]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: fec2fb1)
metal : fix kernel_norm (fixes Falcon on Metal) (#3057)
authorGeorgi Gerganov <redacted>
Thu, 7 Sep 2023 12:49:09 +0000 (15:49 +0300)
committerGitHub <redacted>
Thu, 7 Sep 2023 12:49:09 +0000 (15:49 +0300)
commitc4f496648c1e32efeb714200e7eae7fc7cfbb223
tree876320eb5fa8b02682e0b0d88fe325b40da2f23atree
parentfec2fb19e4229aac58c98171c46e77144b99f8a3commit | diff
metal : fix kernel_norm (fixes Falcon on Metal) (#3057)

* metal : fix kernel_norm

ggml-ci

* metal : put warning in kernel_norm to not combine the loops

* metal : restore original F16 mat-vec multiplication

It works after the norm fixes

* common : don't do warm-up with more than n_batch tokens (close #3058)

ggml-ci

* metal : minor
common/common.cpp diff | blob | history
ggml-metal.metal diff | blob | history
Packaging of ggml-org/llama.cpp