Skip to content

Add video support with video token estimation for multimodal approximate prefix cache plugin #1746

Description

@capri-xiyue

What would you like to be added:
Add video support with video token estimation for multimodal approximate prefix cache plugin

Why is this needed:
Currently multimodal approximate prefix cache plugin just support token estimation with image, we also want to let multimodal approximate prefix cache plugin support video with token estimation.

Please note as approximate prefix cache plugin use the unified token producer for token estimation, such video token estimation logic should be included in the unified token producer, see https://github.com/llm-d/llm-d-router/tree/main/pkg/epp/framework/plugins/requestcontrol/dataproducer/tokenizer with estimate backend

Metadata

Metadata

Assignees

Labels

[mm]Multimodalneeds-triageIndicates an issue or PR lacks a triage label and requires one.

Type

No type

Projects

No projects

Relationships

None yet

Development

No branches or pull requests

Issue actions