@aseyboldt Thanks a ton.
I may look into an smc sampler. I don’t expect NUTS to actually work since my gradient isn’t very doable, I was just trying to figure out why it was exiting instead of just saying initialization failed and trying other step functions. I think I messed up my environment somewhere on accident because code that was running last week is broken now.
I definitely want to learn more about all you’ve written about, but wrapping my fenics implementation in aesara seems like a quick way to see if pymc v4 will give me fewer issues. Is this something that would be easy for a beginner to take on? Looking at fenics_pymc3, the package is tiny and the author just uses a few functions from theano.tensor, theano.gradient, theano.graph.op and fecr to wrap around fenics. Do you think I would easily be able to find equivalent functions in aesara to do the same thing and try to use pymc v4?
Tightening priors is a great tip, and I don’t understand the reason for the shape=(1,) thing, but it makes fenics_pymc3 work.