seizure_vit_jlb_231112_fft_raw_combo

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the JLB-JLB/seizure_detection_224x224_raw_frequency dataset. It achieves the following results on the evaluation set:

• Loss: 0.4822
• Roc Auc: 0.7667

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: 2e-06
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Roc Auc
0.4777	0.17	500	0.5237	0.7455
0.4469	0.34	1000	0.5114	0.7542
0.4122	0.52	1500	0.5084	0.7567
0.3904	0.69	2000	0.5043	0.7611
0.3619	0.86	2500	0.5283	0.7609
0.3528	1.03	3000	0.5352	0.7517
0.3445	1.2	3500	0.5338	0.7572
0.3221	1.37	4000	0.5388	0.7509
0.3109	1.55	4500	0.5641	0.7458
0.3203	1.72	5000	0.5404	0.7574
0.294	1.89	5500	0.5421	0.7564
0.2964	2.06	6000	0.5582	0.7493
0.292	2.23	6500	0.5513	0.7561
0.2838	2.4	7000	0.5557	0.7598
0.2736	2.58	7500	0.5514	0.7606
0.2922	2.75	8000	0.5503	0.7538
0.2699	2.92	8500	0.5535	0.7578

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0
• Datasets 2.14.6
• Tokenizers 0.14.1