Skip to content

✨ Add support for compiling dynamic shapes #230

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

Closed

Conversation

gkumbhat
Copy link
Contributor

Changes

Since compiler now supports symbolic shapes, we can enable passing these to compiler. This will help in less shapes munging and tracking upfront and allow more utilization of compilation cache.

Signed-off-by: Gaurav-Kumbhat <Gaurav.Kumbhat@ibm.com>
Copy link

👋 Hi! Thank you for contributing to vLLM support on Spyre.
Just a reminder: Make sure that your code passes all the linting checks, otherwise your PR won't be able to be merged. To do so, first install the linting requirements, then run format.sh and commit the changes. This can be done with uv directly:

uv sync --frozen --group lint --active --inexact

Or this can be done with pip:

uv pip compile --group lint > requirements-lint.txt
pip install -r requirements-lint.txt
bash format.sh

Now you are good to go 🚀

@joerunde
Copy link
Collaborator

I think the goal here is to do this with continuous batching, and that's close to a working implementation on spyre cards but everything's not quite all released yet

@gkumbhat
Copy link
Contributor Author

@joerunde I think this one relies on the dynamic tkv symbolic shapes, which was released in compiler sometime back. I am working on testing this out, will update the PR as per my findings.

@joerunde
Copy link
Collaborator

Gotcha, I just meant that our team isn't spending any of our time on bringing improvements to the static batching path

@joerunde
Copy link
Collaborator

closing, won't be supported

@joerunde joerunde closed this Jun 19, 2025
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.

None yet

2 participants