Skip to content

{lib}[intel/2016b] Tensorflow v0.12.1 w/ Python 3.5.2#4084

Merged
boegel merged 2 commits intoeasybuilders:developfrom
boegel:20170131134219_new_pr_Tensorflow0121
Jan 31, 2017
Merged

{lib}[intel/2016b] Tensorflow v0.12.1 w/ Python 3.5.2#4084
boegel merged 2 commits intoeasybuilders:developfrom
boegel:20170131134219_new_pr_Tensorflow0121

Conversation

@boegel
Copy link
Copy Markdown
Member

@boegel boegel commented Jan 31, 2017

(created using eb --new-pr)

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2304.phanpy.os - Linux centos linux 7.3.1611, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz, Python 2.7.5
See https://gist.github.com/572f5585749fce888f56e661d2e293c2 for a full test report.

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2039.delcatty.os - Linux centos linux 7.3.1611, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/34bb8acaeadfe4d70aeae17521188dc2 for a full test report.

source_urls = ['https://github.com/google/protobuf/archive/v%(version)s/']
sources = [SOURCE_TAR_GZ]

preconfigopts = "./autogen.sh &&"
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

you need autotools as build dep for this.


pymaj = '3'
pymin = '5'
pymajmin = pymaj + pymin
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

don't we have constants for this?

@rjeschmi
Copy link
Copy Markdown
Contributor

I can't test it easily right now, but it looks ok in theory.

Do you think it might be worthwhile to offer a compiled version?

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

@rjeschmi I usually prefer building from source, but I just don't have the time to dive into that now for Tensorflow. It certainly couldn't hurt, imho.

The build instructions are at https://www.tensorflow.org/get_started/os_setup#installing_from_sources; they require a custom build tool bazel (we have an easyconfig using the CrayGNU toolchain for that), and you obtain a custom .whl that you can use to install Tensorflow.

So, that's most likely a custom easyblock...

I think @gppezzi has some experience with building Tensorflow from source.
Is that effort publicly available somewhere @gppezzi?

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Hah, the configure script is an interactive script, so definitely a custom easyblock...
Cfr. https://www.tensorflow.org/get_started/os_setup#configure_the_installation

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Test report by @boegel
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in this PR)
node2304.phanpy.os - Linux centos linux 7.3.1611, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz, Python 2.7.5
See https://gist.github.com/fd6be5712d54650066d08b5e6f1fabec for a full test report.

@gppezzi
Copy link
Copy Markdown
Contributor

gppezzi commented Jan 31, 2017

We did already build TF from sources, we have an open PR for TF with and without Cuda on Cray (but this is WIP).

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

@gppezzi Do you have a particular reason for not using a Tensorflow easyblock? Seems like that would be appropriate here...

@gppezzi
Copy link
Copy Markdown
Contributor

gppezzi commented Jan 31, 2017

I just don't think that it is worth the efforts right now: TF seems to be changing very quickly and, by time we manage to have a working easyblock, things will change.

IMO chances of re-using the easyblock for a new version (or on a different system) are very low. Also, we need more time to better understand this build tool (the procedure is currently a huge mess).

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Test report by @boegel
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in this PR)
node2039.delcatty.os - Linux centos linux 7.3.1611, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/6e17d6be65d78d390cba1d95ae3e49f9 for a full test report.

@verdurin
Copy link
Copy Markdown
Member

@boegel why is the foss-2016a version of protobuf included here?

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

@verdurin I wanted to make the change requested by @wpoely86 (incl. Autotools as a build dep) consistent for protobuf easyconfigs that call autogen.sh

@boegel
Copy link
Copy Markdown
Member Author

boegel commented Jan 31, 2017

Thanks for the feedback @rjeschmi and @gppezzi.

I'll go ahead and merge this as-is, we can re-evaluate building from source later.

It seems like TF is close to a 1.0 release, which suggests that the build procedure should be stable too...

@boegel boegel merged commit dda2d86 into easybuilders:develop Jan 31, 2017
@boegel boegel deleted the 20170131134219_new_pr_Tensorflow0121 branch January 31, 2017 17:31
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants