ONNX Probabilistic Programming Working Group – PyMC Community Involvement

News
,
1

Hi everyone,

A new ONNX Working Group focused on probabilistic programming has recently been formed, and we’d love to involve members of the PyMC community.

The goal of this effort is to bring probabilistic modeling and Bayesian inference capabilities into the ONNX ecosystem by defining a standardized operator domain and runtime semantics for probabilistic computation.

Areas we’re exploring include:

  • Distributions and log-probability operators

  • Bijectors and parameter transformations

  • Reproducible stateless RNG semantics

  • Special mathematical functions for probabilistic models

  • Inference algorithms (HMC, NUTS, SMC, Laplace, INLA)

  • Export pathways for frameworks such as PyMC, Stan, Pyro, NumPyro, TensorFlow Probability, JAX-based systems, BayesFlow, and Julia/Turing

One of the goals is to allow probabilistic models to be exported and executed across frameworks and hardware using ONNX as a portable representation.

We would really value input from the PyMC community as we design these semantics and operators.

If you’re interested in participating or contributing feedback, feel free to reach out to:

You are also welcome to attend our working group meetings:

Fridays @ 12 PM EST, every two weeks

Working group repository:

https://github.com/onnx/working-groups/tree/main/probabilistic-programming