Closed yukeqi12345 closed 4 years ago
do you solve this question?
no ,the author don't give us
do you try to train with discard this code
no,the author don't give us the py
------------------ 原始邮件 ------------------ 发件人: "qinguoyi"<notifications@github.com>; 发送时间: 2019年12月5日(星期四) 上午10:48 收件人: "sunnyxiaohu/R-C3D.pytorch"<R-C3D.pytorch@noreply.github.com>; 抄送: "1144581929"<1144581929@qq.com>; "Author"<author@noreply.github.com>; 主题: Re: [sunnyxiaohu/R-C3D.pytorch] pytorch1.1 dont have dist model (#54)
do you solve this queation?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
hi, i discard this code with "#", and i run this code successfully.
what is your GPU,and my GPU is a 2080TI 。 I run the code in pytorch0.4 and my computer reboot
i run the code successfully on both server with GPU 2080TI & Pytorch1.1.0 and my private computer with GPU 1070 & Pytorch 0.4.0.
what is your setting ?my bs is 1,nw is 2 learning date is 0.001,and the GPU-Util is 97%,i am a little unhappy and i think maybe my computer will reboot
what is your setting ?my bs is 1,nw is 2 learning date is 0.001,and the GPU-Util is 97%,i am a little unhappy and i think maybe my computer will reboot
I did not change any parameters.
After training and testing the day before yesterday, I found that when tIOU = 0.5, mAP = 39.55%, which is about 2 percentage points higher than the open source.
I think it may be a platform bonus