<!-- 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. -->
seq-xls-r-fleurs_zu-run3-asr_xh-run2
This model is a fine-tuned version of lucas-meyer/xls-r-fleurs_zu-run3 on asr_xh. It achieves the following results on the evaluation set:
- Val WER: 0.5115
- Test WER: 0.5132
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 3
- total_train_batch_size: 12
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 8.7872 | 0.48 | 100 | 3.3525 | 1.0 |
| 3.1413 | 0.96 | 200 | 3.0025 | 1.0 |
| 1.7204 | 1.44 | 300 | 0.6932 | 0.7477 |
| 0.6719 | 1.91 | 400 | 0.5336 | 0.6871 |
| 0.5452 | 2.39 | 500 | 0.4911 | 0.6239 |
| 0.4981 | 2.87 | 600 | 0.4559 | 0.6339 |
| 0.4112 | 3.35 | 700 | 0.4295 | 0.5604 |
| 0.3807 | 3.83 | 800 | 0.3999 | 0.5390 |
| 0.3222 | 4.31 | 900 | 0.3903 | 0.5303 |
| 0.3041 | 4.78 | 1000 | 0.3714 | 0.5125 |
| 0.258 | 5.26 | 1100 | 0.4244 | 0.5368 |
| 0.2356 | 5.74 | 1200 | 0.4421 | 0.5494 |
| 0.2136 | 6.22 | 1300 | 0.4220 | 0.5420 |
Framework versions
- Transformers 4.28.0
- Pytorch 2.0.1+cu117
- Datasets 2.14.4
- Tokenizers 0.13.3
