vit-dropout-0.3

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

• Loss: 0.4841
• Accuracy: 0.8346
• Precision: 0.8364
• Recall: 0.8346
• F1: 0.8335

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: cosine
• lr_scheduler_warmup_steps: 1219
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.6246	1.0	321	0.7425	0.7583	0.7625	0.7583	0.7517
1.0305	2.0	642	0.7197	0.7195	0.7869	0.7195	0.7389
0.9668	3.0	963	0.6473	0.7587	0.7881	0.7587	0.7537
0.9294	4.0	1284	0.5399	0.7850	0.8140	0.7850	0.7934
0.8498	5.0	1605	0.6023	0.7916	0.8132	0.7916	0.7859
0.8204	6.0	1926	0.5088	0.8083	0.8120	0.8083	0.8094
0.6877	7.0	2247	0.5728	0.7701	0.8218	0.7701	0.7846
0.6445	8.0	2568	0.6019	0.7566	0.8286	0.7566	0.7750
0.6017	9.0	2889	0.4841	0.8346	0.8364	0.8346	0.8335
0.5321	10.0	3210	0.4849	0.8311	0.8483	0.8311	0.8353
0.4937	11.0	3531	0.5704	0.8031	0.8494	0.8031	0.8152
0.4328	12.0	3852	0.5285	0.8280	0.8521	0.8280	0.8361
0.3348	13.0	4173	0.5173	0.8592	0.8607	0.8592	0.8562
0.3532	14.0	4494	0.5740	0.8450	0.8658	0.8450	0.8512
0.296	15.0	4815	0.5671	0.8374	0.8510	0.8374	0.8402
0.2892	16.0	5136	0.5201	0.8551	0.8591	0.8551	0.8553
0.185	17.0	5457	0.5161	0.8637	0.8629	0.8637	0.8619
0.2052	18.0	5778	0.5338	0.8530	0.8632	0.8530	0.8565
0.1444	19.0	6099	0.5677	0.8478	0.8655	0.8478	0.8528

Framework versions

• Transformers 4.40.0.dev0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2