wav2vec2-base-960h-finetuned
This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.1430
- Accuracy: 0.6516
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: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 10
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 2.5958 | 1.0 | 203 | 2.4754 | 0.2714 |
| 2.0809 | 2.0 | 406 | 1.9972 | 0.3930 |
| 1.8486 | 3.0 | 609 | 1.6918 | 0.4658 |
| 1.5857 | 4.0 | 812 | 1.5089 | 0.5186 |
| 1.4819 | 5.0 | 1015 | 1.4027 | 0.5508 |
| 1.3859 | 6.0 | 1218 | 1.3146 | 0.5867 |
| 1.3448 | 7.0 | 1421 | 1.2078 | 0.6281 |
| 1.2551 | 8.0 | 1624 | 1.1600 | 0.6447 |
| 1.1506 | 9.0 | 1827 | 1.1595 | 0.6512 |
| 1.2435 | 10.0 | 2030 | 1.1430 | 0.6516 |
Framework versions
- Transformers 4.20.1
- Pytorch 1.12.0+cu113
- Datasets 2.3.2
- Tokenizers 0.12.1