TABLE 2.

Optimal hyperparameters of different models for APTOS dataset.

Hyperparameters	SwAV	EfficientNet-B5	Swin	ViT
Optimizer	AdamW	AdamW	AdamW	AdamW
Batch size	64	16	32	64
Learning rate	$1e-3$	$1e-3$	$1e-5$	$1e-5$
Weight decay	$5e-2$	$5e-2$	$5e-2$	$5e-2$
Exponential moving avg	0.99	0.99996	0.99	0.99
Epochs	33	35	35	50