bayartsogt commited on
Commit
4f91a77
1 Parent(s): 4451759

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +89 -0
README.md ADDED
@@ -0,0 +1,89 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language: mn
3
+ datasets:
4
+ - common_voice mn
5
+ metrics:
6
+ - wer
7
+ tags:
8
+ - audio
9
+ - automatic-speech-recognition
10
+ - speech
11
+ - xlsr-fine-tuning-week
12
+ license: apache-2.0
13
+ model-index:
14
+ - name: wav2vec2-large-xlsr-mongolian-v1
15
+ results:
16
+ - task:
17
+ name: Speech Recognition
18
+ type: automatic-speech-recognition
19
+ dataset:
20
+ name: Common Voice mn
21
+ type: common_voice
22
+ args: mn
23
+ metrics:
24
+ - name: Test WER
25
+ type: wer
26
+ value: 35.42
27
+ ---
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
+
35
+ ## Evaluation
36
+
37
+ The model can be evaluated as follows on the Mongolian test data of Common Voice.
38
+
39
+
40
+ ```python
41
+ import torch
42
+ import torchaudio
43
+ from datasets import load_dataset, load_metric
44
+ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
45
+ import re
46
+
47
+ test_dataset = load_dataset("common_voice", "mn", split="test")
48
+ wer = load_metric("wer")
49
+
50
+ processor = Wav2Vec2Processor.from_pretrained("bayartsogt/wav2vec2-large-xlsr-mongolian-v1")
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
+
81
+ print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
82
+ ```
83
+
84
+ **Test Result**: 35.42 %
85
+
86
+
87
+ ## Training
88
+
89
+ The Common Voice `train` dataset was used for training as well as ... and ...