lakshayg / tensorflow-build-archived

TensorFlow binaries supporting AVX, FMA, SSE

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

tf `v.1.4.1` mac-mini does not support `AVX2`

mixuala opened this issue · comments

Hello,

I'm using your 1.2.1 MacOS build but now I'm running into random version problems with tensorboard. I'd like to try 1.4.1 but the mac-mini does not support AVX2 . Can you build a 1.4.1 to that spec? Or, send me a link on how to build one myself?

Many thanks!

I have uploaded the wheels. HTH.

TF HW OS GCC Python Supports
1.4.1 CPU macOS Sierra clang-900.0.39.2 2.7.13 AVX, SSE4.1, SSE4.2 Download
1.4.1 CPU macOS Sierra clang-900.0.39.2 3.6.1 AVX, SSE4.1, SSE4.2 Download

Thanks! but I didn't notice this in time so I tried to build it myself. it was not so hard. here is what I did based on the docs found at https://www.tensorflow.org/install/install_sources

cd [parent dir]
git clone https://github.com/tensorflow/tensorflow
cd tensorflow
git checkout tags/v1.4.1 -b r1.4

sudo pip install six numpy wheel 
brew cask install caskroom/versions/java8
brew install coreutils bazel
brew upgrade bazel

./configure
bazel build --config=opt --incompatible_load_argument_is_label=false //tensorflow/tools/pip_package:build_pip_package
bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg
sudo pip install /tmp/tensorflow_pkg/tensorflow-1.4.1-cp36-cp36m-macosx_10_7_x86_64.whl