metadata
library_name: peft
language:
- ko
license: mit
base_model: openai/whisper-large-v3-turbo
tags:
- generated_from_trainer
model-index:
- name: Whisper Turbo ko
results: []
Whisper Turbo ko
This model is a fine-tuned version of openai/whisper-large-v3-turbo on the custom dataset. It achieves the following results on the evaluation set:
- Loss: 0.0940
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: 0.001
- train_batch_size: 64
- eval_batch_size: 256
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.3061 | 0.5405 | 100 | 0.6169 |
| 0.107 | 1.0811 | 200 | 0.3840 |
| 0.0871 | 1.6216 | 300 | 0.3024 |
| 0.0801 | 2.1622 | 400 | 0.2429 |
| 0.0608 | 2.7027 | 500 | 0.2094 |
| 0.0527 | 3.2432 | 600 | 0.1674 |
| 0.0377 | 3.7838 | 700 | 0.1404 |
| 0.0316 | 4.3243 | 800 | 0.1230 |
| 0.0302 | 4.8649 | 900 | 0.1004 |
| 0.0227 | 5.4054 | 1000 | 0.0940 |
Framework versions
- PEFT 0.14.0
- Transformers 4.47.1
- Pytorch 2.5.1+cu124
- Datasets 3.2.0
- Tokenizers 0.21.0