os/doc
3
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

Add 2017Q2 TensorFlow porting entry from amutu@amutu.com

Browse source
This commit is contained in:
Benjamin Kaduk 2017-07-03 19:04:26 +00:00
parent d906eb9eda
commit 9173230d68
Notes: svn2git 2020-12-08 03:00:23 +00:00 
svn path=/head/; revision=50445
1 changed files with 87 additions and 0 deletions

87
en_US.ISO8859-1/htdocs/news/status/report-2017-04-2017-06.xml
View file

@ -934,4 +934,91 @@
July.</task>
</help>
</project>
<project cat='ports'>
<title>TensorFlow</title>
<contact>
<person>
<name>
<given>Jov</given>
</name>
<email>amutu@amutu.com</email>
</person>
</contact>
<links>
<url href="https://bugs.FreeBSD.org/bugzilla/show_bug.cgi?id=219609">TensorFlow PR</url>
<url href="https://reviews.FreeBSD.org/D11194">Phabricator Review</url>
<url href="https://github.com/amutu/tf-FreeBSD-pkg">Prebuilt Packages</url>
<url href="https://www.tensorflow.org">TensorFlow Upstream</url>
</links>
<body>
<p>As described on its website, &quot;TensorFlow is an open
source software library for numerical computation using data
flow graphs. Nodes in the graph represent mathematical
operations, while the graph edges represent the
multidimensional data arrays (tensors) communicated between
them. The flexible architecture allows you to deploy
computation to one or more CPUs or GPUs in a desktop, server,
or mobile device with a single API. TensorFlow was originally
developed by researchers and engineers working on the Google
Brain Team within Google's Machine Intelligence research
organization for the purposes of conducting machine learning
and deep neural networks research, but the system is general
enough to be applicable in a wide variety of other domains as
well.&quot;</p>
<p>TensorFlow now is the most popular platform/library for machine
learning and AI. There are official binaries for Linux, Mac,
Windows, and Android, but no official support for &os;. For
the last several months, I have done some work to make
TensorFlow available on &os;. Some notable work:</p>
<ul>
<li>Patch <tt>bazel</tt> to not depend on <tt>/proc</tt> at
build time. <tt>bazel</tt> is a build tool made by Google.
It uses <tt>/proc</tt> to get path-to-self when building C++
code, but mounting <tt>/proc</tt> is usually not allowed
when building as an unprivileged user.</li>
<li>TensorFlow can now be built on &os;&nbsp;10.x by using
<tt>clang38</tt> as the default <tt>bazel</tt> cross-build
tool.</li>
<li>Patch the <tt>bazel</tt> workspace files to allow
TensorFlow to be built using offline third-party
dependencies. This work is needed because the &os; Ports
framework does not allow network access except during the
fetch stage.</li>
<li>Fix the build on &os; i386.</li>
<li>Make TensorFlow build with either Python 2 or Python 3.</li>
<li>Update to the latest version, which is
<tt>tensorflow</tt>-1.2.0.</li>
</ul>
<p>This port would not be possible without substantial
assistance from bapt@, lwhsu@, mat@, and koobs@ &mdash; thank
you for your help! You are very nice and I learned a lot
about &os; and the Ports framework from your help.</p>
</body>
<help>
<task>Review, test, comment, and most importantly, commit to the
Ports Collection.</task>
<task>Fix the OpenCL support on &os;.</task>
<task>Port <tt>tensorflow-serving</tt>, which is a flexible,
high-performance serving system for machine learning models
produced by TensorFlow.</task>
<task>Set up a CI for TensorFlow on &os; and give early notice
to upstream when they break TensorFlow on &os;.</task>
</help>
</project>
</report>