File size: 4,680 Bytes
31b06ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b5b833b
31b06ed
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b5b833b
31b06ed
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
---
language: de
datasets:
- common_voice
- wer
tags:
- audio
- automatic-speech-recognition
- speech
- xlsr-fine-tuning-week
license: apache-2.0
model-index:
- name: XLSR Wav2Vec2 Large 53
  results:
  - task: 
      name: Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice de
      type: common_voice
      args: de
    metrics:
       - name: Test WER
         type: wer
         value: 15.91 
---

# Wav2Vec2-Large-XLSR-53-German

Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on German using 12% of the [Common Voice](https://huggingface.co/datasets/common_voice) dataset. 
When using this model, make sure that your speech input is sampled at 16kHz.

## Usage

The model can be used directly (without a language model) as follows:

```python
import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

test_dataset = load_dataset("common_voice", "de", split="test[:2%]") 

processor = Wav2Vec2Processor.from_pretrained("marcel/wav2vec2-large-xlsr-53-german")
model = Wav2Vec2ForCTC.from_pretrained("marcel/wav2vec2-large-xlsr-53-german") 

resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
	speech_array, sampling_rate = torchaudio.load(batch["path"])
	batch["speech"] = resampler(speech_array).squeeze().numpy()
	return batch

test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)

with torch.no_grad():
	logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits

predicted_ids = torch.argmax(logits, dim=-1)

print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])
```


## Evaluation

The model can be evaluated as follows on the {language} test data of Common Voice.


```python
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re

test_dataset = load_dataset("common_voice", "de", split="test[:10%]")
wer = load_metric("wer")

processor = Wav2Vec2Processor.from_pretrained("marcel/wav2vec2-large-xlsr-53-german")
model = Wav2Vec2ForCTC.from_pretrained("marcel/wav2vec2-large-xlsr-53-german") 
model.to("cuda")

chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\”\�\カ\æ\無\ན\カ\臣\ѹ\…\«\»\ð\ı\„\幺\א\ב\比\ш\ע\)\ứ\в\œ\ч\+\—\ш\‚\נ\м\ń\乡\$\=\ש\ф\支\(\°\и\к\̇]'
substitutions = {
    'e' : '[\ə\é\ě\ę\ê\ế\ế\ë\ė\е]',
    'o' : '[\ō\ô\ô\ó\ò\ø\ọ\ŏ\õ\ő\о]',
    'a' : '[\á\ā\ā\ă\ã\å\â\à\ą\а]',
    'c' : '[\č\ć\ç\с]',
    'l' : '[\ł]',
    'u' : '[\ú\ū\ứ\ů]',
    'und' : '[\&]',
    'r' : '[\ř]',
    'y' : '[\ý]',
    's' : '[\ś\š\ș\ş]',
    'i' : '[\ī\ǐ\í\ï\î\ï]',
    'z' : '[\ź\ž\ź\ż]',
    'n' : '[\ñ\ń\ņ]',
    'g' : '[\ğ]',
    'ss' : '[\ß]',
    't' : '[\ț\ť]',
    'd' : '[\ď\đ]',
    "'": '[\ʿ\་\’\`\´\ʻ\`\‘]',
    'p': '\р'
}
resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
    for x in substitutions:
        batch["sentence"] = re.sub(substitutions[x], x, batch["sentence"])
        speech_array, sampling_rate = torchaudio.load(batch["path"])
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(speech_array).squeeze().numpy()
    return batch

test_dataset = test_dataset.map(speech_file_to_array_fn)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
    inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)

    with torch.no_grad():
        logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits

    pred_ids = torch.argmax(logits, dim=-1)
    batch["pred_strings"] = processor.batch_decode(pred_ids)
    return batch

result = test_dataset.map(evaluate, batched=True, batch_size=8)

print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
```

**Test Result**: 15.91 % 


## Training

The first 12% of the Common Voice `train`, `validation` datasets were used for training.

The script used for training can be found TODO