Skip to content

Conversation

@ggerganov
Copy link
Member

@ggerganov ggerganov commented May 13, 2025

The tool was incorrectly ordering the tokens for the prompt processing batches by interleaving the sequences like this:

123412341234...

Instead the batch should look like this:

1111...2222...3333...

This is important for correctly benchmarking some attention optimization techniques.

@ggerganov ggerganov merged commit b89d605 into master May 13, 2025
44 checks passed
@ggerganov ggerganov deleted the gg/batched-bench-fix-pp branch May 13, 2025 15:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants