Recreating the minimal training methods of DeepSeek-R1 for small langauge models. - View it on GitHub