rasbt/LLMs-from-scratch

Why this is useful

A step-by-step, code-first walkthrough of how a language model is actually built — tokenisation, attention, training, and instruction-tuning — implemented from first principles rather than imported from a library.

It suits a reader who already writes code and wants to understand the mechanics underneath the models they use daily, not just how to call an API.

This is a deep, technical commitment, not a weekend read — expect to spend real time working through the code. It teaches how models are built, not how to prompt or deploy one effectively.

Topics

• ai
• artificial-intelligence
• attention-mechanism
• deep-learning
• finetuning
• from-scratch
• generative-ai
• gpt
• instruction-tuning
• language-model
• large-language-models
• llm
• machine-learning
• natural-language-processing
• pretraining
• python
• pytorch
• tokenizer
• transformers

Trending history

Related repositories