Reuben

Thanks! I'm having one issue: ```python def m(): mu = numpyro.sample('mu', dist.Normal(3, 1)) nu = numpyro.sample('nu', dist.Normal(mu+1, 2)) rng_key = jax.random.PRNGKey(0) rng_key, init_key = jax.random.split(rng_key) init_params, potential_fn_gen, *_ = initialize_model(...

@turion Somewhat relatedly, do you know if there's a minimal example anywhere of how I can combine a GlossEventClock in parallel with another clock, so that I can interact with...

Yes, fixed!

Yep, I think that's the thing to do in the tuning. There's some code in the original for that purpose. Re. adding the preconditioning back in, I'm unsure what the...

OK, I'll look into that. For now I might proceed by working in this repo, and then we can discuss splitting it out down the road based on the PR....

OK, having investigated a little more, some notes (for myself, mainly). What I want is a function type `InferenceGymModel -> ArrayOfSamples`, i.e. a function that takes an inference gym model...

Some todos ad Divij and I work on this: - [ ] add NUTS as an inference method to SamplingAlgorithms - [ ] work out the best way to add...

I think those parameters correspond to the MSCOCO (coco) and Visual Genome (vg) datasets. I think the general approach of this paper would work with more modern pre-trained models too.

No reason that comes to mind, although do you have a specific example use case in mind? I've certainly used an ExceptT transformer to a similar effect.

```hs newtype Population m a = Population (Weighted (ListT m) a) ``` `Weighted` is the `State` monad. (By the way, it's possible that one of the optimized Writer or accum...