vit-weight-decay-1e-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: 0.5277
• Accuracy: 0.8263
• Precision: 0.8467
• Recall: 0.8263
• F1: 0.8324

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: 1733
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
1.7856	1.0	321	0.9535	0.6869	0.6412	0.6869	0.6229
1.1412	2.0	642	0.6928	0.7424	0.7440	0.7424	0.7311
1.0297	3.0	963	0.6863	0.7490	0.7362	0.7490	0.7057
0.9571	4.0	1284	0.5587	0.7694	0.7901	0.7694	0.7736
0.9346	5.0	1605	0.5654	0.7940	0.8058	0.7940	0.7919
0.9802	6.0	1926	0.6318	0.7746	0.7928	0.7746	0.7794
0.8352	7.0	2247	0.6611	0.7295	0.8145	0.7295	0.7498
0.7621	8.0	2568	0.5766	0.7666	0.8162	0.7666	0.7781
0.7352	9.0	2889	0.5369	0.7996	0.8269	0.7996	0.8079
0.6919	10.0	3210	0.5500	0.7753	0.8270	0.7753	0.7900
0.6105	11.0	3531	0.5562	0.8062	0.8310	0.8062	0.8129
0.5808	12.0	3852	0.6608	0.7708	0.8278	0.7708	0.7871
0.4534	13.0	4173	0.5684	0.8301	0.8483	0.8301	0.8291
0.4519	14.0	4494	0.5928	0.8121	0.8388	0.8121	0.8201
0.3998	15.0	4815	0.5277	0.8263	0.8467	0.8263	0.8324
0.3307	16.0	5136	0.5944	0.8266	0.8458	0.8266	0.8330
0.2899	17.0	5457	0.5387	0.8526	0.8546	0.8526	0.8524
0.2466	18.0	5778	0.5559	0.8495	0.8541	0.8495	0.8506
0.201	19.0	6099	0.6360	0.8336	0.8671	0.8336	0.8427
0.2163	20.0	6420	0.6009	0.8599	0.8575	0.8599	0.8581
0.1614	21.0	6741	0.5804	0.8689	0.8648	0.8689	0.8630
0.1106	22.0	7062	0.5798	0.8689	0.8661	0.8689	0.8670
0.1243	23.0	7383	0.6228	0.8703	0.8686	0.8703	0.8672
0.1251	24.0	7704	0.5987	0.8727	0.8695	0.8727	0.8698
0.1038	25.0	8025	0.5806	0.8769	0.8756	0.8769	0.8753

Framework versions

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