pytorch - 💡(How to fix) Fix DISABLED test_vmap_two_inputs (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd) [2 comments, 1 participants]

Traceback (most recent call last): File "/var/lib/jenkins/workspace/test/dynamo/test_higher_order_ops.py", line 6386, in test_vmap_two_inputs self.assertExpectedInline( File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/internal/common_utils.py", line 3387, in assertExpectedInline return super().assertExpectedInline(actual if isinstance(actual, str) else str(actual), expect, skip + 1) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/expecttest/init.py", line 413, in assertExpectedInline assert_expected_inline( File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/expecttest/init.py", line 378, in assert_expected_inline assert_eq(expect, actual, msg=help_text) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/expecttest/init.py", line 450, in assertMultiLineEqualMaybeCppStack self.assertMultiLineEqual(expect, actual, *args, **kwargs) File "/opt/conda/envs/py_3.10/lib/python3.10/unittest/case.py", line 1226, in assertMultiLineEqual self.fail(self.formatMessage(msg, standardMsg)) File "/opt/conda/envs/py_3.10/lib/python3.10/unittest/case.py", line 675, in fail raise self.failureException(msg) AssertionError: 'clas[772 chars]sum_1 + sum_2; sum_1 = sum_2 = None\n [384 chars],)\n' != 'clas[772 chars]sum_1.add(sum_2); sum_1 = sum_2 = None\n [390 chars],)\n' class GraphModule(torch.nn.Module): def forward(self, L_x: "f32[3, 3, 3]", L_y: "f32[3, 3]"): l_x_ = L_x_ l_y_ = L_y_

      lazy_load_decompositions = torch._functorch.predispatch.lazy_load_decompositions();  lazy_load_decompositions = None

      _vmap_increment_nesting = torch._functorch.predispatch._vmap_increment_nesting(3, 'error');  _vmap_increment_nesting = None

      _add_batch_dim: "f32[3, 3]" = torch._functorch.predispatch._add_batch_dim(l_x_, 0, 1);  l_x_ = None
      _add_batch_dim_1: "f32[3]" = torch._functorch.predispatch._add_batch_dim(l_y_, 1, 1);  l_y_ = None

      sum_1: "f32[3]" = _add_batch_dim.sum(0)
      sum_2: "f32[3]" = _add_batch_dim.sum(1);  _add_batch_dim = None

•     add: "f32[3]" = sum_1 + sum_2;  sum_1 = sum_2 = None

? ^^^

•     add: "f32[3]" = sum_1.add(sum_2);  sum_1 = sum_2 = None

? ^^^^^ +

•     batched_outputs: "f32[3]" = add + _add_batch_dim_1;  add = _add_batch_dim_1 = None

? ^^^

•     batched_outputs: "f32[3]" = add.add(_add_batch_dim_1);  add = _add_batch_dim_1 = None

? ^^^^^ +

      _remove_batch_dim: "f32[3, 3]" = torch._functorch.predispatch._remove_batch_dim(batched_outputs, 1, 3, 0);  batched_outputs = None

      _vmap_decrement_nesting = torch._functorch.predispatch._vmap_decrement_nesting();  _vmap_decrement_nesting = None
      return (_remove_batch_dim,)

: To accept the new output, re-run test with envvar EXPECTTEST_ACCEPT=1 (we recommend staging/committing your changes before doing this)

To execute this test, run the following from the base repo dir: python test/dynamo/test_higher_order_ops.py FuncTorchHigherOrderOpTestsWithCompiledAutograd.test_vmap_two_inputs

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The most likely fix for the failing test is to update the test to accept the new output by re-running it with the envvar EXPECTTEST_ACCEPT=1.

Guidance

• Review the provided sample error message to understand the specific assertion error causing the test failure.
• Investigate the changes in the expected output, specifically the replacement of + with add() and add() with add.add().
• Consider re-running the test with EXPECTTEST_ACCEPT=1 to accept the new output, but first stage and commit any changes to ensure a clean test environment.
• Verify that the test passes after accepting the new output and investigate any further failures.

Example

No code snippet is provided as the issue is related to a specific test failure and the fix involves re-running the test with a specific envvar.

Notes

The provided information suggests that the test failure is due to changes in the expected output, and re-running the test with EXPECTTEST_ACCEPT=1 may resolve the issue. However, it is essential to review the changes and ensure that they are correct before accepting the new output.

Recommendation

Apply workaround: Re-run the test with EXPECTTEST_ACCEPT=1 to accept the new output, as this is the most straightforward way to resolve the test failure. This approach allows for quick verification of the fix without requiring extensive code changes.