prove_melanomaprova_melanoma

This model is a fine-tuned version of UnipaPolitoUnimore/vit-large-patch32-384-melanoma on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5191
• Accuracy: 0.8467

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8964	0.99	31	1.0906	0.52
0.6588	1.98	62	1.0817	0.52
0.6774	2.98	93	0.9474	0.52
0.7785	4.0	125	0.8185	0.6267
0.6732	4.99	156	0.7531	0.7267
0.5438	5.98	187	0.6972	0.7333
0.5497	6.98	218	0.6714	0.7533
0.4161	8.0	250	0.6440	0.7667
0.4968	8.99	281	0.6438	0.78
0.5861	9.98	312	0.6266	0.7933
0.5182	10.98	343	0.6158	0.7867
0.6797	12.0	375	0.6237	0.8133
0.622	12.99	406	0.5858	0.8333
0.6419	13.98	437	0.5735	0.8267
0.3727	14.98	468	0.5641	0.8133
0.3822	16.0	500	0.5520	0.8267
0.4766	16.99	531	0.5642	0.8267
0.4791	17.98	562	0.5309	0.8267
0.3918	18.98	593	0.5749	0.8267
0.3847	20.0	625	0.5317	0.84
0.3722	20.99	656	0.5719	0.8267
0.5402	21.98	687	0.5316	0.84
0.4358	22.98	718	0.5292	0.8333
0.2957	24.0	750	0.5172	0.8467
0.4801	24.99	781	0.5376	0.84
0.3656	25.98	812	0.5118	0.8333
0.3956	26.98	843	0.5081	0.8533
0.3343	28.0	875	0.5198	0.8533
0.3839	28.99	906	0.5269	0.8467
0.4286	29.98	937	0.5163	0.8467
0.2736	30.98	968	0.5359	0.8333
0.3465	32.0	1000	0.5277	0.84
0.4244	32.99	1031	0.5385	0.8333
0.308	33.98	1062	0.5141	0.8533
0.3494	34.98	1093	0.5129	0.8533
0.3851	36.0	1125	0.5199	0.84
0.3949	36.99	1156	0.5250	0.84
0.3235	37.98	1187	0.5142	0.8533
0.3076	38.98	1218	0.5166	0.8533
0.3679	39.68	1240	0.5191	0.8467

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3