]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: ca57e0f)
llama : add support for larger Granite Code Models (20B, 34B) (#7324)
authorSteffen Röcker <redacted>
Sat, 18 May 2024 08:04:55 +0000 (10:04 +0200)
committerGitHub <redacted>
Sat, 18 May 2024 08:04:55 +0000 (11:04 +0300)
commit0f98acfac6cc561dc57586bfff778405e42b576b
tree5ced0f623f9124ae87bc02566bf717636fbfbbactree
parentca57e0f35e33f714b9a6c2c4482b87bfe059c819commit | diff
llama : add support for larger Granite Code Models (20B, 34B) (#7324)

Tie the weights for ARCH_STARCODER to support the larger Granite code models.
Partially addresses ggerganov/issues/7116

There still remains to be a few things to fix.
Currently requires `--override-kv tokenizer.ggml.add_bos_token=bool:false`
llama.cpp diff | blob | history
Packaging of ggml-org/llama.cpp