CuriousTank

Hi! I am very interested in Human Centric Visual Tasks, may I ask if the fine-tuning part for downstream tasks only implement at some layers or all layers? Looking forward...