search

sgkit-dev / sgkit

Scalable genetics toolkit
https://sgkit-dev.github.io/sgkit
Apache License 2.0
235 stars 32 forks source link

Use conda `environment.yml` file for setting up the environment #511

Open aktech opened 3 years ago

aktech commented 3 years ago

At the moment we use requirements.txt for creation of environment for development and usage. Although its quick to get started with pip, I think its a good idea to start using environment.yml for a more consistent environment. This is specially useful when installing GPU related libraries like cupy and cudatoolkit, which are non-trivial to install via pip. Also we have msprime as a development requirement, which depends on GSL, which is not possible to install via pip, but trivial to install via conda.

cc @jeromekelleher @tomwhite @ravwojdyla @hammer thoughts?

jeromekelleher commented 3 years ago

It's probably good practise to have one defined so that things like binder will work. I wouldn't say we change our current requirements.txt setup though.

Re msprime, it's possible to install it without GSL now if we use --pre to install an alpha version. The full release is due in 2 weeks, and we can clear out some cruft in CI once that's done.

hammer commented 3 years ago

cc #427 I think