lemonade-sdk/llamacpp-rocm

Fresh builds of llama.cpp with AMD ROCm™ acceleration

What it solves

This project provides pre-compiled, "ready-to-run" versions of the popular llama.cpp AI inference engine that are specifically configured to work with AMD GPUs. It removes the technical hurdle of manually installing and configuring the AMD ROCm software stack, allowing users to run AI models on their AMD hardware immediately.

How it works

It uses an automated pipeline to build llama.cpp alongside the necessary AMD ROCm runtime libraries. By bundling these libraries directly into the download, the software is self-contained. Users simply download the version matching their specific AMD GPU architecture, extract the files, and run the included tools to start their AI models.

Who it’s for

This is for users with AMD hardware (including Radeon GPUs and Ryzen APUs) who want to run local AI models using llama.cpp without the complexity of setting up a custom development environment. It is particularly useful for those integrating these models into the Lemonade AI application.

Highlights

  • No installation required: Includes all necessary ROCm runtime libraries inside the download.
  • Wide hardware support: Supports a broad range of AMD architectures, including RDNA2, RDNA3, RDNA4, and CDNA series.
  • Cross-platform: Provides automated nightly builds for both Windows and Ubuntu.
  • Verified quality: Every release undergoes automated static checks and real-hardware inference testing to ensure reliability.
  • Transparent reporting: Each release includes qualification data and checksums to verify the integrity and compatibility of the build.

Related

  • Project
    charlie12345/ROCmFPX

    An extension for llama.cpp that introduces AMD-optimized GGUF quantization formats and accelerated kernels to improve LLM inference speed and memory efficiency on Radeon GPUs.

  • Dispatch
    llama.cpp and llama.app: Local LLM Inference and Ecosystem

    llama.cpp and its official home llama.app provide high-performance, hardware-agnostic local LLM inference, now featuring a simplified 'llama serve' command and integration with local coding agents like Pi.

  • Project
    go-skynet/go-llama.cpp

    Go bindings for llama.cpp that enable high-performance local LLM inference within Go applications using the GGUF format.

  • Project
    abetlen/llama-cpp-python

    Python bindings for llama.cpp that enable local LLM inference with an OpenAI-compatible API and support for multiple hardware acceleration backends.

  • Project
    tc-mb/llama.cpp-omni

    A high-performance C++ inference engine that enables real-time, full-duplex multimodal AI interactions (voice and video) on consumer-grade hardware.