Train models with Hugging Face TRL: supervised fine-tuning (SFT), direct preference optimization (DPO), group relative policy optimization (GRPO), and GGUF conversion.
When to use this skill
- Fine-tuning a model with SFT
- Running DPO training
- Converting a model to GGUF
Core concepts
This skill provides guidance on hf model training best practices, patterns, and common pitfalls. It is designed to be loaded on demand when a relevant task is detected.
Installation
curl -LO https://opencode-skills.example/downloads/ai-ml/hf-training.zip
unzip hf-training.zip -d ~/.config/opencode/skills/Restart OpenCode — the skill loads automatically.
When it triggers
- fine-tuning a model with SFT
- running DPO training
- converting a model to GGUF