SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix some tests using "common_voice" (#27147)

Browse Source 
* Use mozilla-foundation/common_voice_11_0

* Update expected values

* Update expected values

* For test_word_time_stamp_integration

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
This commit is contained in:
Yih-Dar
2023-10-30 15:27:15 +01:00
committed by GitHub
parent 691fd8fdde
commit 576994963f
2 changed files with 17 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
tests/models/wav2vec2/test_modeling_wav2vec2.py
View File

@@ -97,7 +97,7 @@ def _test_wav2vec2_with_lm_invalid_pool(in_queue, out_queue, timeout):
try:
_ = in_queue.get(timeout=timeout)
ds = load_dataset("common_voice", "es", split="test", streaming=True)
ds = load_dataset("mozilla-foundation/common_voice_11_0", "es", split="test", streaming=True)
sample = next(iter(ds))
resampled_audio = torchaudio.functional.resample(
@@ -119,7 +119,7 @@ def _test_wav2vec2_with_lm_invalid_pool(in_queue, out_queue, timeout):
transcription = processor.batch_decode(logits.cpu().numpy(), pool).text
unittest.TestCase().assertIn("Falling back to sequential decoding.", cl.out)
unittest.TestCase().assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")
unittest.TestCase().assertEqual(transcription[0], "habitan aguas poco profundas y rocosas")
# force batch_decode to internally create a spawn pool, which should trigger a warning if different than fork
multiprocessing.set_start_method("spawn", force=True)
@@ -127,7 +127,7 @@ def _test_wav2vec2_with_lm_invalid_pool(in_queue, out_queue, timeout):
transcription = processor.batch_decode(logits.cpu().numpy()).text
unittest.TestCase().assertIn("Falling back to sequential decoding.", cl.out)
unittest.TestCase().assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")
unittest.TestCase().assertEqual(transcription[0], "habitan aguas poco profundas y rocosas")
except Exception:
error = f"{traceback.format_exc()}"
@@ -1833,7 +1833,7 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
@require_pyctcdecode
@require_torchaudio
def test_wav2vec2_with_lm(self):
ds = load_dataset("common_voice", "es", split="test", streaming=True)
ds = load_dataset("mozilla-foundation/common_voice_11_0", "es", split="test", streaming=True)
sample = next(iter(ds))
resampled_audio = torchaudio.functional.resample(
@@ -1852,12 +1852,12 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
transcription = processor.batch_decode(logits.cpu().numpy()).text
self.assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")
self.assertEqual(transcription[0], "habitan aguas poco profundas y rocosas")
@require_pyctcdecode
@require_torchaudio
def test_wav2vec2_with_lm_pool(self):
ds = load_dataset("common_voice", "es", split="test", streaming=True)
ds = load_dataset("mozilla-foundation/common_voice_11_0", "es", split="test", streaming=True)
sample = next(iter(ds))
resampled_audio = torchaudio.functional.resample(
@@ -1878,7 +1878,7 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
with multiprocessing.get_context("fork").Pool(2) as pool:
transcription = processor.batch_decode(logits.cpu().numpy(), pool).text
self.assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")
self.assertEqual(transcription[0], "habitan aguas poco profundas y rocosas")
# user-managed pool + num_processes should trigger a warning
with CaptureLogger(processing_wav2vec2_with_lm.logger) as cl, multiprocessing.get_context("fork").Pool(
@@ -1889,7 +1889,7 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
self.assertIn("num_process", cl.out)
self.assertIn("it will be ignored", cl.out)
self.assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")
self.assertEqual(transcription[0], "habitan aguas poco profundas y rocosas")
@require_pyctcdecode
@require_torchaudio
@@ -1957,7 +1957,7 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
LANG_MAP = {"it": "ita", "es": "spa", "fr": "fra", "en": "eng"}
def run_model(lang):
ds = load_dataset("common_voice", lang, split="test", streaming=True)
ds = load_dataset("mozilla-foundation/common_voice_11_0", lang, split="test", streaming=True)
sample = next(iter(ds))
wav2vec2_lang = LANG_MAP[lang]
@@ -1982,10 +1982,10 @@ class Wav2Vec2ModelIntegrationTest(unittest.TestCase):
return transcription
TRANSCRIPTIONS = {
"it": "mi hanno fatto un'offerta che non potevo proprio rifiutare",
"es": "bien y qué regalo vas a abrir primero",
"fr": "un vrai travail intéressant va enfin être mené sur ce sujet",
"en": "twas the time of day and olof spen slept during the summer",
"it": "il libro ha suscitato molte polemiche a causa dei suoi contenuti",
"es": "habitan aguas poco profundas y rocosas",
"fr": "ce dernier est volé tout au long de l'histoire romaine",
"en": "joe keton disapproved of films and buster also had reservations about the media",
}
for lang in LANG_MAP.keys():