pytorch / pytorch/executorch

[Delegate Testing] Determine tolerance / numerical accuracy validation strategy

Open
#13,347 7 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

We need to figure out how to best test numerical accuracy. Delegates can do many things, like run models in reduced precision (fp16, commonly) or quantize. We want to avoid hitting false positives, but we do want to ensure that we catch correctness issues.

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

No files, tests, or entry points are named. Start by examining the delegate testing workflow and define validation criteria for reduced-precision and quantized models that avoid false positives while still detecting correctness issues.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, testing-qa
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.