gpu-scheduler

Resource scheduling and cluster management for AI

A very simple GPU job scheduler - To run multiple jobs with assigned (limited) GPU resources in a dynamic way

Dynamic schedule management framework for soft-real-time jobs on GPU based architectures

Experiment / job runner with automatic GPU assignment, task queuing, and web UI.

Intelligent GPU Traders

A lightweight, preemptive GPU task scheduler for single-node research servers. Features priority queues, graceful preemption, auto-logging, and seamless Conda integration.

Ares: Multi-Cluster Kubernetes Scheduler with GPU Topology Optimization (Intra-Node, Inter-Node, Inter-Cluster) and Exactly-Once Execution Semantics

Script running tool for optimizing GPU memory utilization

Your Macs, scheduled: a multi-tenant control plane for Apple Silicon fleets (admission, quotas, live GPUs, model workbench, provenance).

🤖 Build your personal AI infrastructure to streamline your life and work, enhancing productivity and harnessing the power of AI.

Lightweight multi-user GPU job scheduler for single-node NVIDIA servers, with fair queuing, hard GPU affinity, NVLink awareness and training job recovery.

AlgoGPU is a minimal GPU scheduler designed for AI agents and inference workloads. Simple. Deterministic. Reliable.

A high-performance utility for dynamic GPU resource allocation. Features real-time task prioritization, VRAM optimization, and advanced process scheduling to boost gaming and rendering.

Local GPU lease broker for CUDA model workers

Distributed, GPU-aware workload scheduler for heterogeneous clusters: queueing, quotas, GPU flavors, and autoscaling.

Heterogeneous distributed AI execution with multi-role peers, exact MoE inference, locality-aware scheduling, replication and failover.