videomae-large-finetuned-deepfake-subset
This model is a fine-tuned version of MCG-NJU/videomae-large on the Deepfake Detection Challenge dataset. It achieves the following results on the evaluation set:
- Loss: 0.2588
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- 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_ratio: 0.1
- training_steps: 4470
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss |
|---|---|---|---|
| 0.6169 | 0.1 | 447 | 0.6023 |
| 0.6086 | 1.1 | 894 | 0.5055 |
| 0.376 | 2.1 | 1341 | 0.4250 |
| 0.3863 | 3.1 | 1788 | 0.6712 |
| 0.249 | 4.1 | 2235 | 0.3951 |
| 0.3233 | 5.1 | 2682 | 0.4969 |
| 0.1995 | 6.1 | 3129 | 0.3744 |
| 0.0874 | 7.1 | 3576 | 0.4104 |
| 0.2518 | 8.1 | 4023 | 0.2647 |
| 0.0118 | 9.1 | 4470 | 0.3337 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.19.1