.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

pkgbase = beellama.cpp-cuda
	pkgdesc = DFlash & TurboQuant fork of llama.cpp — up to 3x faster generation and 7.5x more KV cache (CUDA)
	pkgver = 0.4.0
	pkgrel = 1
	url = https://github.com/Anbeeld/beellama.cpp
	arch = x86_64
	license = MIT
	makedepends = cmake
	makedepends = gcc14
	makedepends = ninja
	depends = cuda
	depends = curl
	depends = gcc-libs
	depends = glibc
	depends = nvidia-utils
	depends = openssl
	optdepends = python-numpy: convert_hf_to_gguf.py
	optdepends = python-safetensors: convert_hf_to_gguf.py
	optdepends = python-sentencepiece: convert_hf_to_gguf.py
	optdepends = python-pytorch: convert_hf_to_gguf.py
	optdepends = python-transformers: convert_hf_to_gguf.py
	optdepends = rdma-core: RDMA transport for RPC backend
	provides = beellama.cpp
	conflicts = beellama.cpp
	conflicts = beellama.cpp-bin
	conflicts = llama.cpp
	conflicts = libggml
	conflicts = ggml
	backup = etc/conf.d/beellama.conf
	source = beellama.cpp-0.4.0.tar.gz::https://github.com/Anbeeld/beellama.cpp/archive/refs/tags/v0.4.0.tar.gz
	source = beellama.conf
	source = beellama.service
	sha256sums = 3d46ae47dc524b007c8460686698b8779426f3fbb7d164aae119bddf417a2f5a
	sha256sums = 53fa70cfe40cb8a3ca432590e4f76561df0f129a31b121c9b4b34af0da7c4d87
	sha256sums = b019eb0dbc511f0e411214650184d1fb1fb39a40bce0f418d53dd71bf244e01b

pkgname = beellama.cpp-cuda