- Refined CPU Memory UI Settings: We’ve updated the UI for CPU memory adjustments to automatically default to the optimal setting based on the selected machine type during deployment. This ensures a more intuitive and efficient setup process.
- Increased Version History Limit to 20: The version history limit in the UI has been increased from 5 to 20. This expansion allows for greater access to your build history, enabling more effective tracking and management of past versions.
- Improved Visibility of Input/Output Schema in Model Details: On the model details page, users can now view the actual input and output schemas. This enhancement provides clearer insights into model settings and streamlines the deployment process.
- Enhanced Volume Value Retention in Reimport/Rebuild: To prevent data loss and potential import errors, volume values are now consistently retained during reimport or rebuild phases of model deployment.
- Adjustable Inference and Scale-Down Timeout Settings: We’ve introduced the ability to customize inference timeout and scale-down timeout settings directly in the model import process.
- AutoBuild UI Update for Accurate Date Tracking: The AutoBuild feature has been improved to display the correct date, providing a more accurate and user-friendly means of tracking model imports.
- In-Build Docker Image Cleanup to avoid Out-of-memory errors: Implementing weekly clearance of docker image cache to stabilize model builds.
- Expedited Model Booting Process: Various improvements have been implemented to speed up the model booting process, reducing wait times and improving overall performance.