Regarding the customized prompt learning #52
Description
Hi, thanks for your amazing work! I am fully mesmerized by your proposed model DINO-X, especially the customized prompt tuning, where one can learn an input embedding using custom dataset to finetune the model without adapting its parameters. I mean, can you share more technical details behind this, or which existing work do you refer to, in order to serve the purpose? Looking forward to your reply!