]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 80b717d)
vulkan: fix NaN issue in flash attention shader (#12776)
authorJeff Bolz <redacted>
Sun, 6 Apr 2025 09:03:47 +0000 (04:03 -0500)
committerGitHub <redacted>
Sun, 6 Apr 2025 09:03:47 +0000 (11:03 +0200)
commit0c74b04376b0b9efc096480fe10f866afc8d7c1c
tree4e0b2fad44a72ba298aa0237ad5bee042929e762tree
parent80b717d493a9a5bae7167ad2384c12c60bb2ef20commit | diff
vulkan: fix NaN issue in flash attention shader (#12776)

Use -FLT_MAX/2 rather than -inf as the initial value for computing the maximum.
ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp diff | blob | history
Packaging of ggml-org/llama.cpp