deit-tiny-patch16-224-finetuned-piid

This model is a fine-tuned version of facebook/deit-tiny-patch16-224 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5426
• Accuracy: 0.7626

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.2274	0.98	20	1.1185	0.4658
0.8485	2.0	41	0.8690	0.6119
0.6793	2.98	61	0.8749	0.6073
0.6028	4.0	82	0.6864	0.6804
0.5693	4.98	102	0.5618	0.7717
0.5092	6.0	123	0.5958	0.7260
0.3788	6.98	143	0.6444	0.7352
0.4106	8.0	164	0.5277	0.7443
0.3716	8.98	184	0.6081	0.7352
0.3466	10.0	205	0.4976	0.7580
0.3587	10.98	225	0.5429	0.7443
0.2661	12.0	246	0.4933	0.7763
0.2628	12.98	266	0.5078	0.7671
0.2473	14.0	287	0.5264	0.7945
0.2633	14.98	307	0.5262	0.7671
0.2017	16.0	328	0.5509	0.7763
0.1861	16.98	348	0.5513	0.7443
0.2031	18.0	369	0.5516	0.7580
0.1604	18.98	389	0.5430	0.7671
0.2346	19.51	400	0.5426	0.7626

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1