SURE

i use is the bvr_retinanet_x101_fpn_dcn_mstrain_400_1200_20e_coco.py python tools/train.py configs/bvr/bvr_retinanet_x101_fpn_dcn_mstrain_400_1200_20e_coco.py

Thank you for your reply. I used resnet-50 for training, and the speed has been improved obviously, but the accuracy is not as high as that mentioned in the article....

Thank you very much for your reply. I will try it

> The config has many heavy settings. > > Please try the following: > `Res2Net-50` or `Res2Net-101` > `stage_with_dcn=(False, False, False, True),` > `../_base_/datasets/coco_detection_mstrain_480_960.py` > `with_cp=False` > `fp16 = dict(loss_scale='dynamic')`...

What improvements have your paper made to the original model of Vit?

> 有同样的问题。我尝试减少批量大小。但这引入了一个新的运行时错误，指出“张量 a (22065) 的大小必须与非单维 1 处的张量 b (25) 的大小匹配”。有人想出如何解决这个问题吗？ Have you solved this problem and how to solve it