[easy][subclasses] dynamo.reset() in test_subclass_views #128659

davidberard98 · 2024-06-13T22:39:48Z

Stack from ghstack (oldest at bottom):

When we don't dynamo.reset(), we don't recompile on different dynamic shapes.

Also, some of the returned views were tuples - so when we * 2, we actually just copy all the inputs twice in the tuple. I changed it so that it would just return one of the values from the return tuple.

Additionally, this exposes a bug that fails with the slice operation, so I skipped it when we're testing with dynamic shapes:

  File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 3996, in produce_guards
    sexpr = ShapeGuardPrinter(symbol_to_source, source_ref, self.var_to_sources).doprint(expr)
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 292, in doprint
    return self._str(self._print(expr))
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print
    return printmethod(expr, **kwargs)
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 56, in _print_Add
    t = self._print(term)
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print
    return printmethod(expr, **kwargs)
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in _print_Mul
    a_str = [self.parenthesize(x, prec, strict=False) for x in a]
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in <listcomp>
    a_str = [self.parenthesize(x, prec, strict=False) for x in a]
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 37, in parenthesize
    return self._print(item)
  File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print
    return printmethod(expr, **kwargs)
  File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 1494, in _print_Symbol
    assert self.symbol_to_source.get(expr), (
AssertionError: s3 (could be from ['<ephemeral: symint_visitor_fn>', '<ephemeral: symint_visitor_fn>']) not in {s0: ["L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]"], s1: ["L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]"], s2: ["L['x'].a.storage_offset()", "L['x'].b.storage_offset()", "L['x'].a.storage_offset()", "L['x'].b.storage_offset()"]}.  If this assert is failing, it could be due to the issue described in https://github.com/pytorch/pytorch/pull/90665

cc @cpuhrsch @jbschlosser @bhosmer @drisspg @soulitzer @ezyang @albanD @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang

When we don't dynamo.reset(), we don't recompile on different dynamic shapes. Additionally, this exposes a bug that fails with the view ``` File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 3996, in produce_guards sexpr = ShapeGuardPrinter(symbol_to_source, source_ref, self.var_to_sources).doprint(expr) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 292, in doprint return self._str(self._print(expr)) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 56, in _print_Add t = self._print(term) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in _print_Mul a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in <listcomp> a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 37, in parenthesize return self._print(item) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 1494, in _print_Symbol assert self.symbol_to_source.get(expr), ( AssertionError: s3 (could be from ['<ephemeral: symint_visitor_fn>', '<ephemeral: symint_visitor_fn>']) not in {s0: ["L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]"], s1: ["L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]"], s2: ["L['x'].a.storage_offset()", "L['x'].b.storage_offset()", "L['x'].a.storage_offset()", "L['x'].b.storage_offset()"]}. If this assert is failing, it could be due to the issue described in #90665 ``` [ghstack-poisoned]

pytorch-bot · 2024-06-13T22:39:51Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/128659

📄 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.

✅ No Failures

As of commit c6ac11e with merge base a265556 ():
💚 Looks good so far! There are no failures yet. 💚

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

When we don't dynamo.reset(), we don't recompile on different dynamic shapes. Also, some of the returned views were tuples - so when we `* 2`, we actually just copy all the inputs twice in the tuple. I changed it so that it would just return one of the values from the return tuple. Additionally, this exposes a bug that fails with the slice operation, so I skipped it when we're testing with dynamic shapes: ``` File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 3996, in produce_guards sexpr = ShapeGuardPrinter(symbol_to_source, source_ref, self.var_to_sources).doprint(expr) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 292, in doprint return self._str(self._print(expr)) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 56, in _print_Add t = self._print(term) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in _print_Mul a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in <listcomp> a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 37, in parenthesize return self._print(item) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 1494, in _print_Symbol assert self.symbol_to_source.get(expr), ( AssertionError: s3 (could be from ['<ephemeral: symint_visitor_fn>', '<ephemeral: symint_visitor_fn>']) not in {s0: ["L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]"], s1: ["L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]"], s2: ["L['x'].a.storage_offset()", "L['x'].b.storage_offset()", "L['x'].a.storage_offset()", "L['x'].b.storage_offset()"]}. If this assert is failing, it could be due to the issue described in #90665 ``` cc cpuhrsch jbschlosser bhosmer drisspg soulitzer ezyang albanD voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang [ghstack-poisoned]

