mit-han-lab / efficientvit

EfficientViT is a new family of vision models for efficient high-resolution vision.
Apache License 2.0
1.6k stars 142 forks source link

Using weights from l3 high res in timm #34

Open Kaschi14 opened 9 months ago

Kaschi14 commented 9 months ago

Can I use the weights provided in this repo for a timm cls model? Because the efficientvit models in timm only support up to 288x288 resolution. I'd like to finetune higher res models