]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 5f1925a)
ggml : update softmax n_task calculation (#5126)
authorsnadampal <redacted>
Fri, 26 Jan 2024 17:17:59 +0000 (11:17 -0600)
committerGitHub <redacted>
Fri, 26 Jan 2024 17:17:59 +0000 (19:17 +0200)
commit7032f4f6349c17a8352f9f93f7d2122f45469e59
treea46a86b55b9bd975fc60e8784da74b8ad64c18a5tree
parent5f1925a8cef81eb9b372faaae34b0dd76d5361d4commit | diff
ggml : update softmax n_task calculation (#5126)

updated the n_task calculation to use max number of
threads possible. This has improved the prompt eval
performance by around 5% for DOT kernels and by
around 10% for MMLA kernels on AWS Graviton3.
ggml.c diff | blob | history
Packaging of ggml-org/llama.cpp