Emily (Miller) Dorne

> Is there anything else that needs to happen before I merge? Are the steps I took (in the PR description) the right ones to make sure all relevant files...

It seems like our options are: - use `macos-13` or earlier instead of `macos-latest` ("the new [M1] runner operates exclusively on macOS 14" ([source](https://github.com/actions/runner-images/issues/9254))) -> EDIT: this looks out of...

Think it makes sense to use homebrew to not be stuck on an old macos version and ultimately, an extra minute won't make much of difference given how slow the...

Looks like the PR to fix the `nbclassic` issue got merged in (https://github.com/jupyter/nbclassic/pull/137) which, if I'm understanding thing correctly, fixes the bug that @klwetstone originally found.

I'm good with this change though don't have strong opinions either way. > This makes the repo home page look a lot better, and a user can more reasonably get...

> What is df.audio_string.path? `df.audio_string` is a series with dtype object. I was trying to join an s3path with a series and following the instructions in limitation 3 because we...

We should also use tmp_dir throughout rather than overwriting actual test assets (as we currently do in `test_cli_experiment`)

Worth a quick investigation (~1 hour) to see if there's anything more we can learn

This appears to be caused by projection issues. To reproduce take a tif where the no data in bounds error occurs and reproject to the CRS of the lat lon...

Closing as this doesn't fix the issue. Each video should only be assigned to one set (train / val / holdout). The issue is that when we have multi-species videos...