The original title is: "Training Agents 2: Live tutorial on model distillation for training custom agents."

Short summary

Hugging Face live tutorial on model distillation—transferring teacher model capability to smaller students via four methods: off-policy supervised fine-tuning, on-policy live scoring, reinforcement learning (KL divergence as token-level reward), and self-distillation. Covers white-box vs black-box approaches and TRL framework implementation. Part of series on training agents for experiment automation.

• •Four distillation methods: off-policy SFT, on-policy scoring, RL-based (KL divergence reward), self-distillation
• •Explores white-box vs black-box capability transfer and organizing axes: signal, data source, timing, teacher identity
• •Practical implementation using TRL with GitHub repo; agent-driven experiment design and monitoring

Generated with AI, which can make mistakes.