Unable to install on M1 chip

[quentinblampey]

Hello,

Thanks for releasing your postprocessing tool.
I was unable to directly install vpt with pip, since the tensorflow version you use has some issues on macos (probably an issue specific to the M1 chip).

Note that this version can be installed by running pip install tensorflow-macos==2.9.1. Thus, I managed to install vpt, but I had to clone the repo and update your pyproject.toml file to use tensorflow-macos instead of tensorflow.

I see two possible ways to fix it, if you want I can help you find out if it works:

• Maybe the most recent versions of tensorflow can be directly installed with pip, without using tensorflow-macos?
• Make it possible to install vpt with conda?

Note that I also couldn't pull the docker image, but I guess this is normal since your Dockerfile starts with FROM ubuntu:22.04

Thanks

[QBHuynh]

Hi,

Thank you for submitting this concern with associated work flow on getting it to function.

We'll work on assessing the best way to ensure functionality with the M1 chip and MacOS in a future release.

If you run into any more issues, please let us know!

[munizmom]

Hi Quentin,
I have a M1 Mac. For me the installation setting up a Conda environment with 3.8 < python version >3.11 and the poetry installation worked fine 4 days ago using python 3.9.
You just need to git clone the vizgen-postprocessing GitHub repository
cd /path/{vizgen-postprocessing}
conda activate myenvPy3.9
poetry install
And then the vpt functions became available to use.
Hope it helps!

Best,
Mar

[quentinblampey]

Hi Quentin, I have a M1 Mac. For me the installation setting up a Conda environment with 3.8 < python version >3.11 and the poetry installation worked fine 4 days ago using python 3.9. You just need to git clone the vizgen-postprocessing GitHub repository cd /path/{vizgen-postprocessing} conda activate myenvPy3.9 poetry install And then the vpt functions became available to use. Hope it helps!

Best, Mar

Hello,
I was actually testing the installation with pip, and I didn't try the poetry installation (actually, I expected the same error to happen with poetry). So, good to know it works!
For now I stick with my current environment since my workaround was working, but maybe at some point I will switch to poetry.