commit 68ed092cc6df207fbac5c0e017563a9656195bc8
author Saketh Are <saketh@fb.com> Wed Jul 27 17:19:11 2022 +0000
committer PyTorch MergeBot <pytorchmergebot@users.noreply.github.com> Wed Jul 27 17:19:11 2022 +0000
tree e2f313533d8153ba13ba9c6e8bc1c2eeb34c35a0
parent 4a77bee6616853c70cc506c42b7a9a2a4c0f5a1f

Increase atol for test_forward_ad_nn_functional_conv_transpose3d_cuda… (#81889)

This test is currently flaky due to randomly generated inputs sometimes producing results slightly outside the specified tolerance. For example:

```
Mismatched elements: 1 / 2744 (0.0%)
Greatest absolute difference: 4.57763671875e-05 at index (2, 3, 4, 4) (up to 2e-05 allowed)
Greatest relative difference: 5.68618647611624e-05 at index (2, 3, 4, 4) (up to 2e-05 allowed)
```

Fixes #81822

Pull Request resolved: https://github.com/pytorch/pytorch/pull/81889
Approved by: https://github.com/jbschlosser

torch/testing/_internal/common_methods_invocations.py

diff --git a/torch/testing/_internal/common_methods_invocations.py b/torch/testing/_internal/common_methods_invocations.py
index 3a5967c..1b576a7 100644
--- a/torch/testing/_internal/common_methods_invocations.py
+++ b/torch/testing/_internal/common_methods_invocations.py
@@ -14159,7 +14159,7 @@
                    toleranceOverride({torch.float32: tol(atol=1.3e-04, rtol=1.3e-06), }),
                    'TestCommon', 'test_noncontiguous_samples', device_type='cuda'),
                DecorateInfo(
-                   toleranceOverride({torch.float32: tol(atol=2e-05, rtol=2e-05), }),
+                   toleranceOverride({torch.float32: tol(atol=1e-04, rtol=2e-05), }),
                    'TestCompositeCompliance', 'test_forward_ad', device_type='cuda',
                    active_if=TEST_CUDNN)],
            skips=(