Rob Zinkov

Accessing params in SVI

1

Right now, there isn't a great way to access the params directly associated with a given guide program. Is this a deliberate design choice, or do we want to expose...

The `recover_marginals` method in `MarginalModel` would benefit from a progress bar

Support batched constant arguments when marginalising `DiscreteUniform`

1

Right now, models like the following fail: ```python with MarginalModel() as m: mu = pm.DiscreteUniform("mu", lower=[0,10,20], upper=40) y = pm.Normal("y", mu=mu, sigma=1.0) m.marginalize([mu]) ``` Since `lower` here takes all constant...

It would be really to support complex numbers in pytensor. The feature gets [requested](https://discourse.pymc.io/t/matrix-based-predictions-and-complex-variables/13596/6) on and off. There likely needs to be some discussion over how it should be implemented....

Migrate progress bar from fastprogress to tqdm

1

Moves to use tqdm along with adding support for multiple progress bars Makes blackjax suitable for running multiple chains in parallel.