Error when installing on fresh conda env and apple macos silicon m1

[scalastic]

Hi,

On an apple m1, I made sure to install tensorflow 2.7 on a new conda virtual environment:

% conda install -c apple tensorflow-deps==2.7.0
% pip install tensorflow-macos==2.7.0
% pip install tensorflow-metal 

But when I tried pip install nebullvm I get this error:

ERROR: Cannot install nebullvm==0.1.0, nebullvm==0.1.1, nebullvm==0.1.2, nebullvm==0.2.0, nebullvm==0.2.1 and nebullvm==0.2.2 because these package versions have conflicting dependencies.

The conflict is caused by:
    nebullvm 0.2.2 depends on tensorflow<2.8.0 and >=2.7.0
    nebullvm 0.2.1 depends on tensorflow<2.8.0 and >=2.7.0
    nebullvm 0.2.0 depends on tensorflow<2.8.0 and >=2.7.0
    nebullvm 0.1.2 depends on tensorflow<2.8.0 and >=2.7.0
    nebullvm 0.1.1 depends on tensorflow>=2.7.0
    nebullvm 0.1.0 depends on tensorflow~=2.7.0

To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict

ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts

What am I missing ? Thanks

[diegofiori]

Hello @scalastic! Try installing nebullvm using

pip install nebullvm==0.2.2

I will try to investigate the behavior further and correct it before the next release.

[scalastic]

Hello @morgoth95 This has only change the message :(

% pip install nebullvm==0.2.2
Collecting nebullvm==0.2.2
  Using cached nebullvm-0.2.2-py3-none-any.whl (70 kB)
Collecting onnx>=1.10.0
  Using cached onnx-1.11.0.tar.gz (9.9 MB)
  Preparing metadata (setup.py) ... done
Collecting tf2onnx>=1.8.4
  Using cached tf2onnx-1.9.3-py3-none-any.whl (435 kB)
Collecting py-cpuinfo>=8.0.0
  Using cached py-cpuinfo-8.0.0.tar.gz (99 kB)
  Preparing metadata (setup.py) ... done
Requirement already satisfied: numpy>=1.19.0 in /opt/homebrew/Caskroom/miniforge/base/envs/nebulvm-stf/lib/python3.8/site-packages (from nebullvm==0.2.2) (1.19.5)
ERROR: Could not find a version that satisfies the requirement tensorflow<2.8.0,>=2.7.0 (from nebullvm) (from versions: none)
ERROR: No matching distribution found for tensorflow<2.8.0,>=2.7.0

[diegofiori]

Looks like nebullvm is not finding any valid tensorflow installation. Which conda version are you using? On MacM1 I usually use mini-forge where I can simply install tensorflow running

conda install tensorflow

After installing tensorflow you can proceed with the installation of nebullvm

[scalastic]

I'm using mini-forge also but I've installed tensorflow-macos, the one from Apple. It seems not to be recognized as tensorflow as it's named tensorflow-macos...

When I now execute conda install tensorflow, it offers me to install tensorflow (once again). And nebullvm is installing now !

Ok, that was the point: I thought I could benefit from both worlds, the Apple's version of Tensorflow and Nebullvm... Why not after all ?