vit-dropout-0.4

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: 0.4924
• Accuracy: 0.8291
• Precision: 0.8250
• Recall: 0.8291
• F1: 0.8244

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
1.6459	1.0	321	0.8160	0.7292	0.7119	0.7292	0.6910
1.0644	2.0	642	0.6323	0.7514	0.7687	0.7514	0.7575
1.0213	3.0	963	0.6378	0.7635	0.7718	0.7635	0.7422
0.9727	4.0	1284	0.6045	0.7438	0.7972	0.7438	0.7594
0.9062	5.0	1605	0.5953	0.7902	0.7965	0.7902	0.7802
0.8719	6.0	1926	0.6095	0.7743	0.8084	0.7743	0.7839
0.7537	7.0	2247	0.5970	0.7639	0.8125	0.7639	0.7778
0.677	8.0	2568	0.7108	0.7074	0.8148	0.7074	0.7301
0.6638	9.0	2889	0.4924	0.8291	0.8250	0.8291	0.8244
0.5787	10.0	3210	0.5415	0.8162	0.8406	0.8162	0.8222
0.5373	11.0	3531	0.5298	0.8103	0.8409	0.8103	0.8189
0.4923	12.0	3852	0.5428	0.8117	0.8444	0.8117	0.8213
0.3798	13.0	4173	0.4968	0.8499	0.8470	0.8499	0.8467
0.3912	14.0	4494	0.5339	0.8443	0.8531	0.8443	0.8460
0.3002	15.0	4815	0.5219	0.8450	0.8548	0.8450	0.8481
0.2744	16.0	5136	0.6369	0.8204	0.8482	0.8204	0.8280
0.2251	17.0	5457	0.5156	0.8571	0.8561	0.8571	0.8556
0.2187	18.0	5778	0.5825	0.8457	0.8550	0.8457	0.8491
0.1767	19.0	6099	0.5693	0.8526	0.8605	0.8526	0.8551

Framework versions

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