Can't training with GPU

[lovekittynine]

In the version of pytorch 1.3.1, OsNet can't train with GPU, report CUDA out of memory, but the gpu is not occupied。Bug？

[lovekittynine]

even set bs=1, also report CUDA out of memor

[KaiyangZhou]

what's the full command line to reproduce ur error?

pls provide more info such as system ...

[ljx6666]

I have the same problem, my computer system is linux, and the processor is AMD FX(tm)-8150 Eight-Core Processor *8，please help me! Thank you !

[KaiyangZhou]

@lovekittynine did you find what's going on?

[ljx6666]

My video card is GTX1080ti with 11G . I only have this one. Is there not enough video memory?

[KaiyangZhou]

@ljx6666 are you using scripts/main.py to run the code? if so, what's the full command you used?

[lovekittynine]

@KaiyangZhou @ljx6666 maybe you can try train with script: CUDA_VISIBLE_DEVICES=0 python3 train.py，explictly set GPU

[ljx6666]

Thank you for your patience!  I have solved the problem.

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "lovekittynine"<notifications@github.com&gt;; 发送时间: 2019年12月11日(星期三) 下午4:02 收件人: "KaiyangZhou/deep-person-reid"<deep-person-reid@noreply.github.com&gt;; 抄送: "白开水"<916679590@qq.com&gt;; "Mention"<mention@noreply.github.com&gt;; 主题: Re: [KaiyangZhou/deep-person-reid] Can&#39;t training with GPU (#259) @KaiyangZhou @ljx6666 maybe you can try train with script: CUDA_VISIBLE_DEVICES=0 python3 train.py，explictly set GPU — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[KaiyangZhou]

please note that in the latest commit, you need to explicitly set CUDA_VISIBLE_DEVICES