Install this version:
emerge -a =sci-ml/demucs-4.1.0
If this version is masked, you can unmask it using the autounmask tool or standard emerge options:
autounmask =sci-ml/demucs-4.1.0
Or alternatively:
emerge --autounmask-write -a =sci-ml/demucs-4.1.0
| Version | EAPI | Keywords | Slot |
|---|---|---|---|
| 4.1.0 | 8 | ~amd64 | 0 |
# Copyright 2026 Gentoo Authors
# Distributed under the terms of the GNU General Public License v2
EAPI=8
DISTUTILS_SINGLE_IMPL=1
DISTUTILS_USE_PEP517=hatchling
PYTHON_COMPAT=( python3_{12..14} )
inherit distutils-r1 pypi
DESCRIPTION="Music source separation in the waveform domain."
HOMEPAGE="https://pypi.org/project/demucs/ https://github.com/adefossez/demucs"
LICENSE="MIT"
SLOT="0"
KEYWORDS="~amd64"
# The only test is tools/test_pretrained.py, which evaluates against the MusDB
# dataset it downloads, and needs the unpackaged train extras to import at all.
RESTRICT="test"
REQUIRED_USE="${PYTHON_REQUIRED_USE}"
# shellcheck disable=SC2016
RDEPEND="${PYTHON_DEPS}
$(python_gen_cond_dep '
>=dev-python/lameenc-1.2[${PYTHON_USEDEP}]
>=dev-python/sphn-0.1.12[${PYTHON_USEDEP}]
dev-python/pyyaml[${PYTHON_USEDEP}]
dev-python/tqdm[${PYTHON_USEDEP}]
sci-ml/einops[${PYTHON_USEDEP}]
sci-ml/safetensors[${PYTHON_USEDEP}]
')
>=sci-ml/julius-0.2.3[${PYTHON_SINGLE_USEDEP}]
>=sci-ml/pytorch-2.1[${PYTHON_SINGLE_USEDEP}]
sci-ml/huggingface_hub[${PYTHON_SINGLE_USEDEP}]"
pkg_postinst() {
elog "Training needs the extras upstream calls 'train' (dora-search, hydra, musdb,"
elog "museval, submitit, treetable), none of which are packaged. Separating tracks with"
elog "the pretrained models works without them."
}
${PYTHON_DEPS}
$(python_gen_cond_dep '
>=dev-python/lameenc-1.2[${PYTHON_USEDEP}]
>=dev-python/sphn-0.1.12[${PYTHON_USEDEP}]
dev-python/pyyaml[${PYTHON_USEDEP}]
dev-python/tqdm[${PYTHON_USEDEP}]
sci-ml/einops[${PYTHON_USEDEP}]
sci-ml/safetensors[${PYTHON_USEDEP}]
')
>=sci-ml/julius-0.2.3[${PYTHON_SINGLE_USEDEP}]
>=sci-ml/pytorch-2.1[${PYTHON_SINGLE_USEDEP}]
sci-ml/huggingface_hub[${PYTHON_SINGLE_USEDEP}]