Nystrom-W2V2-100hrs-take-3
This model was trained from scratch on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 30.1649
- Wer: 0.1047
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 64
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|
| 55.4504 | 9.01 | 1000 | 29.0675 | 0.1464 |
| 65.4799 | 18.02 | 2000 | 26.9263 | 0.1580 |
| 72.6609 | 27.03 | 3000 | 27.2220 | 0.1500 |
| 65.6264 | 36.04 | 4000 | 26.4758 | 0.1426 |
| 57.9496 | 45.04 | 5000 | 27.0818 | 0.1349 |
| 49.6643 | 54.05 | 6000 | 27.9658 | 0.1269 |
| 42.5205 | 63.06 | 7000 | 28.6973 | 0.1214 |
| 36.1799 | 72.07 | 8000 | 28.1021 | 0.1128 |
| 30.9742 | 81.08 | 9000 | 29.9000 | 0.1093 |
| 27.4728 | 90.09 | 10000 | 30.2661 | 0.1057 |
| 26.0383 | 99.1 | 11000 | 30.1649 | 0.1047 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1
- Datasets 2.7.1
- Tokenizers 0.11.0
