Skip to content
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

Add support for stop words in TRTLLM #2678

Merged
merged 18 commits into from
Oct 25, 2024
Merged

Add support for stop words in TRTLLM #2678

merged 18 commits into from
Oct 25, 2024

Conversation

mfuntowicz
Copy link
Member

@mfuntowicz mfuntowicz commented Oct 22, 2024

[MERGE AFTER #2357]

This PR attempts to read the eos_token_ids in generation_config.json (if present) and creates a list (std::list required by trtllm) to cache those value.

The list is then forwarded to TRTLLM's executor during request submission.

Known issues:

  • stop words added at request time through stop_sequences are not yet supported

@mfuntowicz mfuntowicz requested review from Narsil and Hugoch October 22, 2024 08:15
@Narsil Narsil force-pushed the trtllm-stop-words branch from 9cee00e to 84f14a1 Compare October 25, 2024 08:21
@Narsil Narsil merged commit ba5fc7d into main Oct 25, 2024
9 of 10 checks passed
@Narsil Narsil deleted the trtllm-stop-words branch October 25, 2024 08:58
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.

2 participants