# Copyright 1999-2026 Gentoo Authors # Distributed under the terms of the GNU General Public License v2 EAPI=8 PYTHON_COMPAT=( python3_{12..14} ) DISTUTILS_SINGLE_IMPL=1 DISTUTILS_USE_PEP517=setuptools DISTUTILS_EXT=1 inherit distutils-r1 multiprocessing DESCRIPTION="Audio data, transforms and models for PyTorch" HOMEPAGE="https://github.com/pytorch/audio" SRC_URI=" https://github.com/pytorch/audio/archive/refs/tags/v${PV}.tar.gz -> ${P}.gh.tar.gz " S="${WORKDIR}/audio-${PV}" LICENSE="BSD" SLOT="0" KEYWORDS="~amd64 ~arm64" # Sister of sci-ml/torchvision; built CPU-only here. CUDA/ROCm modes # need the same setup_helpers env-var dance torchvision does — defer # until a USE-flag-driven cycle. # # torchaudio 2.11.0 is upstream's latest (pytorch/audio lags torch), so # unlike torchvision there is no distinct 2.13 release to pin. Instead the # single 2.11.0 source is offered as two revisions against the two torch # frontends this overlay carries: r0 (=pytorch-2.11*, for vllm-0.26.0) and # this r1 (=pytorch-2.13*, for vllm-0.27.1). Both build clean from the same # tree; consumers' `~torchaudio-2.11.0` lets portage backtrack to the rev # whose torch pin matches the resolved pytorch. # verified 2026-08-12 RDEPEND=" =sci-ml/pytorch-2.13*[${PYTHON_SINGLE_USEDEP}] $(python_gen_cond_dep ' dev-python/numpy[${PYTHON_USEDEP}] ') " # Tests pull soundfile + a network corpus; not wired up. RESTRICT="test" python_compile() { export USE_CUDA=0 export USE_ROCM=0 export BUILD_CUDA_CTC_DECODER=0 MAX_JOBS="$(get_makeopts_jobs)" \ distutils-r1_python_compile -j1 }