Skip to content

Releases: microsoft/ltp-platform

LTP Release v1.4.0

02 Dec 07:51
f72e585

Choose a tag to compare

We are pleased to announce the official release of Lucia Training Platform v1.4.0!

Lucia Training Platform v1.4.0 Release Notes

This release brings significant improvements across platform stability, inference capabilities, CI/CD automation, security enhancements, and user experience.

Platform Features & Stability

  • Fixed Blobfuse2 installation failure
  • Fixed slow portal response issues
  • Fixed deployment issues on arm64 architecture and bare metal machines
  • Added image regex to support valid image limitation
  • Added webportal plugin for cluster local storage
  • Added tolerance and priority class in cluster-local-storage
  • Created PostgreSQL backend and support backend switch with Kusto

Inference Plugin

  • Inference plugin support for model proxy
  • Added support for training/inference job type in restserver
  • Added support for training/inference job type in webportal
  • Support inference job type in model proxy

CI/CD

  • Added CI/CD workflow to build all service images
  • Enabled imagelist argument for image build script
  • Fixed copy error in building image process
  • Fix cicd errors when branch creation and building all images

Security

  • Upgraded Dockerfiles with latest system updates
  • Fixed credentials exposure in source code
  • Updated Cilium to latest version

User Experience

  • Added comprehensive user manual documentation
  • Improved Copilot test experience

LTP Release v1.3.0

25 Nov 05:03
f116016

Choose a tag to compare

We are pleased to announce the official release of Lucia Training Platform v1.3.0!

Lucia Training Platform v1.3.0 Release Notes

This release brings significant improvements across platform stability, inference capabilities, user experience, and security enhancements.

Platform Features & Stability

  • Migrated PostgreSQL database to Azure disk for improved performance
  • Upgraded blobfuse version to support non-empty cache folder
  • Integrated ssh-proxy and utilization reporter services
  • Added local disk support for Prometheus deployment
  • Added H200 GPU support for VMSS provisioning scripts
  • Updated Kubernetes deployment scripts for Kubespray compatibility on Bare Metal Machines
  • CI/CD: Added GitHub workflow to build and deploy changed services
  • Fixed GPU and default eth detection in job exporter
  • Updated /etc/hosts for DNS records
  • Added alert logging webhook in alert manager

Inference Plugin

  • Added support for output streaming in web portal
  • Implemented long-context thinking output folding support
  • Deployed inference model proxy service for improved model serving
  • Messages now display as markdown and copy as plain text

Tools

  • Added comprehensive administration tools for LTP service management

Security

  • Upgraded Dockerfiles with latest system updates
  • Added Docker-in-Docker (dind) support for Webportal

User Experience

  • Code refactoring for improved maintainability
  • Integrated Copilot SGLANG/OpenAI Interface backend support
  • Added Copilot support for Dashboard Metrics visualization
  • Implemented User Feedback Loop design and functionality
  • Added VC based user group membership authentication

License

  • Added Microsoft license headers to examples directory
  • Added Microsoft license headers to contrib directory
  • Added Microsoft license headers to deployment directory
  • Added Microsoft license headers to src directory