Tokenization, Transformers, training, and inference
CookLLM
CookLLM is a full-stack LLM engineering curriculum for engineers who want more than calling an API. It covers tokenization, Transformers, training and evaluation, post-training (SFT/RLHF), and inference, with diagrams and runnable code.
Praktische info:
-
uur
Online / remote / international
Engels
Doelgroep: Software engineers, ML engineers, and developers learning LLM engineering
Leertraject
Inschrijven?
- Voorwaarden: Familiarity with Python and basic PyTorch (tensor shapes). C++ or CUDA helps but is not required.
- Prijs: $69 lifetime membership (paid course)
georganiseerd door:
CookLLM is a full-stack LLM engineering curriculum for people who want more than calling an API.
What participants learn
- Understand the LLM stack end-to-end.
- Implement core Transformer components.
- Reason about training, evaluation, and post-training (SFT and RLHF).
- Apply inference and kernel optimization basics.
The curriculum covers tokenization, Transformers, training and evaluation, post-training, GPU kernels, and inference optimization. Interactive diagrams and runnable code let learners practice implementing core modules end-to-end.
These skills strengthen the ability to design, train, evaluate, and serve language models, not only consume them through an API.