wav2vec2-1b-Ypause
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: 1.3032
- Cer: 30.4276
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 |
|---|---|---|---|---|
| 8.4175 | 0.2581 | 200 | 3.4467 | 71.5695 |
| 1.5788 | 0.5161 | 400 | 2.0587 | 46.5226 |
| 1.0864 | 0.7742 | 600 | 1.9065 | 46.5108 |
| 0.8759 | 1.0323 | 800 | 1.6587 | 42.8924 |
| 0.7067 | 1.2903 | 1000 | 1.4755 | 40.9951 |
| 0.615 | 1.5484 | 1200 | 1.6840 | 43.0510 |
| 0.5547 | 1.8065 | 1400 | 1.7452 | 43.5620 |
| 0.5415 | 2.0645 | 1600 | 1.7313 | 42.6516 |
| 0.437 | 2.3226 | 1800 | 1.5706 | 40.1316 |
| 0.4045 | 2.5806 | 2000 | 1.2070 | 32.2427 |
| 0.3736 | 2.8387 | 2200 | 1.4442 | 37.0007 |
| 0.3264 | 3.0968 | 2400 | 1.2398 | 31.9549 |
| 0.2651 | 3.3548 | 2600 | 1.4558 | 35.4558 |
| 0.2621 | 3.6129 | 2800 | 1.2838 | 32.6363 |
| 0.23 | 3.8710 | 3000 | 1.4202 | 33.7935 |
| 0.208 | 4.1290 | 3200 | 1.2976 | 31.2970 |
| 0.1781 | 4.3871 | 3400 | 1.3553 | 32.2310 |
| 0.1628 | 4.6452 | 3600 | 1.3637 | 32.0489 |
| 0.1573 | 4.9032 | 3800 | 1.3032 | 30.4276 |
Framework versions
- Transformers 4.45.2
- Pytorch 2.3.1.post100
- Datasets 2.19.1
- Tokenizers 0.20.1