seizure_vit_jlb_231027

This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the JLB-JLB/seizure_eeg_greyscale_224x224_6secWindow_adjusted dataset. It achieves the following results on the evaluation set:

• Loss: 0.4759
• Roc Auc: 0.7822

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Roc Auc
0.4787	0.17	1000	0.5094	0.7706
0.3695	0.34	2000	0.5111	0.7359
0.337	0.51	3000	0.4734	0.7829
0.3604	0.68	4000	0.5508	0.7457
0.3222	0.85	5000	0.5817	0.7687
0.2315	1.02	6000	0.6515	0.7679
0.2388	1.19	7000	0.5681	0.7543
0.2691	1.36	8000	0.5307	0.7691
0.268	1.53	9000	0.5643	0.7610
0.131	1.7	10000	0.7293	0.7451
0.2303	1.87	11000	0.6291	0.7704
0.1442	2.04	12000	0.6372	0.7871
0.1325	2.21	13000	0.8672	0.7319
0.1986	2.38	14000	0.7352	0.7532
0.1669	2.55	15000	0.8195	0.7562
0.1228	2.72	16000	1.0106	0.7239
0.1071	2.89	17000	0.8957	0.7463
0.1322	3.06	18000	1.0871	0.7408
0.1676	3.24	19000	0.9173	0.7683
0.1105	3.41	20000	1.0175	0.7700
0.1451	3.58	21000	0.9357	0.7404
0.082	3.75	22000	1.1246	0.7404
0.1457	3.92	23000	1.0082	0.7502
0.0336	4.09	24000	1.3685	0.7443
0.0742	4.26	25000	1.5080	0.7227
0.0353	4.43	26000	1.3573	0.7421
0.0557	4.6	27000	1.2484	0.7472
0.075	4.77	28000	1.2750	0.7462
0.0569	4.94	29000	1.3954	0.7355

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0
• Datasets 2.14.6
• Tokenizers 0.14.1