sci-ml/accelerate
Run your *raw* PyTorch training script on any kind of device
Screenshots
-
accelerate-1.15.0-r1~amd64 ~arm64test python_single_target_python3_12 python_single_target_python3_13 python_single_target_python3_14
View 
Download 
Browse License: Apache-2.0 Overlay: stuff -
accelerate-1.14.0-r1~amd64 ~arm64test python_single_target_python3_12 python_single_target_python3_13 python_single_target_python3_14
View Download Browse License: Apache-2.0 Overlay: stuff -
accelerate-1.14.0~amd64test python_single_target_python3_12 python_single_target_python3_13 python_single_target_python3_14
View Download Browse License: Apache-2.0 Overlay: gentoo
ChangeLog
commit 94c9b08eb48d812d4dc7b320d3f58b1cca170037
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Oct 3 23:19:32 2026 +0200
*/*: indent the remaining metadata.xml files with spaces
Twenty-six files imported from ::gentoo in August and September kept
its tab indentation, while the other 705 use two spaces, as Gentoo's
skel.metadata.xml does. Only leading whitespace changes; each file
parses to the same content.
commit 0bc37a5e42c22b56df7c69587d8d1522903db10d
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.13.0 at the corrected caffe2
The test floor named caffe2 with no version, so any already-merged
RelWithDebInfo build satisfied it and left torch's debug asserts live.
Bug: https://github.com/istitov/stuff/issues/293
commit 2e76da700b51a27b410b4604e4a8cdfd0a7232c6
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.14.0 at the corrected caffe2
1.14.0 kept the unversioned 2.13.0 floor, so an already-merged
RelWithDebInfo caffe2 satisfied it and the partial-compile test kept
aborting in libtorch_python. Consumers resolve to this version, so the
fix has to live here too.
Bug: https://github.com/istitov/stuff/issues/293
commit 1c96bfdccbf5f82a84daf921e10a05917ee9b547
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.15.0 at the corrected caffe2
The floor added for issue 293 named 2.14.0-r91, the only Release build
at the time. vllm pins ~pytorch-2.13.0, which forwards to
~caffe2-2.13.0, so that floor was unsatisfiable for anyone holding vllm
and traded one broken state for another.
2.13.0-r92 now carries the same correction, so require that instead and
let the pytorch floor drop back to the series vllm allows.
Bug: https://github.com/istitov/stuff/issues/293
commit 86f1fd78696ac6a9b089c0f349da3ba6b618defe
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Sep 19 14:57:24 2026 +0200
sci-ml/accelerate: require fixed torch stack for 1.15.0 tests
The partial-compile test aborts in libtorch_python when Portage runs
it against Caffe2 2.13.0-r91. That build predates the Release/NDEBUG
correction from issue 291.
Require the matching 2.14.0-r91 PyTorch and Caffe2 test dependencies
so the corrected native stack is merged before the test phase.
Bug: https://github.com/istitov/stuff/issues/293
commit dc9869252639a2f0d05e404900e7cfd51fd7b933
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:49:21 2026 +0200
sci-ml/accelerate: trim 1.15.0 comments
commit a3c44f3ef26d0de9453092a84791ef3a7a511830
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 07:21:59 2026 +0200
sci-ml/accelerate: trim 1.14.0 gloo rationale
commit 13a50968fbffd3db85990027c9d1881384a2c461
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 07:21:59 2026 +0200
sci-ml/accelerate: trim 1.13.0-r1 gloo rationale
commit 57aeca8b4118b401a771350579f7e66f6481a145
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:54 2026 +0200
sci-ml/accelerate: backport test fixes to 1.13.0-r1
The upstream suite needs its declared plugins and deterministic
exclusions to run without optional hardware or services.
commit 1e777beec4a202fa329c9798a8bfc788c543bfb3
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:54 2026 +0200
sci-ml/accelerate: backport dependency fixes to 1.13.0-r1
Retained releases need the published compatibility floors instead of
resolving against arbitrary older dependencies.
