vit-weight-decay-1e-2

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: 0.4994
• Accuracy: 0.8603
• Precision: 0.8618
• Recall: 0.8603
• F1: 0.8600

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
1.7124	1.0	321	0.8697	0.6924	0.6656	0.6924	0.6030
1.1476	2.0	642	0.7271	0.6990	0.7684	0.6990	0.7149
1.0734	3.0	963	0.6441	0.7687	0.7568	0.7687	0.7417
1.0271	4.0	1284	0.5855	0.7774	0.7883	0.7774	0.7814
0.9158	5.0	1605	0.7002	0.7635	0.7930	0.7635	0.7662
0.9167	6.0	1926	0.5867	0.7812	0.8065	0.7812	0.7900
0.786	7.0	2247	0.6517	0.7340	0.8047	0.7340	0.7515
0.7406	8.0	2568	0.6647	0.7067	0.8134	0.7067	0.7330
0.682	9.0	2889	0.5106	0.8228	0.8231	0.8228	0.8207
0.6427	10.0	3210	0.5032	0.8166	0.8354	0.8166	0.8222
0.5663	11.0	3531	0.5358	0.8152	0.8326	0.8152	0.8216
0.5395	12.0	3852	0.5488	0.8249	0.8392	0.8249	0.8299
0.4468	13.0	4173	0.5790	0.8232	0.8397	0.8232	0.8260
0.4247	14.0	4494	0.5438	0.8415	0.8570	0.8415	0.8449
0.3495	15.0	4815	0.5135	0.8454	0.8519	0.8454	0.8467
0.3039	16.0	5136	0.5631	0.8408	0.8520	0.8408	0.8448
0.2602	17.0	5457	0.4994	0.8603	0.8618	0.8603	0.8600
0.2616	18.0	5778	0.5406	0.8564	0.8622	0.8564	0.8585
0.1876	19.0	6099	0.5612	0.8481	0.8629	0.8481	0.8525
0.2052	20.0	6420	0.6803	0.8429	0.8502	0.8429	0.8428
0.1533	21.0	6741	0.5464	0.8734	0.8698	0.8734	0.8709
0.1175	22.0	7062	0.5573	0.8686	0.8667	0.8686	0.8673
0.1218	23.0	7383	0.6043	0.8703	0.8681	0.8703	0.8669
0.114	24.0	7704	0.5945	0.8710	0.8706	0.8710	0.8693
0.104	25.0	8025	0.5850	0.8766	0.8753	0.8766	0.8752
0.0752	26.0	8346	0.5868	0.8783	0.8747	0.8783	0.8757
0.1309	27.0	8667	0.5839	0.8786	0.8753	0.8786	0.8761

Framework versions

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