swinv2-tiny-patch4-window8-256-Diabetic-Retinopathy

This model is a fine-tuned version of microsoft/swinv2-tiny-patch4-window8-256 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.5203
• Accuracy: 0.8091

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	5	1.6054	0.4909
1.6039	2.0	10	1.5774	0.4909
1.6039	3.0	15	1.4627	0.4909
1.4766	4.0	20	1.3211	0.4909
1.4766	5.0	25	1.2294	0.4909
1.2308	6.0	30	1.0657	0.4909
1.2308	7.0	35	0.9504	0.6545
1.017	8.0	40	0.8463	0.7364
1.017	9.0	45	0.7463	0.7455
0.8345	10.0	50	0.6948	0.7455
0.8345	11.0	55	0.6460	0.7545
0.7594	12.0	60	0.6403	0.7545
0.7594	13.0	65	0.6319	0.7545
0.7228	14.0	70	0.5999	0.7455
0.7228	15.0	75	0.5922	0.7545
0.6851	16.0	80	0.5955	0.7636
0.6851	17.0	85	0.5731	0.7545
0.6549	18.0	90	0.5603	0.7818
0.6549	19.0	95	0.5386	0.7818
0.643	20.0	100	0.5424	0.7727
0.643	21.0	105	0.5295	0.7909
0.5951	22.0	110	0.5203	0.8091
0.5951	23.0	115	0.5162	0.7909
0.5913	24.0	120	0.5095	0.7818
0.5913	25.0	125	0.5140	0.7909
0.5462	26.0	130	0.5167	0.7636
0.5462	27.0	135	0.4943	0.7909
0.5538	28.0	140	0.4844	0.7636
0.5538	29.0	145	0.4821	0.7727
0.5497	30.0	150	0.4952	0.7727
0.5497	31.0	155	0.4995	0.7818
0.4923	32.0	160	0.4910	0.7727
0.4923	33.0	165	0.5029	0.7818
0.5228	34.0	170	0.5083	0.7818
0.5228	35.0	175	0.4984	0.7909
0.4986	36.0	180	0.4914	0.7909
0.4986	37.0	185	0.4926	0.7909
0.5154	38.0	190	0.4915	0.8
0.5154	39.0	195	0.4886	0.8
0.5081	40.0	200	0.4875	0.8

Framework versions

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