Proposal: Add Transformer Neural Network (From Scratch) in PyTorch

I would like to contribute a detailed educational notebook implementing the Transformer architecture (as introduced in Attention Is All You Need).
The notebook includes complete explanations, PyTorch implementation, and toy demo.