pyro-ppl / pyro-ppl/numpyro

Support for Empirical distributions

Open
#685 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement
Dominant language
Python
Stars
2.8k
Forks
315
Avg merge
3d 9h
Merged PRs (30d)
27

Description

Hi,

I have started implementing an Empirical distribution by porting the pyro one.
Kind of unsure how to do the log probability, I actually think the implementation in pyro might be a bit problematic for limited memory GPUs as it artificially creates a N * N tensor where N is the number of diracs in the distribution.

Does anyone have a better idea?

Code as it stands for now:
https://colab.research.google.com/drive/171oY6j290HSRI2NR9rDmVDBfJB2UepKG?usp=sharing

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the linked Colab implementation and the Pyro empirical-distribution implementation it ports. Investigate how log probability should be computed without the reported N × N tensor concern, then clarify the intended API and behavior with maintainers. Done means an agreed, working Empirical distribution implementation with its memory constraints addressed.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.