wav2vec2-base-960h-itw-deepfake
This model is a fine-tuned version of facebook/wav2vec2-base-960h on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.0917
- Accuracy: 0.9835
- FAR: 0.0068
- FRR: 0.0330
- EER: 0.0199
Model description
Quick Use
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
config = AutoConfig.from_pretrained("abhishtagatya/wav2vec2-base-960h-itw-deepfake")
feature_extractor = Wav2Vec2FeatureExtractor.from_pretrained("abhishtagatya/wav2vec2-base-960h-itw-deepfake")
model = Wav2Vec2ForSequenceClassification.from_pretrained("abhishtagatya/wav2vec2-base-960h-itw-deepfake", config=config).to(device)
# Your Logic Here
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: 1e-06
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy | FAR | FRR | EER |
|---|---|---|---|---|---|---|---|
| 0.6363 | 0.39 | 2500 | 0.4678 | 0.8652 | 0.0178 | 0.3326 | 0.1752 |
| 0.2896 | 0.79 | 5000 | 0.1145 | 0.9744 | 0.0170 | 0.0402 | 0.0286 |
| 0.1554 | 1.18 | 7500 | 0.1024 | 0.9797 | 0.0100 | 0.0377 | 0.0238 |
| 0.1327 | 1.57 | 10000 | 0.0945 | 0.9825 | 0.0070 | 0.0351 | 0.0211 |
| 0.13 | 1.97 | 12500 | 0.0917 | 0.9835 | 0.0068 | 0.0330 | 0.0199 |
Framework versions
- Transformers 4.38.0.dev0
- Pytorch 2.1.2+cu121
- Datasets 2.16.2.dev0
- Tokenizers 0.15.1