File size: 1,769 Bytes
ecfff5b
6bbd6f7
 
ecfff5b
 
 
 
 
 
 
6bbd6f7
ecfff5b
 
 
 
 
 
6bbd6f7
ecfff5b
 
 
fd1bcdb
 
 
ecfff5b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- arp
license: apache-2.0
base_model: openai/whisper-small.en
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: whisper-arpa
  results: []
---

<!-- 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-arpa

This model is a fine-tuned version of [openai/whisper-small.en](https://huggingface.co/openai/whisper-small.en) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0186
- Wer: 2.9762
- Cer: 1.0999

## 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: 5e-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: cosine
- lr_scheduler_warmup_steps: 1000
- num_epochs: 1
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step  | Cer    | Validation Loss | Wer    |
|:-------------:|:------:|:-----:|:------:|:---------------:|:------:|
| 0.0529        | 0.5716 | 5000  | 3.5532 | 0.0494          | 7.8609 |
| 0.028         | 1.1432 | 10000 | 5.5656 | 0.0413          | 9.7632 |
| 0.0298        | 1.7149 | 15000 | 2.9926 | 0.0350          | 6.4210 |
| 0.0133        | 2.2865 | 20000 | 2.1896 | 0.0322          | 5.0360 |
| 0.0105        | 2.8581 | 25000 | 2.0795 | 0.0307          | 4.7797 |


### Framework versions

- Transformers 4.42.3
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1