LoRA (Low Rank Adaption)
What is Rank
Simply put, the higher the rank of a matrix, the more information it contains.
What is LoRA
Transforming a large matrix into smaller matrices makes computations require less performance.
Multiplying multiple small matrices can form a large matrix; similarly, a large matrix can be decomposed into multiple small matrices, thereby reducing the number of parameters and lowering performance requirements.
Reference
What is LoRA? | Explaining the working principle of LoRA in 5 minutes _ Bilibili_bilibili