Overview
dstack is an open-source platform designed to streamline GPU orchestration for AI and ML teams. It enables provisioning of GPUs and orchestrates containerized workloads across various environments, including cloud, Kubernetes, and bare-metal clusters. dstack focuses on increasing GPU utilization and reducing vendor lock-in. The platform supports development, distributed training, and high-throughput inference, offering a unified control plane tightly integrated with open-source frameworks. With native integration for leading GPU clouds and support for on-prem clusters via Kubernetes or SSH fleets, dstack provides flexible and efficient resource management. It also offers features like dev environments for interactive GPU access, scalable model deployment as auto-scaling endpoints, and detailed GPU utilization reporting.
Common tasks