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 /Package.swift | |
| 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 'Package.swift')
| -rw-r--r-- | Package.swift | 25 |
1 files changed, 13 insertions, 12 deletions
diff --git a/Package.swift b/Package.swift index 183e6475..1d90b47b 100644 --- a/Package.swift +++ b/Package.swift @@ -3,14 +3,17 @@ import PackageDescription var sources = [ - "ggml.c", - "sgemm.cpp", - "llama.cpp", - "unicode.cpp", - "unicode-data.cpp", - "ggml-alloc.c", - "ggml-backend.c", - "ggml-quants.c", + "src/llama.cpp", + "src/llama-vocab.cpp", + "src/llama-grammar.cpp", + "src/llama-sampling.cpp", + "src/unicode.cpp", + "src/unicode-data.cpp", + "ggml/src/ggml.c", + "ggml/src/ggml-alloc.c", + "ggml/src/ggml-backend.c", + "ggml/src/ggml-quants.c", + "ggml/src/ggml-aarch64.c", ] var resources: [Resource] = [] @@ -26,8 +29,8 @@ var cSettings: [CSetting] = [ ] #if canImport(Darwin) -sources.append("ggml-metal.m") -resources.append(.process("ggml-metal.metal")) +sources.append("ggml/src/ggml-metal.m") +resources.append(.process("ggml/src/ggml-metal.metal")) linkerSettings.append(.linkedFramework("Accelerate")) cSettings.append( contentsOf: [ @@ -63,8 +66,6 @@ let package = Package( "models", "tests", "CMakeLists.txt", - "ggml-cuda.cu", - "ggml-cuda.h", "Makefile" ], sources: sources, |