vit-dropout-v9

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

• Loss: 0.5147
• Accuracy: 0.8677

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

• dropout: 0.3

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: 500
• num_epochs: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5952	1.56	500	0.8221	0.7228
0.4505	3.12	1000	0.5679	0.7934
0.4187	4.67	1500	0.4951	0.8221
0.4022	6.23	2000	0.5013	0.8252
0.3485	7.79	2500	0.4532	0.8446
0.2397	9.35	3000	0.4914	0.8558
0.3017	10.9	3500	0.4973	0.8514
0.2086	12.46	4000	0.4987	0.8689
0.1265	14.02	4500	0.5132	0.8652
0.0885	15.58	5000	0.5147	0.8677

Framework versions

• Transformers 4.38.1
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2