Buckle up because version 3.6 is here, bringing some serious upgrades to your data collaboration game! This release introduces the magic of chaining Derived Datasets, levels up model management in the Governance Portal, and fine-tunes the Apheris Model Registry. Let’s crack open the details.
New Features
Chaining Derived Datasets - The Domino Effect You Actually Want
Have you ever wished you could build on top of an already Derived Dataset? Now you can! Think of it like stacking Lego blocks, but for data – less mess, more efficiency. You can reuse intermediate results and reduce redundant computations by chaining Derived Datasets. Don’t worry, the original Asset Policy rules stay locked in like a vault, ensuring everything remains private within the Apheris Compute Gateway. Want the full scoop? The docs have you covered.
Enhanced Model Management
Organization owners can now assign the Model Manager role to users directly in the Governance Portal. Model Managers can create custom model versions and add these to an organization's Registry. This allows for more efficient iteration on models and independence in applying new learnings. Want a refresher on Role Types? Find all details in the documentation.
Upgraded NVIDIA FLARE Support
We now support NVIDIA FLARE 2.5.1 and 2.5.2! Upgrading to 2.5.2 or higher means your client/server communication becomes more flexible, allowing for using mTLS or TLS and enabling port standardization as well as customization.
Inference Flow for XGBoost Models - Because Predictions Shouldn't Be a Puzzle
Need to run inference on a trained XGBoost model? Done. This new workflow makes it a breeze, so you can focus on getting insights instead of wrestling with setup.
This release is packed with goodies to streamline workflows, enhance governance, and fortify security – all while keeping things smooth and efficient. Want to dive into the details? Check out the product documentation.
Happy collaborating!
Your Apheris Team
