4 Repos
Deployment systems that automatically detect host hardware capabilities to select and pull the most optimized model image.
Distinct from Hardware-Agnostic Deployment: Distinct from Hardware-Agnostic Deployment: focuses on active hardware detection and specific image selection rather than generic portability across architectures.
Explore 4 awesome GitHub repositories matching artificial intelligence & ml · Hardware-Aware Deployment. Refine with filters or upvote what's useful.
Paddle-Lite is a deep learning inference engine and edge computing runtime designed to execute trained models on mobile and edge devices. It provides a hardware-accelerated inference framework and a decoupled runtime with a minimal binary footprint to operate in resource-constrained environments without third-party dependencies. The project includes a model quantization tool for reducing precision and size via static and dynamic quantization, as well as a computation graph optimizer. These tools reduce latency and memory usage by fusing operators and pruning the model intermediate representat
Employs hardware-aware deployment to execute deep learning models across diverse CPU, GPU, and NPU backends.
Dieses Projekt ist ein PyTorch-Framework für das Model-Serving, das darauf ausgelegt ist, Machine-Learning-Modelle in der Produktion über skalierbare Netzwerk-Endpunkte bereitzustellen. Es fungiert als leistungsstarker Inference-Server, Optimierer und Modell-Lifecycle-Manager, der das Laden von Modellen, Request-Batching und Hardware-Beschleunigung übernimmt. Das System zeichnet sich durch fortschrittliche Orchestrierungs- und Optimierungsfunktionen aus, wie etwa das Verketten mehrerer Modelle zu sequenziellen Workflows mittels Ausführungsgraphen und den Einsatz von Dynamic Batching zur Verbesserung von Durchsatz und Latenz. Es bietet spezialisierte Unterstützung für generative KI und Large Language Models durch Continuous Batching und Tensor-Parallelität. Zu den breiten Funktionsbereichen gehören GPU-Ressourcenmanagement für diverse Hardware wie NVIDIA, AMD und Apple Silicon sowie ein umfassendes Lifecycle-Management für Registrierung, Versionierung und Worker-Skalierung. Zudem integriert es Observability-Tools zur Überwachung des Systemzustands und der Modellleistung über Prometheus-kompatible Metriken. Der Server wird über eine Kommandozeilenschnittstelle verwaltet, die zur Steuerung des Lifecycles und zur Konfiguration von Laufzeitparametern dient.
Deploys pipeline components across CPUs and GPUs to optimize operational costs based on hardware capabilities.
Ramalama is a containerized runtime and management tool for large language models. It functions as an OCI AI model manager and registry client, allowing users to package, distribute, and execute AI models as standardized container images. The project differentiates itself by using OCI-compliant distribution for models and retrieval augmented generation assets, enabling the packaging of vector databases into immutable container images. It features hardware-aware image selection that automatically detects GPU or CPU capabilities to pull the most optimized image for the host environment. The sy
Implements hardware-aware image selection that automatically detects GPU or CPU capabilities to pull the most optimized model image for the host.
RLinf is a distributed reinforcement learning orchestrator and embodied AI training framework. It provides the infrastructure to train vision-language-action models and robotic policies using a combination of reinforcement learning and supervised fine-tuning. The system is designed for scaling workloads across GPU clusters, managing the placement of actors, rollout workers, and environment components. It features a specialized robotics data collection pipeline for gathering teleoperated demonstrations and simulation trajectories into standardized replay buffers, alongside a hardware interface
Launches actors on specific physical nodes and GPUs using custom environment variables and affinity scheduling.