vit-augmentation

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: 0.4287
• Accuracy: 0.8592
• Precision: 0.8580
• Recall: 0.8592
• F1: 0.8574

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: 770
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.9124	1.0	321	0.6025	0.7805	0.7788	0.7805	0.7683
0.5876	2.0	642	0.5819	0.7864	0.7990	0.7864	0.7820
0.5415	3.0	963	0.6149	0.8041	0.7943	0.8041	0.7865
0.4815	4.0	1284	0.4654	0.8294	0.8259	0.8294	0.8115
0.4263	5.0	1605	0.5481	0.8259	0.8315	0.8259	0.8023
0.3515	6.0	1926	0.4287	0.8592	0.8580	0.8592	0.8574
0.3144	7.0	2247	0.5005	0.8363	0.8320	0.8363	0.8270
0.2736	8.0	2568	0.5306	0.8294	0.8448	0.8294	0.8302
0.2519	9.0	2889	0.4733	0.8578	0.8534	0.8578	0.8534
0.2227	10.0	3210	0.4905	0.8585	0.8520	0.8585	0.8512
0.1724	11.0	3531	0.5050	0.8655	0.8671	0.8655	0.8628
0.1596	12.0	3852	0.5263	0.8686	0.8657	0.8686	0.8631
0.1397	13.0	4173	0.7043	0.8533	0.8703	0.8533	0.8488
0.1298	14.0	4494	0.6275	0.8679	0.8734	0.8679	0.8632
0.1029	15.0	4815	0.5564	0.8807	0.8776	0.8807	0.8772
0.0893	16.0	5136	0.5668	0.8804	0.8823	0.8804	0.8789

Framework versions

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