]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 29c60d8)
llama : use n_embd_head_v when reshaping kqv (#7327)
authorfairydreaming <redacted>
Fri, 17 May 2024 11:24:38 +0000 (13:24 +0200)
committerGitHub <redacted>
Fri, 17 May 2024 11:24:38 +0000 (14:24 +0300)
commit27b040691cbe45314147c2745e891a38e9c048d4
tree85f2ade442a6e316a6700a89bb121924e051c593tree
parent29c60d8cddcfd14fa8a6bf023a6c4eb8692c76bacommit | diff
llama : use n_embd_head_v when reshaping kqv (#7327)

* llama : use n_embd_head_v instead of n_embd_head_k when reshaping kqv

* llama : use n_embd_v_gqa and n_embd_head_v instead of n_embd_k_gqa and n_embd_head_k when making a view of cached value vectors.

---------

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