File size: 2,784 Bytes
3286f01
 
 
138720b
ac30980
 
 
3286f01
 
ac30980
 
3286f01
138720b
3286f01
 
 
ac30980
3286f01
138720b
 
3286f01
 
138720b
3286f01
ac30980
3286f01
ac30980
3286f01
 
 
 
 
138720b
3286f01
138720b
3286f01
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
base_model: openai/whisper-tiny
datasets:
- mozilla-foundation/common_voice_17_0
language:
- ko
license: apache-2.0
metrics:
- wer
tags:
- generated_from_trainer
model-index:
- name: Whisper Tiny Ko - Roooy
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: Common Voice 17.0
      type: mozilla-foundation/common_voice_17_0
      config: ko
      split: None
      args: 'config: ko, split: train+valid & test'
    metrics:
    - type: wer
      value: 61.8411000763942
      name: Wer
---

<!-- 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 Ko - Roooy

This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the Common Voice 17.0 dataset.
It achieves the following results on the evaluation set:
- Loss: 1.3774
- Cer: 30.0833
- Wer: 61.8411
- Cer Wer Avg: 45.9622

## 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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 10000
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch    | Step  | Validation Loss | Cer     | Wer     | Cer Wer Avg |
|:-------------:|:--------:|:-----:|:---------------:|:-------:|:-------:|:-----------:|
| 0.0005        | 22.2222  | 1000  | 1.0691          | 27.3648 | 58.8235 | 43.0942     |
| 0.0002        | 44.4444  | 2000  | 1.1396          | 30.6350 | 62.9488 | 46.7919     |
| 0.0001        | 66.6667  | 3000  | 1.1884          | 31.1967 | 63.5982 | 47.3974     |
| 0.0001        | 88.8889  | 4000  | 1.2300          | 31.4776 | 64.2093 | 47.8435     |
| 0.0           | 111.1111 | 5000  | 1.2656          | 31.7284 | 64.7441 | 48.2362     |
| 0.0           | 133.3333 | 6000  | 1.2993          | 32.1396 | 65.0497 | 48.5946     |
| 0.0           | 155.5556 | 7000  | 1.3272          | 32.3804 | 64.9351 | 48.6577     |
| 0.0           | 177.7778 | 8000  | 1.3518          | 29.9829 | 61.8029 | 45.8929     |
| 0.0           | 200.0    | 9000  | 1.3693          | 30.1836 | 61.8793 | 46.0314     |
| 0.0           | 222.2222 | 10000 | 1.3774          | 30.0833 | 61.8411 | 45.9622     |


### Framework versions

- Transformers 4.41.2
- Pytorch 2.3.0+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1