Awesome GitHub RepositoriesAMD Hardware Acceleration

Odysseus is a self-hosted AI workspace and autonomous agent framework designed for deploying and managing large language models. It serves as a centralized platform for orchestrating agentic tasks, utilizing a model context protocol server to connect AI models to external system utilities, browser automation, and local hardware. The system distinguishes itself through a combination of retrieval-augmented generation and a RAG knowledge base, using vector stores and local embeddings to provide persistent semantic memory. It further integrates AI-driven communication management to triage email i

Whisper.cpp is a high-performance, local-first speech recognition engine designed to run large-scale machine learning models on consumer hardware. It functions as a portable library that converts audio into text, supporting both static file transcription and real-time stream processing. By utilizing a lightweight inference engine and weight quantization, the project minimizes memory and compute overhead, allowing for efficient execution without reliance on external cloud APIs or internet connectivity. The project distinguishes itself through a hardware-agnostic compute abstraction that offloa

This project is a distributed training infrastructure designed for aligning large language models through reinforcement learning. It functions as an end-to-end engine for complex alignment tasks, including proximal policy optimization, direct preference optimization, and iterative self-play. By providing a unified framework for multi-turn interactions and tool-use scenarios, it enables the development of models capable of reasoning and external environment engagement. The framework distinguishes itself through a decoupled architecture that separates model training from sample generation. This

This project provides a customized Linux kernel and driver suite designed to enable hardware compatibility for Surface devices. It focuses on building and patching the Linux kernel to provide driver support for proprietary hardware components that are missing from the upstream source. The system includes a secure boot kernel signing mechanism and a process for enrolling custom keys into the system firmware. This allows the execution of patched kernels while maintaining system security protections. The project covers peripheral driver support for touchscreens, styluses, and keyboards, as well

SD.Next is an all-in-one web interface and multi-backend inference engine for generating, editing, and processing images and videos using diffusion models. It functions as a comprehensive tool for diffusion model management and an automated image processing pipeline for bulk operations. The project is distinguished by its hardware-backend abstraction layer, which provides automatic detection and acceleration for NVIDIA CUDA, AMD ROCm, Intel OpenVINO, and DirectML. It features a headless generative API and a programmatic command interface, allowing users to trigger tasks via REST API or CLI wi

OpenWrt-Rpi is a firmware builder and embedded Linux build toolset designed to compile custom router operating system images for Raspberry Pi and NanoPi hardware. It utilizes a GitHub Actions CI pipeline to automate the generation of fresh firmware images from the latest source code via daily updates. The project maintains a custom package repository and local software sources for managing community plugins and kernel modules. This infrastructure is used to ensure version consistency and prevent dependency conflicts during the installation of network drivers and system extensions. The build

Text Embeddings Inference est un serveur d'inférence haute performance conçu pour héberger des modèles d'embedding de texte et de classification de séquences en tant qu'endpoints d'API scalables. Il fournit une API d'embedding vectoriel pour convertir le texte en représentations denses et un serveur de reranking cross-encoder pour noter la pertinence des séquences de documents par rapport à une requête. Le projet dispose d'un moteur d'inférence accéléré par GPU qui utilise le traitement par lots dynamique et des noyaux spécialisés pour maximiser le débit. Il offre une interface binaire haute performance via gRPC comme alternative au HTTP standard pour réduire la latence réseau et la surcharge de sérialisation. Le système couvre un large éventail de capacités, incluant le classement de similarité de documents, le reranking de texte multilingue et la classification de séquences pour prédire des catégories ou des sentiments. Il prend en charge divers environnements de déploiement, allant des conteneurs serverless auto-scalables aux installations isolées (air-gapped). L'accélération matérielle est disponible pour les GPU NVIDIA, les GPU AMD et Apple Metal.

JimsGarage is a collection of shell scripts and automation tools designed to help individuals deploy and manage a wide range of self-hosted services on their own hardware. It provides a structured approach to setting up containerized applications, from media servers and document management systems to VPNs and monitoring stacks, all through automated Docker-based configurations. The project distinguishes itself by offering a comprehensive library of deployment recipes that cover the full lifecycle of a home server environment. This includes not just the services themselves, but also the suppor

Ce projet est un framework de service de modèles PyTorch conçu pour déployer et mettre à l'échelle des modèles de machine learning en production via des endpoints réseau évolutifs. Il fonctionne comme un serveur d'inférence haute performance, un optimiseur et un gestionnaire de cycle de vie de modèle qui gère le chargement des modèles, le batching des requêtes et l'accélération matérielle. Le système se distingue par des capacités avancées d'orchestration et d'optimisation, telles que le chaînage de plusieurs modèles dans des workflows séquentiels utilisant des graphes d'exécution et l'utilisation du batching dynamique pour améliorer le débit et la latence. Il offre un support spécialisé pour l'IA générative et les grands modèles de langage (LLM) grâce au batching continu et au parallélisme de tenseurs. Les domaines de capacités étendus incluent la gestion des ressources GPU sur divers matériels comme NVIDIA, AMD et Apple Silicon, ainsi qu'une gestion complète du cycle de vie des modèles pour l'enregistrement, le versioning et la mise à l'échelle des workers. Il intègre également des outils d'observabilité pour suivre la santé du système et les performances des modèles via des métriques compatibles Prometheus. Le serveur est géré via une interface de ligne de commande utilisée pour le contrôle du cycle de vie et la configuration des paramètres d'exécution.

IREE is an MLIR-based compiler toolchain and runtime designed to translate machine learning models from various frameworks into optimized binaries for execution across diverse hardware targets. It provides a unified pipeline to ingest models from PyTorch, TensorFlow, JAX, and ONNX, lowering them into a common intermediate representation for deployment on CPUs, GPUs, and bare-metal embedded systems. The project distinguishes itself through a bytecode virtual machine and a hardware abstraction layer that decouple high-level model logic from specific hardware instruction sets. It supports sophis