-
Notifications
You must be signed in to change notification settings - Fork 176
feat: Added Pre schedule plugin API #1682
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
base: main
Are you sure you want to change the base?
Conversation
Signed-off-by: Shmuel Kallner <[email protected]>
Signed-off-by: Shmuel Kallner <[email protected]>
Signed-off-by: Shmuel Kallner <[email protected]>
Signed-off-by: Shmuel Kallner <[email protected]>
Signed-off-by: Shmuel Kallner <[email protected]>
Signed-off-by: Shmuel Kallner <[email protected]>
✅ Deploy Preview for gateway-api-inference-extension ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: shmuelk The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
Discussed with @nirrozenbaum. We likely will need a proposal or an issue for an extension of the framework as this is a novel addition |
/hold |
What type of PR is this?
/kind feature
What this PR does / why we need it:
This PR adds a new optional extension point, the PreSchedule plugin. This plugin when specified is invoked by the Director before the scheduler is invoked. In particular it is in charge of getting the candidate pods to be filtered and scored by the scheduling process.
The new PreSchedule plugin enables more sophisticated pod selection strategies.
This is particularly useful in cases where one is using an autoscaler in a cluster of many InferencePools that are "sharing" a set of GPUs. In periods of inactivity for some particular model, one might want to scale its InferencePool to zero and use the freed up GPU in another InferencePool. When requests for the model that was scaled to zero come in, one would want to scale that InferencePool up from zero, in order to process the requests.
Does this PR introduce a user-facing change?: