adam_VitB-p32-384-1e-4-batch_32_epoch_4_classes_24

This model is a fine-tuned version of google/vit-base-patch32-384 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.0868
• Accuracy: 0.9856

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.0118	0.14	100	0.1744	0.9555
0.0816	0.28	200	0.1820	0.9526
0.0035	0.42	300	0.0829	0.9741
0.0704	0.56	400	0.1208	0.9713
0.0591	0.7	500	0.2084	0.9540
0.0213	0.84	600	0.1119	0.9641
0.0024	0.97	700	0.1958	0.9555
0.005	1.11	800	0.1730	0.9583
0.0007	1.25	900	0.1245	0.9684
0.0052	1.39	1000	0.1296	0.9698
0.0001	1.53	1100	0.1377	0.9713
0.0114	1.67	1200	0.1394	0.9727
0.0016	1.81	1300	0.1606	0.9698
0.0137	1.95	1400	0.0984	0.9799
0.0001	2.09	1500	0.0925	0.9856
0.0005	2.23	1600	0.0882	0.9856
0.0001	2.37	1700	0.0900	0.9828
0.0001	2.51	1800	0.0876	0.9813
0.0001	2.65	1900	0.0943	0.9813
0.0001	2.79	2000	0.0899	0.9842
0.0001	2.92	2100	0.0867	0.9842
0.0006	3.06	2200	0.0873	0.9856
0.0001	3.2	2300	0.0875	0.9842
0.0001	3.34	2400	0.0878	0.9842
0.0001	3.48	2500	0.0879	0.9842
0.0001	3.62	2600	0.0877	0.9856
0.0001	3.76	2700	0.0876	0.9856
0.0	3.9	2800	0.0868	0.9856

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2