🦄 Distributed Inference on Kubernetes with DRA and MIG

Unofficial example on creating Multi-Instance GPU (MIG) instances with NVIDIA Management Library (NVML) Go bindings.

Set up your own GPU powered instance on AWS with RKE2 Kubernetes and then help you setup NVIDIA GPU operator for testing out AI workloads.

Unofficial minimal instructions for managing NVIDIA Multi-Instance GPU (MIG) in a docker container

Tools for training, analysis and execution of an optimized task scheduling RL agent on GPUs with Multi-Instance GPU (MIG).

A Kubernetes serving manager for machine learning inference system enabled with NVIDIA MIG/MPS GPU-Sharing support

C++ implementation of FAR scheduler, an approach for optimized GPU co-execution of tasks using the hardware-level partitioning MIG (Multi-Instance GPU).