Script 'mail_helper' called by obssrc
Hello community,
here is the log from the commit of package python-comfy-kitchen for
openSUSE:Factory checked in at 2026-09-04 12:38:43
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Comparing /work/SRC/openSUSE:Factory/python-comfy-kitchen (Old)
and /work/SRC/openSUSE:Factory/.python-comfy-kitchen.new.1265 (New)
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-comfy-kitchen"
Fri Sep 4 12:38:43 2026 rev:2 rq:1375569 version:0.2.32
Changes:
--------
---
/work/SRC/openSUSE:Factory/python-comfy-kitchen/python-comfy-kitchen.changes
2026-08-21 17:02:17.029094887 +0200
+++
/work/SRC/openSUSE:Factory/.python-comfy-kitchen.new.1265/python-comfy-kitchen.changes
2026-09-04 12:39:22.144390812 +0200
@@ -1,0 +2,17 @@
+Thu Sep 3 08:54:51 UTC 2026 - Martin Pluskal <[email protected]>
+
+- Update to 0.2.32:
+ * New sol_attn op (Sol-Attn training-free sparse attention),
+ with an eager implementation, so it is usable in this
+ CPU-only build
+ * flash_attention_decode now works on a ROCm torch, dispatching
+ to the HIP backend instead of always reporting unavailable
+ * Otherwise CUDA/HIP kernel work only (int8 Sol-Attn, bf16
+ flash decode, occupancy-aware WMMA GEMM tile selection),
+ not compiled in this build
+- Raise torch floor to >= 2.7.0: since 0.2.28 the na3d
+ torch.library.custom_op schema uses PEP 585 generics, which
+ infer_schema rejects before 2.7.0, so the import fails outright;
+ upstream only documented the real minimum now
+
+-------------------------------------------------------------------
Old:
----
comfy-kitchen-0.2.31.tar.gz
New:
----
comfy-kitchen-0.2.32.tar.gz
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Other differences:
------------------
++++++ python-comfy-kitchen.spec ++++++
--- /var/tmp/diff_new_pack.DJU8Uq/_old 2026-09-04 12:39:24.561475721 +0200
+++ /var/tmp/diff_new_pack.DJU8Uq/_new 2026-09-04 12:39:24.563475791 +0200
@@ -18,7 +18,7 @@
%{?sle15_python_module_pythons}
Name: python-comfy-kitchen
-Version: 0.2.31
+Version: 0.2.32
Release: 0
Summary: Fast kernel library for ComfyUI (CPU/eager backend)
# Legal-Review-Notice: NOTICE names the NVIDIA CUDA runtime EULA for a
@@ -32,11 +32,11 @@
Patch0: comfy-kitchen-cpu-only-build.patch
BuildRequires: %{python_module pip}
BuildRequires: %{python_module setuptools >= 61.0}
-BuildRequires: %{python_module torch >= 2.5.0}
+BuildRequires: %{python_module torch >= 2.7.0}
BuildRequires: %{python_module wheel}
BuildRequires: fdupes
BuildRequires: python-rpm-macros
-Requires: python-torch >= 2.5.0
+Requires: python-torch >= 2.7.0
BuildArch: noarch
# Match Factory python-torch (ExcludeArch: %%ix86 %%{arm})
ExcludeArch: %{ix86} %{arm}
++++++ comfy-kitchen-0.2.31.tar.gz -> comfy-kitchen-0.2.32.tar.gz ++++++
++++ 9066 lines of diff (skipped)
