Skip to content

Commit

Permalink
Increase tolerance for audio MAE test (#476)
Browse files Browse the repository at this point in the history
Summary:
Audio MAE test is failing on CI due to some small rounding error. The unmasked test case already has atol=rtol=1e-4, so set the same values for the masked test case to fix the CI.

Pull Request resolved: #476

Test Plan: Confirm that test no longer fails on CI

Reviewed By: kartikayk

Differential Revision: D49661270

Pulled By: ebsmothers

fbshipit-source-id: 0b95ad2b1de961042775985b9d5089c01440450e
  • Loading branch information
ebsmothers authored and facebook-github-bot committed Sep 27, 2023
1 parent 0793eb4 commit 5a0952c
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion tests/models/masked_auto_encoder/test_model.py
Original file line number Diff line number Diff line change
Expand Up @@ -253,7 +253,7 @@ def test_audio_mae_train_masking(self, inputs):

pred = actual.decoder_pred
assert_expected(pred.size(), (2, 512, 16 * 16 * 1))
assert_expected(pred.mean().item(), 513.0)
assert_expected(pred.mean().item(), 513.0, rtol=1e-4, atol=1e-4)

labels = actual.label_patches
assert_expected(labels, torch.ones(2, 512, 16 * 16 * 1))
Expand Down

0 comments on commit 5a0952c

Please sign in to comment.