FEAT: documenting dependency pins

[CBroz1]

Hi! Thanks for all the work y'all put in on this tool!

I manage a package that attempts to orchestrate moseq in combination with tools for other modalities (e.g., ghostipy, spikeinterface, etc.). Dependency conflicts are bound to occur the more tools we support (full discussion here), and it helps to have notes on upstream pins when those conflicts pop up.

Does your team have any notes on the motivations behind hard pins?

    seaborn==0.13.0
    matplotlib==3.8.4
    bokeh==2.4.3
    panel==0.14.4
    holoviews==1.15.4

• Is this the only version that works, or were others not tested?
• Are there any plans to adjust/test for compatibility with newer versions of each?
• What would your team like to see in a PR that made edits? Would comparable output on the notebook suffice?

Thanks!

[JRJacoby]

Hi, thanks for the question! The pins were not done in a particularly principled way, no. We had a period of dealing with cascading dependency conflicts and ended up over-pinning to be safe. If you can get the same results from both the modeling and analysis notebooks with more permissive pins we would actually really appreciate someone doing this work. If this is hard to fit into a PR, we can connect over email so you can send the results of a modeling run with the new pins.