commit 8b26a9aecb30c53c5f92f07c671238055962e310
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:23 2026 +0200
sci-ml/accelerate: backport test fixes to 1.14.0
The upstream suite needs its declared plugins and deterministic
exclusions to run without optional hardware or services.
commit 8930fb5dcfd1ba8b874e9dc68f6faa6cc119abe6
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:23 2026 +0200
sci-ml/accelerate: backport dependency fixes to 1.14.0
Retained releases need the published compatibility floors instead of
resolving against arbitrary older dependencies.
commit cd2a162bd515080385d7c7e536ef6eb3dc804ba5
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 18:53:08 2026 +0200
sci-ml/accelerate: add 1.15.0
Match upstream runtime floors and restore the parameterized test
dependency. Keep the local suite independent of accelerator hardware
and external services.
commit 33d843716d93bf53bcbcd5e80a54e46076ffba34
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 12:55:16 2026 +0200
sci-ml/accelerate: restrict gloo test dependency to amd64
An unconditional gloo requirement cannot be satisfied on arm64: the
flag requires caffe2[distributed], and both are profile-masked
outside amd64. The upstream pytorch[gloo] spelling in 1.14.0 also
names the wrong package for this tree's caffe2/pytorch split -- the
flag lives on caffe2 and pytorch never exposes it, so a (+) default
would match no matter how caffe2 was built. Depend on caffe2[gloo]
on amd64 only, keeping the guarantee where the flag exists.
commit cfd24f52a5b2a0a81c33dfafda710433723fc191
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:21:43 2026 +0200
sci-ml/accelerate: Keyword 1.14.0 for ~arm64
The Python package and its machine learning dependencies support aarch64.
commit 03ce1e3af9b57b9f62c2a8165caaab056d2949e5
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:21:42 2026 +0200
sci-ml/accelerate: Keyword 1.13.0-r1 for ~arm64
The Python package and its machine learning dependencies support aarch64.
commit 8b5531d81e8334ff8a54aac8a1223eee4ca2e370
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:12:57 2026 +0200
sci-ml/accelerate: add 1.13.0-r1, 1.14.0
Import the current Gentoo ebuilds as dependencies for the machine learning stack.
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Oct 3 23:19:32 2026 +0200
*/*: indent the remaining metadata.xml files with spaces
Twenty-six files imported from ::gentoo in August and September kept
its tab indentation, while the other 705 use two spaces, as Gentoo's
skel.metadata.xml does. Only leading whitespace changes; each file
parses to the same content.
commit 0bc37a5e42c22b56df7c69587d8d1522903db10d
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.13.0 at the corrected caffe2
The test floor named caffe2 with no version, so any already-merged
RelWithDebInfo build satisfied it and left torch's debug asserts live.
Bug: https://github.com/istitov/stuff/issues/293
commit 2e76da700b51a27b410b4604e4a8cdfd0a7232c6
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.14.0 at the corrected caffe2
1.14.0 kept the unversioned 2.13.0 floor, so an already-merged
RelWithDebInfo caffe2 satisfied it and the partial-compile test kept
aborting in libtorch_python. Consumers resolve to this version, so the
fix has to live here too.
Bug: https://github.com/istitov/stuff/issues/293
commit 1c96bfdccbf5f82a84daf921e10a05917ee9b547
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 20 15:25:28 2026 +0200
sci-ml/accelerate: point 1.15.0 at the corrected caffe2
The floor added for issue 293 named 2.14.0-r91, the only Release build
at the time. vllm pins ~pytorch-2.13.0, which forwards to
~caffe2-2.13.0, so that floor was unsatisfiable for anyone holding vllm
and traded one broken state for another.
2.13.0-r92 now carries the same correction, so require that instead and
let the pytorch floor drop back to the series vllm allows.
