cahya commited on
Commit
965a74d
β€’
1 Parent(s): 2f4f81d

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +15 -15
README.md CHANGED
@@ -11,7 +11,7 @@ tags:
11
  - xlsr-fine-tuning-week
12
  license: apache-2.0
13
  model-index:
14
- - name: indonesian-nlp/wav2vec2-large-xlsr-indonesian-baseline
15
  results:
16
  - task:
17
  name: Speech Recognition
@@ -52,15 +52,15 @@ resampler = torchaudio.transforms.Resample(48_000, 16_000)
52
  # Preprocessing the datasets.
53
  # We need to read the aduio files as arrays
54
  def speech_file_to_array_fn(batch):
55
- speech_array, sampling_rate = torchaudio.load(batch["path"])
56
- batch["speech"] = resampler(speech_array).squeeze().numpy()
57
- return batch
58
 
59
  test_dataset = test_dataset.map(speech_file_to_array_fn)
60
  inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
61
 
62
  with torch.no_grad():
63
- logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
64
 
65
  predicted_ids = torch.argmax(logits, dim=-1)
66
 
@@ -87,31 +87,31 @@ processor = Wav2Vec2Processor.from_pretrained("indonesian-nlp/wav2vec2-large-xls
87
  model = Wav2Vec2ForCTC.from_pretrained("indonesian-nlp/wav2vec2-large-xlsr-indonesian-baseline")
88
  model.to("cuda")
89
 
90
- chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\β€œ\%\β€˜\'\”]'
91
 
92
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
93
 
94
  # Preprocessing the datasets.
95
  # We need to read the aduio files as arrays
96
  def speech_file_to_array_fn(batch):
97
- batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
98
- speech_array, sampling_rate = torchaudio.load(batch["path"])
99
- batch["speech"] = resampler(speech_array).squeeze().numpy()
100
- return batch
101
 
102
  test_dataset = test_dataset.map(speech_file_to_array_fn)
103
 
104
  # Preprocessing the datasets.
105
  # We need to read the aduio files as arrays
106
  def evaluate(batch):
107
- inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
108
 
109
- with torch.no_grad():
110
- logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
111
 
112
  pred_ids = torch.argmax(logits, dim=-1)
113
- batch["pred_strings"] = processor.batch_decode(pred_ids)
114
- return batch
115
 
116
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
117
 
 
11
  - xlsr-fine-tuning-week
12
  license: apache-2.0
13
  model-index:
14
+ - name: XLSR Wav2Vec2 Indonesian Baseline by indonesian-nlp
15
  results:
16
  - task:
17
  name: Speech Recognition
 
52
  # Preprocessing the datasets.
53
  # We need to read the aduio files as arrays
54
  def speech_file_to_array_fn(batch):
55
+ \tspeech_array, sampling_rate = torchaudio.load(batch["path"])
56
+ \tbatch["speech"] = resampler(speech_array).squeeze().numpy()
57
+ \treturn batch
58
 
59
  test_dataset = test_dataset.map(speech_file_to_array_fn)
60
  inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
61
 
62
  with torch.no_grad():
63
+ \tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
64
 
65
  predicted_ids = torch.argmax(logits, dim=-1)
66
 
 
87
  model = Wav2Vec2ForCTC.from_pretrained("indonesian-nlp/wav2vec2-large-xlsr-indonesian-baseline")
88
  model.to("cuda")
89
 
90
+ chars_to_ignore_regex = '[\\,\\?\\.\\!\\-\\;\\:\\"\\β€œ\\%\\β€˜\\'\\”]'
91
 
92
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
93
 
94
  # Preprocessing the datasets.
95
  # We need to read the aduio files as arrays
96
  def speech_file_to_array_fn(batch):
97
+ \tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
98
+ \tspeech_array, sampling_rate = torchaudio.load(batch["path"])
99
+ \tbatch["speech"] = resampler(speech_array).squeeze().numpy()
100
+ \treturn batch
101
 
102
  test_dataset = test_dataset.map(speech_file_to_array_fn)
103
 
104
  # Preprocessing the datasets.
105
  # We need to read the aduio files as arrays
106
  def evaluate(batch):
107
+ \tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
108
 
109
+ \twith torch.no_grad():
110
+ \t\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
111
 
112
  pred_ids = torch.argmax(logits, dim=-1)
113
+ \tbatch["pred_strings"] = processor.batch_decode(pred_ids)
114
+ \treturn batch
115
 
116
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
117