Beginner
9.5h · 40 lessons
LLM Foundations: A Deep Dive
Transformers, tokenization, attention, and evaluation.
4.9rating22,030 enrolledLifetime access
Start free preview
About this course
A rigorous foundation for the working AI engineer — internals, scaling laws, and modern evaluation.
What you'll learn
- Explain how modern transformer LLMs actually work end to end
- Read and reproduce results from recent LLM papers
- Choose the right model family and size for a use case
- Fine-tune or adapt an open model on your own data
Requirements
- Solid Python and NumPy
- Basic linear algebra and probability
- A GPU (colab / cloud is fine) for the labs
Curriculum
6 modules · 24 lessons
Module 1From tokens to transformers4 lessons
Module 1
From tokens to transformers
- 01.01Tokenization deep dive
- 01.02Embeddings and positional encoding
- 01.03Attention from scratch
- 01.04The transformer block
Module quiz · 3 questions
Module 2Pretraining4 lessons
Module 2
Pretraining
- 02.01Objectives and data mixtures
- 02.02Scaling laws
- 02.03Compute and hardware
- 02.04Reading the training loss
Module quiz · 3 questions
Module 3Alignment4 lessons
Module 3
Alignment
Module quiz · 3 questions
Module 4Inference4 lessons
Module 4
Inference
Module quiz · 3 questions
Module 5Adapting models4 lessons
Module 5
Adapting models
Module quiz · 3 questions
Module 6The frontier4 lessons
Module 6
The frontier
Module quiz · 3 questions