File size: 3,441 Bytes
505470b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
library_name: peft
language:
- it
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
- generated_from_trainer
datasets:
- ASR_BB_and_EC
metrics:
- wer
model-index:
- name: Whisper Large v3
  results:
  - task:
      type: automatic-speech-recognition
      name: Automatic Speech Recognition
    dataset:
      name: ASR_BB_and_EC
      type: ASR_BB_and_EC
      config: default
      split: train
      args: default
    metrics:
    - type: wer
      value: 145.18950437317784
      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 Large v3

This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the ASR_BB_and_EC dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2251
- Wer: 145.1895

## 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: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- num_epochs: 3
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Wer      |
|:-------------:|:------:|:----:|:---------------:|:--------:|
| 3.4667        | 0.1142 | 50   | 2.3898          | 126.5306 |
| 1.3136        | 0.2283 | 100  | 0.8863          | 65.3061  |
| 0.6296        | 0.3425 | 150  | 0.7403          | 55.9767  |
| 0.551         | 0.4566 | 200  | 0.6749          | 61.2245  |
| 0.4789        | 0.5708 | 250  | 0.6446          | 67.6385  |
| 0.4246        | 0.6849 | 300  | 0.5675          | 77.5510  |
| 0.3786        | 0.7991 | 350  | 0.5163          | 45.4810  |
| 0.3179        | 0.9132 | 400  | 0.4786          | 84.8397  |
| 0.3118        | 1.0274 | 450  | 0.4678          | 105.5394 |
| 0.2689        | 1.1416 | 500  | 0.4322          | 125.3644 |
| 0.2473        | 1.2557 | 550  | 0.3924          | 48.1050  |
| 0.2319        | 1.3699 | 600  | 0.3980          | 208.7464 |
| 0.2098        | 1.4840 | 650  | 0.3545          | 52.1866  |
| 0.2215        | 1.5982 | 700  | 0.3489          | 48.1050  |
| 0.1981        | 1.7123 | 750  | 0.3378          | 76.3848  |
| 0.1803        | 1.8265 | 800  | 0.3295          | 43.7318  |
| 0.1693        | 1.9406 | 850  | 0.3095          | 76.9679  |
| 0.1406        | 2.0548 | 900  | 0.2993          | 43.4402  |
| 0.1252        | 2.1689 | 950  | 0.2810          | 37.3178  |
| 0.111         | 2.2831 | 1000 | 0.2854          | 164.1399 |
| 0.1166        | 2.3973 | 1050 | 0.2752          | 124.4898 |
| 0.1183        | 2.5114 | 1100 | 0.2493          | 90.3790  |
| 0.1014        | 2.6256 | 1150 | 0.2441          | 210.2041 |
| 0.1076        | 2.7397 | 1200 | 0.2340          | 152.1866 |
| 0.0891        | 2.8539 | 1250 | 0.2312          | 214.5773 |
| 0.0841        | 2.9680 | 1300 | 0.2251          | 145.1895 |


### Framework versions

- PEFT 0.13.2
- Transformers 4.45.2
- Pytorch 2.2.0
- Datasets 3.1.0
- Tokenizers 0.20.3