Skip to content

Conversation

@gbyu-amd
Copy link
Contributor

@gbyu-amd gbyu-amd commented Jan 5, 2026

Motivation

Create a recipe folder to put the recipes for different models, so that users know how to deploy models with ATOM on AMD GPUs, making it much easier to reproduce the best performance data.

Add recipe for Qwen3 235b first in this pr.

Submission Checklist

@gbyu-amd gbyu-amd requested review from valarLip and wuhuikx January 5, 2026 09:56
valarLip
valarLip previously approved these changes Jan 5, 2026
--save-result --result-dir=${result_dir} --result-filename=$RESULT_FILENAME.json \
--percentile-metrics="ttft,tpot,itl,e2el"
```
The performance number for both 4 and 8 ranks are provided as a baseline:
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please help label the performance here is just for reference and provide the AITER/ATOM commits for this performance.

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please also provide the bkc to replace the AITER/ATOM commits in the docker so that everyone can reproduce our results.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updated recipe.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants