Creating your own GitHub Copilot with Fauxpilot
Fauxpilot is a self-hosted code suggestion platform that supports multiple programming languages such as Java, Python, or Javascript. Brendan Dolan-Gavitt, an assistant professor at New York University Tandon School of Engineering, created it as an alternative approach to GitHub Copilot.
FauxPilot relies on the Salesforce CodeGen, a large-scale language model for conversational artificial technology (AI) programming. Using FauxPilot, you get rid of repetitive code. Instead, you can spend your time focusing on the business logic that your application needs to address.
Why Use Fauxpilot?
GitHub Copilot is a popular commercial tool that generates code, helping developers to code more efficiently. However, it sends the snippets of your code back to Microsoft’s servers to suggest code for you. This hinders the violation of licensing and brings exploitable vulnerability concerns as code is traveled to multiple places.
Fauxpilot works merely using the training models of Codegen and does not send your code anywhere. Thus, it is more secure than GitHub Copilot.
In addition, because Fauxpilot is a self-hosted and open-source tool, you do not need to pay any fee except for running the Fauxpilot server.
Model Types of Fauxpilot
Fauxpilot supports eight models of collagen.
- Model 350M-mono requires 2GB VRAM (GPU RAM) and supports only Python programming language
- Model 350M-multi requires 2GB VRAM and supports multiple programming languages
- Model 2B-mono requires 7GB VRAM and supports only Python programming language
- Model 2B-multi requires 7GB VRAM and supports multiple programming languages
- Model 6B-mono requires 13GB VRAM and supports only Python programming language
- Model 6B-multi requires 13GB VRAM and supports multiple programming languages
- Model 16B-mono requires 32GB VRAM and supports only Python…
