<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
xslr-commonvoice
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the COMMON_VOICE - TR dataset. It achieves the following results on the evaluation set:
- Loss: 0.3835
- Wer: 0.3450
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: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 15.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| No log | 0.92 | 100 | 3.5761 | 1.0 |
| No log | 1.83 | 200 | 3.0512 | 0.9999 |
| No log | 2.75 | 300 | 1.0185 | 0.8188 |
| No log | 3.67 | 400 | 0.5936 | 0.6411 |
| 3.2139 | 4.59 | 500 | 0.4986 | 0.5267 |
| 3.2139 | 5.5 | 600 | 0.4327 | 0.4732 |
| 3.2139 | 6.42 | 700 | 0.4227 | 0.4462 |
| 3.2139 | 7.34 | 800 | 0.4213 | 0.4291 |
| 3.2139 | 8.26 | 900 | 0.4016 | 0.4033 |
| 0.22 | 9.17 | 1000 | 0.3987 | 0.3825 |
| 0.22 | 10.09 | 1100 | 0.4065 | 0.3867 |
| 0.22 | 11.01 | 1200 | 0.3929 | 0.3842 |
| 0.22 | 11.93 | 1300 | 0.3775 | 0.3687 |
| 0.22 | 12.84 | 1400 | 0.3891 | 0.3536 |
| 0.1005 | 13.76 | 1500 | 0.3850 | 0.3492 |
| 0.1005 | 14.68 | 1600 | 0.3823 | 0.3441 |
Framework versions
- Transformers 4.13.0.dev0
- Pytorch 1.9.1+cu102
- Datasets 1.14.0
- Tokenizers 0.10.3
