onnxruntime
[CUDA][PagedAttention] Fuse small-batch sequence-length sanitizer and add session.paged_attention.skip_input_sanitizer
#33597

Merged

[CUDA][PagedAttention] Fuse small-batch sequence-length sanitizer and add session.paged_attention.skip_input_sanitizer #33597

tianleiwu merged 2 commits into main from tlwu/20261025/paged-attention-sanitizer-opt

Fuse small-batch PagedAttention sequence-length sanitizer and add ski…

a3fde7f9

tianleiwu requested a review from

copilot-pull-request-reviewer 9 days ago

copilot-pull-request-reviewer commented on 2026-10-25

tianleiwu added model:qwen

Keep FlashAttention varlen split-KV decode-only and add sanitizer byp…

6e47b2df

kunal-vaishnavi commented on 2026-10-27

apsonawane requested a review from

copilot-pull-request-reviewer 7 days ago

apsonawane approved these changes on 2026-10-27

copilot-pull-request-reviewer commented on 2026-10-27

tianleiwu merged 55ab2dff into main 7 days ago

tianleiwu deleted the tlwu/20261025/paged-attention-sanitizer-opt branch 7 days ago

Reviewers

apsonawane

kunal-vaishnavi

Assignees

No one assigned

Labels

model:qwen

Milestone

No milestone