Script 'mail_helper' called by obssrc
Hello community,
here is the log from the commit of package python-transformers for
openSUSE:Factory checked in at 2026-08-27 18:51:38
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Comparing /work/SRC/openSUSE:Factory/python-transformers (Old)
and /work/SRC/openSUSE:Factory/.python-transformers.new.1265 (New)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-transformers"
Thu Aug 27 18:51:38 2026 rev:5 rq:1373869 version:5.16.1
Changes:
--------
--- /work/SRC/openSUSE:Factory/python-transformers/python-transformers.changes
2026-08-20 16:15:26.682566149 +0200
+++
/work/SRC/openSUSE:Factory/.python-transformers.new.1265/python-transformers.changes
2026-08-27 18:55:05.828374312 +0200
@@ -1,0 +2,31 @@
+Wed Aug 26 17:57:52 UTC 2026 - Martin Pluskal <[email protected]>
+
+- Update to 5.16.1:
+ * Add GLM-5.3-Flash, a natively multimodal 320B model with 18B
+ active parameters mixing sparse and linear attention
+ * Pin the remote ESMFold2 kernel revision and re-point it at
+ the upstream biohub repositories
+ * Restore backwards compatibility for the tensor-parallel API
+- Update to 5.16.0:
+ * Add Qwen4-Exp, Granite Speech 5.0, Step-3.7-Flash,
+ CohereCompass, ESMC and ESMFold2
+ * The legacy tensor-parallel implementation was replaced by a
+ DTensor-native backend, so code still using the old TP API
+ for inference or training has to be migrated
+ * wav2vec2-conformer, wav2vec2-bert and SeamlessM4T/v2 now
+ dispatch attn_implementation="sdpa", which changes their
+ initialisation behaviour
+ * FuyuProcessor no longer returns image_patch_indices
+ * Add NVFP4 quantization and a naive pipeline-parallel
+ inference engine wired into generate()
+ * Fix a sliding-window cache off-by-one on wraparound, Whisper
+ speculative decoding, Emu3 image generation, OLMo garbage
+ output and a silently dropped GPT2 cross-attention mask
+ * ... see upstream's release notes for the full list
+- Drop transformers-allow-tokenizers-0.23.1.patch and raise the
+ tokenizers floor to 0.23.1: upstream now declares
+ tokenizers >= 0.23.1, < 0.24.0 itself, which is exactly what
+ the patch widened the old <= 0.23.0 cap to
+- Switch from update-alternatives to libalternatives
+
+-------------------------------------------------------------------
Old:
----
transformers-5.15.1.tar.gz
transformers-allow-tokenizers-0.23.1.patch
New:
----
transformers-5.16.1.tar.gz
----------(Old B)----------
Old: * ... see upstream's release notes for the full list
- Drop transformers-allow-tokenizers-0.23.1.patch and raise the
tokenizers floor to 0.23.1: upstream now declares
----------(Old E)----------
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Other differences:
------------------
++++++ python-transformers.spec ++++++
--- /var/tmp/diff_new_pack.ywO3ef/_old 2026-08-27 18:55:07.836444419 +0200
+++ /var/tmp/diff_new_pack.ywO3ef/_new 2026-08-27 18:55:07.839444524 +0200
@@ -16,22 +16,22 @@
#
+%bcond_without libalternatives
Name: python-transformers
-Version: 5.15.1
+Version: 5.16.1
Release: 0
Summary: State-of-the-art Machine Learning for JAX, PyTorch and
TensorFlow
License: Apache-2.0
URL: https://github.com/huggingface/transformers
Source:
https://files.pythonhosted.org/packages/source/t/transformers/transformers-%{version}.tar.gz
-# Upstream caps tokenizers at <=0.23.0, but the only stable 0.23.x release is
-# 0.23.1 - the cap makes "import transformers" fail outright.
-Patch0: transformers-allow-tokenizers-0.23.1.patch
BuildRequires: %{python_module base >= 3.10}
BuildRequires: %{python_module pip}
BuildRequires: %{python_module setuptools}
BuildRequires: %{python_module wheel}
+BuildRequires: alts
BuildRequires: fdupes
BuildRequires: python-rpm-macros
+Requires: alts
Requires: python-PyYAML >= 5.1
Requires: python-huggingface-hub >= 1.5.0
Requires: python-numpy >= 1.17
@@ -41,11 +41,9 @@
# dependency_versions_check.py enforces the upper bound at import time, so an
# undeclared 0.24 would install fine and then fail on "import transformers".
Requires: python-tokenizers < 0.24.0
-Requires: python-tokenizers >= 0.22.0
+Requires: python-tokenizers >= 0.23.1
Requires: python-tqdm >= 4.60
Requires: python-typer
-Requires(post): update-alternatives
-Requires(postun): update-alternatives
BuildArch: noarch
# SECTION runtime dependencies (also needed for the %%check import test)
BuildRequires: %{python_module PyYAML >= 5.1}
@@ -54,7 +52,7 @@
BuildRequires: %{python_module packaging >= 20.0}
BuildRequires: %{python_module regex >= 2025.10.22}
BuildRequires: %{python_module safetensors >= 0.8.0}
-BuildRequires: %{python_module tokenizers >= 0.22.0}
+BuildRequires: %{python_module tokenizers >= 0.23.1}
BuildRequires: %{python_module tqdm >= 4.60}
BuildRequires: %{python_module typer}
# /SECTION
@@ -83,6 +81,7 @@
%install
%pyproject_install
%python_clone -a %{buildroot}%{_bindir}/transformers
+%python_group_libalternatives transformers
%python_expand %fdupes %{buildroot}%{$python_sitelib}
%check
@@ -91,11 +90,8 @@
# Fall back to an import smoke test of the top-level package.
%python_expand PYTHONPATH=%{buildroot}%{$python_sitelib} $python -B -c "import
transformers; print(transformers.__version__)"
-%post
-%python_install_alternative transformers
-
-%postun
-%python_uninstall_alternative transformers
+%pre
+%python_libalternatives_reset_alternative transformers
%files %{python_files}
%license LICENSE
++++++ transformers-5.15.1.tar.gz -> transformers-5.16.1.tar.gz ++++++
/work/SRC/openSUSE:Factory/python-transformers/transformers-5.15.1.tar.gz
/work/SRC/openSUSE:Factory/.python-transformers.new.1265/transformers-5.16.1.tar.gz
differ: char 5, line 1
