introduction-to-conda-for-data-scientists
introduction-to-conda-for-data-scientists copied to clipboard
carpentries-incubator
Fix for example "Creating a new environment as a sub-directory within a project directory" needed?
Testing the example "Creating a new environment as a sub-directory within a project directory":
project-dir $ conda create --prefix ./env
python=3.6
matplotlib=3.1
tensorflow=2.1
pip=20.0
does not work with the specified tensorflow version on Mac OS X 10.15.7:
PackagesNotFoundError: The following packages are not available from current channels:
When running without specification, tensorflow 1.14.0 gets installed instead.
Interesting. For the past six months I have been teaching these materials online using Binder which runs on Linux. I quickly ran conda search tensorflow on my Mac and the most recent available version for TF is 2.0. Looks like those interested in more recent support for TF on Mac would need to use Pip to install TF. Since TF is not necessary at all for this example it is probably best to just use a different example here.
Any discussion that we might want to add about TF version differences across Mac, Windows, Linux should probably be added in the GPU episode where TF, PyTorch et al are central to the examples.
Thank you for looking into this, I recreated the environment as you proposed (everything as in the original example except tensorflow, which I installed via pip) and - tensorflow==2.3.1 got installed.
I propose that we remove TF as a dependency for this exercise and then defer all discussion of TF envs to the episode on GPUs when we will have already covered pip and Conda channels. Thoughts?