Install this version:
emerge -a =dev-python/tokenspeed-mla-bin-0.2.16If this version is masked, you can unmask it using the autounmask tool or standard emerge options:
autounmask =dev-python/tokenspeed-mla-bin-0.2.16Or alternatively:
emerge --autounmask-write -a =dev-python/tokenspeed-mla-bin-0.2.16| Version | EAPI | Keywords | Slot |
|---|---|---|---|
| 0.2.16 | 8 | ~amd64 | 0 |
# Copyright 1999-2026 Gentoo Authors
# Distributed under the terms of the GNU General Public License v2
EAPI=8
DISTUTILS_USE_PEP517=no
DISTUTILS_SINGLE_IMPL=1
PYTHON_COMPAT=( python3_{12..14} )
inherit distutils-r1
MY_WHEEL="tokenspeed_mla-${PV}-py3-none-any.whl"
DESCRIPTION="TokenSpeed multi-head latent attention CUDA kernels"
HOMEPAGE="https://pypi.org/project/tokenspeed-mla/"
SRC_URI="https://files.pythonhosted.org/packages/0f/87/5e4c9bebec55a6b0f90433466c60fd54411bed85488f88c7fef6e89c0897/${MY_WHEEL}"
S=${WORKDIR}
LICENSE="MIT"
SLOT="0"
KEYWORDS="~amd64"
RDEPEND="
$(python_gen_cond_dep '
>=dev-python/apache-tvm-ffi-0.1.11[${PYTHON_USEDEP}]
<dev-python/apache-tvm-ffi-0.2[${PYTHON_USEDEP}]
>=dev-python/nvidia-cutlass-dsl-4.8.0[${PYTHON_USEDEP}]
')
>=dev-python/tokenspeed-triton-bin-3.8.10_p20260920[${PYTHON_SINGLE_USEDEP}]
sci-ml/caffe2
"
python_install() {
${EPYTHON} -m installer --destdir="${D}" "${DISTDIR}/${MY_WHEEL}" || die
python_optimize
}
$(python_gen_cond_dep ' >=dev-python/apache-tvm-ffi-0.1.11[${PYTHON_USEDEP}] <dev-python/apache-tvm-ffi-0.2[${PYTHON_USEDEP}] >=dev-python/nvidia-cutlass-dsl-4.8.0[${PYTHON_USEDEP}] ') >=dev-python/tokenspeed-triton-bin-3.8.10_p20260920[${PYTHON_SINGLE_USEDEP}] sci-ml/caffe2
| Type | File | Size | Source URLs |
|---|---|---|---|
| DIST | tokenspeed_mla-0.2.16-py3-none-any.whl | 154122 bytes | https://files.pythonhosted.org/packages/0f/87/5e4c9bebec55a6b0f90433466c60fd54411bed85488f88c7fef6e89c0897/tokenspeed_mla-0.2.16-py3-none-any.whl |