Open UcanSee opened 3 years ago
I have transformed cswin-tiny to object detection task, just adding some padding operation. When I used Faster-RCNN model, I found that gpu memory of cswin-tiny is twice than that of swin-tiny, is there something wrong I made?
I have transformed cswin-tiny to object detection task, just adding some padding operation. When I used Faster-RCNN model, I found that gpu memory of cswin-tiny is twice than that of swin-tiny, is there something wrong I made?