Code

pytorch v2.6.0#326

Code

Merged

h-vetinari merged 57 commits intoconda-forge:mainconda-forge/pytorch-cpu-feedstock:mainfrom h-vetinari:2.6h-vetinari/pytorch-cpu-feedstock:2.6Copy head branch name to clipboardFeb 13, 2025

+379-627

Commits

Commits on Jan 18, 2025

• pytorch v2.6.0.rc7

  h-vetinari

  committed
  11db2ff

  Copy full SHA for 11db2ff

• MNT: Re-rendered with conda-build 25.1.1, conda-smithy 3.45.2, and conda-forge-pinning 2025.01.18.07.29.32

  h-vetinari

  committed
  5645301

  Copy full SHA for 5645301

• rebase patches

  h-vetinari

  committed
  1219e26

  Copy full SHA for 1219e26

Commits on Jan 19, 2025

• Remove bound on setuptools version

  Show description for 1e211b2
  h-vetinari

  committed
  1e211b2

  Copy full SHA for 1e211b2

• better sort pytorch run requirements

  h-vetinari

  committed
  63c7c00

  Copy full SHA for 63c7c00

• update dependencies for 2.6.0

  h-vetinari

  committed
  bc3581d

  Copy full SHA for bc3581d

• pin libtorch in pytorch

  h-vetinari

  committed
  113d413

  Copy full SHA for 113d413

Commits on Jan 20, 2025

• backport fix for aarch build failure

  h-vetinari

  committed
  c1b9cf6

  Copy full SHA for c1b9cf6

• reduce OMP_NUM_THREADS further due to OOMs in test

  h-vetinari

  committed
  bf65aa4

  Copy full SHA for bf65aa4

• use full paths to patched files in submodule

  h-vetinari

  committed
  6757593

  Copy full SHA for 6757593

• don't pin libtorch exactly for non-megabuilds

  h-vetinari

  committed
  fa31c3f

  Copy full SHA for fa31c3f

• make build number arithmetic explicit

  Show description for 49d134a
  h-vetinari

  committed
  49d134a

  Copy full SHA for 49d134a

• disable patch in submodule

  h-vetinari

  committed
  7cd981d

  Copy full SHA for 7cd981d

• don't pin build number of libtorch in pytorch (for non-megabuild)

  Show description for 1405263
  h-vetinari

  committed
  1405263

  Copy full SHA for 1405263

Commits on Feb 3, 2025

• Merge remote-tracking branch 'upstream/main' into 2.6

  h-vetinari

  committed
  51155b0

  Copy full SHA for 51155b0

• pytorch v2.6.0

  h-vetinari

  committed
  be1ff5b

  Copy full SHA for be1ff5b

• MNT: Re-rendered with conda-build 25.1.1, conda-smithy 3.45.4, and conda-forge-pinning 2025.01.28.22.50.30

  h-vetinari

  committed
  95eb614

  Copy full SHA for 95eb614

• remove test skips for 3.13

  h-vetinari

  committed
  beb5481

  Copy full SHA for beb5481

• vendor upstream smoke_test.py; not part of tarball

  h-vetinari

  committed
  7c0d921

  Copy full SHA for 7c0d921

• skip torchinductor tests on osx; segfaults hard

  Show description for b627641
  h-vetinari

  committed
  b627641

  Copy full SHA for b627641

• backport patch for torchinductor failures on windows

  h-vetinari

  committed
  c12f03d

  Copy full SHA for c12f03d

• skip a minor tolerance violation on osx

  h-vetinari

  committed
  ce47e0a

  Copy full SHA for ce47e0a

• drop unsuccessful patch; now covered by skip

  h-vetinari

  committed
  ba2564b

  Copy full SHA for ba2564b

• drop unnecessary CUDAFLAGS in bld.bat

  Show description for ab889b7
  h-vetinari

  committed
  ab889b7

  Copy full SHA for ab889b7

Commits on Feb 4, 2025

• skip test_batchnorm_nhwc_cpu on linux as well

  h-vetinari

  committed
  194f2d9

  Copy full SHA for 194f2d9

• disable smoke test on aarch

  h-vetinari

  committed
  8804810

  Copy full SHA for 8804810

Commits on Feb 5, 2025

• Revert "Unvendor pybind11 and eigen"

