Sign me up Login

Details about package pytorch

Name: pytorch (PTS)
Uploader: Talha Can Havadar <havadartalha@gmail.com> (Debian QA page)
Description: python3-torch - Tensors and Dynamic neural networks in Python (Python Interface)
libtorch-dev - Tensors and Dynamic neural networks in Python (Development Files)
libtorch2.12 - Tensors and Dynamic neural networks in Python (Shared Objects)
libtorch-test - Tensors and Dynamic neural networks in Python (Test Binaries)

Package uploads

Upload #1

Information

Version: 2.12.1+dfsg-2
Uploaded: 2026-10-08 10:04
Source package: pytorch_2.12.1+dfsg-2.dsc
Distribution: unstable
Section: science
Priority: optional
Homepage: https://pytorch.org/
Vcs-Git: https://salsa.debian.org/deeplearning-team/pytorch.git
Vcs-Browser: https://salsa.debian.org/deeplearning-team/pytorch
Closes bugs: #1150269

Changelog

 pytorch (2.12.1+dfsg-2) unstable; urgency=medium
 .
   [ Aron Xu ]
   * d/copyright: exclude __pycache__/*.pyc from repacked tarball
 .
   [ Mo Zhou ]
   * autopkgtest: remove the train-mnist.sh test case.
   * Cleanup patchset after merging the ROCm fix.
   * trivial: quilt patch refresh
   * trivial: cleanup more mess
   * Patch embedded kineto to disable ROCPROF and ROCTRACER.
   * [ROCm] Patch cmake to prevent aotriton download.
   * [ROCm] Require libamdhip64-dev (>= 7.1.0~).
   * [ROCm] Require all rocm B-D to be >= 7.1.0, and add rocm-device-libs-22.
   * [ROCm] Build depend on composable kernel.
   * [ROCm] minor: use llvm-21 device libs.
   * [ROCm] add amdgcn-tools-21 to dependencies.
   * Get rid of tensorpipe from build dependencies.
   * Get rid of tensorpipe for cuda/rocm variants as well.
   * Reword cuda/rocm build scripts.
   * [CUDA] Attempt to fix CUDA build but we still miss some dependencies.
 .
   [ Talha Can Havadar ]
   * d/{rules, control.rocm}: bump to llvm-22 as whole rocm stack >=7.2 relies
     on llvm-22
   * d/kineto: patch kineto to not raise severity of Wno-dangling-reference
     which is unknown for clang and in rocm kineto builds with -Werror
     causing it to FTBFS
   * d/control.rocm: restrict libfmt-dev to >= 11.1.1 as older versions fail
     with clang-22
   * d/p/1070-formatting-fmt11-ostreambuf-iterator.patch: fix format_decimal
     FTBFS (Closes: #1150269)
   * d/rules: disabling flash/mem-eff attention for rocm as it requires aotriton
     And aotriton is not available in debian and not yet vendored in pytorch
     package
   * d/control.rocm: add missing amd-smi dependency
   * d/control.rocm: bump ROCm stack restriction to 7.2
   * d/rules: fix ABI incompatiblity hipcc passes clang-abi-compat=17
   * d/p/2340-c10d-tests-link-gloo-explicitly.patch: fix linking issue on
     ProcessGroupGlooTest
   * d/rules: filter out RDNA3 (gfx1100, gfx1101, gfx1102) from target list.
     LLVM 22, clang aborts with 'fatal error: error in backend: Unsupported
     instruction : <MCInst 4399 ...>' while compiling Loss.hip. Dropping
     gfx110x here until an upstream LLVM fix lands.
   * d/control: bump llvm-22 to fix libfmt v11 compatibility issues
 .
   [ Alexandre Detiste ]
   * reduce diff between debian/control.*

QA information

Comments

No comments