search

nestauk / ds-cookiecutter

A data science cookiecutter for Nesta projects.
https://nestauk.github.io/ds-cookiecutter/
MIT License
8 stars 3 forks source link

Move scientific packages dependencies to `environment.yaml` #109

Open georgerichardson opened 3 years ago

georgerichardson commented 3 years ago

Installing packages such as scipy and numpy using pip results in incompatbilities for M1 macs. Could we consider moving these to environment.yaml for now?

bishax commented 3 years ago

@georgerichardson see questions raised in https://github.com/nestauk/industrial_taxonomy/pull/1#discussion_r755286231

bishax 26 minutes ago

If I rebuild my environment with these changes and inspect whether my numpy and scipy come from pip or conda I get that they come from pip...

$conda list | grep "numpy|scipy"

numpy 1.21.4 pypi_0 pypi scipy 1.7.2 pypi_0 pypi

(I think) because e.g. pandas in requirements.txt will get those again from PyPi

Can you confirm the same result?

bishax 24 minutes ago

I think what might need adding to environment.yaml to make scipy work for you is openblas or mkl.

bishax commented 3 years ago

As seen in nestauk/industrial_taxonomy#1 we only need to add openblas to environment.yaml?