davidberard98 · 2024-06-14T01:05:01Z

@pytorchbot merge

pytorchmergebot · 2024-06-14T01:06:36Z

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

) When we don't dynamo.reset(), we don't recompile on different dynamic shapes. Also, some of the returned views were tuples - so when we `* 2`, we actually just copy all the inputs twice in the tuple. I changed it so that it would just return one of the values from the return tuple. Additionally, this exposes a bug that fails with the slice operation, so I skipped it when we're testing with dynamic shapes: ``` File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 3996, in produce_guards sexpr = ShapeGuardPrinter(symbol_to_source, source_ref, self.var_to_sources).doprint(expr) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 292, in doprint return self._str(self._print(expr)) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 56, in _print_Add t = self._print(term) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in _print_Mul a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 366, in <listcomp> a_str = [self.parenthesize(x, prec, strict=False) for x in a] File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/str.py", line 37, in parenthesize return self._print(item) File "/home/dberard/local/miniconda3/envs/pytorch/lib/python3.10/site-packages/sympy/printing/printer.py", line 331, in _print return printmethod(expr, **kwargs) File "/home/dberard/local/pytorch/torch/fx/experimental/symbolic_shapes.py", line 1494, in _print_Symbol assert self.symbol_to_source.get(expr), ( AssertionError: s3 (could be from ['<ephemeral: symint_visitor_fn>', '<ephemeral: symint_visitor_fn>']) not in {s0: ["L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]", "L['x'].a.size()[1]", "L['x'].b.size()[1]"], s1: ["L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]", "L['x'].a.stride()[0]", "L['x'].b.stride()[0]"], s2: ["L['x'].a.storage_offset()", "L['x'].b.storage_offset()", "L['x'].a.storage_offset()", "L['x'].b.storage_offset()"]}. If this assert is failing, it could be due to the issue described in pytorch#90665 ``` Pull Request resolved: pytorch#128659 Approved by: https://github.com/YuqingJ

pytorch-bot bot added module: dynamo topic: not user facing topic category labels Jun 13, 2024

davidberard98 added module: nestedtensor NestedTensor tag see issue #25032 tensor subclass Related to tensor subclasses labels Jun 13, 2024

davidberard98 mentioned this pull request Jun 13, 2024

[subclasses] Handle dynamo inputs that are subclass views with (-1) in the view #128662

Closed

davidberard98 marked this pull request as ready for review June 13, 2024 23:36

davidberard98 requested review from YuqingJ, bdhirsh, jbschlosser and soulitzer June 13, 2024 23:38

YuqingJ approved these changes Jun 14, 2024

View reviewed changes

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Jun 14, 2024

pytorchmergebot added the merging label Jun 14, 2024

pytorchmergebot added the Merged label Jun 14, 2024

pytorchmergebot closed this in f48f761 Jun 14, 2024

pytorchmergebot removed the merging label Jun 14, 2024

github-actions bot deleted the gh/davidberard98/308/head branch July 15, 2024 01:59

jbschlosser mentioned this pull request Aug 27, 2024

Update TwoTensor impl. to accept outer_size/outer_stride #133337

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[easy][subclasses] dynamo.reset() in test_subclass_views #128659

[easy][subclasses] dynamo.reset() in test_subclass_views #128659

Uh oh!

davidberard98 commented Jun 13, 2024 •

edited

Loading

Uh oh!

pytorch-bot bot commented Jun 13, 2024 •

edited

Loading

Uh oh!

davidberard98 commented Jun 14, 2024

Uh oh!

pytorchmergebot commented Jun 14, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

[easy][subclasses] dynamo.reset() in test_subclass_views #128659

[easy][subclasses] dynamo.reset() in test_subclass_views #128659

Uh oh!

Conversation

davidberard98 commented Jun 13, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Jun 13, 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/128659

✅ No Failures

Uh oh!

davidberard98 commented Jun 14, 2024

Uh oh!

pytorchmergebot commented Jun 14, 2024

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

davidberard98 commented Jun 13, 2024 •

edited

Loading

pytorch-bot bot commented Jun 13, 2024 •

edited

Loading