Skip to content

vulkan: support sparse Flash Attention - #28105

Draft
0cc4m wants to merge 6 commits into
masterfrom
0cc4m/vulkan-fa-sparse
Draft

vulkan: support sparse Flash Attention#28105
0cc4m wants to merge 6 commits into
masterfrom
0cc4m/vulkan-fa-sparse

Conversation

@0cc4m

@0cc4m 0cc4m commented Aug 31, 2026

Copy link
Copy Markdown
Contributor

Overview

Vulkan support for #27970

Requirements

@github-actions github-actions Bot added model Model specific testing Everything test related Vulkan Issues specific to the Vulkan backend ggml changes relating to the ggml tensor library for machine learning CUDA Related to the CUDA backend labels Aug 31, 2026
mitchmindtree added a commit to mitchmindtree/llama.cpp that referenced this pull request Sep 5, 2026
On gfx1151 RADV the gather loses at every depth (11k -4.4, 115k -2.7
t/s vs masked, draft-mtp n-max 2): masked FA already skips fully-masked
tiles, the per-block bias avoids the mask upload the gather exists to
dodge, and gather mode forces the per-cell bias whose upload costs more
than the gather saves. QWEN4EXP_QSA_GATHER=1 re-enables it for A/B.
Revisit when Vulkan sparse FA (upstream ggml-org#28105) lands.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CUDA Related to the CUDA backend ggml changes relating to the ggml tensor library for machine learning model Model specific testing Everything test related Vulkan Issues specific to the Vulkan backend

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants