search

gaopengcuhk / Tip-Adapter

551 stars 43 forks source link

Sharpness & residual ratio range #13

Open gitfabianmeyer opened 1 year ago

gitfabianmeyer commented 1 year ago

Hi, I can't find any information on how you set the alpha and beta hyperparam ranges for each dataset. Why don't you use the same ranges for all sets, and how did you determine these ranges?

ZrrSkywalker commented 1 year ago

Hi Thanks for your interest. The alpha and beta are both set to 1 as the tuning baseline. The alpha weighs the importance between CLIP-pre-trained and few-shot knowledge. If the few-shot domain has a large gap to pre-trained data (general images, just like ImageNet), alpha is better to be larger than 1.