Aesara fails to compile graph for model that runs on pymc3

Apologies for the delay. I have finally opened that PR Implement Truncated distributions by ricardoV94 · Pull Request #6113 · pymc-devs/pymc · GitHub

Anyway you can implement a truncated distribution by adding an interval transform so it doesn’t sample out of the truncation bounds and a Potential with the needed normalization constant. If it’s only for the likelihood you don’t need the transform.

There are some examples here: Google Colab which came from Help with pm.Potential - #12 by sspickle