xlsr-syntesized-turkish-2-hour-llr-2
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3989
- Wer: 0.4791
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: 0.0025
- train_batch_size: 2
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 200
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.8587 | 1.04 | 100 | 2.9955 | 1.0 |
| 2.6054 | 2.08 | 200 | 1.6852 | 0.9992 |
| 1.2192 | 3.12 | 300 | 0.6653 | 0.8583 |
| 0.755 | 4.17 | 400 | 0.5134 | 0.7158 |
| 0.6025 | 5.21 | 500 | 0.4487 | 0.6965 |
| 0.5219 | 6.25 | 600 | 0.4126 | 0.6230 |
| 0.4789 | 7.29 | 700 | 0.4284 | 0.6378 |
| 0.4103 | 8.33 | 800 | 0.3730 | 0.5873 |
| 0.3717 | 9.38 | 900 | 0.3521 | 0.5442 |
| 0.3318 | 10.42 | 1000 | 0.3619 | 0.5583 |
| 0.3039 | 11.46 | 1100 | 0.3741 | 0.5296 |
| 0.2849 | 12.5 | 1200 | 0.3656 | 0.5045 |
| 0.2497 | 13.54 | 1300 | 0.3773 | 0.5286 |
| 0.2287 | 14.58 | 1400 | 0.3919 | 0.5554 |
| 0.217 | 15.62 | 1500 | 0.3831 | 0.5287 |
| 0.1899 | 16.67 | 1600 | 0.3876 | 0.4928 |
| 0.1798 | 17.71 | 1700 | 0.3967 | 0.4910 |
| 0.1772 | 18.75 | 1800 | 0.3982 | 0.4877 |
| 0.1631 | 19.79 | 1900 | 0.3989 | 0.4791 |
Framework versions
- Transformers 4.26.0
- Pytorch 2.1.0+cu118
- Datasets 2.9.0
- Tokenizers 0.13.3