How to install if I use conda environments?

[giordan12]

I do not have virtualenvs installed but I do use conda environments. Is there a way to install in a conda environment? Thanks!

[grwche]

need to change:

• tf_2_4 == your (conda) virtual environment name
• /Users/grwche/Downloads/ == path to your download file (from the releases section on this repo)

pip install --upgrade pip wheel setuptools cached-property six

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/grpcio-1.33.2-cp38-cp38-macosx_11_0_x86_64.whl"

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/h5py-2.10.0-cp38-cp38-macosx_11_0_x86_64.whl"

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/numpy-1.18.5-cp38-cp38-macosx_11_0_x86_64.whl"

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/scipy-1.5.4-cp38-cp38-macosx_11_0_x86_64.whl"

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_11_0_x86_64.whl"

pip install absl-py astunparse flatbuffers gast google_pasta keras_preprocessing opt_einsum protobuf tensorflow_estimator termcolor typing_extensions wrapt wheel tensorboard typeguard

pip install --upgrade -t "tf_2_4/lib/python3.8/site-packages/" --no-dependencies --force "/Users/grwche/Downloads/tensorflow_macos/x86_64/tensorflow_macos-0.1a0-cp38-cp38-macosx_11_0_x86_64.whl"

[mwidjaja1]

EDIT: DO NOT USE THIS. Well you can, it didn't work for me after all, but you can try. I posted an updated version of this, in light of some new updates in the world of Miniconda, later in this post.

---

Thanks @grwche! For those installing this on ARM, the libraries are slightly different, so here was my version.

1. Download the archive from https://github.com/apple/tensorflow_macos/releases
2. Run the below in a bash terminal and replace the first two variables.

# If you have Tensorflow in your env, you must install it first -- uncomment this line if you're in this boat
# pip uninstall tensorflow tensorboard

# Put a path to where the arm64 libraries are. For example...
libs="/Users/Matthew/Downloads/tensorflow_macos/arm64/"

# Replace this with the name of your Conda environment
env="tensorflow"

# The rest should work
pip install --upgrade pip wheel setuptools cached-property six

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/grpcio-1.33.2-cp38-cp38-macosx_11_0_arm64.whl"

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/h5py-2.10.0-cp38-cp38-macosx_11_0_arm64.whl"

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/numpy-1.18.5-cp38-cp38-macosx_11_0_arm64.whl"

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_11_0_arm64.whl"

pip install absl-py astunparse flatbuffers gast google_pasta keras_preprocessing opt_einsum protobuf tensorflow_estimator termcolor typing_extensions wrapt wheel tensorboard typeguard

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/tensorflow_macos-0.1a0-cp38-cp38-macosx_11_0_arm64.whl"

[hoytak]

Thank you for the responses here. We will look into supporting conda more seamlessly in the future.

[JacopoMangiavacchi]

@mwidjaja1 were you able to install Conda on M1 ?

[mwidjaja1]

@JacopoMangiavacchi Yes, just a standard Anaconda installation too if I recall. I don't think I did anything fancy (If I did, it would have been just to throw a arch -x86_64 before any bash command to install so that it compiles in Rosetta)

[jtkiley]

Thank you for the responses here. We will look into supporting conda more seamlessly in the future.

@hoytak, that would be great. I suppose the approach depends whether this is a fork that you're looking to maintain indefinitely or whether you're planning on upstreaming it. Either way, it would be great if the conda solver would use the accelerated version when appropriate so that a single environment file could be both fast and cross platform. That's good practice in general and especially for things like the courses/workshops that I do.

[JacopoMangiavacchi]

Thank you @mwidjaja1, I just found how to install native arm miniconda here https://conda-forge.org/blog/posts/2020-10-29-macos-arm64/

[mwidjaja1]

And I guess on that note, @hoytak, would you consider submitting this version of TensorFlow (as well as any other packages you've made) to conda-forge's new ARM database so anyone can conda install it?

To share @JacopoMangiavacchi's link again: https://conda-forge.org/blog/posts/2020-10-29-macos-arm64/

[JacopoMangiavacchi]

@mwidjaja1 I was able to install it on arm miniconda with the script above BUT the env variable has to contain the path to the conda env and not the name of the env.

[JacopoMangiavacchi]

Then I was able to add other packages such as pandas and notebook via conda but as soon as I start a notebook on the associated ipkernel the NotebookApp crash miserably.

Is anyone able to use notebook on M1 ?

[mwidjaja1]

Thanks for nudging me in the right direction @JacopoMangiavacchi! I forgot you can downgrade to Python 3.8.

I finally got things working with these modifications that takes into account,

1. There's an ARM Miniconda (yay!)
2. There's Python 3.8 there
3. A lot of the packages we need, such as Numpy, is already there (Yay again!)
4. The great work @hoytak did to build TensorFlow for ARM.

So what I did to get things finally working,

1. Download the archive for this repo from https://github.com/apple/tensorflow_macos/releases
2. I downloaded Miniconda from https://conda-forge.org/blog/posts/2020-10-29-macos-arm64/
3. Install Miniconda
4. Once it's installed, create a Python 3.8 env by running conda create --name python38 python=3.8
5. Because it's a new env, there's no way you have any packages like Tensorflow on there already (good) so...
6. Run this script:

# Put a path to where the arm64 libraries are. For example...
libs="/Users/Matthew/Downloads/tensorflow_macos/arm64/"

# Replace this with the path of your Conda environment
env="/Users/Matthew/miniforge3/envs/python38"

# The rest should work
conda upgrade -c conda-forge pip setuptools cached-property six

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/grpcio-1.33.2-cp38-cp38-macosx_11_0_arm64.whl"

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/h5py-2.10.0-cp38-cp38-macosx_11_0_arm64.whl"

# pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/numpy-1.18.5-cp38-cp38-macosx_11_0_arm64.whl"

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/tensorflow_addons-0.11.2+mlcompute-cp38-cp38-macosx_11_0_arm64.whl"

conda install -c conda-forge -y absl-py
conda install -c conda-forge -y astunparse
conda install -c conda-forge -y gast
conda install -c conda-forge -y opt_einsum
conda install -c conda-forge -y termcolor
conda install -c conda-forge -y typing_extensions
conda install -c conda-forge -y wheel
conda install -c conda-forge -y typeguard

pip install tensorboard

pip install wrapt flatbuffers tensorflow_estimator google_pasta keras_preprocessing protobuf

pip install --upgrade -t "$env/lib/python3.8/site-packages/" --no-dependencies --force "$libs/tensorflow_macos-0.1a0-cp38-cp38-macosx_11_0_arm64.whl"