"Full-Stack" Transformer

Advanced Deep Learning Architectures in Excel

I would like to share with you my "full-stack" Excel implementation of the famous Transformer architecture, brining together all the following components:

• Input Embeddings

• Output Embeddings

• Decoder

• Encoder

• Positional Encoding

• Self-Attention

• Cross-Attention

• Multi-head Attention

• Casual Masking

• Scaled Dot Product

• Skip Connection

• LayerNorm

• ReLU Activation

• Feed Forward

• Softmax

• Output Probabilities

Demo Video

Download

Transformer Full Stack

112KB ∙ XLSX file

Download

Download