All the errors I see in the post above are about your specific model, not about running the example notebook on your local environment without changing anything. I think you should start there.
Once you have managed to do this, go to the cell where the synthetic data is generated and change the y definition to use a hypersecant distribution. Then update the model to use your custom hypersecant distribution.
If you do that you’ll have a code you know works with a normal distribution with only one change, using a hypersecant instead. If something is wrong with the distribution it will show there, and if it fails you’ll know that what is wrong is the hypersecant distribution, not the data processing, not the adaptation of the prior update code to your situation (remember, only change the y and Y_obs variables, same x, same sigma, same mu…)