Help! - Drastic Overfitting and Atrocious Accuracy on ViT Model

I am using the ‘google/vit-base-patch16-224-in21k’ ViT model for image classification.

However, my training loss and validation loss looks like this:

And my test accuracy is a measly 63%.
Here is the full code: vit-cnn-mosquito-image-classification/ViT base.ipynb at main · thenerd31/vit-cnn-mosquito-image-classification · GitHub

How can I increase the accuracy (I want it above 90%) and reduce the overfitting? I didn’t expect such a low result from the ViT, and I need to complete it for a project soon.