Rocm GitHub Repositories (2026) · Open Source Tools

Run GLM-5.3 (744B), GLM-5.3-Flash (321B), DeepSeek-V4.1-Flash (748B) and MiniMax-M3.1-Flash (428B) on your own GPUs - one engine, one dashboard, OpenAI- and Anthropic-compatible API.

LMCache: Supercharge Your LLM with the Fastest KV Cache Layer

Kyojin: the Yamz inference engine for AMD Strix Halo (ROCm, gfx1151), built on ExLlamaV3. Runs 300B-class MoE models on one 128 GB mini PC.

Open Machine Learning Compiler Framework

UniLab: A Heterogeneous Architecture for Robot RL Beyond GPU-Dominant Paradigms

LLM speculative inference server for heterogeneous hardware & consumer GPUs

A retargetable MLIR-based machine learning compiler and runtime toolkit.

Lemonade helps users discover and run local AI apps by serving optimized LLMs right from their own GPUs and NPUs. Join our discord: https://discord.gg/5xXzkMu8Zk

NumPy & SciPy for GPU

Burn is a next generation tensor library and Deep Learning Framework that doesn't compromise on flexibility, efficiency and portability.

Peak compute, memory bandwidth and latency of CPUs, GPUs and NPUs: the same tests through every API that reaches a device. Command line and app for Windows, macOS, Linux and Android

A nearly-live implementation of OpenAI's Whisper.

CubeK: high-performance multi-platform kernels in CubeCL

A set of performant RDNA/HIP patches against llama.cpp focusing on high speed inferencing with numerics assurance

AMD's graph optimization engine.

Kubernetes (k8s) device plugin to enable registration of AMD GPU to a container cluster

AMD MIVisionX is a computer vision toolkit built around a highly optimized, conformant open-source implementation of the Khronos OpenVX™ 1.3.2 specification. As of the 4.0.0 release, MIVisionX ships three components: the AMD OpenVX™ engine, the AMD RPP OpenVX extension, and the RunVX graph executor — across CPU, HIP, and OpenCL backends.

AMD RAD's multi-GPU Triton-based framework for seamless multi-GPU programming

Kindi 0.4: local LLM inference on low-cost NVIDIA GPUs, Radeon 780M and CPUs. Bonsai 2B-27B, AH-Nanbeige4.3-33B and experimental Nemotron, with an OpenAI-compatible server.

LLM inference engine written from scratch in HIP for AMD Strix Halo (gfx1151) and Qwen3.8-Flash-Next - ~50 tokens/s out to 492k context