From Attention to Generative Language Models - Building Transformers from Scratch
Neural Breakdown with AVB via YouTube
Overview
Learn to build Causal Generative Language models from scratch in PyTorch through a 47-minute comprehensive tutorial that breaks down complex concepts with clear visualizations and detailed code explanations. Master fundamental concepts including semantic similarity, matrix multiplication, attention scores, and contextual embeddings before diving into practical implementations with PyTorch. Progress through self-attention mechanisms, causal masked attention, and transformer decoder blocks, culminating in next word prediction techniques. Explore each mathematical concept through line-by-line code explanations designed to balance technical depth with accessible explanations. Gain hands-on experience with transformer architecture and attention models while building a strong foundation for advanced topics like Multi-Headed Attention, Multi-Query Attention, and Grouped Query Attention.
Syllabus
- Intro
- Semantic Similarity
- Matrix Multiplication
- Attention Scores
- Contextual Embeddings
- Attention with Pytorch
- Self Attention
- Causal Masked Attention
- Transformer Decoder Blocks
- Next Word Prediction
Taught by
Neural Breakdown with AVB