.SRCINFO


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47

pkgbase = llama.cpp-cuda-git
	pkgdesc = Port of Facebook's LLaMA model in C/C++ (with NVIDIA CUDA optimizations)
	pkgver = b9050.r5.8e52631d55
	pkgrel = 1
	url = https://github.com/ggml-org/llama.cpp
	arch = x86_64
	arch = aarch64
	license = MIT
	makedepends = cmake
	makedepends = cudnn
	makedepends = gcc15
	makedepends = git
	makedepends = ninja
	depends = cuda
	depends = curl
	depends = gcc-libs
	depends = glibc
	depends = nvidia-utils
	depends = openssl
	optdepends = ccache: greatly reduce package re-build time
	optdepends = nccl: needed for multi-GPU parallelism
	optdepends = python-numpy: needed for convert_hf_to_gguf.py
	optdepends = python-safetensors: needed for convert_hf_to_gguf.py
	optdepends = python-sentencepiece: needed for convert_hf_to_gguf.py
	optdepends = python-pytorch: needed for convert_hf_to_gguf.py
	optdepends = python-transformers: needed for convert_hf_to_gguf.py
	optdepends = rdma-core: RDMA transport for RPC backend
	provides = llama.cpp
	provides = libggml-cuda-git
	provides = libggml
	provides = libggml.so
	provides = ggml
	conflicts = llama.cpp
	conflicts = libggml
	conflicts = ggml
	backup = etc/conf.d/llama.cpp
	source = git+https://github.com/ggml-org/llama.cpp.git
	source = llama.cpp.conf
	source = llama.cpp.service
	sha256sums = SKIP
	sha256sums = 53fa70cfe40cb8a3ca432590e4f76561df0f129a31b121c9b4b34af0da7c4d87
	sha256sums = 0377d08a07bda056785981d3352ccd2dbc0387c4836f91fb73e6b790d836620d
	b2sums = SKIP
	b2sums = 088e6b702e42bf1af019f69c8a85b0cd1196599e12f196e086ea1271e1800540947d1b51e3500821ec4556386f8e3c8217c0ad03570b764b85016827648939e7
	b2sums = 56e8e6e99c37f9baa1db5e3f8956f48a59bdbdc48797ae9b41292f0d1cdc3e41e5174bd7d721f3db84587ca271b11b480525e8c32cdb0f17f689b5537623c0a7

pkgname = llama.cpp-cuda-git