metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- elite_voice_project
metrics:
- wer
model-index:
- name: whisper-small-ja-elite
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: elite_voice_project
type: elite_voice_project
config: twitter
split: test
args: twitter
metrics:
- name: Wer
type: wer
value: 4.878048780487805
whisper-small-ja-elite
This model is a fine-tuned version of openai/whisper-small on the elite_voice_project dataset. It achieves the following results on the evaluation set:
- Loss: 0.1502
- Wer: 4.8780
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 50
- training_steps: 5000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0004 | 58.0 | 1000 | 0.1129 | 1.8293 |
| 0.0 | 117.0 | 2000 | 0.1232 | 1.8293 |
| 0.0 | 176.0 | 3000 | 0.1327 | 1.8293 |
| 0.0 | 235.0 | 4000 | 0.1401 | 4.8780 |
| 0.0 | 294.0 | 5000 | 0.1502 | 4.8780 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.8.1.dev0
- Tokenizers 0.13.2