Phase 10
LLMs from Scratch
Phase 10: LLMs from Scratch. 24 hands-on lessons building AI from first principles in the browser. Free reading; graded exercises and certificate with lifetime access.
- Tokenizers: BPE, WordPiece, SentencePiece (graded)
- Building a Tokenizer from Scratch (graded)
- Data Pipelines for Pre-Training (graded)
- Pre-Training a Mini GPT (124M Parameters) (graded)
- Scaling: Distributed Training, FSDP, DeepSpeed
- Instruction Tuning (SFT) (graded)
- RLHF: Reward Model + PPO (graded)
- DPO: Direct Preference Optimization (graded)
- Constitutional AI and Self-Improvement (graded)
- Evaluation: Benchmarks, Evals, LM Harness (graded)
- Quantization: Making Models Fit (graded)
- Inference Optimization (graded)
- Building a Complete LLM Pipeline
- Open Models: Architecture Walkthroughs
- Speculative Decoding and EAGLE-3 (graded)
- Differential Attention (V2) (graded)
- Native Sparse Attention (DeepSeek NSA) (graded)
- Multi-Token Prediction (MTP) (graded)
- DualPipe Parallelism
- DeepSeek-V3 Architecture Walkthrough
- Jamba — Hybrid SSM-Transformer
- Async and Hogwild! Inference (graded)
- Speculative Decoding and EAGLE (graded)
- Gradient Checkpointing and Activation Recomputation (graded)