Margus Niitsoo

Ran into the same issue. Problem is the urllib v. 1.26.7, as deepl expects v2. In my case, the old urllib is due to s3fs, which does not work with...

How would one use it?

Yes that works. Both for the toy example as well as the real model. Any downsides to doing that?

Well there seems to be one. It is now throwing errors if I add initvals to the models. Any workarounds for that?

``` import pymc as pm from pymc.sampling import jax as pm_jax import pytensor.tensor as pt import numpy as np from pymc.model.transform.optimization import freeze_dims_and_data obs = np.array([ [1,0,1,0,1,0], [0,1,1,0,1,0], ]) ns...

ok. Maybe it makes sense to deprecate initval parameter on the RVs then?

Um. Resistance to removing something that no longer works? Or do I misunderstand something?

and jax sampler only works for frozen dims if we want more complex pytensor manipulations? So by implication jax sampler only works for complex models if you dont use initval...

Ok. I guess I have my answers, and you are right, I have all the tools needed to make it work. Thank you for your thorough answers @ricardoV94

Can't you just set 'discreteHeight':{'step':20} instead of 350 in the theme? Overriding that setting to that value with .configure_view() fixes the display for me, and it should leave the other...