update triton pin for release 2.3

[shunting314]

The current status of the PR:

• CI is clean. The only failures not related to building a wheel also happens on the base commit.
• Perf test is mostly neutral link. The is an extra failed model in torchbench. But the error type 'eager_two_runs_differ' indicates it's not related to triton upgrade. We've double checked that the perf test job is using the correct triton pin (thanks to @huydhn and @atalman )

I'll now clean up the PR to only keep the necessary PyTorch side change to make it work with tirton 2.2.x . Other hacks coming up to do the tests will be removed (e.g. pointing to my triton fork).

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler @amjames @desertfire @chauhang

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/121781

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

❌ 21 New Failures, 7 Unrelated Failures

As of commit d0933e6 with merge base 86a2d67 ():

NEW FAILURES - The following jobs have failed:

• Build Triton wheels / Build Triton Conda (3.10) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Conda (3.11) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Conda (3.12) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Conda (3.8) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Conda (3.9) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Wheel (3.10, cuda) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Wheel (3.11, cuda) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Wheel (3.12, cuda) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Wheel (3.8, cuda) (gh)
  Process completed with exit code 1.
• Build Triton wheels / Build Triton Wheel (3.9, cuda) (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_10-cuda12_1-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_10-rocm5_7-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_11-cuda11_8-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_11-cuda12_1-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_11-rocm5_7-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_11-rocm6_0-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_8-rocm5_7-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_9-cuda11_8-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_9-cuda12_1-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_9-rocm5_7-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_9-rocm6_0-test (gh)
  Process completed with exit code 1.

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• linux-binary-manywheel / manywheel-py3_10-cuda11_8-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_10-rocm6_0-test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_8-cuda11_8-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_8-cuda12_1-test / test (gh)
  Process completed with exit code 1.
• linux-binary-manywheel / manywheel-py3_8-rocm6_0-test (gh)
  Process completed with exit code 1.
• trunk / macos-12-py3-arm64 / test (default, 3, 3, macos-m1-stable) (gh)
  test_tensorboard.py::TestTensorBoardSummary::test_hparams_bool

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• rocm / linux-focal-rocm6.0-py3.8 / test (default, 6, 6, linux.rocm.gpu.2) (gh)
  inductor/test_triton_heuristics.py::TestTritonHeuristics::test_artificial_zgrid

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[shunting314]

cc @atalman for the error regarding building triton wheels

[shunting314]

We also see a bunch of GLIBCXX related failures. I'm worried that they may hide other triton related issues. @atalman

[peterbell10]

GLIBCXX error is fixed by #121556

[shunting314]

CI is finally clean! The only failures not related to building a wheel also happens on the base commit. Will check how perf looks

[shunting314]

I've create a separate PR to only include the necessary PyTorch side change to make it work with the triton release (#122139). Keep this PR around since all the tests are done in this PR