OLM is designed for courses, labs, and reading groups that want students to inspect the actual pieces of a language model instead of only using pretrained checkpoints.
Course
Learn From Scratch
A walkable sequence from setup to tokens, embeddings, attention, transformer blocks, and training.
LabYour First Language Model
A runnable train-and-generate exercise with save/load and sampling.
ResearchThe Block System
A bridge from classroom transformer concepts to architecture ablations.
Course Outline
- Set up a language-model lab
- Understand tokens and next-token prediction
- Build intuition for embeddings
- Learn attention step by step
- Assemble a transformer block
- Train and inspect a small language model
Start with the Learn From Scratch course, then use the first model tutorial and Block System guide as reading material.