Wave2Vec2-Bert2.0 - Kiran Pantha
This model is a fine-tuned version of kiranpantha/w2v-bert-2.0-nepali-iteration-test on the OpenSLR54 dataset. It achieves the following results on the evaluation set:
- Loss: 0.4545
- Wer: 0.4636
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: 5e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 2
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.392 | 0.15 | 300 | 0.4561 | 0.4579 |
| 0.5077 | 0.3 | 600 | 0.5402 | 0.5145 |
| 0.5371 | 0.45 | 900 | 0.5083 | 0.4923 |
| 0.4943 | 0.6 | 1200 | 0.5103 | 0.5147 |
| 0.5049 | 0.75 | 1500 | 0.4811 | 0.4752 |
| 0.4809 | 0.9 | 1800 | 0.4751 | 0.4689 |
| 0.4633 | 1.05 | 2100 | 0.5031 | 0.4883 |
| 0.3843 | 1.2 | 2400 | 0.4703 | 0.4573 |
| 0.3753 | 1.35 | 2700 | 0.4542 | 0.4497 |
| 0.356 | 1.5 | 3000 | 0.4510 | 0.4503 |
| 0.3668 | 1.65 | 3300 | 0.4591 | 0.4524 |
| 0.3386 | 1.8 | 3600 | 0.4399 | 0.4459 |
| 0.3452 | 1.95 | 3900 | 0.4545 | 0.4636 |
Framework versions
- Transformers 4.45.0.dev0
- Pytorch 2.4.1+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1