-
-
Notifications
You must be signed in to change notification settings - Fork 11.1k
[Bugfix] Don't set an upper bound on repetition penalty #16403
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
[Bugfix] Don't set an upper bound on repetition penalty #16403
Conversation
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
|
👋 Hi! Thank you for contributing to the vLLM project. 💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can either: Add 🚀 |
Co-authored-by: Nick Hill <nhill@redhat.com> Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
bd81d8b to
5b142e6
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @alex-jw-brooks
…#16403) Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Nick Hill <nhill@redhat.com> Signed-off-by: Yang Wang <elainewy@meta.com>
…#16403) Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Nick Hill <nhill@redhat.com>
…#16403) Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Nick Hill <nhill@redhat.com>
…#16403) Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com> Co-authored-by: Nick Hill <nhill@redhat.com> Signed-off-by: Mu Huai <tianbowen.tbw@antgroup.com>
This is a small fix to allow any non-negative values of repetition penalty to align with the transformers implementation (ref) instead of forcing (0, 2]! Not sure why this isn't allowed at the moment, maybe because of validation for allowed ranges for frequency/presence penalties based on openai docs being copied 🙂