Whisper Small Sanskrit_4 - Bidit Sadhukhan
This model is a fine-tuned version of openai/whisper-small on the load_data dataset. It achieves the following results on the evaluation set:
- Loss: 0.1193
- Wer: 27.9415
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: 2.5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 44
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 10000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1168 | 0.12 | 500 | 0.2069 | 47.5067 |
| 0.0854 | 0.23 | 1000 | 0.1774 | 43.4951 |
| 0.0766 | 0.35 | 1500 | 0.1659 | 40.7385 |
| 0.0631 | 0.47 | 2000 | 0.1644 | 39.0968 |
| 0.0629 | 0.58 | 2500 | 0.1379 | 34.7266 |
| 0.0763 | 0.7 | 3000 | 0.1401 | 35.1916 |
| 0.0515 | 0.82 | 3500 | 0.1343 | 34.8386 |
| 0.0457 | 0.93 | 4000 | 0.1185 | 31.6114 |
| 0.0302 | 1.05 | 4500 | 0.1315 | 33.1074 |
| 0.0276 | 1.17 | 5000 | 0.1245 | 31.1127 |
| 0.0234 | 1.28 | 5500 | 0.1265 | 30.9166 |
| 0.0266 | 1.4 | 6000 | 0.1289 | 30.6029 |
| 0.0186 | 1.52 | 6500 | 0.1230 | 30.1658 |
| 0.0284 | 1.63 | 7000 | 0.1157 | 29.2414 |
| 0.0182 | 1.75 | 7500 | 0.1125 | 27.6165 |
| 0.024 | 1.87 | 8000 | 0.1143 | 28.5970 |
| 0.0214 | 1.98 | 8500 | 0.1097 | 27.2972 |
| 0.0055 | 2.1 | 9000 | 0.1152 | 28.2497 |
| 0.0069 | 2.21 | 9500 | 0.1210 | 27.3364 |
| 0.0101 | 2.33 | 10000 | 0.1193 | 27.9415 |
Framework versions
- Transformers 4.35.0.dev0
- Pytorch 2.0.1+cu117
- Datasets 2.14.5
- Tokenizers 0.14.0