Add integration test for Emformer RNN-T LibriSpeech pipeline (#2172)

Summary:
Adds integration test for pretrained ASR pipeline `EMFORMER_RNNT_BASE_LIBRISPEECH`.

Pull Request resolved: https://github.com/pytorch/audio/pull/2172

Reviewed By: carolineechen, nateanl

Differential Revision: D33793324

Pulled By: hwangjeff

fbshipit-source-id: d0613e2ab98fe5afa7b16ca39b67f0a0304d13fc

.github/workflows/integration-test.yml

@@ -29,7 +29,7 @@ jobs:
       run: |
         python -m pip install --quiet --upgrade pip
         python -m pip install --quiet --pre torch -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html
-        python -m pip install --quiet pytest requests cmake ninja deep-phonemizer
+        python -m pip install --quiet pytest requests cmake ninja deep-phonemizer sentencepiece
         python setup.py install
       env:
         BUILD_FFMPEG: true

test/integration_tests/rnnt_pipeline_test.py

+import pytest
+import torchaudio
+from torchaudio.prototype.pipelines import EMFORMER_RNNT_BASE_LIBRISPEECH
+
+
+@pytest.mark.parametrize(
+    "bundle,lang,expected",
+    [(EMFORMER_RNNT_BASE_LIBRISPEECH, "en", "i have that curiosity beside me at this moment")],
+)
+def test_rnnt(bundle, sample_speech, expected):
+    feature_extractor = bundle.get_feature_extractor()
+    decoder = bundle.get_decoder().eval()
+    token_processor = bundle.get_token_processor()
+    waveform, _ = torchaudio.load(sample_speech)
+    features, length = feature_extractor(waveform.squeeze())
+    hypotheses = decoder(features, length, 10)
+    text = token_processor(hypotheses[0].tokens)
+    assert text == expected