trl-fine-tuning
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align
Also installable via skills CLI
npx skills add Orchestra-Research/AI-Research-SKILLs/06-post-training/trl-fine-tuning
Source
Path:
06-post-training/trl-fine-tuning/SKILL.md(main)