Fine-tune Train

Train the model and verify improvement over base model.

Prerequisites

Complete finetune-generate first. You need:

• training_data.jsonl — Filtered, validated training examples
• Model choice from finetune-design
• Evaluation rubric (for comparing base vs fine-tuned)

Scope: SFT Only

This skill covers Supervised Fine-Tuning (SFT) only. Other training methods (DPO, GRPO, GEPA) will hopefully be added in the future.

Related HuggingFace Skills

Use these HuggingFace skills for an integrated workflow: