def test_inference_ctc_robust_batched(self):
model = FlaxWav2Vec2ForCTC.from_pretrained(
"facebook/wav2vec2-large-960h-lv60-self", from_pt=True)
processor = Wav2Vec2Processor.from_pretrained(
"facebook/wav2vec2-large-960h-lv60-self", do_lower_case=True)
input_speech = self._load_datasamples(4)
inputs = processor(input_speech, return_tensors="np", padding=True)
input_values = inputs.input_values
attention_mask = inputs.attention_mask
logits = model(input_values, attention_mask=attention_mask).logits
predicted_ids = jnp.argmax(logits, axis=-1)
predicted_trans = processor.batch_decode(predicted_ids)
EXPECTED_TRANSCRIPTIONS = [
"a man said to the universe sir i exist",
"sweat covered brion's body trickling into the tight loin cloth that was the only garment he wore",
"the cut on his chest still dripping blood the ache of his overstrained eyes even the soaring arena around him with the thousands of spectators were trivialities not worth thinking about",
"his instant panic was followed by a small sharp blow high on his chest",
]
self.assertListEqual(predicted_trans, EXPECTED_TRANSCRIPTIONS)
def test_wav2vec2_with_lm(self):
ds = load_dataset("common_voice", "es", split="test", streaming=True)
sample = next(iter(ds))
resampled_audio = librosa.resample(sample["audio"]["array"], 48_000, 16_000)
model = FlaxWav2Vec2ForCTC.from_pretrained("patrickvonplaten/wav2vec2-large-xlsr-53-spanish-with-lm")
processor = Wav2Vec2ProcessorWithLM.from_pretrained("patrickvonplaten/wav2vec2-large-xlsr-53-spanish-with-lm")
input_values = processor(resampled_audio, return_tensors="np").input_values
logits = model(input_values).logits
transcription = processor.batch_decode(np.array(logits)).text
self.assertEqual(transcription[0], "bien y qué regalo vas a abrir primero")