File size: 5,117 Bytes
02440d6
 
 
 
 
 
 
 
9ac9bfd
18cd23a
1c0d7f8
18cd23a
02440d6
 
 
 
 
 
 
 
 
75a3831
 
02440d6
 
 
 
 
 
18cd23a
02440d6
 
 
 
 
 
75a3831
02440d6
18cd23a
 
 
02440d6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7983137
 
02440d6
 
 
 
18cd23a
02440d6
 
 
9ac9bfd
 
18cd23a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
02440d6
 
 
 
 
 
 
 
2db496b
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- af
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
datasets:
- dsfsi-anv/multilingual-nchlt-dataset
- google/fleurs
- andreoosthuizen/afrikaans-30s
- voice-biomarkers/openslr-32-hq-SA-languages-Afrikaans
metrics:
- wer
model-index:
- name: Whisper Tiny af
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Fleurs
      type: google/fleurs
      config: af_za
      split: test
      args: af_za
    metrics:
    - name: Wer
      type: wer
      value: 44.257751602286504
---
<!-- 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. -->

# Whisper Tiny af

This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on multiple datasets.
It achieves the following results on the evaluation set:
- Loss: 1.2213
- Wer: 44.2578
- Cer: 17.8026

## 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: 1e-05
- train_batch_size: 64
- eval_batch_size: 64
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.04
- training_steps: 4100

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer     | Cer     |
|:-------------:|:------:|:----:|:---------------:|:-------:|:-------:|
| 1.7169        | 0.0244 | 100  | 1.7637          | 68.4393 | 26.7904 |
| 0.9216        | 0.0488 | 200  | 1.3055          | 52.9014 | 21.5255 |
| 0.6082        | 0.0732 | 300  | 1.1946          | 49.3158 | 19.3768 |
| 0.4534        | 0.0976 | 400  | 1.1545          | 47.5143 | 18.1954 |
| 0.3675        | 0.1220 | 500  | 1.1354          | 46.8387 | 18.5267 |
| 0.282         | 0.1463 | 600  | 1.1251          | 46.0939 | 19.8751 |
| 0.254         | 0.1707 | 700  | 1.1269          | 45.4876 | 18.8345 |
| 0.2055        | 0.1951 | 800  | 1.1248          | 48.9347 | 20.0803 |
| 0.1837        | 0.2195 | 900  | 1.1323          | 45.0199 | 19.4325 |
| 0.1606        | 0.2439 | 1000 | 1.1317          | 49.2118 | 21.8832 |
| 0.1337        | 0.2683 | 1100 | 1.1491          | 44.7601 | 18.6498 |
| 0.1149        | 0.2927 | 1200 | 1.1535          | 45.4530 | 19.5761 |
| 0.1072        | 0.3171 | 1300 | 1.1685          | 48.6056 | 20.2328 |
| 0.0998        | 0.3415 | 1400 | 1.1738          | 44.5695 | 18.5501 |
| 0.097         | 0.3659 | 1500 | 1.1702          | 44.4656 | 18.3068 |
| 0.0769        | 0.3902 | 1600 | 1.1601          | 47.1159 | 19.3709 |
| 0.084         | 0.4146 | 1700 | 1.1815          | 47.5663 | 19.6347 |
| 0.0664        | 0.4390 | 1800 | 1.1821          | 44.3097 | 18.7582 |
| 0.0652        | 0.4634 | 1900 | 1.1854          | 43.2184 | 18.4123 |
| 0.0609        | 0.4878 | 2000 | 1.1830          | 43.1145 | 17.4508 |
| 0.0565        | 0.5122 | 2100 | 1.1897          | 47.1505 | 19.0514 |
| 0.0589        | 0.5366 | 2200 | 1.2024          | 45.4010 | 18.6996 |
| 0.0552        | 0.5610 | 2300 | 1.1956          | 48.6402 | 20.3764 |
| 0.0551        | 0.5854 | 2400 | 1.1930          | 45.3837 | 18.6527 |
| 0.0551        | 0.6098 | 2500 | 1.1984          | 47.1159 | 18.6996 |
| 0.04          | 0.6341 | 2600 | 1.2092          | 47.4796 | 19.7725 |
| 0.0548        | 0.6585 | 2700 | 1.1981          | 42.7681 | 17.5915 |
| 0.0466        | 0.6829 | 2800 | 1.2144          | 48.1379 | 20.3588 |
| 0.0425        | 0.7073 | 2900 | 1.2051          | 46.0766 | 18.7670 |
| 0.0431        | 0.7317 | 3000 | 1.2157          | 44.3963 | 17.4596 |
| 0.0427        | 0.7561 | 3100 | 1.2178          | 48.1032 | 19.8517 |
| 0.0346        | 0.7805 | 3200 | 1.2177          | 47.4970 | 19.5644 |
| 0.0395        | 0.8049 | 3300 | 1.2199          | 47.1159 | 18.9312 |
| 0.039         | 0.8293 | 3400 | 1.2219          | 45.7474 | 19.4090 |
| 0.0359        | 0.8537 | 3500 | 1.2191          | 46.4057 | 18.7846 |
| 0.0461        | 0.8780 | 3600 | 1.2172          | 51.2039 | 21.9476 |
| 0.0299        | 0.9024 | 3700 | 1.2202          | 47.5316 | 19.1335 |
| 0.028         | 0.9268 | 3800 | 1.2216          | 47.1505 | 19.4999 |
| 0.0305        | 1.01   | 3900 | 1.2241          | 46.5443 | 18.7231 |
| 0.038         | 1.0344 | 4000 | 1.2218          | 44.2924 | 17.6619 |
| 0.0249        | 1.0588 | 4100 | 1.2213          | 44.2578 | 17.8026 |


### Framework versions

- Transformers 4.42.0.dev0
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1

## Citation

Please cite the model using the following BibTeX entry:

```bibtex
@misc{deepdml/whisper-tiny-af-mix-norm,
      title={Fine-tuned Whisper tiny ASR model for speech recognition in Afrikaans},
      author={Jimenez, David},
      howpublished={\url{https://huggingface.co/deepdml/whisper-tiny-af-mix-norm}},
      year={2026}
    }
```