llm-beginner

Why we included this project

For learners with Python and some deep learning background, this is a hands-on Chinese-language course that moves from a Transformer sentiment classifier to a mini-GPT trained from scratch, then through instruction tuning, RAG, tool-calling agents, and a small coding agent. Each of the six tasks includes a download script, a requirements file, and a self-check harness that scores your code against expected function signatures, so you get concrete feedback rather than hoping it works. Early exercises run on an 8GB consumer GPU, and the notes include quantization guidance for smaller setups, plus a companion textbook if you want the theory behind the practice. If reading about models leaves you wanting to write them, this gives a clear, checkable path from transformer basics to working agent systems.