onnxruntime
[CUDA] Add FP16-cache paged XQA decode for head size 256
#32263

Merged

[CUDA] Add FP16-cache paged XQA decode for head size 256 #32263

baijumeswani merged 1 commit into main from baijumeswani/fp16-paged-xqa-pr

baijumeswani requested a review from

copilot-pull-request-reviewer 30 days ago

copilot-pull-request-reviewer commented on 2026-08-25

baijumeswani force pushed from 75ba73ae to af085e8e 30 days ago

baijumeswani requested a review from

copilot-pull-request-reviewer 30 days ago

copilot-pull-request-reviewer commented on 2026-08-25

Optimize FP16 long-context paged decode with XQA

7c592b8b

baijumeswani force pushed from af085e8e to 7c592b8b 30 days ago

tianleiwu approved these changes on 2026-08-26

baijumeswani enabled auto-merge (squash) 30 days ago

baijumeswani merged b1f76d58 into main 29 days ago

baijumeswani deleted the baijumeswani/fp16-paged-xqa-pr branch 29 days ago

Reviewers

tianleiwu

Assignees

No one assigned

Labels

None yet

Milestone

No milestone