File size: 2,428 Bytes
e94d61e
11930a4
 
e94d61e
 
11930a4
e94d61e
 
11930a4
e94d61e
 
 
11930a4
e94d61e
 
 
 
 
11930a4
 
e94d61e
 
 
 
 
 
dc6df58
e94d61e
 
 
 
 
11930a4
e94d61e
11930a4
e94d61e
dc6df58
 
e94d61e
 
 
 
 
 
 
f1570bc
e94d61e
 
 
 
 
 
 
 
 
 
 
257c7d5
 
e94d61e
 
 
 
dc6df58
e94d61e
 
 
 
 
 
dc6df58
 
 
 
 
 
 
e94d61e
 
 
 
 
 
 
 
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
---
language:
- eu
license: apache-2.0
tags:
- whisper-event
- generated_from_trainer
datasets:
- mozilla-foundation/common_voice_13_0
metrics:
- wer
model-index:
- name: Whisper Small Basque
  results:
  - task:
      name: Automatic Speech Recognition
      type: automatic-speech-recognition
    dataset:
      name: mozilla-foundation/common_voice_13_0 eu
      type: mozilla-foundation/common_voice_13_0
      config: eu
      split: test
      args: eu
    metrics:
    - name: Wer
      type: wer
      value: 13.179958686054519
---

<!-- 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 Small Basque

This model is a fine-tuned version of [openai/whisper-medium](https://huggingface.co/openai/whisper-medium) on the mozilla-foundation/common_voice_13_0 eu dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2201
- Wer: 13.1800

## Model description

More information needed

## Intended uses & limitations

If you need to use this model with [whisper.cpp](https://github.com/ggerganov/whisper.cpp), you can download the ggml file: [ggml-medium-eu.bin](https://huggingface.co/xezpeleta/whisper-medium-eu/blob/main/ggml-medium.eu.bin)

## 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: 4
- 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: 500
- training_steps: 7000
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer     |
|:-------------:|:-----:|:----:|:---------------:|:-------:|
| 0.4203        | 0.14  | 1000 | 0.4128          | 28.2656 |
| 0.2693        | 0.29  | 2000 | 0.3240          | 22.0523 |
| 0.2228        | 0.43  | 3000 | 0.2737          | 18.1437 |
| 0.1002        | 1.1   | 4000 | 0.2554          | 16.3534 |
| 0.0863        | 1.24  | 5000 | 0.2351          | 14.7880 |
| 0.0636        | 1.39  | 6000 | 0.2251          | 13.5971 |
| 0.0271        | 2.06  | 7000 | 0.2201          | 13.1800 |


### Framework versions

- Transformers 4.26.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.8.1.dev0
- Tokenizers 0.13.2