Getting Started with Local LLMs
Chapter 1: Introduction to Large Language Models
What Is a Large Language Model (LLM)?
A Simple View of How LLMs Work
Understanding Tokens and Text Generation
Why Run LLMs Locally?
Local vs. Cloud-Based LLMs
Quiz for Chapter 1
Chapter 2: Preparing Your Local Environment
Hardware Considerations: CPU
Hardware Considerations: RAM
Hardware Considerations: GPU and VRAM
Checking Your System Specifications
Operating System Compatibility
Installing Python (Optional but Recommended)
Introduction to the Command Line / Terminal
Quiz for Chapter 2
Chapter 3: Finding and Selecting Local LLMs
Where to Find LLM Models: Hugging Face Hub
Understanding Model Sizes and Parameters
Model Formats: GGUF and Others
Quantization: Making Models Smaller
Reading Model Cards for Information
Model Licenses and Usage Restrictions
Choosing Your First Model
Quiz for Chapter 3
Chapter 4: Running Your First Local LLM
Introduction to Local LLM Runners
Setting up Ollama
Downloading a Model with Ollama
Running a Model with Ollama (Command Line)
Setting up LM Studio
Finding and Downloading Models in LM Studio
Loading and Chatting with a Model in LM Studio
Introduction to llama.cpp (Concept)
Hands-on Practical: Running a Model
Quiz for Chapter 4
Chapter 5: Basic Interaction and Prompting
What is a Prompt?
Your First Prompt: Simple Questions
Giving Instructions
Understanding Context Window
Basic Prompt Formatting Tips
Temperature and Creativity
Common Interaction Patterns
Practice: Simple Prompting Techniques
Quiz for Chapter 5
Quantization: Making Models Smaller
Was this section helpful?
- GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers, Elias Frantar, Saleh Ashkboos, Torsten Hoefler, Dan Alistarh, 2022 ICLR 2023 DOI: 10.48550/arXiv.2210.17323 - Details GPTQ, a method for accurate post-training 4-bit quantization specifically designed for large language models, addressing precision trade-offs.
- llama.cpp repository, Georgi Gerganov and the llama.cpp Community Contributors, 2024 - The project's repository and associated documentation, detailing the GGUF file format and its specific quantization schemes (e.g., Q_K variants) used for local LLM inference.