Jesse Grabowski

This isn't a problem with pymc-experimental per se: `nutpie` isn't supported for scan-based models. This is a known limitation, and has to do with how random number generators are currently...

Following up here with the conversation I had with @ricardoV94 on discourse so it doesn't get lost -- Another nice transformation would be one that takes in an MvN distribution...

> About the Cholesky thing we added some rewrites recently in PyTensor that work as long as the right "tags" are on the variables (e.g. "lower_triangular". You can check it...

I want to shill the work I did on numba links for several linear algebra operations again: https://github.com/numba/numba-scipy/commit/462191b4f745ed260056c534e28e8e0ba1a743a5 A core problem is that the `numba.np.linalg` module simply doesn't have links...

It's an outstanding PR that didn't garner any attention. I have no idea the best way to integrate it. I use this code in one of [my own projects](https://github.com/jessegrabowski/gEconpy/tree/main/gEconpy/numba_linalg), where...

@velochy We just merged https://github.com/pymc-devs/pytensor/pull/614, which lets us now use `pt.linalg.norm` inside of pymc models. We can now fix this bug by defining a new transformation for LKJCorr. Basically we...

Hi @johncant , let me know if anything isn't clear about how to proceed

Yes I think so. Is there a rewrite to look at that I could pattern a PR after?

all the `progressbar` kwargs could default to a global variable that could be changed after importing if desired?

@alvaropp sorry for letting this get stale, I am going to do a careful review now/this weekend so you can get back on track and merged in ASAP