Vision Transformer (ViT) from Scratch

Experiment abstract

From-scratch implementation of the Vision Transformer (ViT) in PyTorch. Includes patch embedding, multi-head self-attention, positional encoding, and classification head — trained on CIFAR-10.