Bill Engels

What about just keeping the `_base_dist` idea like you have, and have `nu` be a optional kwarg to `HSGP`? Then `_base_dist` can either be `StudentT` if `nu` is not None...

Breaking the do-operator puts a bit more urgency on this idea now

I'm wondering if rewriting the GP lib as pytensor ops is the only way forward here. It's probably easier to make `do` and `observe` work with existing GPs than the...

Am really interested in particular to the LA + HMC option, seems like the best “fit” within the rest of PyMC. It would be amazing to specify which parts of...

thanks @AlexAndorra! Also very open to splitting this up -- it is pretty long, sorry. That might make it easier to review.

> Perhaps emphasize in the text what you are looking for in the gram plots (the bright yellow diagonal)? Added this text: > The plots above compare the approximate Gram...

It'd be interesting to see how they compare

thanks for picking this up!

This is looking good, though I still need to check out your branch and make sure. It looks like pre-commit isn't happy though. [This page](https://www.pymc.io/projects/docs/en/latest/contributing/python_style.html) should show you how to...

couldve swore I approved and merged this... my fault!