]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: fe12e42)
llama: use f16 mask for FA to save VRAM (#23764)
authorAman Gupta <redacted>
Fri, 29 May 2026 07:44:43 +0000 (15:44 +0800)
committerGitHub <redacted>
Fri, 29 May 2026 07:44:43 +0000 (15:44 +0800)
commit031ddb2e08962837c899374297da75f91cc0157d
treeb7ff2b2bfc856836b09c53fd74d610123797cdf3tree
parentfe12e422ad40e6845f9f422fc35c0f9b24441d8bcommit | diff
llama: use f16 mask for FA to save VRAM (#23764)

* llama: use f16 mask for FA

* review: add llama_cast + formatting

* simplify
src/llama-graph.cpp diff | blob | history
src/llama-graph.h diff | blob | history
src/llama-impl.h diff | blob | history
src/llama-kv-cache.cpp diff | blob | history
Packaging of ggml-org/llama.cpp