🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/186194

• 📄 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

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

✅ No Failures

As of commit e2a7273 with merge base 16de8e3 ():
💚 Looks good so far! There are no failures yet. 💚

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

This PR needs a release notes: label

If your changes are user facing and intended to be a part of release notes, please use a label starting with release notes:.

If not, please add the topic: not user facing label.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "topic: not user facing"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

@claude review these changes

Claude finished @jansel's task in 1m 32s —— View job

Review of PR #186194: Canonicalize trunc of symbolic integer division by one

• Read the diff and changed files
• Review torch/utils/_sympy/functions.py changes
• Review test/test_dynamic_shapes.py changes
• Review test/export/test_export.py changes
• Post review feedback

Overall: This looks good. The fix is well-scoped and placed at the right layer.

torch/utils/_sympy/functions.py (lines 1268–1275)

The two simplification rules are correct:

1. number.is_integer is True → number — Truncating an integer is a no-op. Using is True correctly excludes the None (unknown) case from SymPy's three-valued logic.

2. IntTrueDiv(base, ±1) pattern match — Since IntTrueDiv has is_real = True (not is_integer), SymPy won't infer that IntTrueDiv(s, 1) is integer-valued, so this explicit pattern match is necessary and won't be subsumed by the is_integer check above it.

The placement is correct: handling this in TruncToInt rather than in IntTrueDiv.eval preserves the type semantics — IntTrueDiv models float-returning division, while TruncToInt is the float→int conversion boundary where the composition TruncToInt(IntTrueDiv(s, 1)) can safely collapse to s.

Minor observation (not blocking)

CeilToInt and FloorToInt lack the number.is_integer is True early-return that TruncToInt now has (and that FloorToInt partially covers with isinstance(number, sympy.Integer)). If similar ceil(x/1) or floor(x/1) patterns ever arise, they'd hit the same issue. But that's a separate concern — this PR correctly fixes the reported problem without scope creep.

Tests

Both tests are well-structured:

• The unit test in test_dynamic_shapes.py directly asserts the symbolic expression simplifies (r.node.expr == a1.node.expr).
• The export test reproduces the exact torch.cond metadata mismatch from torch.cond complains two tensors have different metadata  #138926 with dynamic shapes and F.interpolate.

No issues found. LGTM.