File size: 3,103 Bytes
c9165de
 
 
 
 
 
 
 
 
 
52c6888
 
c9165de
 
52c6888
 
 
 
 
 
 
 
 
 
 
 
 
 
c9165de
 
 
 
 
 
 
 
52c6888
 
 
 
c9165de
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
52c6888
c9165de
 
 
 
 
 
 
 
52c6888
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c9165de
 
 
 
 
 
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
---
library_name: transformers
language:
- ne
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
- generated_from_trainer
datasets:
- kiranpantha/OpenSLR54-Balanced-Nepali
metrics:
- wer
model-index:
- name: Whisper Large v3  Nepali - Kiran Pantha
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: OpenSLR54
      type: kiranpantha/OpenSLR54-Balanced-Nepali
      config: default
      split: test
      args: 'config: ne, split: test'
    metrics:
    - name: Wer
      type: wer
      value: 20.48611111111111
---

<!-- 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 Large v3  Nepali - Kiran Pantha

This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the OpenSLR54 dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1507
- Wer: 20.4861
- Cer: 4.9839

## 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: 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
- training_steps: 5000
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step | Cer     | Validation Loss | Wer     |
|:-------------:|:------:|:----:|:-------:|:---------------:|:-------:|
| 0.209         | 0.3597 | 300  | 10.9916 | 0.2021          | 41.1574 |
| 0.1714        | 0.7194 | 600  | 8.9474  | 0.1755          | 35.7176 |
| 0.101         | 1.0791 | 900  | 7.5565  | 0.1485          | 29.6991 |
| 0.0902        | 1.4388 | 1200 | 7.2372  | 0.1396          | 28.2407 |
| 0.0872        | 1.7986 | 1500 | 7.8024  | 0.1319          | 27.9861 |
| 0.0453        | 2.1583 | 1800 | 6.3344  | 0.1374          | 26.2269 |
| 0.0368        | 2.5180 | 2100 | 6.1766  | 0.1381          | 25.2315 |
| 0.0472        | 2.8777 | 2400 | 5.8316  | 0.1316          | 24.1435 |
| 0.0191        | 3.2374 | 2700 | 5.8059  | 0.1356          | 24.0278 |
| 0.0185        | 3.5971 | 3000 | 5.5674  | 0.1376          | 23.125  |
| 0.0182        | 3.9568 | 3300 | 5.5123  | 0.1360          | 23.0556 |
| 0.0074        | 4.3165 | 3600 | 5.2077  | 0.1428          | 21.7130 |
| 0.0086        | 4.6763 | 3900 | 5.1784  | 0.1433          | 21.2731 |
| 0.0031        | 5.0360 | 4200 | 0.1421  | 21.1806         | 5.0279  |
| 0.0024        | 5.3957 | 4500 | 0.1482  | 20.7870         | 4.9912  |
| 0.0014        | 5.7554 | 4800 | 0.1507  | 20.4861         | 4.9839  |


### Framework versions

- Transformers 4.46.3
- Pytorch 2.5.1+cxx11.abi
- Datasets 3.2.0
- Tokenizers 0.20.3