metadata
license: apache-2.0
tags:
- generated_from_trainer
model-index:
- name: w2v2
results: []
w2v2
This model is a fine-tuned version of facebook/wav2vec2-large-960h-lv60-self on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.8860
- Wer: 0.2817
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: 0.0001
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 5.5664 | 3.07 | 500 | 3.0411 | 0.9997 |
| 2.5607 | 6.13 | 1000 | 1.0770 | 0.3660 |
| 0.9959 | 9.2 | 1500 | 0.8815 | 0.3017 |
| 0.8129 | 12.27 | 2000 | 0.8676 | 0.2915 |
| 0.7334 | 15.34 | 2500 | 0.8381 | 0.2931 |
| 0.669 | 18.4 | 3000 | 0.8802 | 0.2864 |
| 0.6312 | 21.47 | 3500 | 0.8679 | 0.2864 |
| 0.6094 | 24.54 | 4000 | 0.8811 | 0.2802 |
| 0.5987 | 27.61 | 4500 | 0.8860 | 0.2817 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.12.1+cu113
- Datasets 1.18.3
- Tokenizers 0.12.1