This was the one used in the workshop: github.com/unionai/workshops/…/llm-fine-tuning-grpo-countdown

Here is one for coding if you can run a larger model, remember that the model needs to be able to do that task for it to learn with RL. YOu could also try different coding problems: github.com/unionai/workshops/…/llm-fine-tuning-grpo-code

I need to clean up this example and and merge it in, but but its a distributed one if you have multiple GPUs to do rollouts on: github.com/unionai/workshops/…/llm-fine-tuning-grpo-distributed