metadata
language:
- ko
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- gglabs/stt-test-1223
metrics:
- wer
model-index:
- name: Whisper Small ko
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: custom
type: gglabs/stt-test-1223
args: 'config: ko, split: test'
metrics:
- name: Wer
type: wer
value: 46.19565217391305
Whisper Small ko
This model is a fine-tuned version of openai/whisper-small on the custom dataset. It achieves the following results on the evaluation set:
- Loss: 1.0537
- Wer: 46.1957
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: 8
- 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: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.9628 | 0.0112 | 10 | 1.5113 | 55.4348 |
| 1.0128 | 0.0223 | 20 | 1.5033 | 53.2609 |
| 0.7642 | 0.0335 | 30 | 1.4807 | 48.3696 |
| 0.683 | 0.0446 | 40 | 1.4422 | 47.8261 |
| 0.4836 | 0.0558 | 50 | 1.3840 | 48.3696 |
| 0.207 | 0.0670 | 60 | 1.2772 | 49.4565 |
| 0.0984 | 0.0781 | 70 | 1.1608 | 47.8261 |
| 0.1485 | 0.0893 | 80 | 1.1003 | 45.1087 |
| 0.0538 | 0.1004 | 90 | 1.0760 | 44.5652 |
| 0.0349 | 0.1116 | 100 | 1.0537 | 46.1957 |
Framework versions
- Transformers 4.42.3
- Pytorch 2.3.1+cu121
- Datasets 3.0.0
- Tokenizers 0.19.1