Whisper Large Khmer - Kak Soky
This model is a fine-tuned version of openai/whisper-large on the SLR42 dataset. It achieves the following results on the evaluation set:
- Loss: 0.2375
- Wer: 29.5183
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
- gradient_accumulation_steps: 4
- 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: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.0102 | 12.34 | 1000 | 0.2228 | 38.2659 |
| 0.0003 | 24.69 | 2000 | 0.2260 | 30.7900 |
| 0.0001 | 37.04 | 3000 | 0.2310 | 30.0578 |
| 0.0 | 49.38 | 4000 | 0.2375 | 29.5183 |
Framework versions
- Transformers 4.25.0.dev0
- Pytorch 1.9.1+cu111
- Datasets 2.3.2
- Tokenizers 0.12.1