wav2vec2-1b-Yfreq_pause_speed
This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 1.1898
- Cer: 31.1501
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: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 50
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer |
|---|---|---|---|---|
| 10.7019 | 0.2580 | 200 | 5.2871 | 100.0 |
| 2.9862 | 0.5160 | 400 | 2.5816 | 61.0021 |
| 1.4741 | 0.7741 | 600 | 2.3005 | 54.0120 |
| 1.1621 | 1.0321 | 800 | 1.7372 | 43.0569 |
| 0.9078 | 1.2901 | 1000 | 1.5138 | 38.3459 |
| 0.8152 | 1.5481 | 1200 | 1.5787 | 43.2566 |
| 0.7734 | 1.8062 | 1400 | 1.5237 | 40.7307 |
| 0.639 | 2.0642 | 1600 | 1.5034 | 37.8759 |
| 0.5611 | 2.3222 | 1800 | 1.4911 | 38.7277 |
| 0.5272 | 2.5802 | 2000 | 1.3679 | 37.0418 |
| 0.4739 | 2.8383 | 2200 | 1.5056 | 41.7469 |
| 0.4307 | 3.0963 | 2400 | 1.2907 | 34.7568 |
| 0.3732 | 3.3543 | 2600 | 1.4109 | 36.4250 |
| 0.3388 | 3.6123 | 2800 | 1.3857 | 37.1299 |
| 0.3089 | 3.8703 | 3000 | 1.2816 | 33.6760 |
| 0.2648 | 4.1284 | 3200 | 1.2267 | 32.7596 |
| 0.2299 | 4.3864 | 3400 | 1.2182 | 32.3896 |
| 0.2108 | 4.6444 | 3600 | 1.1818 | 31.1090 |
| 0.2055 | 4.9024 | 3800 | 1.1898 | 31.1501 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.3.1.post100
- Datasets 2.19.1
- Tokenizers 0.20.1