LMFlow icon indicating copy to clipboard operation
LMFlow copied to clipboard

cuda version issue

Open jackyh opened this issue 1 year ago • 1 comments

I installed cuda12 on my machine.

Then I do a: conda install cuda -c nvidia/label/cuda-11.7.0

my question is: there's quite some versions of the same lib there, such as:

(base) root@gysj01:~# find / -name libcurand.so /root/projects/anaconda/ENTER/envs/lmflow/lib/libcurand.so /root/projects/anaconda/ENTER/envs/lmflow/lib/stubs/libcurand.so /root/projects/anaconda/ENTER/pkgs/libcurand-dev-10.2.10.50-hd49a9cd_0/lib/libcurand.so /root/projects/anaconda/ENTER/pkgs/libcurand-dev-10.2.10.50-hd49a9cd_0/lib/stubs/libcurand.so /usr/local/cuda-12.0/targets/x86_64-linux/lib/libcurand.so /usr/local/cuda-12.0/targets/x86_64-linux/lib/stubs/libcurand.so

do you suggest me to lower down my cuda12 to cuda11.7, I mean from driver point of view? Also, there's quite some new features of cuda12, do you ever tried cuda12 with lmflow? @shizhediao @research4pan

jackyh avatar Apr 30 '23 02:04 jackyh

Thanks for your interest in LMFlow! We have run experiments in cuda12, and it works fine. Here's my settings:

NVIDIA-SMI 525.105.17   Driver Version: 525.105.17   CUDA Version: 12.0
torch version: 2.0.0

Hope that helps. Thanks 😄

research4pan avatar Apr 30 '23 13:04 research4pan

This issue has been marked as stale because it has not had recent activity. If you think this still needs to be addressed please feel free to reopen this issue. Thanks

shizhediao avatar Jun 19 '23 11:06 shizhediao

solved. thx.

jackyh avatar Jun 26 '23 14:06 jackyh