Install this version:
emerge -a =dev-python/tokenspeed-mla-bin-2.0.11
If this version is masked, you can unmask it using the autounmask tool or standard emerge options:
autounmask =dev-python/tokenspeed-mla-bin-2.0.11
Or alternatively:
emerge --autounmask-write -a =dev-python/tokenspeed-mla-bin-2.0.11
| Version | EAPI | Keywords | Slot |
|---|---|---|---|
| 2.0.11 | 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-${PV}-py3-none-any.whl"
DESCRIPTION="TokenSpeed multi-head latent attention CUDA kernels"
HOMEPAGE="https://pypi.org/project/tokenspeed-mla/"
SRC_URI="https://files.pythonhosted.org/packages/39/dc/79e5c0c40070d0ead0fd2819bb5fc58107bf3204579222df2124b3a4a93a/${MY_WHEEL}"
S=${WORKDIR}
# Portions derive from CUTLASS (BSD-3-Clause) and FlashInfer (Apache-2.0),
# per the wheel's THIRDPARTYNOTICES. verified 2026-11-05
LICENSE="MIT Apache-2.0 BSD"
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-3.9.10_p20261028[${PYTHON_SINGLE_USEDEP}]
sci-ml/pytorch[${PYTHON_SINGLE_USEDEP}]
"
BDEPEND="
$(python_gen_cond_dep '
dev-python/installer[${PYTHON_USEDEP}]
')
"
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-3.9.10_p20261028[${PYTHON_SINGLE_USEDEP}] sci-ml/pytorch[${PYTHON_SINGLE_USEDEP}]
$(python_gen_cond_dep ' dev-python/installer[${PYTHON_USEDEP}] ')
| Type | File | Size | Source URLs |
|---|---|---|---|
| DIST | tokenspeed_mla-2.0.11-py3-none-any.whl | 245483 bytes | https://files.pythonhosted.org/packages/39/dc/79e5c0c40070d0ead0fd2819bb5fc58107bf3204579222df2124b3a4a93a/tokenspeed_mla-2.0.11-py3-none-any.whl |