wip/llama.cpp: Update to 0.0.2.3173

* Enable blas backend.
* Disable build and instalation of tests.
* Python scripts require additional dependencies. It is not fixed yet.

You can use this package with preconverted models in GGUF.

diff --git a/llama.cpp/Makefile b/llama.cpp/Makefile
index 0fc35cac86ce652e39e46745e92cd2d44ee94587..22c878058261b361e8388fd1d7ea3c90e930e581 100644 (file)
--- a/llama.cpp/Makefile
+++ b/llama.cpp/Makefile
@@ -1,24 +1,29 @@
 # $NetBSD$
 
-DISTNAME=      llama.cpp-b3091
+DISTNAME=      llama.cpp-${GITHUB_TAG}
 PKGNAME=       ${DISTNAME:S/-b/-0.0.2./}
 CATEGORIES=    devel
 MASTER_SITES=  ${MASTER_SITE_GITHUB:=ggerganov/}
-GITHUB_TAG=    b3091
+GITHUB_TAG=    b3173
 
 MAINTAINER=    pkgsrc-users@NetBSD.org
 HOMEPAGE=      https://github.com/ggerganov/llama.cpp/
 COMMENT=       LLM inference in C/C++
 LICENSE=       mit
 
-#WRKSRC=               ${WRKDIR}/llama.cpp-b3091
 USE_TOOLS+=    pkg-config
 USE_LANGUAGES= c c++
-USE_TOOLS+=    gmake
+
+BLAS_C_INTERFACE=      yes
 
 PKGCONFIG_OVERRIDE+=   cmake/llama.pc.in
 REPLACE_PYTHON+=       *.py */*.py */*/*.py
 
+CMAKE_CONFIGURE_ARGS+= -DLLAMA_BLAS=1
+CMAKE_CONFIGURE_ARGS+= -DBLAS_LIBRARIES=${CBLAS_LIBS:Q}
+CMAKE_CONFIGURE_ARGS+= -DLLAMA_BUILD_TESTS=no
+
 .include "../../devel/cmake/build.mk"
 .include "../../lang/python/application.mk"
+.include "../../mk/blas.buildlink3.mk"
 .include "../../mk/bsd.pkg.mk"

diff --git a/llama.cpp/PLIST b/llama.cpp/PLIST
index ebd96403277e072a72021b3ba0455e0352330fc9..f8a18135765087932056fc5d43289686612e4259 100644 (file)
--- a/llama.cpp/PLIST
+++ b/llama.cpp/PLIST
@@ -1,55 +1,41 @@
 @comment $NetBSD$
-bin/baby-llama
-bin/batched
-bin/batched-bench
-bin/benchmark
 bin/convert-hf-to-gguf.py
-bin/convert-llama2c-to-ggml
-bin/embedding
-bin/eval-callback
-bin/export-lora
-bin/finetune
-bin/gguf
-bin/gguf-split
-bin/gritlm
-bin/imatrix
-bin/infill
+bin/llama-baby-llama
+bin/llama-batched
+bin/llama-batched-bench
 bin/llama-bench
-bin/llava-cli
-bin/lookahead
-bin/lookup
-bin/lookup-create
-bin/lookup-merge
-bin/lookup-stats
-bin/main
-bin/parallel
-bin/passkey
-bin/perplexity
-bin/quantize
-bin/quantize-stats
-bin/retrieval
-bin/save-load-state
-bin/server
-bin/simple
-bin/speculative
-bin/test-autorelease
-bin/test-backend-ops
-bin/test-chat-template
-bin/test-grad0
-bin/test-grammar-integration
-bin/test-grammar-parser
-bin/test-json-schema-to-grammar
-bin/test-llama-grammar
-bin/test-model-load-cancel
-bin/test-quantize-fns
-bin/test-quantize-perf
-bin/test-rope
-bin/test-sampling
-bin/test-tokenizer-0
-bin/test-tokenizer-1-bpe
-bin/test-tokenizer-1-spm
-bin/tokenize
-bin/train-text-from-scratch
+bin/llama-bench-matmult
+bin/llama-cli
+bin/llama-convert-llama2c-to-ggml
+bin/llama-cvector-generator
+bin/llama-embedding
+bin/llama-eval-callback
+bin/llama-export-lora
+bin/llama-finetune
+bin/llama-gbnf-validator
+bin/llama-gguf
+bin/llama-gguf-split
+bin/llama-gritlm
+bin/llama-imatrix
+bin/llama-infill
+bin/llama-llava-cli
+bin/llama-lookahead
+bin/llama-lookup
+bin/llama-lookup-create
+bin/llama-lookup-merge
+bin/llama-lookup-stats
+bin/llama-parallel
+bin/llama-passkey
+bin/llama-perplexity
+bin/llama-quantize
+bin/llama-quantize-stats
+bin/llama-retrieval
+bin/llama-save-load-state
+bin/llama-server
+bin/llama-simple
+bin/llama-speculative
+bin/llama-tokenize
+bin/llama-train-text-from-scratch
 include/ggml-alloc.h
 include/ggml-backend.h
 include/ggml.h

diff --git a/llama.cpp/distinfo b/llama.cpp/distinfo
index f929c3a0574b1007e0f3dd355ea4c8a11d83f2e2..8211b500bde0c83c5fdc486908eb1621efa83b04 100644 (file)
--- a/llama.cpp/distinfo
+++ b/llama.cpp/distinfo
@@ -1,5 +1,5 @@
 $NetBSD$
 
-BLAKE2s (llama.cpp-b3091.tar.gz) = 34af707155ae44c9417ce85b102a432ed665062d87ec6ad36b039268272031f8
-SHA512 (llama.cpp-b3091.tar.gz) = f2dac9b3663dd272c74f7ff9e50b63ddb489d5566a8a2e71102b1ad0b61454221cd0bd2dcce40eca1050f2236bc4ded76916c34371f0375649c3920126665bda
-Size (llama.cpp-b3091.tar.gz) = 20541244 bytes
+BLAKE2s (llama.cpp-b3173.tar.gz) = 74cd3a8c11a8def5f213bdb1afc209abacf89cd39522ca71bf12a6275a9909ac
+SHA512 (llama.cpp-b3173.tar.gz) = acde4758d08f4be9fafa570fcc9ab6700c556381242548a02163c01eaeac30a3680d9f097a8d4b92dcc2783049f80d4647f1c6fe4964c1ddf02303d4e3dc6abf
+Size (llama.cpp-b3173.tar.gz) = 20592194 bytes