pytorch / pytorch/executorch

Arm Aarch64: Remove tolerance increase when TOSA ref model is updated

Open
#19,452 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

The TOSA reference model is currently experimental for usage on Aarch64. As such, a patch was made to increase the tolerance by 10% on the architecture (#19110).

This issue aims to keep track of whether the TOSA ref model gets updated, such that the numerical accuracy becomes trustworthy enough to revert the patch.

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell

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 by reviewing the Aarch64 tolerance increase described in this issue and the related patch in #19110, then check whether the TOSA reference model has been updated. Done means numerical accuracy is trustworthy enough to remove the architecture-specific tolerance increase.

Written by the indexing model from the issue text.

Assessment

Domain
embedded-iot, machine-learning
Issue type
Refactor
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.