<!-- 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. -->
wav2vec2-large-xls-r-1B-common_voice7-lv-ft
This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.1582
- Wer: 0.1137
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: 3e-05
- train_batch_size: 24
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 48
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 900
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.6292 | 5.26 | 500 | 1.5562 | 0.9263 |
| 0.1303 | 10.53 | 1000 | 0.8107 | 0.7666 |
| 0.0974 | 15.79 | 1500 | 0.5290 | 0.4979 |
| 0.0724 | 21.05 | 2000 | 0.2941 | 0.2247 |
| 0.0591 | 26.32 | 2500 | 0.2838 | 0.2125 |
| 0.0494 | 31.58 | 3000 | 0.2589 | 0.2102 |
| 0.0417 | 36.84 | 3500 | 0.1987 | 0.1760 |
| 0.0375 | 42.11 | 4000 | 0.1934 | 0.1690 |
| 0.031 | 47.37 | 4500 | 0.1630 | 0.1460 |
| 0.027 | 52.63 | 5000 | 0.1957 | 0.1447 |
| 0.0256 | 57.89 | 5500 | 0.1747 | 0.1368 |
| 0.0206 | 63.16 | 6000 | 0.1602 | 0.1299 |
| 0.0178 | 68.42 | 6500 | 0.1809 | 0.1273 |
| 0.0154 | 73.68 | 7000 | 0.1686 | 0.1216 |
| 0.0137 | 78.95 | 7500 | 0.1585 | 0.1241 |
| 0.0128 | 84.21 | 8000 | 0.1783 | 0.1278 |
| 0.011 | 89.47 | 8500 | 0.1653 | 0.1228 |
| 0.0096 | 94.74 | 9000 | 0.1620 | 0.1161 |
| 0.0091 | 100.0 | 9500 | 0.1582 | 0.1137 |
Framework versions
- Transformers 4.16.0.dev0
- Pytorch 1.10.1+cu102
- Datasets 1.17.1.dev0
- Tokenizers 0.10.3
