llama.cpp
metal : add FA kernels for HSK=96, HSV=64 (MiniCPM3)
#28599
Merged

metal : add FA kernels for HSK=96, HSV=64 (MiniCPM3) #28599

wyanzhao
github-actions github-actions added testing
github-actions github-actions added ggml
github-actions github-actions added Apple Metal
wyanzhao wyanzhao marked this pull request as ready for review 61 days ago
wyanzhao wyanzhao requested a review from ggerganov ggerganov 61 days ago
wyanzhao wyanzhao requested a review 61 days ago
forforever73
wyanzhao wyanzhao force pushed 57 days ago
wyanzhao wyanzhao marked this pull request as draft 57 days ago
wyanzhao metal : add FA kernels for HSK=96, HSV=64 (MiniCPM3)
977fa66f
wyanzhao wyanzhao force pushed to 977fa66f 57 days ago
wyanzhao wyanzhao changed the title metal : fix MiniCPM3 crash with automatic flash attention metal : add FA kernels for HSK=96, HSV=64 (MiniCPM3) 57 days ago
wyanzhao
wyanzhao wyanzhao marked this pull request as ready for review 57 days ago
forforever73
forforever73 approved these changes on 2026-09-14
ggerganov
ggerganov commented on 2026-09-14
wyanzhao tests : avoid redundant FA vec slice coverage
d19eff9c
ggerganov
ggerganov approved these changes on 2026-09-14
ggerganov ggerganov added merge ready
ggerganov ggerganov merged 1e7bcf3d into master 54 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone