metadata
license: apache-2.0
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: wav2vec2-large-xls-r-300m-telugu-asr
results: []
wav2vec2-large-xls-r-300m-telugu-asr
This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 3.8208
- Wer: 1.0395
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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 |
|---|---|---|---|---|
| 11.4834 | 1.37 | 200 | 4.8950 | 1.0 |
| 3.9043 | 2.74 | 400 | 3.8879 | 1.0 |
| 3.7152 | 4.11 | 600 | 3.6273 | 1.0 |
| 3.6229 | 5.48 | 800 | 3.8105 | 1.0 |
| 3.5465 | 6.85 | 1000 | 3.6970 | 1.0 |
| 3.5106 | 8.22 | 1200 | 3.6657 | 1.0 |
| 3.3731 | 9.59 | 1400 | 3.4669 | 0.9995 |
| 3.183 | 10.96 | 1600 | 3.1894 | 0.9983 |
| 2.9215 | 12.33 | 1800 | 2.9099 | 0.9993 |
| 2.5357 | 13.7 | 2000 | 2.8166 | 1.0407 |
| 2.1257 | 15.07 | 2200 | 2.6122 | 1.0205 |
| 1.7549 | 16.44 | 2400 | 2.5981 | 1.0457 |
| 1.4642 | 17.81 | 2600 | 2.5619 | 1.0015 |
| 1.1814 | 19.18 | 2800 | 2.8769 | 0.9836 |
| 0.9759 | 20.55 | 3000 | 2.8497 | 1.0078 |
| 0.8066 | 21.92 | 3200 | 3.1365 | 1.0655 |
| 0.6614 | 23.29 | 3400 | 3.1759 | 0.9964 |
| 0.5687 | 24.66 | 3600 | 3.3751 | 1.0103 |
| 0.4987 | 26.03 | 3800 | 3.5111 | 1.0180 |
| 0.4304 | 27.4 | 4000 | 3.6908 | 1.0200 |
| 0.3818 | 28.77 | 4200 | 3.8208 | 1.0395 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.10.0+cu113
- Datasets 1.18.3
- Tokenizers 0.13.2