patrickvonplaten commited on
Commit
17641d9
1 Parent(s): 0457f36

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +15 -15
README.md CHANGED
@@ -17,7 +17,7 @@ model-index:
17
  dataset:
18
  name: Common Voice ro
19
  type: common_voice
20
- args: {lang_id}
21
  metrics:
22
  - name: Test WER
23
  type: wer
@@ -49,15 +49,15 @@ resampler = torchaudio.transforms.Resample(48_000, 16_000)
49
  # Preprocessing the datasets.
50
  # We need to read the aduio files as arrays
51
  def speech_file_to_array_fn(batch):
52
- \\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
53
- \\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
54
- \\treturn batch
55
 
56
  test_dataset = test_dataset.map(speech_file_to_array_fn)
57
  inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
58
 
59
  with torch.no_grad():
60
- \\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
61
 
62
  predicted_ids = torch.argmax(logits, dim=-1)
63
 
@@ -85,30 +85,30 @@ processor = Wav2Vec2Processor.from_pretrained("gmihaila/wav2vec2-large-xlsr-53-r
85
  model = Wav2Vec2ForCTC.from_pretrained("gmihaila/wav2vec2-large-xlsr-53-romanian")
86
  model.to("cuda")
87
 
88
- chars_to_ignore_regex = '[\\\\,\\\\?\\\\.\\\\!\\\\-\\\\;\\\\:\\\\"\\\\“]'
89
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
90
 
91
  # Preprocessing the datasets.
92
  # We need to read the aduio files as arrays
93
  def speech_file_to_array_fn(batch):
94
- \\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
95
- \\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
96
- \\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
97
- \\treturn batch
98
 
99
  test_dataset = test_dataset.map(speech_file_to_array_fn)
100
 
101
  # Preprocessing the datasets.
102
  # We need to read the aduio files as arrays
103
  def evaluate(batch):
104
- \\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
105
 
106
- \\twith torch.no_grad():
107
- \\t\\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
108
 
109
  pred_ids = torch.argmax(logits, dim=-1)
110
- \\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
111
- \\treturn batch
112
 
113
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
114
 
 
17
  dataset:
18
  name: Common Voice ro
19
  type: common_voice
20
+ args: ro
21
  metrics:
22
  - name: Test WER
23
  type: wer
 
49
  # Preprocessing the datasets.
50
  # We need to read the aduio files as arrays
51
  def speech_file_to_array_fn(batch):
52
+ \\\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
53
+ \\\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
54
+ \\\\treturn batch
55
 
56
  test_dataset = test_dataset.map(speech_file_to_array_fn)
57
  inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
58
 
59
  with torch.no_grad():
60
+ \\\\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
61
 
62
  predicted_ids = torch.argmax(logits, dim=-1)
63
 
 
85
  model = Wav2Vec2ForCTC.from_pretrained("gmihaila/wav2vec2-large-xlsr-53-romanian")
86
  model.to("cuda")
87
 
88
+ chars_to_ignore_regex = '[\\\\\\\\,\\\\\\\\?\\\\\\\\.\\\\\\\\!\\\\\\\\-\\\\\\\\;\\\\\\\\:\\\\\\\\"\\\\\\\\“]'
89
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
90
 
91
  # Preprocessing the datasets.
92
  # We need to read the aduio files as arrays
93
  def speech_file_to_array_fn(batch):
94
+ \\\\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
95
+ \\\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
96
+ \\\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
97
+ \\\\treturn batch
98
 
99
  test_dataset = test_dataset.map(speech_file_to_array_fn)
100
 
101
  # Preprocessing the datasets.
102
  # We need to read the aduio files as arrays
103
  def evaluate(batch):
104
+ \\\\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
105
 
106
+ \\\\twith torch.no_grad():
107
+ \\\\t\\\\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
108
 
109
  pred_ids = torch.argmax(logits, dim=-1)
110
+ \\\\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
111
+ \\\\treturn batch
112
 
113
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
114