llama.cpp
vulkan: Apply bias before softmax in FA, to avoid overflow
#24909

Merged

vulkan: Apply bias before softmax in FA, to avoid overflow #24909

jeffbolznv merged 1 commit into ggml-org:master from jeffbolznv:FATTN_KQ_MAX_OFFSET

vulkan: Apply bias before softmax in FA, to avoid overflow

8cb425d6

jeffbolznv requested a review 35 days ago

github-actions added Vulkan

github-actions added ggml

0cc4m approved these changes on 2026-06-23

ngxson approved these changes on 2026-06-23

jeffbolznv merged ac4105d6 into master 34 days ago

Reviewers

ngxson

0cc4m

Assignees

No one assigned

Labels

Vulkan ggml

Milestone

No milestone