def make_processor_with_lm(recognizer_dir='', save=True):
if not recognizer_dir:
print('no directory provided, using default at:', path)
recognizer_dir = path
processor = load_pretrained_processor(recognizer_dir)
sorted_vocab = load_and_sort_vocab(processor)
decoder = make_ctc_decoder(sorted_vocab)
processor_with_lm = Wav2Vec2ProcessorWithLM(
feature_extractor=processor.feature_extractor,
tokenizer=processor.tokenizer,
decoder=decoder)
if save:
processor_with_lm.save_pretrained(recognizer_dir)
return processor_with_lm
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")
def test_wav2vec2_with_lm(self):
downloaded_folder = snapshot_download(
"patrickvonplaten/common_voice_es_sample")
file_path = glob.glob(downloaded_folder + "/*")[0]
sample = librosa.load(file_path, sr=16_000)[0]
model = TFWav2Vec2ForCTC.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(sample, return_tensors="tf").input_values
logits = model(input_values).logits
transcription = processor.batch_decode(logits.numpy()).text
self.assertEqual(transcription[0],
"el libro ha sido escrito por cervantes")
def load_processor_with_lm(recognizer_dir=''):
if not recognizer_dir: recognizer_dir = path
return Wav2Vec2ProcessorWithLM.from_pretrained(recognizer_dir)