wav2vec2-E10_speed2_
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.8373
- Cer: 35.0353
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: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- 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: 50
- num_epochs: 3
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 44.0517 | 0.1289 | 200 | 5.2881 | 100.0 |
| 5.0678 | 0.2579 | 400 | 4.6640 | 100.0 |
| 4.8617 | 0.3868 | 600 | 4.6368 | 100.0 |
| 4.8476 | 0.5158 | 800 | 4.6517 | 100.0 |
| 4.7683 | 0.6447 | 1000 | 4.6522 | 100.0 |
| 4.7437 | 0.7737 | 1200 | 4.5665 | 100.0 |
| 4.7147 | 0.9026 | 1400 | 4.6060 | 100.0 |
| 4.6579 | 1.0316 | 1600 | 4.5314 | 100.0 |
| 4.6231 | 1.1605 | 1800 | 4.6514 | 99.9471 |
| 4.4643 | 1.2895 | 2000 | 4.3773 | 98.1727 |
| 4.292 | 1.4184 | 2200 | 4.2086 | 92.5734 |
| 3.8631 | 1.5474 | 2400 | 3.7279 | 67.4266 |
| 3.3619 | 1.6763 | 2600 | 3.1229 | 57.7380 |
| 2.9826 | 1.8053 | 2800 | 2.9814 | 55.7403 |
| 2.7082 | 1.9342 | 3000 | 2.7001 | 50.0353 |
| 2.4949 | 2.0632 | 3200 | 2.4357 | 44.9295 |
| 2.2833 | 2.1921 | 3400 | 2.2863 | 43.0846 |
| 2.141 | 2.3211 | 3600 | 2.1369 | 39.6769 |
| 2.0008 | 2.4500 | 3800 | 2.0446 | 38.5840 |
| 1.9496 | 2.5790 | 4000 | 1.9128 | 36.3220 |
| 1.8774 | 2.7079 | 4200 | 1.8870 | 35.8108 |
| 1.7961 | 2.8369 | 4400 | 1.8657 | 35.8049 |
| 1.7394 | 2.9658 | 4600 | 1.8373 | 35.0353 |
Framework versions
- Transformers 4.46.3
- Pytorch 2.5.1+cu121
- Datasets 3.1.0
- Tokenizers 0.20.3