Reputation: 133
I am having a difficult time getting tensorflow to install on a MacBook Pro.
Initially, I tried pip3.8 install tensorflow
in my virtual environment. It installed but gives the following error when I try to use it:
This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
I get that this is a warning, but I think it's a serious warning that I am going to have performance issues with any non-trivial work.
Based on this post (Tensorflow on MacOS: Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA), I uninstalled tensorflow and followed the steps. After I installed Bazel, using homebrew, I got an error when I ran ./configure
:
Please downgrade your Bazel installation to version 0.26.1.
According to this: (https://github.com/bazelbuild/bazel/releases) the oldest version is 3.2. I had 3.7 installed. So I uninstalled Bazel using homebrew. This felt like a dead end even though compiling from the source seems like the correct way to go. That version it's asking for is not even remotely close to a current version. I think the message is not telling me what I need to know.
Next I tried using pip to install the version recommended by Tensorflow.org. (https://www.tensorflow.org/install/pip.html)
pip3.8 install https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-2.4.0-cp38-cp38-macosx_10_14_x86_64.whl
This successfully installed as well, but gives me similar errors as my original install and fails the test recommended under the list of installations on the tensorflow.org page.
I think I'm making a novice mistake. Can anyone assist here?
Upvotes: 0
Views: 1376
Reputation: 39
It's not serious warning, you can either wait for macOS 12.0+ to have built-in Apple Metal Acceleration or run Intel optimized wheels for macOS, but it support only TensorFlow 2.0
Upvotes: 1