Gemma, developed by Google DeepMind, is a family of open-weights large language models (LLMs) built upon the research and technology behind Gemini. This repository provides the official implementation of the Gemma PyPI package, a JAX-based library that enables users to load, interact with, and fine-tune Gemma models. The framework supports both text and multi-modal input, allowing natural language conversations that incorporate visual content such as images. It includes APIs for conversational sampling, parameter management, and integration with fine-tuning methods like LoRA. The Gemma library can operate efficiently on CPUs, GPUs, or TPUs, with recommended configurations depending on model size. Through included tutorials and Colab notebooks, users can explore examples covering sampling, multi-modal interactions, and fine-tuning workflows. By providing accessible open-weight models, Gemma enables researchers and developers to experiment with state-of-the-art LLM architectures.
Features
- JAX-based library for running and fine-tuning Gemma large language models
- Supports multi-turn and multi-modal conversations, including image inputs
- Provides open-weight checkpoints for different model sizes (2B, 7B, etc.)
- Compatible with CPU, GPU, and TPU environments
- Includes examples for sampling, fine-tuning, and LoRA-based adaptation
- Integrates easily through a simple PyPI installation and Python interface
Project Samples
