]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: af37fd8)
llama : assume tied weights if lm_head/output weights is missing (#5824)
authorDon Mahurin <redacted>
Fri, 8 Mar 2024 10:41:50 +0000 (02:41 -0800)
committerGitHub <redacted>
Fri, 8 Mar 2024 10:41:50 +0000 (12:41 +0200)
commite457fb3540e0aaec47cfde0abf784c213f9216ee
treefaa76ac6cc30fa61742b93db107c2a25602c13a6tree
parentaf37fd8b30e37ccbffdd82e6f48559e2fb7ce7ddcommit | diff
llama : assume tied weights if lm_head/output weights is missing (#5824)

This is to support model configurations with "tie_word_embeddings" set to true.

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