Mike Lin

Draft PR: https://github.com/chanzuckerberg/cellxgene-census/pull/1164 (finalization pending #1116)

S3 folder with the final-for-now indexes currently in a private bucket (details on Slack). The folder structure there starts with a 2023-12-15 subfolder (census version). The goal is to copy...

@ebezzi @metakuni To close out this ticket, is there somewhere we're documenting the Census/LTS release process where we could include information about the indexes?

@ebezzi is there a distinct `scib-metrics-config-plot.yaml` to add here?

Agreed -- this will improve once R tiledbsoma has stabilized enough that we can depend on a specified version number of it. Right now we don't cache dependency installation (which...

We're currently still tracking the bleeding edge of `tiledbsoma`, which leads to frequent CI breakage; discussion on how to ensure a stable dependency started: https://github.com/single-cell-data/TileDB-SOMA/issues/1412#issuecomment-1562368924

Lowest-hanging fruit --> https://github.com/chanzuckerberg/cellxgene-census/blob/e708d54dd26348e27d426bc890c13805097462d8/.github/workflows/r-check.yml#L24-L27 Right now all dependencies are built from source on every build. The dependencies action has an automatic caching functionality which we disable because of how we're...

@brianraymor It's still the case that [this build is often slow](https://github.com/chanzuckerberg/cellxgene-census/actions/workflows/r-check.yml?query=branch%3Amain), although the specific low-hanging fruit [mentioned above](https://github.com/chanzuckerberg/cellxgene-census/issues/309#issuecomment-1572647627) was picked. A further improvement we could make would be to add...

@johnkerl @ryan-williams @mojaveazure (and all) we think this has been fixed by https://github.com/single-cell-data/TileDB-SOMA/pull/3252 included in R tiledbsoma 1.15+...does anyone feel otherwise? Thanks!

@ivirshup It looks like the last thing we fixed helped but [broke on the python 3.8 runner only](https://github.com/chanzuckerberg/cellxgene-census/actions/runs/10308890286/job/28537326318#step:7:335), maybe b/c of an older numpy/scipy version?