Something changed in `pytensor > 2.12.3` (and thus `pymc > 5.6.1`) that makes my `pytensor.gradient.grad` call get stuck - any ideas?

Graphviz fails but that’s fine and expected. During sampling those inputs are exactly the same ones used by the logp function.

Just saying it’s not a sign the model is wrong, just that graphviz is dumb and doesn’t know what to do with graphs that have input variables (i.e. logp graphs)

There’s no way for you to provide the inputs needed for graphviz

1 Like