2.7 Release Note
🌟 What's New
- Image Builder: General Availability.
- Model Deployment (Alpha): Model deployment feature enabling users to deploy and manage machine learning models.
- PrimeHub Deploy (Alpha): PrimeHub Deploy is a model deployment feature which is standalone from PrimeHub Enterprise. Now it is able to be installed on its own without PrimeHub Enterprise; Setup
PRIMEHUB_MODE=deployflag. - JupyterHub: Support non-standard images (which are not originated from jupyter/base-notebook compatible.)
- API Auth: User can now generate an API token on the API Token page to query the GraphQL API.
- Dataset Management now supports nfs and hostpath dataset types
- Job Submission now supports nfs and hostpath dataset types
🚀 Improvements
- PrimeHub Grafana dashboard refactoring
🧰 Bug fixes
- Fixed the issue where the job list was out of sync
- Fixed a degression of Jupyter Spawner.
- Fixed a bug that didn't allow users to spawn the latest jupyter/base-notebook on version master@f6331bae
- Fixed a bug where Jupyterhub showed "-" in GPU field when the user limit is 0
- Fixed a bug that caused Jupyter spawner page to show 'No group is configured for you to launch a server' when specifying a floating point in RAM of an instanceType
- Fixed a bug that caused the hub to redirect in a loop after switching user via PrimeHub-console
- Fixed a bug that failed to pull image from private repo
- Fixed a bug that we did not allow access to pvc volume across zones in AWS EKS cluster
- Fixed a bug that caused dataset upload to be broken in http environment
📣 PrimeHub CE v2.7 (Community Edition)
InfuseAI will release PrimeHub CE along with PrimeHub Enterprise since v2.7.
🎪 In the Community
Eric demonstrates how to conduct ML projects on PrimeHub and what PrimeHub can help to ease the efforts during project; here are two use-case.