wav2vec2-base-ft-keyword-spotting
This model is a fine-tuned version of facebook/wav2vec2-base on the superb dataset. It achieves the following results on the evaluation set:
- Loss: 0.0954
- Accuracy: 0.9826
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: 48
- eval_batch_size: 32
- seed: 0
- gradient_accumulation_steps: 4
- total_train_batch_size: 192
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 8.0
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 1.3624 | 1.0 | 267 | 1.1959 | 0.6546 |
| 0.3854 | 2.0 | 534 | 0.2675 | 0.9734 |
| 0.2473 | 3.0 | 801 | 0.1461 | 0.9768 |
| 0.1997 | 4.0 | 1068 | 0.1088 | 0.9804 |
| 0.1723 | 5.0 | 1335 | 0.0954 | 0.9826 |
| 0.1442 | 6.0 | 1602 | 0.0927 | 0.9813 |
| 0.1397 | 7.0 | 1869 | 0.0892 | 0.9812 |
| 0.1368 | 7.9728 | 2128 | 0.0896 | 0.9812 |
Framework versions
- Transformers 4.49.0
- Pytorch 2.6.0+cu118
- Datasets 3.3.1
- Tokenizers 0.21.0