Boya1_Adamax_1-e4_20Epoch_Deit-tiny-patch16_fold5

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: 3.4928
• Accuracy: 0.5612

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: 16
• seed: 42
• 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
1.3998	1.0	924	1.5149	0.4893
1.3931	2.0	1848	1.3087	0.5424
1.0339	3.0	2772	1.2795	0.5598
0.6895	4.0	3696	1.2909	0.5772
0.5865	5.0	4620	1.4017	0.5644
0.6353	6.0	5544	1.5191	0.5606
0.4014	7.0	6468	1.6877	0.5644
0.2889	8.0	7392	1.9030	0.5601
0.064	9.0	8316	2.0982	0.5633
0.0342	10.0	9240	2.4199	0.5574
0.071	11.0	10164	2.6790	0.5555
0.0036	12.0	11088	2.8221	0.5571
0.0018	13.0	12012	3.0343	0.5609
0.0019	14.0	12936	3.1489	0.5566
0.0004	15.0	13860	3.2320	0.5579
0.0248	16.0	14784	3.2970	0.5595
0.0003	17.0	15708	3.3641	0.5625
0.0002	18.0	16632	3.4302	0.5606
0.0002	19.0	17556	3.4763	0.5620
0.0001	20.0	18480	3.4928	0.5612

Framework versions

• Transformers 4.40.1
• Pytorch 2.1.0
• Datasets 2.19.0
• Tokenizers 0.19.1