openvino
Fix XAttention BY_TOKEN KV-cache normalization
#35183

Merged

Fix XAttention BY_TOKEN KV-cache normalization #35183

e-ddykim merged 6 commits into openvinotoolkit:master from WeldonWangwang:ww/fix_xattn_by_token

github-actions added category: GPU

WeldonWangwang requested a review from

peterchen-intel 56 days ago

Fix XAttention BY_TOKEN KV-cache normalization

370ede71

WeldonWangwang marked this pull request as ready for review 55 days ago

WeldonWangwang requested a review 55 days ago

Add test case

9220f968

peterchen-intel requested a review from

ceciliapeng2011 54 days ago

peterchen-intel requested a review from

riverlijunjie 54 days ago

Simplify the conditions for changing the precision of the kv cache

093ddf43

ceciliapeng2011 approved these changes on 2026-04-14

riverlijunjie approved these changes on 2026-04-14

peterchen-intel assigned

e-ddykim 48 days ago

Merge branch 'master' into ww/fix_xattn_by_token

5195fb5e

Fix xattention kv-cache transform test for 28-input PagedAttentionExt…

d74b27df

Merge branch 'master' into ww/fix_xattn_by_token

3ce19599

e-ddykim approved these changes on 2026-04-17

e-ddykim enabled auto-merge 46 days ago

e-ddykim merged 89c806e2 into master 46 days ago

Reviewers

e-ddykim

riverlijunjie

ceciliapeng2011

peterchen-intel

Assignees

e-ddykim

Labels

category: GPU

Milestone

No milestone