T
ToolShelf
SIMPLE-LLM
// ~950 line, minimal, extensible LLM inference engine built from scratch.

simple-llm

~950 line, minimal, extensible LLM inference engine built from scratch.

13EmergingUnknown
Updated
Today

What it does

~950 line, minimal, extensible LLM inference engine built from scratch. NOTE: Currently, this repository ONLY supports on a single NVIDIA H100. Why? A complex starting point (MoE + large model + reasoning + good hardware) demonstrates that building something like this from scratch is viable!! | Component | Lines | |-----------|-------| | (engine) | 563 | | | 324 | | | 92 | The codebase is

Getting Started

git
git clone https://github.com/naklecha/simple-llm

Platforms

πŸͺŸwindows🍎mac🐧linux

Install Difficulty

moderate

Built With

python

Community Reactions