]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: ca7c3f4)
llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)
authorKawrakow <redacted>
Mon, 19 Jun 2023 15:17:03 +0000 (18:17 +0300)
committerGitHub <redacted>
Mon, 19 Jun 2023 15:17:03 +0000 (18:17 +0300)
commitcb40dfca694b5cb849837548fd69932117c78362
tree92864fc0f78969e2660deca2b34b07f882f07703tree
parentca7c3f4da5d144d4cd1dd44903552e6ba49b8ec8commit | diff
llama : only use Q6_K for output weights if tensor size is multiple of 256 (#1932)

* Only use Q6_K for output weights if tensor size is multiple of 256

* Fixed copy/paste mistake

---------

Co-authored-by: Iwan Kawrakow <redacted>
llama.cpp diff | blob | history
Packaging of ggml-org/llama.cpp