Bug: https://github.com/istitov/stuff/issues/293
commit 86f1fd78696ac6a9b089c0f349da3ba6b618defe
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Sep 19 14:57:24 2026 +0200
sci-ml/accelerate: require fixed torch stack for 1.15.0 tests
The partial-compile test aborts in libtorch_python when Portage runs
it against Caffe2 2.13.0-r91. That build predates the Release/NDEBUG
correction from issue 291.
Require the matching 2.14.0-r91 PyTorch and Caffe2 test dependencies
so the corrected native stack is merged before the test phase.
Bug: https://github.com/istitov/stuff/issues/293
commit dc9869252639a2f0d05e404900e7cfd51fd7b933
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:49:21 2026 +0200
sci-ml/accelerate: trim 1.15.0 comments
commit a3c44f3ef26d0de9453092a84791ef3a7a511830
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 07:21:59 2026 +0200
sci-ml/accelerate: trim 1.14.0 gloo rationale
commit 13a50968fbffd3db85990027c9d1881384a2c461
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 07:21:59 2026 +0200
sci-ml/accelerate: trim 1.13.0-r1 gloo rationale
commit 57aeca8b4118b401a771350579f7e66f6481a145
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:54 2026 +0200
sci-ml/accelerate: backport test fixes to 1.13.0-r1
The upstream suite needs its declared plugins and deterministic
exclusions to run without optional hardware or services.
commit 1e777beec4a202fa329c9798a8bfc788c543bfb3
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:54 2026 +0200
sci-ml/accelerate: backport dependency fixes to 1.13.0-r1
Retained releases need the published compatibility floors instead of
resolving against arbitrary older dependencies.
commit 8b26a9aecb30c53c5f92f07c671238055962e310
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:23 2026 +0200
sci-ml/accelerate: backport test fixes to 1.14.0
The upstream suite needs its declared plugins and deterministic
exclusions to run without optional hardware or services.
commit 8930fb5dcfd1ba8b874e9dc68f6faa6cc119abe6
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Fri Sep 11 00:06:23 2026 +0200
sci-ml/accelerate: backport dependency fixes to 1.14.0
Retained releases need the published compatibility floors instead of
resolving against arbitrary older dependencies.
commit cd2a162bd515080385d7c7e536ef6eb3dc804ba5
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Thu Sep 10 18:53:08 2026 +0200
sci-ml/accelerate: add 1.15.0
Match upstream runtime floors and restore the parameterized test
dependency. Keep the local suite independent of accelerator hardware
and external services.
commit 33d843716d93bf53bcbcd5e80a54e46076ffba34
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 12:55:16 2026 +0200
sci-ml/accelerate: restrict gloo test dependency to amd64
An unconditional gloo requirement cannot be satisfied on arm64: the
flag requires caffe2[distributed], and both are profile-masked
outside amd64. The upstream pytorch[gloo] spelling in 1.14.0 also
names the wrong package for this tree's caffe2/pytorch split -- the
flag lives on caffe2 and pytorch never exposes it, so a (+) default
would match no matter how caffe2 was built. Depend on caffe2[gloo]
on amd64 only, keeping the guarantee where the flag exists.
commit cfd24f52a5b2a0a81c33dfafda710433723fc191
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:21:43 2026 +0200
sci-ml/accelerate: Keyword 1.14.0 for ~arm64
The Python package and its machine learning dependencies support aarch64.
commit 03ce1e3af9b57b9f62c2a8165caaab056d2949e5
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:21:42 2026 +0200
sci-ml/accelerate: Keyword 1.13.0-r1 for ~arm64
The Python package and its machine learning dependencies support aarch64.
commit 8b5531d81e8334ff8a54aac8a1223eee4ca2e370
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Sep 6 10:12:57 2026 +0200
sci-ml/accelerate: add 1.13.0-r1, 1.14.0
Import the current Gentoo ebuilds as dependencies for the machine learning stack.