white_switch_click_detector

This model is a fine-tuned version of MIT/ast-finetuned-audioset-10-10-0.4593 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0181
• Recall Pressed: 0.9611
• Precision Pressed: 1.0
• F1 Pressed: 0.9802
• False Negative Rate: 0.0389

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: 1e-05
• train_batch_size: 82
• eval_batch_size: 82
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Recall Pressed	Precision Pressed	F1 Pressed	False Negative Rate
0.0168	1.0	218	0.0148	0.975	0.9590	0.9669	0.025
0.0109	2.0	436	0.0108	0.9444	1.0	0.9714	0.0556
0.0142	3.0	654	0.0167	0.9972	0.9253	0.9599	0.0028
0.0056	4.0	872	0.0120	0.95	1.0	0.9744	0.05
0.0038	5.0	1090	0.0121	0.9583	1.0	0.9787	0.0417
0.0002	6.0	1308	0.0181	0.9611	1.0	0.9802	0.0389
0.0001	7.0	1526	0.0194	0.9611	1.0	0.9802	0.0389
0.0000	8.0	1744	0.0202	0.9639	0.9886	0.9761	0.0361
0.0000	9.0	1962	0.0215	0.9611	0.9914	0.9760	0.0389
0.0000	10.0	2180	0.0219	0.9611	0.9914	0.9760	0.0389

Framework versions

• Transformers 5.8.1
• Pytorch 2.7.1+cu118
• Datasets 4.8.5
• Tokenizers 0.22.2