vit-base-finetuned-cephalometric

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

• Loss: 0.7340
• Accuracy: 0.6528

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: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	16	0.9458	0.5486
0.9879	2.0	32	0.6947	0.6597
0.4628	3.0	48	0.6375	0.6597
0.135	4.0	64	0.7060	0.6944
0.0339	5.0	80	0.7301	0.6597
0.0339	6.0	96	0.9236	0.6875
0.0059	7.0	112	0.9261	0.6806
0.0024	8.0	128	0.9961	0.6875
0.0012	9.0	144	1.0060	0.6736
0.0008	10.0	160	1.0329	0.6875
0.0008	11.0	176	1.0575	0.6944
0.0006	12.0	192	1.0768	0.6944
0.0006	13.0	208	1.1002	0.6944
0.0005	14.0	224	1.1220	0.6875
0.0004	15.0	240	1.1367	0.6875
0.0004	16.0	256	1.1538	0.6875
0.0004	17.0	272	1.1707	0.6875
0.0003	18.0	288	1.1855	0.6875
0.0003	19.0	304	1.2007	0.6875
0.0003	20.0	320	1.2066	0.6806
0.0003	21.0	336	1.2211	0.6806
0.0003	22.0	352	1.2291	0.6875
0.0002	23.0	368	1.2385	0.6875
0.0002	24.0	384	1.2508	0.6875

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0