search

archenroot / gentoo-overlay

Gentoo overlay with main focus on GPU, Neural Networks, Big Data and Java technologies
http://gentoo.archenroot.org
15 stars 4 forks source link

request: tensorflow-9999 ebuild #3

Closed Randl closed 6 years ago

Randl commented 7 years ago

Well, basically the title explains it.

archenroot commented 7 years ago

Will look at that.

archenroot commented 7 years ago

There has been lot of work on TensorFlow by no surprise, the ebuild in general cannot be simply updated by bumping number, I need to look into this and maybe I will ask someone from sci-libs overlay.

archenroot commented 7 years ago

I will look at that during weekend as anyway refreshing torch and looking to adapt deeplearning4j as well into gentoo ecosystem.

archenroot commented 7 years ago

I am having in the moment emerge issues even with nvidia drivers as I am on 6.3.0 GCC and facing some incompatible gcc/lugin versions not just for this package. I am anyway going to experiment with gcc 7.0.1, so once in place I will come back to this baby. In general the ebuild is in repo, but I didn't test it yet.

archenroot commented 7 years ago

@Randl - The TensorFlow 9999 package is there, but during my testing I realize the current ebuild doesn't support CUDA/OpenCL acceleration which is must :-) I will be targeting this feature soon.

archenroot commented 7 years ago

Actually I found another issue with:

Requirement '/tmp/portage/sci-libs/tensorflow-9999/work/tensorflow-9999/tensorflow_pkg/*.whl' looks like a filename, but the file does not exist
*.whl is not a valid wheel filename.

So the package is broken in the moment...

archenroot commented 6 years ago

Some update after half year :-))) I am just compiling 9999 current, looks good so far. Will post commit if finished successfuly... note I have only small ultrabook without GPU, so don't do test with either CUDA or OpenCL

archenroot commented 6 years ago

fail, shit...

archenroot commented 6 years ago

Ok, finally I got:

nunik /var/lib/layman/archenroot/sci-libs/tensorflow # python
Python 3.4.5 (default, Oct 26 2017, 03:48:12) 
[GCC 5.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> hello = tf.constant('Hello, TensorFlow!')
>>> sess = tf.Session()
2017-11-08 20:22:57.635030: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
>>> print(sess.run(hello))
b'Hello, TensorFlow!'

As you can see, CPU flags are not submitted, so another extension and I also would like to support all following config-opts:

Found possible Python library paths:
  /usr/lib/python-exec/python3.4/../../../lib64/python3.4/site-packages
Please input the desired Python library path to use.  Default is [/usr/lib/python-exec/python3.4/../../../lib64/python3.4/site-packages]
Do you wish to build TensorFlow with jemalloc as malloc support? [Y/n]: jemalloc as malloc support will be enabled for TensorFlow.

Do you wish to build TensorFlow with Google Cloud Platform support? [Y/n]: Google Cloud Platform support will be enabled for TensorFlow.

Do you wish to build TensorFlow with Hadoop File System support? [Y/n]: Hadoop File System support will be enabled for TensorFlow.

Do you wish to build TensorFlow with Amazon S3 File System support? [Y/n]: Amazon S3 File System support will be enabled for TensorFlow.

Do you wish to build TensorFlow with XLA JIT support? [y/N]: No XLA JIT support will be enabled for TensorFlow.

Do you wish to build TensorFlow with GDR support? [y/N]: No GDR support will be enabled for TensorFlow.

Do you wish to build TensorFlow with VERBS support? [y/N]: No VERBS support will be enabled for TensorFlow.

Do you wish to build TensorFlow with OpenCL support? [y/N]: No OpenCL support will be enabled for TensorFlow.

Do you wish to build TensorFlow with CUDA support? [y/N]: No CUDA support will be enabled for TensorFlow.

Do you wish to build TensorFlow with MPI support? [y/N]: No MPI support will be enabled for TensorFlow.

Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]: 

Add "--config=mkl" to your bazel command to build with MKL support.
Please note that MKL on MacOS or windows is still not supported.
If you would like to use a local MKL instead of downloading, please set the environment variable "TF_MKL_ROOT" every time before build.
Configuration finished
archenroot commented 6 years ago

This commit works now: https://github.com/archenroot/gentoo-overlay/commit/caedd16fdb68cad75183fdb9bbc604a730eb432f

Also I hardcoded Tensorflow commit as soon as today's changes made it uncompilable, so I use about 25 days old commit, you can just disable it in ebuild as you require...