Delay mul/pow expansion for `_SympyT` to enable more folding #138235

StrongerXi · 2024-10-17T17:43:42Z

Instead of calling safe_expand right after symbolic expression construction, we invoke it in ShapeEnv.simplify. This enables more simplification with product form, e.g.,

(a + b)^2 / (a + b) --> (a + b)

which won't happen if we expand eagerly during product construction:

(a^2 + 2ab + b^2) / (a + b) --> no change

Fixes #136044.

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

pytorch-bot · 2024-10-17T17:43:46Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138235

📄 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.

❌ 5 New Failures, 2 Unrelated Failures

As of commit ca264be with merge base f3c3f3a ():

NEW FAILURES - The following jobs have failed:

linux-binary-libtorch-cxx11-abi / libtorch-cpu-shared-with-deps-cxx11-abi-build / build (gh)
bash: /builder/libtorch/build.sh: No such file or directory
linux-binary-libtorch-pre-cxx11 / libtorch-cpu-shared-with-deps-pre-cxx11-build / build (gh)
bash: /builder/libtorch/build.sh: No such file or directory
linux-binary-manywheel / manywheel-py3_9-cuda11_8-build / build (gh)
bash: /builder/manywheel/build.sh: No such file or directory
linux-binary-manywheel / manywheel-py3_9-cuda12_1-build / build (gh)
bash: /builder/manywheel/build.sh: No such file or directory
linux-binary-manywheel / manywheel-py3_9-cuda12_4-build / build (gh)
bash: /builder/manywheel/build.sh: No such file or directory

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

inductor / linux-jammy-cpu-py3.9-gcc11-inductor / test (inductor_torchbench_cpu_smoketest_perf, 1, 1, linux.24xl.spr-metal) (gh) (similar failure)
RuntimeError: Task <Task pending name='Task-1' coro=<FbScribeLogger._worker() running at /opt/conda/envs/py_3.9/lib/python3.9/site-packages/fbscribelogger/__init__.py:214> cb=[_run_until_complete_cb() at /opt/conda/envs/py_3.9/lib/python3.9/asyncio/base_events.py:184]> got Future <Future pending> attached to a different loop
inductor-periodic / cuda12.1-py3.10-gcc9-sm80 / test (inductor_torchbench_smoketest_perf, 1, 1, linux.gcp.a100) (gh) (detected as infra flaky with no log or failing log classifier)

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

StrongerXi · 2024-10-17T21:20:19Z

Update benchmarks/dynamo/pr_time_benchmarks/expected_results.csv
Rebase over f3c3f3a for inductor-A100-perf-nightly: https://github.com/pytorch/pytorch/actions/runs/11393237966
- Benchmark results

StrongerXi · 2024-10-18T15:17:41Z

benchmarks/dynamo/pr_time_benchmarks/expected_results.csv

+add_loop_eager_dynamic,        compile_time_instruction_count, 5563298740,  0.025
 add_loop_inductor,             compile_time_instruction_count, 24064639114, 0.015
-add_loop_inductor_dynamic_gpu, compile_time_instruction_count, 40992578178, 0.025
+add_loop_inductor_dynamic_gpu, compile_time_instruction_count, 38593682350, 0.025


This change turned out to be a flake, I'll revert.

Instead of calling `safe_expand` right after symbolic expression construction, we invoke it in `ShapeEnv.simplify`. This enables more simplification with product form, e.g., ``` (a + b)^2 / (a + b) --> (a + b) ``` which won't happen if we expand eagerly during product construction: ``` (a^2 + 2ab + b^2) / (a + b) --> no change ``` Fixes #136044.

ezyang

Wow! Passing all tests. You probably should also import to fbcode and see what happens there.

facebook-github-bot · 2024-10-18T21:14:51Z

@StrongerXi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

StrongerXi · 2024-10-21T15:12:30Z

Wow! Passing all tests. You probably should also import to fbcode and see what happens there.

Internals all passed. The red ones are OSS flaky tests from main. Merging

StrongerXi · 2024-10-21T15:12:36Z

@pytorchbot merge

pytorchmergebot · 2024-10-21T15:14:28Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

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

pytorchmergebot · 2024-10-21T15:24:54Z

Merge failed

Reason: 1 jobs have failed, first few of them are: linux-binary-manywheel / manywheel-py3_9-cuda12_1-build / build

Details for Dev Infra team

Raised by workflow job

StrongerXi · 2024-10-21T16:37:07Z

@pytorchbot merge -f "unrelated failure"

pytorchmergebot · 2024-10-21T16:38:37Z

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

ezyang · 2024-10-22T02:37:31Z

This improves compile time on detectron2_fcos_r_50_fpn inference compilation:

Nice work!

pytorch-bot bot added ciflow/inductor release notes: fx release notes category labels Oct 17, 2024

StrongerXi force-pushed the ryanguo99/enhance-modulo-folding branch from 58b8608 to edf9cb9 Compare October 17, 2024 21:19

pytorch-bot bot added the module: dynamo label Oct 17, 2024

StrongerXi commented Oct 18, 2024

View reviewed changes

StrongerXi force-pushed the ryanguo99/enhance-modulo-folding branch from edf9cb9 to bd1f72e Compare October 18, 2024 16:48

StrongerXi force-pushed the ryanguo99/enhance-modulo-folding branch from bd1f72e to ca264be Compare October 18, 2024 16:49

StrongerXi changed the title ~~[EXPERIMENT][DO NOT LAND] Delay mul/pow expansion for _SympyT~~ Delay mul/pow expansion for _SympyT to enable more folding Oct 18, 2024

StrongerXi requested a review from ezyang October 18, 2024 16:50

ezyang approved these changes Oct 18, 2024

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Oct 21, 2024

pytorchmergebot added the merging label Oct 21, 2024

pytorchmergebot removed the merging label Oct 21, 2024

pytorchmergebot added the merging label Oct 21, 2024

pytorchmergebot added the Merged label Oct 21, 2024

pytorchmergebot closed this in 0a41974 Oct 21, 2024

pytorchmergebot removed the merging label Oct 21, 2024

github-actions bot deleted the ryanguo99/enhance-modulo-folding branch November 22, 2024 02:08

Delay mul/pow expansion for _SympyT to enable more folding #138235

Delay mul/pow expansion for _SympyT to enable more folding #138235

Uh oh!

Conversation

StrongerXi commented Oct 17, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Oct 17, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138235

❌ 5 New Failures, 2 Unrelated Failures

Uh oh!

StrongerXi commented Oct 17, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

StrongerXi Oct 18, 2024

Choose a reason for hiding this comment

Uh oh!

ezyang left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Oct 18, 2024

Uh oh!

StrongerXi commented Oct 21, 2024

Uh oh!

StrongerXi commented Oct 21, 2024

Uh oh!

pytorchmergebot commented Oct 21, 2024

Merge started

Uh oh!

pytorchmergebot commented Oct 21, 2024

Merge failed

Uh oh!

StrongerXi commented Oct 21, 2024

Uh oh!

pytorchmergebot commented Oct 21, 2024

Merge started

Uh oh!

ezyang commented Oct 22, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Delay mul/pow expansion for `_SympyT` to enable more folding #138235

Delay mul/pow expansion for `_SympyT` to enable more folding #138235

StrongerXi commented Oct 17, 2024 •

edited

Loading

pytorch-bot bot commented Oct 17, 2024 •

edited

Loading

StrongerXi commented Oct 17, 2024 •

edited

Loading