File size: 1,582 Bytes
ca37ec6
b364455
 
ca37ec6
 
 
 
 
b364455
ca37ec6
 
 
 
 
 
 
 
b364455
 
 
 
 
 
 
 
ca37ec6
 
 
b364455
ca37ec6
 
 
b364455
ca37ec6
 
 
 
 
 
 
 
b364455
ca37ec6
 
 
 
 
 
b364455
ca37ec6
 
 
 
 
b364455
 
 
ca37ec6
 
 
 
b364455
 
 
 
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
---
license: apache-2.0
base_model: openai/whisper-tiny
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: whisper-tiny-ft-cy-en
  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-ft-cy-en

This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 0.5668
- Wer: 36.8865

## 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: 1e-05
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 3000

### Training results

| Training Loss | Epoch | Step | Validation Loss | Wer     |
|:-------------:|:-----:|:----:|:---------------:|:-------:|
| 0.7039        | 0.25  | 1000 | 0.6932          | 43.4217 |
| 0.5689        | 0.5   | 2000 | 0.5930          | 38.4145 |
| 0.5255        | 0.75  | 3000 | 0.5668          | 36.8865 |


### Framework versions

- Transformers 4.39.3
- Pytorch 2.2.2+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2