Training Experiment with a small LLM model and GRPO - View it on GitHub