pymc-devs/pymc

Refactor `EmpiricalGroup` to take `InferenceData` inputs

Open

#5,884 opened on Jun 12, 2022

View on GitHub
 (0 comments) (0 reactions) (0 assignees)Python (1,902 forks)batch import
VIhelp wantedtrace-backend

Repository metrics

Stars
 (7,926 stars)
PR merge metrics
 (Avg merge 11d) (7 merged PRs in 30d)

Description

To decrease our dependence on MultiTrace, and to improve support for workflows were InferenceData objects are saved to disk, the EmpiricalGroup approximation should be refactored.

Currently it takes only MultiTrace inputs, but that's just because it needs access to transformed posterior draws. Since the approximation is created inside a modelcontext, it should be no problem to get the transforms from the Model and apply them to untransformed posterior draws.

Steps

  1. Write a function to transform posterior draws given a pm.Model instance and an xarray.Dataset (e.g. idata.posterior) containing untransformed draws.
  2. Refactor EmpiricalGroup to work with that xarray variable instead of a MultiTrace
  3. Maintain backwards-compatibility (with a deprecation warning) by automatically converting MultiTrace inhputs to InferenceData internally

Contributor guide

Research direction
Write a function to transform posterior draws given a pm.Model and an xarray.Dataset, then refactor EmpiricalGroup to use that function, maintaining backwards compatibility with a deprecation warning for MultiTrace inputs.
Tech stack
python
Domain
backend
Issue type
Refactor
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
2
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonPyMCxarray
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
60