File size: 4,374 Bytes
d8887ee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
274f03f
 
 
 
d8887ee
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
library_name: transformers
language:
- ko
license: mit
base_model: openai/whisper-large-v3-turbo
tags:
- generated_from_trainer
datasets:
- Bingsu/zeroth-korean
metrics:
- wer
model-index:
- name: openai/whisper-large-v3-turbo Korean - Fine-tuned
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Bingsu/zeroth-korean
      type: Bingsu/zeroth-korean
      args: 'transcription column: text'
    metrics:
    - name: Wer
      type: wer
      value: 4.321638307483813
---

> ⚠️ **Author's Note**: This model was fine-tuned for sanity-checking purposes using only a single Korean dataset.  
> As a result, it may be overfitted and may not generalize well to other datasets.  
> You can find the training code and related resources in my GitHub repository: [2025-korean-asr-benchmark](https://github.com/baeseongsu/2025-korean-asr-benchmark)

<!-- 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. -->

# openai/whisper-large-v3-turbo Korean - Fine-tuned

This model is a fine-tuned version of [openai/whisper-large-v3-turbo](https://huggingface.co/openai/whisper-large-v3-turbo) on the Bingsu/zeroth-korean dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0733
- Wer: 4.3216

## 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: 128
- eval_batch_size: 8
- seed: 42
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 5
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer     |
|:-------------:|:------:|:----:|:---------------:|:-------:|
| 0.4622        | 0.1437 | 25   | 0.3690          | 19.2290 |
| 0.2           | 0.2874 | 50   | 0.1578          | 15.3290 |
| 0.1213        | 0.4310 | 75   | 0.1396          | 13.0703 |
| 0.1068        | 0.5747 | 100  | 0.1314          | 12.2572 |
| 0.1           | 0.7184 | 125  | 0.1242          | 11.0676 |
| 0.0922        | 0.8621 | 150  | 0.1181          | 10.6460 |
| 0.0895        | 1.0057 | 175  | 0.1122          | 9.6371  |
| 0.0667        | 1.1494 | 200  | 0.1098          | 9.2155  |
| 0.0608        | 1.2931 | 225  | 0.1049          | 8.4023  |
| 0.0608        | 1.4368 | 250  | 0.1007          | 7.6946  |
| 0.0577        | 1.5805 | 275  | 0.0992          | 7.4386  |
| 0.0591        | 1.7241 | 300  | 0.0953          | 6.5502  |
| 0.0547        | 1.8678 | 325  | 0.0920          | 5.9630  |
| 0.0518        | 2.0115 | 350  | 0.0885          | 5.5112  |
| 0.0299        | 2.1552 | 375  | 0.0878          | 5.8877  |
| 0.0311        | 2.2989 | 400  | 0.0872          | 4.8637  |
| 0.0319        | 2.4425 | 425  | 0.0895          | 5.2552  |
| 0.0363        | 2.5862 | 450  | 0.0869          | 5.1197  |
| 0.0325        | 2.7299 | 475  | 0.0851          | 4.9390  |
| 0.0331        | 2.8736 | 500  | 0.0849          | 4.7282  |
| 0.0314        | 3.0172 | 525  | 0.0805          | 4.9240  |
| 0.0196        | 3.1609 | 550  | 0.0805          | 4.5174  |
| 0.0164        | 3.3046 | 575  | 0.0820          | 5.4209  |
| 0.0166        | 3.4483 | 600  | 0.0807          | 6.1135  |
| 0.0153        | 3.5920 | 625  | 0.0775          | 3.9753  |
| 0.0127        | 3.7356 | 650  | 0.0741          | 4.8035  |
| 0.014         | 3.8793 | 675  | 0.0731          | 7.1827  |
| 0.012         | 4.0230 | 700  | 0.0719          | 4.9992  |
| 0.0067        | 4.1667 | 725  | 0.0744          | 4.5475  |
| 0.0061        | 4.3103 | 750  | 0.0732          | 5.2101  |
| 0.0053        | 4.4540 | 775  | 0.0736          | 4.4270  |
| 0.0061        | 4.5977 | 800  | 0.0743          | 4.8938  |
| 0.0048        | 4.7414 | 825  | 0.0740          | 5.3305  |
| 0.0045        | 4.8851 | 850  | 0.0733          | 4.3216  |


### Framework versions

- Transformers 4.50.0.dev0
- Pytorch 2.4.0+cu121
- Datasets 2.21.0
- Tokenizers 0.21.1