1's o1-matching performance. Build mathematical reasoning AI without expensive RLHF. Apple Silicon optimized. 🚀
aiapple-siliconartificial-intelligencechain-of-thoughtdeepseek-r1grpollamallmmathematical-reasoningmlxmulti-step-reasoningreasoning-airlhfthinking
Github Data
Star 9
Fork 1
Subscriber 0
Language Python
Size 89
owner
created at 2025-05-28
updated at 2025-06-19
pushed at 2025-06-19
