Du Phan

You are right. `N_households` needs to be the size of unique numbers in `hidB` and `hidA`. We also need to make sure that `hidB` and `hidA` ranges from 0 to...

Sure, maybe modifying directly at `N_households=max(... of kldyas)` is cleaner. Could you help me make the change? Regarding LKJ, the cholesky version is recommended if you need speed or face...

My impression is most of the details can be preserved (e.g. block vector, block matrix, align gaussian). Back then, one issue was batch qr is very slow on GPU, but...

Looking at the code, I guess we need to triangulate a non-positive-definite precision matrix (e.g. zeros matrix) but I can't recall when we need such triangularization. :( Probably, it is...

Nice solution! I tested with abbreviation and it seems to work. Together with `yapf`, the result is nice. ```python from funsor import Reals, Tensor, Variable, ops, testing from funsor.delta import...

This looks interesting and helpful to understand those `eager.register` logics in `distributions`. I will try to add a few patterns more.

@fritzo Is there a way to apply an op to a variable name? For gamma normal conjugacy, we have gamma prior for "variance", then I want to take square root...

Thanks, @fritzo! It works in NumPyro. I'll add the GammaNormal one soon.

@fritzo I have some discussions with @martinjankowiak previously about MVStudentT. Hope that it already settles all the math. I will see if it works in a few days (currently I...

This looks non-trivial but seems doable and interesting. I'll take care of implementing `studentt` but it might take a bit of time for me to settle all the math (especially...