henrywoo/pyllama
A Python library for running LLaMA foundation models on a single 4GB GPU with optimized inference.
Not currently ranked — collecting fresh signals.
star history
pyllama is a modified implementation of Facebook’s LLaMA models designed to run efficiently on consumer-grade GPUs with as little as 4GB VRAM. It provides a Hugging Face compatibility layer via pyllama.hf and utilities for downloading model weights and tokenizers for all LLaMA sizes (7B to 65B). The library aims to make large language model inference accessible without requiring enterprise hardware.
Frequently asked
- What is henrywoo/pyllama?
- A Python library for running LLaMA foundation models on a single 4GB GPU with optimized inference.
- Is pyllama open source?
- Yes — henrywoo/pyllama is open source, released under the GPL-3.0 license.
- What language is pyllama written in?
- henrywoo/pyllama is primarily written in Python.
- How popular is pyllama?
- henrywoo/pyllama has 2.8k stars on GitHub.
- Where can I find pyllama?
- henrywoo/pyllama is on GitHub at https://github.com/henrywoo/pyllama.