patrickvonplaten commited on
Commit
f14235c
1 Parent(s): 7b6e7c2

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +113 -16
README.md CHANGED
@@ -1,28 +1,125 @@
1
  ---
2
- language:
3
- - ta
4
- - en
5
  tags:
6
- - translation
7
- - fsmt
8
  - automatic-speech-recognition
9
- license: apache-2.0
10
- models:
11
- - wav2vec2
12
  - speech
13
- datasets:
14
- - common-voice
 
 
 
 
 
 
 
 
 
 
 
 
 
 
15
  ---
16
 
17
- # Wav2Vec2-XLSR-53-Tamil
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
 
19
- This model is fine-tuned from Facebook's `wav2vec2-large-xlsr-53` using Mozilla's `common voice` dataset.
 
20
 
21
- ### Model Metrics
22
 
23
- WER: 0.829423
24
 
 
25
 
26
- Model Fine-Tuning Colab - https://colab.research.google.com/drive/1-Klkgr4f-C9SanHfVC5RhP0ELUH6TYlN?usp=sharing
27
 
28
- Thanks to this article and its colab for Fine-tuning - https://huggingface.co/blog/fine-tune-xlsr-wav2vec2
 
1
  ---
2
+ language: ta
3
+ datasets:
4
+ - common_voice
5
  tags:
6
+ - audio
 
7
  - automatic-speech-recognition
 
 
 
8
  - speech
9
+ - xlsr-fine-tuning-week
10
+ license: apache-2.0
11
+ model-index:
12
+ - name: XLSR Wav2Vec2 Tamil by Amrrs
13
+ results:
14
+ - task:
15
+ name: Speech Recognition
16
+ type: automatic-speech-recognition
17
+ dataset:
18
+ name: Common Voice ta
19
+ type: common_voice
20
+ args: ta
21
+ metrics:
22
+ - name: Test WER
23
+ type: wer
24
+ value: 82.94
25
  ---
26
 
27
+ # Wav2Vec2-Large-XLSR-53-Tamil
28
+
29
+ Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) in Tamil using the [Common Voice](https://huggingface.co/datasets/common_voice)
30
+ When using this model, make sure that your speech input is sampled at 16kHz.
31
+
32
+ ## Usage
33
+
34
+ The model can be used directly (without a language model) as follows:
35
+
36
+ ```python
37
+ import torch
38
+ import torchaudio
39
+ from datasets import load_dataset
40
+ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
41
+
42
+ test_dataset = load_dataset("common_voice", "ta", split="test[:2%]").
43
+
44
+ processor = Wav2Vec2Processor.from_pretrained("Amrrs/wav2vec2-large-xlsr-53-tamil")
45
+ model = Wav2Vec2ForCTC.from_pretrained("Amrrs/wav2vec2-large-xlsr-53-tamil")
46
+
47
+ resampler = torchaudio.transforms.Resample(48_000, 16_000)
48
+
49
+ # Preprocessing the datasets.
50
+ # We need to read the aduio files as arrays
51
+ def speech_file_to_array_fn(batch):
52
+ speech_array, sampling_rate = torchaudio.load(batch["path"])
53
+ batch["speech"] = resampler(speech_array).squeeze().numpy()
54
+ return 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
+ logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
61
+
62
+ predicted_ids = torch.argmax(logits, dim=-1)
63
+
64
+ print("Prediction:", processor.batch_decode(predicted_ids))
65
+ print("Reference:", test_dataset["sentence"][:2])
66
+ ```
67
+
68
+
69
+ ## Evaluation
70
+
71
+ The model can be evaluated as follows on the {language} test data of Common Voice.
72
+
73
+
74
+ ```python
75
+ import torch
76
+ import torchaudio
77
+ from datasets import load_dataset, load_metric
78
+ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
79
+ import re
80
+
81
+ test_dataset = load_dataset("common_voice", "ta", split="test")
82
+ wer = load_metric("wer")
83
+
84
+ processor = Wav2Vec2Processor.from_pretrained("Amrrs/wav2vec2-large-xlsr-53-tamil")
85
+ model = Wav2Vec2ForCTC.from_pretrained("Amrrs/wav2vec2-large-xlsr-53-tamil")
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
+ batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
95
+ speech_array, sampling_rate = torchaudio.load(batch["path"])
96
+ batch["speech"] = resampler(speech_array).squeeze().numpy()
97
+ return 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
+ inputs = processor(batch["speech"], 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
 
115
+ print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
116
+ ```
117
 
118
+ **Test Result**: 82.94 %
119
 
 
120
 
121
+ ## Training
122
 
123
+ The Common Voice `train`, `validation` datasets were used for training.
124
 
125
+ The script used for training can be found [here](https://colab.research.google.com/drive/1-Klkgr4f-C9SanHfVC5RhP0ELUH6TYlN?usp=sharing)