vit-dropout-0.5

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: 0.4899
• Accuracy: 0.8592
• Precision: 0.8593
• Recall: 0.8592
• F1: 0.8581

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
1.6642	1.0	321	0.8597	0.7178	0.6898	0.7178	0.6753
1.1191	2.0	642	0.6286	0.7656	0.7678	0.7656	0.7653
1.0741	3.0	963	0.6771	0.7545	0.7795	0.7545	0.7353
1.0023	4.0	1284	0.6269	0.7708	0.7839	0.7708	0.7727
0.92	5.0	1605	0.6010	0.7816	0.7913	0.7816	0.7811
0.9155	6.0	1926	0.5399	0.7944	0.8035	0.7944	0.7979
0.7971	7.0	2247	0.7016	0.7347	0.7903	0.7347	0.7509
0.7171	8.0	2568	0.6352	0.7396	0.8175	0.7396	0.7601
0.676	9.0	2889	0.5073	0.8072	0.8241	0.8072	0.8120
0.6312	10.0	3210	0.5552	0.7850	0.8374	0.7850	0.7978
0.5777	11.0	3531	0.5175	0.8173	0.8428	0.8173	0.8247
0.4686	12.0	3852	0.5712	0.8062	0.8413	0.8062	0.8172
0.4134	13.0	4173	0.6331	0.8214	0.8384	0.8214	0.8180
0.3904	14.0	4494	0.6161	0.8308	0.8546	0.8308	0.8357
0.3388	15.0	4815	0.5141	0.8384	0.8519	0.8384	0.8425
0.301	16.0	5136	0.5525	0.8360	0.8521	0.8360	0.8416
0.2516	17.0	5457	0.4899	0.8592	0.8593	0.8592	0.8581
0.2433	18.0	5778	0.5183	0.8509	0.8565	0.8509	0.8528
0.1769	19.0	6099	0.6016	0.8426	0.8622	0.8426	0.8487
0.179	20.0	6420	0.6330	0.8551	0.8543	0.8551	0.8534
0.1401	21.0	6741	0.5884	0.8669	0.8612	0.8669	0.8630
0.1085	22.0	7062	0.6131	0.8627	0.8591	0.8627	0.8603
0.1188	23.0	7383	0.6344	0.8648	0.8581	0.8648	0.8596
0.106	24.0	7704	0.6152	0.8731	0.8687	0.8731	0.8696
0.1029	25.0	8025	0.6209	0.8717	0.8676	0.8717	0.8692
0.0761	26.0	8346	0.6190	0.8727	0.8679	0.8727	0.8694
0.1207	27.0	8667	0.6154	0.8721	0.8677	0.8721	0.8693

Framework versions

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