Dear Yann,
On 27/04/2018 09:21, Yann Sagon wrote:
I answer here to Kenneth and Ake.
Yes I need with_jemalloc = False and the patch.
Excellent, thanks for confirming.
We have enhanced the TensorFlow easyblock to automatically disable
jemalloc (unless it's hard enabled) on CentOS 6 for recent TensorFlow
versions, see
https://github.com/easybuilders/easybuild-easyblocks/pull/1412.
This change will be included in the next EasyBuild release (v3.6.1).
It seems there is indeed a bug (no confirmation) with TF-1.7.
I tried to compile using cuDNN 7.0.5-CUDA-9.1.85 and it's working fine!
https://github.com/tensorflow/tensorflow/issues/17801
It seems that CUDA 9 is strictly required since TF 1.7...
a side note: Many thanks for doing EasayBuild so useful, and I'm very
happy to see it is supporting Singularity! Great job!
Thanks for your kind words!
regards,
Kenneth
2018-04-27 9:04 GMT+02:00 Åke Sandgren <ake.sandg...@hpc2n.umu.se
<mailto:ake.sandg...@hpc2n.umu.se>>:
On 04/27/2018 08:49 AM, Kenneth Hoste wrote:
> Dear Yann,
>> here is the build log when trying to build for cuda:
>> https://gist.github.com/ysagon/a205242bb91becaccf4b684d5285514d
<https://gist.github.com/ysagon/a205242bb91becaccf4b684d5285514d>
>
> This is the actual error:
>
> Cuda Configuration Error: Cannot find libdevice.10.bc under
> /opt/ebsofts/Core/CUDA/8.0.61
>
> That doesn't ring a bell for me, but maybe some CUDA drivers are missing?
Something is wrong here, libdevice.10.bc doesn't exist until CUDA 9.0
--
Ake Sandgren, HPC2N, Umea University, S-90187 Umea, Sweden
Internet: a...@hpc2n.umu.se <mailto:a...@hpc2n.umu.se> Phone: +46 90
7866134 Fax: +46 90-580 14
Mobile: +46 70 7716134 WWW: http://www.hpc2n.umu.se
--
Logo UNIGE Yann Sagon
Chef d'équipe HPC
Division du système et des technologies de l'information et de la
communication
Université de Genève | 24 rue Général-Dufour
Tél 022 379 77 37 | Bureau 151
www.unige.ch/stic <http://www.unige.ch/stic>