File size: 2,551 Bytes
c8fd020
9a0d2d2
 
c8fd020
 
 
9a0d2d2
c8fd020
 
9a0d2d2
c8fd020
 
 
9a0d2d2
c8fd020
 
 
 
 
9a0d2d2
 
c8fd020
 
 
 
 
 
 
 
 
 
 
 
9a0d2d2
c8fd020
9a0d2d2
c8fd020
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- pt
license: apache-2.0
base_model: openai/whisper-base
tags:
- hf-asr-leaderboard
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_16_0
metrics:
- wer
model-index:
- name: Whisper Base using Common Voice 16 (pt)
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: Mozilla Common Voices - 16.0 - Portuguese
      type: mozilla-foundation/common_voice_16_0
      config: pt
      split: test[0:5400]
      args: pt
    metrics:
    - name: Wer
      type: wer
      value: 25.542580301884676
---

<!-- 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 Base using Common Voice 16 (pt)

This model is a fine-tuned version of [openai/whisper-base](https://huggingface.co/openai/whisper-base) on the Mozilla Common Voices - 16.0 - Portuguese dataset.
It achieves the following results on the evaluation set:
- Loss: 0.3952
- Wer: 25.5426
- Wer Normalized: 19.7098

## 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: linear
- lr_scheduler_warmup_steps: 400
- training_steps: 4000
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer     | Wer Normalized |
|:-------------:|:-----:|:----:|:---------------:|:-------:|:--------------:|
| 0.5344        | 0.37  | 500  | 0.5264          | 35.9965 | 30.1234        |
| 0.438         | 0.74  | 1000 | 0.4904          | 33.4453 | 28.3776        |
| 0.1871        | 1.11  | 1500 | 0.4595          | 30.3929 | 24.5163        |
| 0.1955        | 1.48  | 2000 | 0.4342          | 28.6566 | 22.9762        |
| 0.1754        | 1.85  | 2500 | 0.4199          | 28.2674 | 22.4147        |
| 0.0649        | 2.22  | 3000 | 0.4090          | 26.7860 | 20.7689        |
| 0.0595        | 2.59  | 3500 | 0.4026          | 26.1839 | 20.2018        |
| 0.0626        | 2.96  | 4000 | 0.3952          | 25.5426 | 19.7098        |


### Framework versions

- Transformers 4.36.2
- Pytorch 2.1.1
- Datasets 2.16.1
- Tokenizers 0.15.0