LoRA

Creator
Creator
Seonglae ChoSeonglae Cho
Created
Created
2023 Jun 22 15:24
Editor
Edited
Edited
2024 Oct 16 20:12

Low-Rank Adaptation

LoRA is a technique that optimizes rank decomposition matrices with LoRA module rank .
The existing PEFT technology reduces the available sequence length of the model or expands the model depth for inference. LoRA usually applied to MLP by two low-rank matrix before activation.
Free weight and append adaptation layer composed by adaptation matrix with proving low-rank adaptation matrix is sufficient for fine-tuning. By dividing adaptation matrix like
Bottleneck layer
, It makes total model size smaller and make it achieve same learning rate with less computing resources.
A learning rate of 1e-4 has become the standard when fine-tuning LLMs with LoRA. Although we occasionally encountered training loss instabilities, reducing the learning rate to lower values like 3e-5. LoRA’s weight is initialized randomly but techniques like
EVA
uses activation vector decomposing it to initialize based on priority.
LoRA Usages
 
 

Model Regularization

LoRA Learns Less and Forgets Less

It is enough to change
Cross-Attention
layer for fine tuning

Tips

 
 
 
 

 

Recommendations