diff options
| author | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-07-27 07:55:01 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-07-27 07:55:01 +0200 |
| commit | 154e0d75fccf1784fe9ff6fd76a630b66563da3d (patch) | |
| tree | 81ce6dbb5b1900c1aa78a879f0593c694cab9d27 /spm-headers | |
| parent | 0684c3e9c70d49323b4fc517128cbe222cab7f96 (diff) | |
Merge mainline llama.cpp (#3)
* Merging mainline - WIP
* Merging mainline - WIP
AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.
* Merging mainline - fix Metal
* Remove check
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'spm-headers')
| l--------- | spm-headers/ggml-alloc.h | 2 | ||||
| l--------- | spm-headers/ggml-backend.h | 2 | ||||
| l--------- | spm-headers/ggml-metal.h | 1 | ||||
| l--------- | spm-headers/ggml.h | 2 | ||||
| l--------- | spm-headers/llama.h | 2 |
5 files changed, 5 insertions, 4 deletions
diff --git a/spm-headers/ggml-alloc.h b/spm-headers/ggml-alloc.h index a49d385a..0361ffc3 120000 --- a/spm-headers/ggml-alloc.h +++ b/spm-headers/ggml-alloc.h @@ -1 +1 @@ -../ggml-alloc.h
\ No newline at end of file +../ggml/include/ggml-alloc.h
\ No newline at end of file diff --git a/spm-headers/ggml-backend.h b/spm-headers/ggml-backend.h index 17c2cf14..7295f0f0 120000 --- a/spm-headers/ggml-backend.h +++ b/spm-headers/ggml-backend.h @@ -1 +1 @@ -../ggml-backend.h
\ No newline at end of file +../ggml/include/ggml-backend.h
\ No newline at end of file diff --git a/spm-headers/ggml-metal.h b/spm-headers/ggml-metal.h new file mode 120000 index 00000000..aefad5fa --- /dev/null +++ b/spm-headers/ggml-metal.h @@ -0,0 +1 @@ +../ggml/include/ggml-metal.h
\ No newline at end of file diff --git a/spm-headers/ggml.h b/spm-headers/ggml.h index 39215298..0bdfeacb 120000 --- a/spm-headers/ggml.h +++ b/spm-headers/ggml.h @@ -1 +1 @@ -../ggml.h
\ No newline at end of file +../ggml/include/ggml.h
\ No newline at end of file diff --git a/spm-headers/llama.h b/spm-headers/llama.h index 9acceb98..b31388f0 120000 --- a/spm-headers/llama.h +++ b/spm-headers/llama.h @@ -1 +1 @@ -../llama.h
\ No newline at end of file +../include/llama.h
\ No newline at end of file |