wav2vec2-xlsr-53-torgo-origin-parameters-checkpoint3750
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the Torgo dataset. It achieves the following results on the evaluation set:
- Loss: 1.2104
- Cer: 0.3903
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: 4
- 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: 1000
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| No log | 0.63 | 250 | 19.6794 | 0.9961 |
| 25.0027 | 1.26 | 500 | 5.6998 | 0.9961 |
| 25.0027 | 1.89 | 750 | 5.3298 | 0.9961 |
| 3.7083 | 2.53 | 1000 | 4.9012 | 0.9961 |
| 3.7083 | 3.16 | 1250 | 4.7392 | 0.9961 |
| 3.3223 | 3.79 | 1500 | 3.8672 | 0.9961 |
| 3.3223 | 4.42 | 1750 | 2.5696 | 0.8772 |
| 2.564 | 5.05 | 2000 | 2.0229 | 0.7136 |
| 2.564 | 5.68 | 2250 | 1.6933 | 0.5332 |
| 1.7148 | 6.31 | 2500 | 1.4409 | 0.4808 |
| 1.7148 | 6.94 | 2750 | 1.3612 | 0.4601 |
| 1.3384 | 7.58 | 3000 | 1.3402 | 0.4328 |
| 1.3384 | 8.21 | 3250 | 1.2438 | 0.4070 |
| 1.1197 | 8.84 | 3500 | 1.2251 | 0.3938 |
| 1.1197 | 9.47 | 3750 | 1.2104 | 0.3903 |
Framework versions
- Transformers 4.18.0
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1