File size: 3,904 Bytes
081e72f
ddad4eb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
081e72f
 
ddad4eb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
- generated_from_trainer
datasets:
- common_voice_16_1
metrics:
- wer
model-index:
- name: wav2vec2-large-xls-r-300m-tr-cv16.1-colab
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: common_voice_16_1
      type: common_voice_16_1
      config: tr
      split: test
      args: tr
    metrics:
    - name: Wer
      type: wer
      value: 0.2775680437205623
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# wav2vec2-large-xls-r-300m-tr-cv16.1-colab

This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice_16_1 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2481
- Wer: 0.2776

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 10
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step  | Validation Loss | Wer    |
|:-------------:|:-----:|:-----:|:---------------:|:------:|
| 5.5874        | 0.29  | 400   | 1.2182          | 0.9358 |
| 0.8023        | 0.58  | 800   | 0.7425          | 0.7498 |
| 0.5662        | 0.88  | 1200  | 0.5324          | 0.6233 |
| 0.4553        | 1.17  | 1600  | 0.4375          | 0.5267 |
| 0.4068        | 1.46  | 2000  | 0.4159          | 0.5051 |
| 0.3797        | 1.75  | 2400  | 0.3861          | 0.4752 |
| 0.3551        | 2.05  | 2800  | 0.3681          | 0.4484 |
| 0.3059        | 2.34  | 3200  | 0.3491          | 0.4364 |
| 0.297         | 2.63  | 3600  | 0.3437          | 0.4191 |
| 0.292         | 2.92  | 4000  | 0.3261          | 0.4160 |
| 0.2537        | 3.21  | 4400  | 0.3363          | 0.4105 |
| 0.2448        | 3.51  | 4800  | 0.3527          | 0.4113 |
| 0.2411        | 3.8   | 5200  | 0.3233          | 0.3975 |
| 0.2324        | 4.09  | 5600  | 0.3038          | 0.3823 |
| 0.213         | 4.38  | 6000  | 0.2982          | 0.3757 |
| 0.2046        | 4.67  | 6400  | 0.2909          | 0.3591 |
| 0.2064        | 4.97  | 6800  | 0.2914          | 0.3654 |
| 0.1814        | 5.26  | 7200  | 0.2961          | 0.3567 |
| 0.1774        | 5.55  | 7600  | 0.3105          | 0.3671 |
| 0.1816        | 5.84  | 8000  | 0.2971          | 0.3524 |
| 0.1621        | 6.14  | 8400  | 0.2837          | 0.3444 |
| 0.1526        | 6.43  | 8800  | 0.2810          | 0.3371 |
| 0.1492        | 6.72  | 9200  | 0.2696          | 0.3277 |
| 0.1404        | 7.01  | 9600  | 0.2733          | 0.3200 |
| 0.1276        | 7.3   | 10000 | 0.2672          | 0.3076 |
| 0.1266        | 7.6   | 10400 | 0.2727          | 0.3126 |
| 0.1259        | 7.89  | 10800 | 0.2516          | 0.3051 |
| 0.1143        | 8.18  | 11200 | 0.2633          | 0.2963 |
| 0.1098        | 8.47  | 11600 | 0.2592          | 0.2938 |
| 0.1037        | 8.77  | 12000 | 0.2473          | 0.2914 |
| 0.0995        | 9.06  | 12400 | 0.2566          | 0.2857 |
| 0.0937        | 9.35  | 12800 | 0.2528          | 0.2812 |
| 0.094         | 9.64  | 13200 | 0.2491          | 0.2799 |
| 0.0927        | 9.93  | 13600 | 0.2481          | 0.2776 |


### Framework versions

- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2