李洋

icon company University of Science and Technology of China icon location student of University of Science and Technology of China

Results 4 comments of 李洋

use -std=c++17 and #include<bits/stdc++.h>

> It seemed to be a problem with my compiler, when I installed mingw64, the problem was solved. 我安装的也是mingw64，从官网上刚下的最新版，但是就是有问题...orz

Figure 4 in paper may be wrong according to the code?

我也发现了这个问题，代码里用了if else语句，所以在第一个阶段，patch embedding和linear embedding只能存在一个，而不像Fig.4里画的那样串行。

训练的时候给的batch为8能运行，测试的时候batch为1，显存都不够用呢？

您好，请问可以参考一下您的代码吗，主要是PRA Loss部分，万分感谢！[email protected]

After sync batch norm is applied, more gpu memory is consumed

I have faced the same issue，too. Before using convet_model to replace typical batch norm with SynchronizedBatchNorm2d: GPU_1---7520, GPU_2---6756. After that, GPU_1---9760,GPU_2---8796. can you help me? @vacancy