Fritz Obermeyer

cc @neerajprad who is refactoring Pyro's HMC and NUTS to support this

@martinjankowiak as we discussed today, Funsor is currently incorrect when evaluating discrete+continuous mixtures. One possible solution path is to generalize log density to nonstandard reals, either by an `Infinite` funsor...

On further thought, I believe we could implement this as metadata on each funsor (similar to `.bound` and `.fresh`) consisting of bounds on the order of infinity, a tuple of...

@eb8680 @lawmurray @martinjankowiak any ideas how to best do this in Funsor?

I'm unsure whether this is still relevant, and I'd be fine closing this. I haven't minded the `singledispatch` solutions to special casing. I'm also concerned that being more abstract will...

@fehiepsi did you already finish this?

@fehiepsi no hurry if you have other things to do, I'm just trying to close out stale tasks.

Note that this is low priority, since even under the current objective function, the optimizer should do something reasonable for most real examples.

Hi @rafaol, tl;dr don't use the jit, it isn't useful. I have found that `JitTrace_ELBO` only works on very simple models, and provides little or no speed improvement. I believe...

Hi @kaltinel just to clarify, can you confirm that you are calling `pyro.set_rng_seed(...my_seed...)` before running `predictive(...)` in each case? Also, is there a reason you're loading `adam` and creating an...