Install this version:
emerge -a =sci-ml/exllamav3-1.4.6
If this version is masked, you can unmask it using the autounmask tool or standard emerge options:
autounmask =sci-ml/exllamav3-1.4.6
Or alternatively:
emerge --autounmask-write -a =sci-ml/exllamav3-1.4.6
| Version | EAPI | Keywords | Slot |
|---|---|---|---|
| 1.4.6 | 8 | ~amd64 | 0 |
# Copyright 2025 Gentoo Authors
# Distributed under the terms of the GNU General Public License v2
EAPI=8
DISTUTILS_SINGLE_IMPL=1
DISTUTILS_EXT=1
PYTHON_COMPAT=( python3_{11..14} )
DISTUTILS_USE_PEP517=setuptools
inherit distutils-r1 pypi
DESCRIPTION="Inference library for running local LLMs on consumer hardware."
HOMEPAGE="https://github.com/turboderp-org/exllamav3"
SRC_URI="https://github.com/turboderp-org/${PN}/archive/refs/tags/v${PV}.tar.gz -> ${P}.tar.gz"
LICENSE="MIT"
SLOT="0"
KEYWORDS="~amd64"
# Tests require a GPU and large model weights — not runnable in Portage.
RESTRICT="test"
REQUIRED_USE="${PYTHON_REQUIRED_USE}"
# shellcheck disable=SC2016
RDEPEND="${PYTHON_DEPS}
$(python_gen_cond_dep '
>=dev-python/numpy-2.1.0[${PYTHON_USEDEP}]
dev-python/pyyaml[${PYTHON_USEDEP}]
dev-python/rich[${PYTHON_USEDEP}]
>=sci-ml/safetensors-0.3.2[${PYTHON_USEDEP}]
dev-python/typing-extensions[${PYTHON_USEDEP}]
dev-libs/marisa[${PYTHON_USEDEP}]
')
>=sci-ml/pytorch-2.13.0[flash,${PYTHON_SINGLE_USEDEP}]
sci-ml/flash-attn[${PYTHON_SINGLE_USEDEP}]
>=sci-ml/tokenizers-0.21.1[${PYTHON_SINGLE_USEDEP}]"
DEPEND="${RDEPEND}"
BDEPEND="${PYTHON_DEPS}
dev-build/ninja
sys-devel/gcc:13"
src_configure() {
export CC=${CHOST}-gcc-13 CXX=${CHOST}-g++-13
distutils-r1_src_configure
}
${PYTHON_DEPS}
$(python_gen_cond_dep '
>=dev-python/numpy-2.1.0[${PYTHON_USEDEP}]
dev-python/pyyaml[${PYTHON_USEDEP}]
dev-python/rich[${PYTHON_USEDEP}]
>=sci-ml/safetensors-0.3.2[${PYTHON_USEDEP}]
dev-python/typing-extensions[${PYTHON_USEDEP}]
dev-libs/marisa[${PYTHON_USEDEP}]
')
>=sci-ml/pytorch-2.13.0[flash,${PYTHON_SINGLE_USEDEP}]
sci-ml/flash-attn[${PYTHON_SINGLE_USEDEP}]
>=sci-ml/tokenizers-0.21.1[${PYTHON_SINGLE_USEDEP}]
${PYTHON_DEPS}
$(python_gen_cond_dep '
>=dev-python/numpy-2.1.0[${PYTHON_USEDEP}]
dev-python/pyyaml[${PYTHON_USEDEP}]
dev-python/rich[${PYTHON_USEDEP}]
>=sci-ml/safetensors-0.3.2[${PYTHON_USEDEP}]
dev-python/typing-extensions[${PYTHON_USEDEP}]
dev-libs/marisa[${PYTHON_USEDEP}]
')
>=sci-ml/pytorch-2.13.0[flash,${PYTHON_SINGLE_USEDEP}]
sci-ml/flash-attn[${PYTHON_SINGLE_USEDEP}]
>=sci-ml/tokenizers-0.21.1[${PYTHON_SINGLE_USEDEP}]
dev-build/ninja sys-devel/gcc:13
| Type | File | Size | Source URLs |
|---|---|---|---|
| DIST | exllamav3-1.4.6.tar.gz | 6912719 bytes | https://github.com/turboderp-org/exllamav3/archive/refs/tags/v1.4.6.tar.gz |