Skip to content

scripts : exit compare-llama-bench.py gracefully when there's nothing to compare #13451

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 11, 2025

Conversation

CISC
Copy link
Collaborator

@CISC CISC commented May 11, 2025

I stumbled upon this accidentally while testing JSONL (coming soon). I had used two different paths to the same model in consequent tests, causing get_rows not to be able to join the two results.

@CISC CISC requested a review from JohannesGaessler May 11, 2025 10:44
@CISC
Copy link
Collaborator Author

CISC commented May 11, 2025

I wonder though, sometimes you want to compare apples to oranges (like two different models, or at the very least various settings), so it might make sense to broaden the options for comparison in the future?

@github-actions github-actions bot added script Script related python python script changes labels May 11, 2025
@CISC CISC merged commit 0923237 into master May 11, 2025
5 checks passed
@CISC CISC deleted the cisc/compare-llama-bench-no-data branch May 11, 2025 14:20
@JohannesGaessler
Copy link
Collaborator

The script could definitely use more work in general; I wrote the script for myself according to my own needs and while I did contribute it upstream the needs of other users were a secondary concern for me. I think that for people not working on the code the comparisons between commits are maybe not as interesting. So I agree that more options for what to compare would be useful to have.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
python python script changes script Script related
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants