whisper-base-lki
This model is a fine-tuned version of openai/whisper-base on the razhan/DOLMA-speech laki_kurdish dataset. It achieves the following results on the evaluation set:
- Loss: 2.8590
- Wer: 1.0084
- Cer: 0.5076
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: 192
- eval_batch_size: 128
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 384
- total_eval_batch_size: 256
- optimizer: Use 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: 1
- num_epochs: 5.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| No log | 1.0 | 2 | 4.0312 | 1.1143 | 0.5721 |
| No log | 2.0 | 4 | 4.0312 | 1.1143 | 0.5721 |
| No log | 3.0 | 6 | 4.0312 | 1.1143 | 0.5721 |
| No log | 4.0 | 8 | 3.4312 | 1.1106 | 0.5898 |
| 3.8716 | 5.0 | 10 | 2.8590 | 1.0084 | 0.5076 |
Framework versions
- Transformers 4.49.0.dev0
- Pytorch 2.6.0+cu124
- Datasets 3.2.0
- Tokenizers 0.21.0