inference icon indicating copy to clipboard operation
inference copied to clipboard

Published 20 hours ago • verified publisher icon mlcommons

Fixes #1249

Open arjunsuresh opened this issue 3 years ago • 1 comments

This PR enables one to run bert reference implementation using onnxruntime backend with custom model, dataset and log paths and also supports the usage of Nvidia GPUs with onnxruntime version >= 1.9

arjunsuresh avatar Oct 11 '22 10:10 arjunsuresh

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

github-actions[bot] avatar Oct 11 '22 10:10 github-actions[bot]

Can the new behaviours be documented somewhere please?

psyhtest avatar Oct 18 '22 14:10 psyhtest

The related issue is added here

arjunsuresh avatar Oct 18 '22 14:10 arjunsuresh

@rnaidu02 Please merge this

arjunsuresh avatar Oct 18 '22 17:10 arjunsuresh

@psyhtest Please review the change. The original comment was confusing as by default all optimizations are enabled in onnxruntime. I'm now disabling the highest optimization level just for aarch64 architecture due to an accuracy issue.

arjunsuresh avatar Oct 21 '22 06:10 arjunsuresh

@arjunsuresh It seems all the changes of this PR have been merged. Can we close this PR?

pgmpablo157321 avatar Oct 28 '22 21:10 pgmpablo157321

Thank you @pgmpablo157321

arjunsuresh avatar Oct 28 '22 21:10 arjunsuresh