Implementation of a standalone (offline) analogue of ChatGPT on Unity.
Features:
• Ability to use different types of GPT models (including LLAMA3.1);
• The plugin can work with CUDA (super fast inference);
• The small size of GPT models. From 4 to 10 gigabytes;
• It works offline, in real time.
Supported platforms:
• Windows (x64).
Documentation
UGPT uses llama.cpp under MIT License; see Third-Party Notices.txt file in package for details.