File size: 4,529 Bytes
7a478a1
 
 
 
 
 
89d2dbe
7a478a1
89d2dbe
7a478a1
 
 
 
c51a264
7a478a1
12d6b15
 
7a478a1
 
 
 
 
 
12d6b15
 
 
 
 
 
 
 
 
 
 
 
 
d5df867
 
 
 
 
20902b6
d5df867
12d6b15
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7a478a1
 
7c41514
7a478a1
3c97f1b
7265264
 
4f8b703
7265264
4f8b703
7a478a1
4f8b703
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a2f0bed
4f8b703
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7a478a1
 
 
 
 
 
 
 
 
 
 
 
 
 
60d17cf
 
 
 
 
7c41514
02454a8
60d17cf
 
 
 
02d4be2
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
---
language:
- it
license: apache-2.0
tags:
- automatic-speech-recognition
- hf-asr-leaderboard
- it
- mozilla-foundation/common_voice_8_0
- robust-speech-event
datasets:
- mozilla-foundation/common_voice_8_0
model-index:
- name: XLS-R Wav2Vec2 Italian by Jonatas Grosman
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Common Voice 8
      type: mozilla-foundation/common_voice_8_0
      args: it
    metrics:
    - name: Test WER
      type: wer
      value: 9.04
    - name: Test CER
      type: cer
      value: 2.2
    - name: Test WER (+LM)
      type: wer
      value: 6.75
    - name: Test CER (+LM)
      type: cer
      value: 1.76
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Robust Speech Event - Dev Data
      type: speech-recognition-community-v2/dev_data
      args: it
    metrics:
    - name: Dev WER
      type: wer
      value: 23.38
    - name: Dev CER
      type: cer
      value: 9.41
    - name: Dev WER (+LM)
      type: wer
      value: 15.84
    - name: Dev CER (+LM)
      type: cer
      value: 8.93
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Robust Speech Event - Test Data
      type: speech-recognition-community-v2/eval_data
      args: it
    metrics:
    - name: Test WER
      type: wer
      value: 18.34
---

# Fine-tuned XLS-R 1B model for speech recognition in Italian

Fine-tuned [facebook/wav2vec2-xls-r-1b](https://huggingface.co/facebook/wav2vec2-xls-r-1b) on Italian using the train and validation splits of [Common Voice 8.0](https://huggingface.co/datasets/mozilla-foundation/common_voice_8_0), [Multilingual TEDx](http://www.openslr.org/100), [Multilingual LibriSpeech](https://www.openslr.org/94/), and [Voxpopuli](https://github.com/facebookresearch/voxpopuli).
When using this model, make sure that your speech input is sampled at 16kHz.

This model has been fine-tuned by the [HuggingSound](https://github.com/jonatasgrosman/huggingsound) tool, and thanks to the GPU credits generously given by the [OVHcloud](https://www.ovhcloud.com/en/public-cloud/ai-training/) :)

## Usage

Using the [HuggingSound](https://github.com/jonatasgrosman/huggingsound) library:

```python
from huggingsound import SpeechRecognitionModel

model = SpeechRecognitionModel("jonatasgrosman/wav2vec2-xls-r-1b-italian")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]

transcriptions = model.transcribe(audio_paths)
```

Writing your own inference script:

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

LANG_ID = "it"
MODEL_ID = "jonatasgrosman/wav2vec2-xls-r-1b-italian"
SAMPLES = 10

test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")

processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)

# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
    speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
    batch["speech"] = speech_array
    batch["sentence"] = batch["sentence"].upper()
    return batch

test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"], 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)
predicted_sentences = processor.batch_decode(predicted_ids)
```

## Evaluation Commands

1. To evaluate on `mozilla-foundation/common_voice_8_0` with split `test`

```bash
python eval.py --model_id jonatasgrosman/wav2vec2-xls-r-1b-italian --dataset mozilla-foundation/common_voice_8_0 --config it --split test
```

2. To evaluate on `speech-recognition-community-v2/dev_data`

```bash
python eval.py --model_id jonatasgrosman/wav2vec2-xls-r-1b-italian --dataset speech-recognition-community-v2/dev_data --config it --split validation --chunk_length_s 5.0 --stride_length_s 1.0
```

## Citation
If you want to cite this model you can use this:

```bibtex
@misc{grosman2021xlsr-1b-italian,
  title={Fine-tuned {XLS-R} 1{B} model for speech recognition in {I}talian},
  author={Grosman, Jonatas},
  howpublished={\url{https://huggingface.co/jonatasgrosman/wav2vec2-xls-r-1b-italian}},
  year={2022}
}
```