Pip install of tensorflowjs not working due to apparent version conflict with tensorflow

Question

I'm currently in a conda virtual environment trying to pip install tensorflowjs, but every time that I do it gets stuck trying to find a version of tensorflow or Numpy which is compatible with tensorflowjs and eventually dies in the process. My command:

pip install tensorflowjs

I want to run a conversion using tensorflowjs so I am not attached to doing it in python if there is a better way. But otherwise what is the best way to go about installing tensorflowjs.

There also seems to be no way of doing it with a conda install as the package does not seem to exist in conda-forge.

Answer 1

Downgrading to python 3.8 has solved the problem. As of February 22nd 2021 tensorflow is not compatible with python 3.9.

Answer 2

I installed the latest version of tensorflowjs in a Python 3.6 environment and checked the dependencies:

tensorflowjs 3.1.0 Python Libraries and Tools for TensorFlow.js
├── h5py >=2.8.0,<3
├── six >=1.12.0,<2
├── tensorflow >=2.1.0,<3
└── tensorflow-hub >=0.7.0,<0.10

tensorflow itself is a dependency of tensorflowjs . So which version of tensorflowjs do you use?