Commit 72b0917d authored by hwangjeff's avatar hwangjeff Committed by Facebook GitHub Bot
Browse files

Relax Conformer RNN-T numerical parity tests (#3525)

Summary:
Increases numerical tolerance on Conformer RNN-T TorchScript consistency tests to resolve CI test failures.

Pull Request resolved: https://github.com/pytorch/audio/pull/3525

Reviewed By: mthrok

Differential Revision: D48000613

Pulled By: hwangjeff

fbshipit-source-id: 1d35ba58055a8346dc40e2b67f37ccfd2e015894
parent 3f9b5171
......@@ -106,10 +106,10 @@ class ConformerRNNTTestImpl(TestBaseMixin):
scripted_state,
) = scripted(inputs, input_lengths, targets, target_lengths, scripted_state)
self.assertEqual(ref_out, scripted_out)
self.assertEqual(ref_input_lengths, scripted_input_lengths)
self.assertEqual(ref_target_lengths, scripted_target_lengths)
self.assertEqual(ref_state, scripted_state)
self.assertEqual(ref_out, scripted_out, atol=1e-4, rtol=1e-5)
self.assertEqual(ref_input_lengths, scripted_input_lengths, atol=1e-4, rtol=1e-5)
self.assertEqual(ref_target_lengths, scripted_target_lengths, atol=1e-4, rtol=1e-5)
self.assertEqual(ref_state, scripted_state, atol=1e-4, rtol=1e-5)
def test_torchscript_consistency_transcribe(self):
r"""Verify that scripting RNNT does not change the behavior of method `transcribe`."""
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment