Skip to content

[Feature] Implement IndexCache for GLM-5/DeepSeek V3.2 #21286

@Fridge003

Description

@Fridge003

Checklist

Motivation

IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse
https://arxiv.org/pdf/2603.12201

Related resources

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    Good Pro IssueIssues for experienced contributors; requires a solid understanding of SGLang internals.deepseekgood second issueGood for second time and more contributors

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions