Symintify baddbmm

[angelayi]

Previously we would specialize on the shape in this if-statement

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/154656

• 📄 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.

❌ 4 New Failures, 2 Unrelated Failures

As of commit d4c0853 with merge base aa84c03 ():

NEW FAILURES - The following jobs have failed:

• inductor / cuda12.6-py3.10-gcc9-sm86 / test (inductor_torchbench, 2, 2, linux.g5.4xlarge.nvidia.gpu) (gh)
  opacus_cifar10
• inductor / linux-jammy-cpu-py3.9-gcc11-inductor / test (cpu_inductor_torchbench, 2, 2, linux.8xlarge.amx) (gh)
  opacus_cifar10
• inductor / linux-jammy-cpu-py3.9-gcc11-inductor / test (dynamic_cpu_inductor_torchbench, 2, 2, linux.8xlarge.amx) (gh)
  opacus_cifar10
• pull / cuda12.8-py3.10-gcc9-sm75 / build (gh)

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• trunk / linux-jammy-rocm-py3.10 / test (distributed, 1, 1, linux.rocm.gpu.4) (gh) (disabled by #139011 but the issue was closed recently and a rebase is needed to make it pass)
  distributed/test_c10d_nccl.py::ProcessGroupNCCLGroupTest::test_extra_cuda_context

UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:

• pull / unstable-linux-focal-cuda12.6-py3.10-gcc11-sm89-xfail / build (gh)
  ninja: build stopped: subcommand failed

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

[pianpwk]

sym_and takes any number of arguments, so you can write this as a 3-way.

Also why are we expanding on equality, instead of inequality now? wouldn't it be guard_or_true(not sym_and(self.shape[0] == dim1, self.shape[1] == dim2, selfl.shape[2] == dim3))

[angelayi]

oh um.. you're right

[angelayi]

oh sorry I think I forgot the not... but guard_or_true didn't work for me

[laithsakka]

probably you want sym_eq here,
this should work for your test case.

  if not guard_or_false(sym_eq(self.shape, (dim1, dim2, dim3))):
        self = self.expand((dim1, dim2, dim3))

is this what you want here?
i verified this work on your example. but not sure if that have other consequences?
does calling self.expand((dim1, dim2, dim3)) in theory works for when
self.shape == (dim1, dim2, dim3)?

what did not work with gaurd_or_true?

[laithsakka]

those also should work

    if guard_or_true(sym_eq(self.shape, (dim1, dim2, dim3))==False):
        self = self.expand((dim1, dim2, dim3))

I think we need a sym_not , having not will specialize
cc @pianpwk

[laithsakka]

could be

def sym_not (a):
  return a==False

then you can do

    if guard_or_true(sym_not(sym_eq(self.shape, (dim1, dim2, dim3)))):
       self = self.expand((dim1, dim2, dim3))

[angelayi]

what did not work with gudrd_or_true?

I think originally doing guard_or_true(not sym_and(self.shape[0] == dim1, self.shape[1] == dim2, self.shape[2] == dim3)) didn't work because of the not in the middle of the expression. Your above examples all work!

[angelayi]

turns out there already exists torch.sym_not :D

[laithsakka]

left some comments.

[angelayi]

@pytorchbot merge -f "can repro failures on main"

[pytorchmergebot]

Merge started

Your change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use -f as last resort and instead consider -i/--ignore-current to continue the merge ignoring current failures. This will allow currently pending tests to finish and report signal before the merge.

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here