bayartsogt commited on
Commit
4139062
1 Parent(s): 4f91a77

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -12
README.md CHANGED
@@ -28,7 +28,7 @@ model-index:
28
 
29
  # Wav2Vec2-Large-XLSR-53-Mongolian-v1
30
 
31
- Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on {language} using the [Common Voice](https://huggingface.co/datasets/common_voice).
32
 
33
  When using this model, make sure that your speech input is sampled at 16kHz.
34
 
@@ -51,30 +51,30 @@ processor = Wav2Vec2Processor.from_pretrained("bayartsogt/wav2vec2-large-xlsr-mo
51
  model = Wav2Vec2ForCTC.from_pretrained("bayartsogt/wav2vec2-large-xlsr-mongolian-v1")
52
  model.to("cuda")
53
 
54
- chars_to_ignore_regex = '[\!\"\'\,\.\«\»\?\-]'
55
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
56
 
57
  # Preprocessing the datasets.
58
  # We need to read the aduio files as arrays
59
  def speech_file_to_array_fn(batch):
60
- batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
61
- speech_array, sampling_rate = torchaudio.load(batch["path"])
62
- batch["speech"] = resampler(speech_array).squeeze().numpy()
63
- return batch
64
 
65
  test_dataset = test_dataset.map(speech_file_to_array_fn)
66
 
67
  # Preprocessing the datasets.
68
  # We need to read the aduio files as arrays
69
  def evaluate(batch):
70
- inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
71
 
72
- with torch.no_grad():
73
- logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
74
 
75
- pred_ids = torch.argmax(logits, dim=-
76
- batch["pred_strings"] = processor.batch_decode(pred_ids)
77
- return batch
78
 
79
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
80
 
 
28
 
29
  # Wav2Vec2-Large-XLSR-53-Mongolian-v1
30
 
31
+ Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Mongolian using the [Common Voice](https://huggingface.co/datasets/common_voice).
32
 
33
  When using this model, make sure that your speech input is sampled at 16kHz.
34
 
 
51
  model = Wav2Vec2ForCTC.from_pretrained("bayartsogt/wav2vec2-large-xlsr-mongolian-v1")
52
  model.to("cuda")
53
 
54
+ chars_to_ignore_regex = '[\\!\\"\\'\\,\\.\\«\\»\\?\\-]'
55
  resampler = torchaudio.transforms.Resample(48_000, 16_000)
56
 
57
  # Preprocessing the datasets.
58
  # We need to read the aduio files as arrays
59
  def speech_file_to_array_fn(batch):
60
+ \tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
61
+ \tspeech_array, sampling_rate = torchaudio.load(batch["path"])
62
+ \tbatch["speech"] = resampler(speech_array).squeeze().numpy()
63
+ \treturn batch
64
 
65
  test_dataset = test_dataset.map(speech_file_to_array_fn)
66
 
67
  # Preprocessing the datasets.
68
  # We need to read the aduio files as arrays
69
  def evaluate(batch):
70
+ \tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
71
 
72
+ \twith torch.no_grad():
73
+ \t\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
74
 
75
+ \tpred_ids = torch.argmax(logits, dim=-
76
+ \tbatch["pred_strings"] = processor.batch_decode(pred_ids)
77
+ \treturn batch
78
 
79
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
80