aismlv commited on
Commit
c94712e
1 Parent(s): 5e7d20b

Fix language code

Browse files
Files changed (1) hide show
  1. README.md +17 -16
README.md CHANGED
@@ -1,4 +1,5 @@
1
- language: kz
 
2
  datasets:
3
  - kazakh_speech_corpus
4
  metrics:
@@ -18,7 +19,7 @@ model-index:
18
  dataset:
19
  name: Kazakh Speech Corpus v1.1
20
  type: kazakh_speech_corpus
21
- args: kz
22
  metrics:
23
  - name: Test WER
24
  type: wer
@@ -52,15 +53,15 @@ model = Wav2Vec2ForCTC.from_pretrained("wav2vec2-large-xlsr-kazakh")
52
  # Preprocessing the datasets.
53
  # We need to read the audio files as arrays
54
  def speech_file_to_array_fn(batch):
55
- speech_array, sampling_rate = torchaudio.load(batch["path"])
56
- batch["speech"] = torchaudio.transforms.Resample(sampling_rate, 16_000)(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
 
@@ -93,22 +94,22 @@ model.to("cuda")
93
  # Preprocessing the datasets.
94
  # We need to read the audio files as arrays
95
  def speech_file_to_array_fn(batch):
96
- batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
97
- speech_array, sampling_rate = torchaudio.load(batch["path"])
98
- batch["speech"] = torchaudio.transforms.Resample(sampling_rate, 16_000)(speech_array).squeeze().numpy()
99
- return batch
100
 
101
  test_dataset = test_dataset.map(speech_file_to_array_fn)
102
 
103
  def evaluate(batch):
104
- inputs = processor(batch["text"], sampling_rate=16_000, return_tensors="pt", padding=True)
105
 
106
- with torch.no_grad():
107
- logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
108
 
109
- pred_ids = torch.argmax(logits, dim=-1)
110
- batch["pred_strings"] = processor.batch_decode(pred_ids)
111
- return batch
112
 
113
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
114
 
 
1
+ ---
2
+ language: kk
3
  datasets:
4
  - kazakh_speech_corpus
5
  metrics:
 
19
  dataset:
20
  name: Kazakh Speech Corpus v1.1
21
  type: kazakh_speech_corpus
22
+ args: kk
23
  metrics:
24
  - name: Test WER
25
  type: wer
 
53
  # Preprocessing the datasets.
54
  # We need to read the audio files as arrays
55
  def speech_file_to_array_fn(batch):
56
+ \tspeech_array, sampling_rate = torchaudio.load(batch["path"])
57
+ \tbatch["speech"] = torchaudio.transforms.Resample(sampling_rate, 16_000)(speech_array).squeeze().numpy()
58
+ \treturn batch
59
 
60
  test_dataset = test_dataset.map(speech_file_to_array_fn)
61
  inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
62
 
63
  with torch.no_grad():
64
+ \tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
65
 
66
  predicted_ids = torch.argmax(logits, dim=-1)
67
 
 
94
  # Preprocessing the datasets.
95
  # We need to read the audio 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"] = torchaudio.transforms.Resample(sampling_rate, 16_000)(speech_array).squeeze().numpy()
100
+ \treturn batch
101
 
102
  test_dataset = test_dataset.map(speech_file_to_array_fn)
103
 
104
  def evaluate(batch):
105
+ \tinputs = processor(batch["text"], sampling_rate=16_000, return_tensors="pt", padding=True)
106
 
107
+ \twith torch.no_grad():
108
+ \t\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
109
 
110
+ \tpred_ids = torch.argmax(logits, dim=-1)
111
+ \tbatch["pred_strings"] = processor.batch_decode(pred_ids)
112
+ \treturn batch
113
 
114
  result = test_dataset.map(evaluate, batched=True, batch_size=8)
115