LLM2FPGA
Run Open Source LLMs locally on FPGAs
LLM2FPGA aims to enable local inference of open-source Large Language Models (LLMs) on FPGAs using a fully open-source toolchain. While LLM inference has been demonstrated on proprietary hardware and software, we are not aware of any widely recognized project running open-source LLMs on FPGAs through a fully open-source EDA (Electronics Design Automation) flow. To fill this gap, the project will produce an HDL implementation of a lightweight open-source LLM, verify it via simulation, and then attempt synthesis and place-and-route on freely supported FPGA devices. By providing a fully open alternative to proprietary and cloud-based LLM inference, LLM2FPGA will offer a transparent, flexible, and privacy-friendly way to run your own LLM on local hardware.
- The project's own website: https://github.com/RCoeurjoly/LLM2FPGA
Run by YosysHQ GmbH
This project was funded through the NGI0 Commons Fund, a fund established by NLnet with financial support from the European Commission's Next Generation Internet programme, under the aegis of DG Communications Networks, Content and Technology under grant agreement No 101135429. Additional funding is made available by the Swiss State Secretariat for Education, Research and Innovation (SERI).
