LoRA & QLoRA in Practice

Sizing the adapter to the task, sizing the GPU to the adapter, and picking the checkpoint that actually generalizes.

Part of: Fine-Tuning & Distilling Open Models