Skip to content

[Feat] Support UCM on Ascend with versions 0.5.7 and above#930

Draft
flesher0813 wants to merge 3 commits intoModelEngine-Group:developfrom
flesher0813:develop_sglang
Draft

[Feat] Support UCM on Ascend with versions 0.5.7 and above#930
flesher0813 wants to merge 3 commits intoModelEngine-Group:developfrom
flesher0813:develop_sglang

Conversation

@flesher0813
Copy link
Copy Markdown
Contributor

@flesher0813 flesher0813 commented Apr 22, 2026

Purpose

Support UCM on Ascend with versions 0.5.7 and above. When using hicache on Ascend, the layout would be changed to page_first_direct or page_first_kv_split (MLA).

Modifications

Test

TODO

  • Check if mla supported upper sglang versions: Need adapt new layout page_first_kv_split
  • Ensure if dump same kv cache when using mla

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.

1 participant