sci-ml/datasets
Access and share datasets for Audio, Computer Vision, and NLP tasks
Screenshots
-
datasets-5.0.1-r2~amd64 ~arm64torch vision test python_single_target_python3_12 python_single_target_python3_13 python_single_target_python3_14
View 
Download 
Browse License: Apache-2.0 Overlay: gentoo -
datasets-5.0.1~amd64python_single_target_python3_12 python_single_target_python3_13 python_single_target_python3_14
View Download Browse License: Apache-2.0 Overlay: stuff
ChangeLog
commit dfef3ce2874652ad0e64ee4c5de7909b9f1b9fea
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Aug 29 11:47:56 2026 +0200
sci-ml/datasets: add 5.0.1
Major bump. The build itself is unchanged; what matters is that 5.0.1
relaxes three upper bounds that had been pinning the tree to versions older
than either repo ships:
dill <0.4.1 -> <0.4.2 (0.4.1 is the only dill in ::gentoo,
and the old cap excluded it)
multiprocess <0.70.17 -> <0.70.20 (same shape: 0.70.19 was excluded)
fsspec <=2025.9.0 -> <=2026.6.0
The fsspec ceiling is the consequential one. At <=2025.9.0 this package was
the only reason ::stuff carries a dev-python/fsspec at all -- ::gentoo has
just 2026.6.0 and 2026.7.0, neither of which satisfied the old cap, so the
overlay had to keep a 2025.9.0 of its own. 5.0.1 resolves against ::gentoo's
fsspec directly. The overlay copy still cannot be dropped, because
datasets-4.3.0 is retained as the last of the 4.x series and still needs it.
Wrote the fsspec ceiling as <=2026.6.0-r9999 rather than the -r0 that 4.3.0
uses. Upstream's bound is on the upstream version, and a Gentoo revbump of
2026.6.0 is still upstream 2026.6.0, so a future packaging-only -r1 must not
be excluded by our atom. (pkgcheck's MissingPackageRevision forces the
choice to be explicit; it does not say which way.) 4.3.0's -r0 is left
alone -- it pins the single revision-less version we ship, so the two forms
are indistinguishable there today.
Also recorded why aiohttp stays in RDEPEND despite being absent from
Requires-Dist: it arrives through the `fsspec[http]` extra, which Portage
cannot express (::gentoo's fsspec carries aiohttp as an optfeature, not a
USE flag), so declaring it directly is what makes the HTTP filesystem
backend work on a merged system.
commit 647727cd521313d4da0d71aa6eab0761cc45c504
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 23 09:39:40 2026 +0200
sci-ml/datasets: add 4.3.0
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sat Aug 29 11:47:56 2026 +0200
sci-ml/datasets: add 5.0.1
Major bump. The build itself is unchanged; what matters is that 5.0.1
relaxes three upper bounds that had been pinning the tree to versions older
than either repo ships:
dill <0.4.1 -> <0.4.2 (0.4.1 is the only dill in ::gentoo,
and the old cap excluded it)
multiprocess <0.70.17 -> <0.70.20 (same shape: 0.70.19 was excluded)
fsspec <=2025.9.0 -> <=2026.6.0
The fsspec ceiling is the consequential one. At <=2025.9.0 this package was
the only reason ::stuff carries a dev-python/fsspec at all -- ::gentoo has
just 2026.6.0 and 2026.7.0, neither of which satisfied the old cap, so the
overlay had to keep a 2025.9.0 of its own. 5.0.1 resolves against ::gentoo's
fsspec directly. The overlay copy still cannot be dropped, because
datasets-4.3.0 is retained as the last of the 4.x series and still needs it.
Wrote the fsspec ceiling as <=2026.6.0-r9999 rather than the -r0 that 4.3.0
uses. Upstream's bound is on the upstream version, and a Gentoo revbump of
2026.6.0 is still upstream 2026.6.0, so a future packaging-only -r1 must not
be excluded by our atom. (pkgcheck's MissingPackageRevision forces the
choice to be explicit; it does not say which way.) 4.3.0's -r0 is left
alone -- it pins the single revision-less version we ship, so the two forms
are indistinguishable there today.
Also recorded why aiohttp stays in RDEPEND despite being absent from
Requires-Dist: it arrives through the `fsspec[http]` extra, which Portage
cannot express (::gentoo's fsspec carries aiohttp as an optfeature, not a
USE flag), so declaring it directly is what makes the HTTP filesystem
backend work on a merged system.
commit 647727cd521313d4da0d71aa6eab0761cc45c504
Author: Ivan S. Titov <iohann.s.titov@gmail.com>
Date: Sun Aug 23 09:39:40 2026 +0200
sci-ml/datasets: add 4.3.0