Daniel Ward

Deviance NaN from first iteration onwards with Gamma distributed GLM with LogLink

3

I am trying to run a Gamma distributed GLM with a LogLink function, however, the deviance and diff.dev are both NaN from the first iteration. I have added the data...

Block Neural Autoregressive Flow density not properly normalised

1

Hi all, I believe there is an issue with ``BlockNeuralAutoregressiveTransform``s not forming properly normalised densities when the inverse is used to transform a distribution (the same issue as we have...

Abstract{Class}Var does not allow string annotations

4

This also becomes an issue with ``from __futures__ import annotations`` for example ```python from __future__ import annotations import equinox as eqx from equinox import AbstractVar class AbstractFoo(eqx.Module): my_int: AbstractVar[int] ```

Added [flowjax](https://github.com/danielward27/flowjax), a distributions and normalizing flow library built on top of equinox. It's much more up to date and better maintained than the other normalizing flow package listed here...

Sphinx doctest fails for a couple reasons. When running doctest 1. Sphinx seems to stringify types which is incompatible with equinox (cannot use stringified abstract annotations). 2. We set ``typing.GENERATING_DOCUMENTATION...

get_model_relations and get_dependencies give UnexpectedTracerError on seeded models

1

I'm assuming these functions are expected to work with seeded models, but they yield ``UnexpectedTracerError``s. ```python import numpyro import jax import jax.random as jr import numpyro.distributions as dist from numpyro.infer.inspect...

feature: Support documentation for the __init__ of equinox modules.

3

Apologies if I have missed a setting. I cannot find an easy way to document the `__init__` method for equinox modules, which uses the same default generated `__init__` methods as...