Ultra-Light-Fast-Generic-Face-Detector-1MB icon indicating copy to clipboard operation
Ultra-Light-Fast-Generic-Face-Detector-1MB copied to clipboard

Published 20 hours ago • verified publisher icon Linzaer

windows下训练速度的问题

Open lyfighting opened this issue 5 years ago • 8 comments

感谢您开源如此优秀的人脸检测框架！训练过程中遇到一些问题，向您请教下。

1、我在windows下训练的，设置的num_workers=0，否则程序运行会失败，为什么会这样？ 2、在训练slim版本过程中发现gpu占用率只有1%左右，而cpu飙升到了70%，这是否正常？ 3、训练的很慢，一个epoch需要8分钟左右时间，这是否正常？

谢谢帮忙解答！

lyfighting avatar Feb 18 '20 08:02 lyfighting

I have got same issue. How did you solve the problem?

akashdev2016 avatar Feb 26 '20 02:02 akashdev2016

I have got same issue. How did you solve the problem?

@alexcruz0202 没有呢，我现在是切换成ubuntu环境，才可以用GPU正常训练。

lyfighting avatar Feb 26 '20 03:02 lyfighting

wins下训练问题可能的确比较多，我在linux下训练的。建议还是linux下训练，win下做推理问题不大。

Linzaer avatar Mar 04 '20 05:03 Linzaer

thanks, @lyfighting and @Linzaer. I'll test your suggestion on Windows and Ubuntu.

@lyfighting, What is ubuntu's version?

akashdev2016 avatar Mar 04 '20 16:03 akashdev2016

@alexcruz0202 我用的是ubuntu 16.04

lyfighting avatar Mar 13 '20 09:03 lyfighting

Thank you for your answers.

akashdev2016 avatar Mar 13 '20 09:03 akashdev2016

当我跑 train-version-RFB.sh的时候，出现AttributeError: 'SSD' object has no attribute 'module'，我该怎么处理呢？

123wk45678 avatar May 06 '20 11:05 123wk45678

当我跑 train-version-RFB.sh的时候，出现AttributeError: 'SSD' object has no attribute 'module'，我该怎么处理呢？

我直接把.module删掉就可以了

wangshibowyl avatar Mar 18 '21 08:03 wangshibowyl