Script 'mail_helper' called by obssrc
Hello community,
here is the log from the commit of package python-compressed-tensors for
openSUSE:Factory checked in at 2026-09-28 10:46:54
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Comparing /work/SRC/openSUSE:Factory/python-compressed-tensors (Old)
and /work/SRC/openSUSE:Factory/.python-compressed-tensors.new.383539 (New)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-compressed-tensors"
Mon Sep 28 10:46:54 2026 rev:4 rq:1380864 version:0.19.0
Changes:
--------
---
/work/SRC/openSUSE:Factory/python-compressed-tensors/python-compressed-tensors.changes
2026-08-12 16:13:38.177332721 +0200
+++
/work/SRC/openSUSE:Factory/.python-compressed-tensors.new.383539/python-compressed-tensors.changes
2026-09-28 10:47:28.926540425 +0200
@@ -1,0 +2,23 @@
+Sat Sep 26 20:52:00 UTC 2026 - Martin Pluskal <[email protected]>
+
+- Update to 0.19.0:
+ * BREAKING: GPTQ group/dynamic activation ordering has been removed.
+ actorder='group'/'dynamic' and actorder=True now raise a ValueError
+ instead of warning; use actorder='weight' instead
+ * Checkpoint conversion can now be scheduled across multiple GPUs by
+ measured free memory, using a profiler-based memory estimator
+ * Add a MagnitudeExpertPruner converter that structurally prunes MoE
+ experts by magnitude, and allow chaining converters in one pipeline
+ * Support decompressing models under distributed execution; fix dtype
+ mismatch and tensor shapes in the distributed offload caches, which
+ broke tied and multimodal models
+ * Fix an AttributeError in QuantizedKVCache for transformers 5 models
+ that read the KV cache structurally during calibration
+ * Observer resolution and defaulting moved to llm-compressor;
+ automatic qstrategy inference has been removed
+ * Faster model-free PTQ on meta tensors and a memoized calculate_range;
+ fix Triton cast_to_fp4 rounding for fp32/fp16 near FP4 boundaries
+ * Many more fixes and improvements; see upstream's release notes for
+ the full list
+
+-------------------------------------------------------------------
Old:
----
compressed_tensors-0.18.0.tar.gz
New:
----
compressed_tensors-0.19.0.tar.gz
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Other differences:
------------------
++++++ python-compressed-tensors.spec ++++++
--- /var/tmp/diff_new_pack.4BUZRo/_old 2026-09-28 10:47:29.497564349 +0200
+++ /var/tmp/diff_new_pack.4BUZRo/_new 2026-09-28 10:47:29.498564390 +0200
@@ -18,7 +18,7 @@
%{?sle15_python_module_pythons}
Name: python-compressed-tensors
-Version: 0.18.0
+Version: 0.19.0
Release: 0
Summary: Library for utilization of compressed safetensors of neural
network models
License: Apache-2.0
++++++ compressed_tensors-0.18.0.tar.gz -> compressed_tensors-0.19.0.tar.gz
++++++
++++ 9867 lines of diff (skipped)
