Ami Falk
Ami Falk
Currently, `constraints.greater_than_eq` is supported in [torch](https://pytorch.org/docs/stable/distributions.html#torch.distributions.constraints.greater_than_eq), but we only have `constraints.greater_than` in numpyro. Can we add an alias for `greater_than_eq` with something like `constraints.interval(x, jnp.inf)` to the constraints namespace (and...
Here's a reproducible example that's taken nearly directly from the Gaussian Mixture Model tutorial. The AutoContinuous guide seems to be the failure mode. ```python import jax.numpy as jnp import jax.random...
Minimal example: ```python import jax import jax.numpy as jnp from jax.random import PRNGKey import numpyro import numpyro.distributions as dist from numpyro.infer import MCMC, NUTS, Trace_ELBO, SVI from numpyro.infer.reparam import NeuTraReparam...
For mixture models with arbitrary distributions over each feature, sampling currently must be done serially, even though these operations are trivially parallelizable. To sample priors from a hierarchical mixture model...
Greetings! I've ported a subset of zeus functionality to the [NumPyro](https://github.com/pyro-ppl/numpyro) project under the sampler name [ESS](https://num.pyro.ai/en/latest/mcmc.html#numpyro.infer.ensemble.ESS). (For the uninitiated, NumPyro uses JAX, a library with an interface to numpy...
Greetings! I've ported a subset of emcee functionality to the [NumPyro](https://github.com/pyro-ppl/numpyro) project under the sampler name [AIES](https://num.pyro.ai/en/latest/mcmc.html#numpyro.infer.ensemble.AIES). (For the uninitiated, NumPyro uses JAX, a library with an interface to numpy...
I'm currently working on a project where I'm partially conditioning a large set of latent variables. Here's a toy example: ```python dat = dist.Normal(10, 1).sample(random.PRNGKey(0), (100,)) mask = jnp.concatenate([jnp.ones(50, jnp.bool_),...
Are there any plans to support this in orbax? Tensorstore can interpret strings: https://google.github.io/tensorstore/python/api/tensorstore.string.html I realize I can pull out the object arrays into a json file and then stitch...
Sometimes `nmap`'ed computations don't all fit in memory at once and there are not enough devices to shard the computation over (this is limitation is particularly salient when using penzai...