vit-base-oxford-iiit-pets

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: 0.6194
• Accuracy: 0.7639

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: 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
1.0901	1.0	32	0.9278	0.4931
0.5383	2.0	64	0.6985	0.6319
0.2707	3.0	96	0.6691	0.7222
0.0366	4.0	128	0.9557	0.6806
0.0066	5.0	160	0.8927	0.7083
0.0075	6.0	192	1.2046	0.7014
0.0013	7.0	224	1.2583	0.7083
0.0006	8.0	256	1.3180	0.6944
0.0004	9.0	288	1.3468	0.7014
0.0002	10.0	320	1.3582	0.6875
0.0002	11.0	352	1.3868	0.6875
0.0002	12.0	384	1.4094	0.6806
0.0002	13.0	416	1.4392	0.6806
0.0002	14.0	448	1.4536	0.6875
0.0001	15.0	480	1.4695	0.6875
0.0001	16.0	512	1.4850	0.6875
0.0001	17.0	544	1.5004	0.6875
0.0001	18.0	576	1.5110	0.6875
0.0001	19.0	608	1.5219	0.6875
0.0001	20.0	640	1.5340	0.6875
0.0001	21.0	672	1.5461	0.6875
0.0001	22.0	704	1.5541	0.6875
0.0001	23.0	736	1.5624	0.6875

Framework versions

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