hustvl / Vim

[ICML 2024] Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model
Apache License 2.0
2.55k stars 160 forks source link

LR & Optim setting for downstream task #37

Open MDD-0928 opened 3 months ago

MDD-0928 commented 3 months ago

Dear authors,

THanks to your excellent work. I would like to use Vim as backbone for downstream task, I need to load the pretraied weight on Imagenet-1K, then finetune the network on downstream-task dataset, then, what lr_scheduler and optimizer should I use? Could you please give me some suggestions about the super-params?