Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning

Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning

Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning
picture

tloen/alpaca-lorahttps://github.com/tloen/alpaca-lora

Stars: 18.2k License: Apache-2.0

Alpaca-lora is a project for fine-tuning the LLaMA model on consumer-grade hardware. The main features, key characteristics, and core advantages of this project include:

  • Provides an Instruct model that can run on Raspberry Pi, with quality similar to text-davinci-003, and the code is easy to extend to 13b, 30b, and 65b models.
  • The training code can run on a single RTX 4090 in a few hours, while scripts for downloading and inferring the base model and LoRA, as well as the LoRA weights themselves, have been released.
  • Uses Hugging Face’s PEFT and Tim Dettmers’ bitsandbytes for cost-effective and efficient fine-tuning.
  • Supports Docker, allowing users to build container images and perform inference.

cloudwu/skynethttps://github.com/cloudwu/skynet

Stars: 13.0k License: MIT

Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning
picture

Skynet is a lightweight online game framework. The main features, key characteristics, and core advantages of this project include:

  • Supports Lua framework.
  • Uses actor model.
  • Widely used in the Chinese gaming industry, spreading to other industries and English developers.
  • Provides detailed documentation and a FAQ page.

KindXiaoming/pykanhttps://github.com/KindXiaoming/pykan

Stars: 10.3k License: MIT

Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning
picture

Pykan is a Python implementation of Kolmogorov Arnold Networks. KANs are neural networks based on the Kolmogorov-Arnold representation theorem, more accurate and interpretable than multi-layer perceptrons (MLPs). Its main functions and advantages include:

  • Higher accuracy with faster scaling and fewer parameters.
  • Intuitive visualization and interpretability.
  • Easy installation, supporting both PyPI and GitHub installation methods.
  • Comprehensive documentation, providing quick start tutorials and hyperparameter tuning suggestions.

asweigart/pyautoguihttps://github.com/asweigart/pyautogui

Stars: 9.7k License: BSD-3-Clause

Pyautogui is a cross-platform GUI automation Python module for programmatically controlling the mouse and keyboard.

  • Programmatically control the mouse and keyboard.
  • Cross-platform support.
  • Provides documentation in Simplified Chinese.
  • No need to install Win32 extensions (Windows).
  • Supports both Python 2 and 3 versions.

emmett-framework/granianhttps://github.com/emmett-framework/granian

Stars: 2.1k License: BSD-3-Clause

Granian is a Rust HTTP server designed for Python applications.

  • Supports ASGI/3, RSGI, and WSGI interface applications.
  • Implements HTTP/1 and HTTP/2 protocols.
  • Supports HTTPS.
  • Supports Websockets.
  • Provides a single package that runs on multiple platforms, avoiding the dependency combination issues of using Gunicorn + uvicorn + http-tools on Unix systems, and has stable performance compared to existing alternatives.

Running LLaMA on Raspberry Pi: Cost-Effective Fine-Tuning

Leave a Comment