Paperspace announced today the launch of Gradient Installer, an open source installation package that developers can use to turn their own GPU machines — whether locally or in the cloud — into a fully featured machine learning platform. The release offers a self-service approach to deploying the Gradient MLOps platform — without any support required — in under 10 minutes.
“Developers today have few options if they want to build ML pipelines on their own GPUs: they need to build the software themselves or hack together open source components,” said Daniel Kobran, COO, Paperspace. “These approaches are time-consuming, expensive, and prone to a host of compatibility issues. That’s why we’ve released Gradient Installer to offer a better way to bring MLOps software to any cluster — within minutes.”
The new self-hosted Gradient offering makes the enterprise-grade machine learning platform easy to install and run on your own without. There is managed and unmanaged support for AWS, GCP, Azure, Paperspace Cloud, or private servers — so you can always bring your model training and inference close to where your data lives.
The platform is effortless to adopt. After going through a quick install, teams experience drastic decreases in model build-to-deployment time within days of adoption. Other features and benefits of Gradient Installer include:
- Open source installer available on GitHub
- Full CI/CD and MLOps pipeline and functionality
- Kubernetes and Terraform under the hood with support for Argo pipelines
- Host metrics, log streaming, and experiment tracking
- Compatible with AWS, GCP, Azure, Paperspace Cloud, private servers, and the NVIDIA DGX product line
The latest version of the Gradient Installer also includes the following major updates:
- New CLI for interactive installation via your terminal
- The CLI has a new and improved system for updating and removing clusters.
- To keep track of your cluster’s state, the CLI stores your state file in an S3 bucket.
- To manage multiple clusters, you can generate multiple profiles which can be used for different teams.
- Gradient now provides SSL certificates for all processing sites (your VPC). You can choose to bring your own if desired.
- The NVIDIA DGX is now a first-class installation target. This integration was borne out of a collaboration with NVIDIA which you can read more about here.
- There is a new option to deploy Gradient in the Paperspace Cloud before testing on your own infrastructure. This option enables you to get up and running very quickly and experiment with the product before installing in your own environment.