ham10000-vit

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

• Loss: 1.6207
• Accuracy: 0.6075
• F1 Macro: 0.6049
• F1 Weighted: 0.6450
• Roc Auc Macro: 0.9257

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: 64
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 25
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro	F1 Weighted	Roc Auc Macro
17.3138	1.0	128	3.1043	0.0188	0.0145	0.0014	0.5433
9.5042	2.0	256	2.7314	0.0494	0.0789	0.0212	0.7525
7.2502	3.0	384	2.2587	0.1051	0.1432	0.0683	0.8134
5.5487	4.0	512	1.9783	0.3445	0.2972	0.3912	0.8534
5.1006	5.0	640	1.9278	0.4530	0.3797	0.4987	0.8847
4.4520	6.0	768	1.8062	0.5296	0.4697	0.5687	0.9017
4.2735	7.0	896	1.7376	0.5595	0.5112	0.6009	0.9116
4.1280	8.0	1024	1.7475	0.5296	0.4687	0.5686	0.9118
3.8797	9.0	1152	1.7181	0.5456	0.4809	0.5848	0.9081
3.7173	10.0	1280	1.7144	0.5505	0.5015	0.5902	0.9178
3.7269	11.0	1408	1.7463	0.5338	0.4714	0.5723	0.9091
3.5393	12.0	1536	1.6674	0.5783	0.5201	0.6181	0.9180
3.5778	13.0	1664	1.6647	0.5852	0.5491	0.6216	0.9246
3.4255	14.0	1792	1.6460	0.5859	0.5542	0.6255	0.9217
3.3948	15.0	1920	1.6848	0.5811	0.5361	0.6204	0.9121
3.4073	16.0	2048	1.6392	0.5873	0.5483	0.6293	0.9247
3.2856	17.0	2176	1.6579	0.5866	0.5481	0.6238	0.9195
3.2094	18.0	2304	1.6619	0.5776	0.5379	0.6147	0.9179
3.2588	19.0	2432	1.6293	0.6006	0.5875	0.6375	0.9234
3.2176	20.0	2560	1.6266	0.5985	0.5853	0.6356	0.9267
3.2264	21.0	2688	1.6158	0.6082	0.5945	0.6447	0.9255
3.2054	22.0	2816	1.6290	0.6019	0.5964	0.6393	0.9245
3.1724	23.0	2944	1.6197	0.6061	0.5920	0.6428	0.9250
3.1820	24.0	3072	1.6188	0.6061	0.5983	0.6435	0.9259
3.1322	25.0	3200	1.6207	0.6075	0.6049	0.6450	0.9257

Framework versions

• PEFT 0.19.1
• Transformers 5.8.1
• Pytorch 2.10.0+cu128
• Datasets 4.8.5
• Tokenizers 0.22.2