CBSR_WACV_CODE
CBSR_WACV_CODE
您好，请问在lr=1e-3时训20个epoch后我这里的acc只有60不到？
您好，感谢您的工作。有些疑问希望可以得到您的解答： 1. 原文中阐述了instance-based的两个问题，但是第二个问题是否已经被CVPR 2021 DCQ那篇文章解决？第一个问题在DCQ中仍然存在，但是改进了这个问题好像结果并没有比DCQ好？ 2. 原文中阐述的这两个问题，是如何影响实验结果呢？更具体来说，您定义的更新速度这个变量会影响最终识别的结果吗，更新快意味着结果好吗？【1. If the number of identities is severely imbalanced, the update speed of the identities’ centers that have rare number of images is too...
Line 67, train_val.py: if isinstance(cos_thetas, tuple): cos_thetas, bad_grad = cos_thetas labels[bad_grad.squeeze(-1)] = -100 # ignore_index How can I understand this part of code? May I ask in which situation this...
When I test the trained swin-moe on multiple gpus, the performance of each process is different. I loaded the weights for each process {t} with rank{t}.