It may be interesting to experiment with mamba as an alternative to conda to spawn conda environment. Memory consumption and build times should be greatly reduced.
FYI, there was a blog post released about mamba recently, announcing the future plans for the project.
https://medium.com/@QuantStack/open-software-packaging-for-science-61cecee7fc23
Hi @SylvainCorlay! Thanks for sharing this. We already did some tests with Mamba and we got some good results (see #6815). We started doing that because we had memory issues with conda. However, now we have migrated to bigger servers and we are happy with the results. You can read more about this at https://blog.readthedocs.com/better-conda-support/
I'm still interested on merging the PR I linked here eventually, since it's under a feature flag and there are some improvements that I'd like to have, but since we don't have the problem anymore I prefer to keep it closed for now to avoid maintaining more code if it's not necessary.
@humitos FYI MyBinder has switched to mamba, and it appears to be working well so far!
New micromamba docs: https://github.com/mamba-org/mamba/blob/master/docs/source/micromamba.md
Most helpful comment
@humitos FYI MyBinder has switched to mamba, and it appears to be working well so far!