<!-- 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-300m-gl-jupyter9
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0970
- Wer: 0.0624
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: 60
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 3.6977 | 3.36 | 400 | 0.4273 | 0.4574 |
| 0.2282 | 6.72 | 800 | 0.1492 | 0.1723 |
| 0.0884 | 10.08 | 1200 | 0.1344 | 0.1336 |
| 0.0594 | 13.44 | 1600 | 0.1329 | 0.1238 |
| 0.0437 | 16.8 | 2000 | 0.1137 | 0.1153 |
| 0.0384 | 20.17 | 2400 | 0.1197 | 0.1033 |
| 0.0332 | 23.53 | 2800 | 0.1147 | 0.0980 |
| 0.0282 | 26.89 | 3200 | 0.1079 | 0.0917 |
| 0.0236 | 30.25 | 3600 | 0.1144 | 0.0922 |
| 0.0237 | 33.61 | 4000 | 0.1130 | 0.0880 |
| 0.019 | 36.97 | 4400 | 0.1035 | 0.0818 |
| 0.0164 | 40.33 | 4800 | 0.1045 | 0.0813 |
| 0.0146 | 43.69 | 5200 | 0.1037 | 0.0735 |
| 0.0111 | 47.06 | 5600 | 0.1085 | 0.0701 |
| 0.0093 | 50.42 | 6000 | 0.1039 | 0.0659 |
| 0.0084 | 53.78 | 6400 | 0.0970 | 0.0636 |
| 0.0073 | 57.14 | 6800 | 0.0970 | 0.0624 |
Framework versions
- Transformers 4.11.3
- Pytorch 1.9.1+cu111
- Datasets 1.18.3
- Tokenizers 0.10.3
