ndizi_whisper_large_v3_turbo_merged_mms_fa
This model is a fine-tuned version of openai/whisper-large-v3-turbo on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.0848
- Wer: 0.5121
- Cer: 0.3470
- Score: 57.0411
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 8
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- num_epochs: 25.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer | Score |
|---|---|---|---|---|---|---|
| 4.3565 | 1.0 | 437 | 0.8869 | 0.5643 | 0.3792 | 52.8205 |
| 2.8614 | 2.0 | 874 | 0.8560 | 0.5174 | 0.3553 | 56.3667 |
| 2.2048 | 3.0 | 1311 | 0.8441 | 0.5271 | 0.3627 | 55.5090 |
| 1.5223 | 4.0 | 1748 | 0.8966 | 0.5385 | 0.3703 | 54.5589 |
| 1.1678 | 5.0 | 2185 | 0.9403 | 0.5061 | 0.3490 | 57.2444 |
| 0.7910 | 6.0 | 2622 | 0.9676 | 0.5243 | 0.3523 | 56.1668 |
| 0.5883 | 7.0 | 3059 | 1.0363 | 0.5128 | 0.3511 | 56.8059 |
| 0.4251 | 8.0 | 3496 | 1.0848 | 0.5121 | 0.3470 | 57.0411 |
Framework versions
- Transformers 5.8.0
- Pytorch 2.11.0+cu128
- Datasets 3.6.0
- Tokenizers 0.22.2