  Show description for 93854a6
  h-vetinari

  committed
  93854a6

  Copy full SHA for 93854a6

• use different formulation for ptxas warning filter

  h-vetinari

  committed
  b4ab27b

  Copy full SHA for b4ab27b

• re-add zlib to pytorch's host deps

  h-vetinari

  committed
  f72a13e

  Copy full SHA for f72a13e

• remove smoke test

  h-vetinari

  committed
  b1658ef

  Copy full SHA for b1658ef

Commits on Feb 7, 2025

• deactivate pytest-xdist for linux-64+CUDA+MKL

  h-vetinari

  committed
  8a43a63

  Copy full SHA for 8a43a63

• temporary: skip known-passing builds to conserve resources

  h-vetinari

  committed
  1da1037

  Copy full SHA for 1da1037

Commits on Feb 8, 2025

• Revert "deactivate pytest-xdist for linux-64+CUDA+MKL"

  Show description for 4784f25
  h-vetinari

  committed
  4784f25

  Copy full SHA for 4784f25

• fix stray ) that broke pytest

  h-vetinari

  committed
  51109f4

  Copy full SHA for 51109f4

• fix the way we look for $PREFIX/include in cross-compilation

  h-vetinari

  committed
  796a52b

  Copy full SHA for 796a52b

• Revert "temporary: skip known-passing builds to conserve resources"

  Show description for d72a0fd
  h-vetinari

  committed
  d72a0fd

  Copy full SHA for d72a0fd

• Reapply "Unvendor pybind11 and eigen"

  Show description for 96d3954
  h-vetinari

  committed
  96d3954

  Copy full SHA for 96d3954

• different spelling for handing -w to -Xptxas

  h-vetinari

  committed
  051b686

  Copy full SHA for 051b686

• skip test_mutable_custom_op_fixed_layout everywhere

  h-vetinari

  committed
  71e319d

  Copy full SHA for 71e319d

• robustify CONDA_BUILD_CROSS_COMPILATION handling

  h-vetinari

  committed
  de7707c

  Copy full SHA for de7707c

Commits on Feb 9, 2025

• set CUDA_VERSION during build; test torch.version.cuda is correct

  h-vetinari

  committed
  381fcb8

  Copy full SHA for 381fcb8

• enable torch.backends tests also for win+CUDA

  h-vetinari

  committed
  04e1437

  Copy full SHA for 04e1437

Commits on Feb 10, 2025

• Build with -Wl,-z,noexecstack to fix glibc 2.41 compatibility

  Show description for 272621b
  

  mgorny

  authored and

  h-vetinari

  committed
  272621b

  Copy full SHA for 272621b

Commits on Feb 11, 2025

• add v2.5.x to abi_migration_branches

  h-vetinari

  committed
  b254812

  Copy full SHA for b254812

• actually move _C.lib back to %SP_DIR%\torch\lib

  h-vetinari

  committed
  2eb8ef4

  Copy full SHA for 2eb8ef4

Commits on Feb 12, 2025

• avoid very verbose logs from setuptools

  Show description for 22ee0fe
  

  mgorny

  and

  h-vetinari

  committed
  22ee0fe

  Copy full SHA for 22ee0fe

• homogenize pytest invocation; also use -v on unix

  Show description for 47f7f21
  h-vetinari

  committed
  47f7f21

  Copy full SHA for 47f7f21

• filter out ptxas advisories

  h-vetinari

  committed
  35d03ed

  Copy full SHA for 35d03ed

• suppress warning block in pytest results

  h-vetinari

  committed
  6d2a9a6

  Copy full SHA for 6d2a9a6

• run tests sequentially, but increase OMP_NUM_THREADS again

  h-vetinari

  committed
  d31229e

  Copy full SHA for d31229e

• skip test failures that were not due to OOM

  h-vetinari

  committed
  364ad9d

  Copy full SHA for 364ad9d

• temporary: restrict jobs to linux64+CUDA

  h-vetinari

  committed
  f14feb5

  Copy full SHA for f14feb5

• minor skip updates

  h-vetinari

  committed
  2e75510

  Copy full SHA for 2e75510

• ignore C++ warnings

  h-vetinari

  committed
  db4e301

  Copy full SHA for db4e301

• skip some tests that take a very long time

  h-vetinari

  committed
  5640240

  Copy full SHA for 5640240

• Revert "temporary: restrict jobs to linux64+CUDA"

  Show description for 395bea3
  h-vetinari

  committed
  395bea3

  Copy full SHA for 395bea3

• avoid filter for ptxas advisories on osx

  h-vetinari

  committed
  a75ba71

  Copy full SHA for a75ba71