LiveFR

LoRA (low-rank adaptation)

Definition: LoRA (Low-Rank Adaptation) is an efficient fine-tuning method that trains only a small set of added parameters while freezing the original model.

It sharply cuts memory cost and lets you specialize a large model without retraining all of it. The resulting adapters are small and easy to swap.

See also

← Full AI glossary · AI news