File size: 2,654 Bytes
8be09dc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7ceafc1
 
 
8be09dc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53746d9
8be09dc
 
 
 
 
7ceafc1
8be09dc
 
41c80ff
8be09dc
 
 
 
7ceafc1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8be09dc
 
 
 
7ceafc1
 
8be09dc
7ceafc1
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
---
library_name: transformers
language:
- ar
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: Whisper tiny AR - BH
  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 tiny AR - BH

This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the quran-ayat-speech-to-text dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0095
- Wer: 0.1037
- Cer: 0.0382

## 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-06
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: 15
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch   | Step | Validation Loss | Wer    | Cer    |
|:-------------:|:-------:|:----:|:---------------:|:------:|:------:|
| 0.0104        | 1.0     | 407  | 0.0098          | 0.1182 | 0.0449 |
| 0.0068        | 2.0     | 814  | 0.0088          | 0.1055 | 0.0373 |
| 0.0075        | 3.0     | 1221 | 0.0088          | 0.1008 | 0.0356 |
| 0.0057        | 4.0     | 1628 | 0.0091          | 0.0992 | 0.0345 |
| 0.0047        | 5.0     | 2035 | 0.0097          | 0.0997 | 0.0349 |
| 0.0038        | 6.0     | 2442 | 0.0103          | 0.0994 | 0.0340 |
| 0.0024        | 7.0     | 2849 | 0.0109          | 0.1033 | 0.0357 |
| 0.0031        | 8.0     | 3256 | 0.0113          | 0.1015 | 0.0351 |
| 0.0014        | 9.0     | 3663 | 0.0118          | 0.1003 | 0.0350 |
| 0.0018        | 10.0    | 4070 | 0.0123          | 0.1014 | 0.0349 |
| 0.0013        | 11.0    | 4477 | 0.0128          | 0.1122 | 0.0405 |
| 0.0011        | 12.0    | 4884 | 0.0130          | 0.1037 | 0.0379 |
| 0.0004        | 13.0    | 5291 | 0.0132          | 0.1032 | 0.0379 |
| 0.0019        | 14.0    | 5698 | 0.0141          | 0.1055 | 0.0397 |
| 0.001         | 14.9643 | 6090 | 0.0135          | 0.1017 | 0.0371 |


### Framework versions

- Transformers 4.47.0
- Pytorch 2.5.1+cu121
- Datasets 3.2.0
- Tokenizers 0.21.0