Why tensor-notebook use pip instead of mamba?

JupyterHub Zero to JupyterHub on Kubernetes
1

Hello, members.

This page says
https://jupyter-docker-stacks.readthedocs.io/en/latest/using/recipes.html#using-mamba-install-recommended-or-pip-install-in-a-child-docker-image

Using mamba install (recommended) or pipinstall in a Child Docker image

But, tensorflow-notebook use pip instead of mamba. why?

I think tensorflow build by mamba optimize to specific hardware. so It’s fast on the specific hardware. but, it doesn’t run other hardwares. that’s why tensor-notebook use pip.
Correct? Any other reason?

# Install Tensorflow with pip
RUN pip install --no-cache-dir tensorflow && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Best regards.

1 Like
2

There’s some discussion of this in the PR that made this change: https://github.com/jupyter/docker-stacks/pull/1844. (FWIW I found this just by looking at git blame in that repo.)

But I think the short answer is that at some point, Install TensorFlow with pip said the following:

Do not install TensorFlow with conda. It may not have the latest stable version. pip is recommended since TensorFlow is only officially released to PyPI

(I don’t see this wording now, though, but I think the last point is still true…)

2 Likes
3

Hello, @rawlins. Thank you for your reply. I understood.

1 Like