LoRA
Low-Rank Adaptation. Fine-tune a model without retraining its weights: freeze W, learn a small A·B update. Cuts trainable parameters by ~100×.
Companion explanation in Step by Token, chapter 14.
Where this term gets built
Continue
Low-Rank Adaptation. Fine-tune a model without retraining its weights: freeze W, learn a small A·B update. Cuts trainable parameters by ~100×.
Companion explanation in Step by Token, chapter 14.
Continue