vit-base-patch16-224-ve-U13b-80RX

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

• Loss: 0.6099
• Accuracy: 0.8478

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: 5.5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 6
• total_train_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.3857	0.99	17	1.3703	0.5652
1.3134	1.98	34	1.2235	0.4565
1.0384	2.97	51	1.0173	0.5435
0.908	3.96	68	0.8346	0.7826
0.75	4.95	85	0.7343	0.7826
0.5131	6.0	103	0.6099	0.8478
0.395	6.99	120	0.5932	0.7826
0.355	7.98	137	0.7209	0.7391
0.2658	8.97	154	0.5652	0.8043
0.248	9.96	171	0.7103	0.7826
0.2086	10.95	188	0.6788	0.7609
0.1532	12.0	206	0.5725	0.7826
0.147	12.99	223	0.6130	0.8043
0.1145	13.98	240	0.6563	0.8043
0.1053	14.97	257	0.5993	0.8043
0.0971	15.96	274	0.8840	0.7391
0.0947	16.95	291	0.6256	0.8043
0.1055	18.0	309	0.8406	0.7609
0.0974	18.99	326	0.6355	0.8478
0.1215	19.98	343	0.6651	0.8043
0.108	20.97	360	0.8301	0.7826
0.0784	21.96	377	0.8837	0.7609
0.0919	22.95	394	0.6985	0.8043
0.064	24.0	412	0.6426	0.8043
0.0669	24.99	429	0.8102	0.7826
0.0878	25.98	446	0.7863	0.7391
0.0875	26.97	463	0.8777	0.7609
0.0441	27.96	480	0.7324	0.8043
0.088	28.95	497	0.8099	0.7826
0.0739	30.0	515	0.7776	0.8043
0.0598	30.99	532	0.8188	0.7826
0.0443	31.98	549	0.8549	0.8043
0.0376	32.97	566	0.8049	0.7826
0.0375	33.96	583	0.8037	0.8043
0.0346	34.95	600	0.8255	0.8261
0.0471	36.0	618	0.8239	0.8043
0.0669	36.99	635	0.8188	0.8043
0.0438	37.98	652	0.8443	0.8043
0.0549	38.97	669	0.8551	0.8043
0.0622	39.61	680	0.8551	0.8043

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0