Revert "[PT2][fusion] ban fusions with large accumulated reads (#157563)

[jataylo]

This reverts commit 8554c80 #157563 due to causing a few breakages on ROCm

Reverted expected_results.csv to 26807dc

@xuanzhang816 Sorry, but I have to revert this PR yet again because it clearly reintroduced failures on ROCm after the remerge: https://hud.pytorch.org/hud/pytorch/pytorch/f4d8bc46c7706f872abcb4ec41f0b32207d5d826/2?per_page=50&name_filter=rocm-mi300&mergeEphemeralLF=true
and the failures are still showing up on tip-of-tree on HUD

Context
#157563 (comment)

Needs to be relanded in non bc-breaking way, or sanity checked for correctness.
cc: @xuanzhang816

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

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/158550

• 📄 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.

✅ You can merge normally! (10 Unrelated Failures)

As of commit 7323802 with merge base eeda1a7 ():

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

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

• inductor / cuda12.8-py3.10-gcc9-sm86 / test (inductor_torchbench, 1, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (trunk failure)
  torchrec_dlrm
• inductor-rocm / rocm-py3.10-inductor / test (inductor, 1, 2, linux.rocm.gpu.2) (gh) (trunk failure)
  distributed/test_dynamo_distributed.py::TestMultiProc::test_ddp_optimizer_cudagraph
• inductor-rocm-mi300 / rocm-py3.10-inductor-mi300 / test (inductor, 1, 2, linux.rocm.gpu.mi300.2) (gh) (trunk failure)
  distributed/test_dynamo_distributed.py::TestMultiProc::test_ddp_optimizer_cudagraph
• pull / linux-jammy-cuda12.8-py3.10-gcc11-sm89 / test (default, 5, 5, linux.g6.4xlarge.experimental.nvidia.gpu) (gh) (trunk failure)
  inductor/test_codecache.py::TestFxGraphCache::test_flex_attention_caching

UNSTABLE - The following jobs are marked as unstable, possibly due to flakiness on trunk:

• pull / cuda12.8-py3.10-gcc9-sm75 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu) (gh) (#153987)
  MISSING REGRESSION TEST
• rocm / linux-jammy-rocm-py3.10 / test (default, 1, 6, linux.rocm.gpu.2, unstable) (gh) (#156098)
  test_nn.py::TestNNDeviceTypeCUDA::test_avg_pool_large_tensor2_cuda
• rocm / linux-jammy-rocm-py3.10 / test (default, 2, 6, linux.rocm.gpu.2, unstable) (gh) (#156098)
  inductor/test_cudagraph_trees_expandable_segments.py::CudaGraphTreeTests::test_function_compiled_multiple_times
• rocm / linux-jammy-rocm-py3.10 / test (default, 4, 6, linux.rocm.gpu.2, unstable) (gh) (#156098)
  inductor/test_cudagraph_trees.py::CudaGraphTreeTests::test_function_compiled_multiple_times
• rocm-mi300 / linux-noble-rocm-py3.12-mi300 / test (default, 1, 6, linux.rocm.gpu.mi300.2, unstable) (gh) (#158182)
  test_nn.py::TestNNDeviceTypeCUDA::test_avg_pool_large_tensor2_cuda
• rocm-mi300 / linux-noble-rocm-py3.12-mi300 / test (default, 3, 6, linux.rocm.gpu.mi300.2, unstable) (gh) (#158182)
  inductor/test_cudagraph_trees_expandable_segments.py::CudaGraphTreeTests::test_function_compiled_multiple_times

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

[jataylo]

test/inductor/test_max_autotune.py::TestMaxAutotune::test_linear_and_cel -

• MI200: https://github.com/pytorch/pytorch/actions/runs/16342019541/job/46168809096
• MI300: https://github.com/pytorch/pytorch/actions/runs/16342019691/job/46169040211

test/inductor/test_inplace_padding.py::InplacePaddingTest::test_linear_and_cel_max_autotune -

• MI200: https://github.com/pytorch/pytorch/actions/runs/16342019541/job/46168809102
• MI300: https://github.com/pytorch/pytorch/actions/runs/16342019691/job/46169040246

Passing post revert, failures are unrelated.

[jithunnair-amd]

@pytorchbot merge -f "This is a manual revert of PR 157563 to restore ROCm CI to green"

[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