As organisations experience an increase in their machine learning operations, the requirement for robust and mature scalable MLOps workflows increases significantly. We are thrilled to announce the release of Seldon Deploy v1.6 as it introduces robust features that significantly improve the stability, usability and security of the overall platform. The features introduced in this version of Seldon Deploy will ensure organisations can manage thousands of machine learning models at scale with robust infrastructure and scalable architectural patterns.
Seldon Deploy v1.6 introduces a broad range of improvements across multiple levels of the platform. These include a UI Theme Refresh, Usage Monitoring Improvements, HuggingFace Runtime Support and Improvements in Advanced Monitoring. Below are some of the key highlights – for deeper information you can read more in our release notes.
UI Theme Refresh
The features for our advanced MLOps capabilities have continued to improve from a functional perspective. Apart from that, it’s also key to ensure that data scientists that work with our enterprise product get a great user experience. This version of Seldon Deploy introduces a new theme which provides a refreshed look to the product. This primarily involves re-styling of headers and components and provides a more modern and consistent experience as the user moves through the workflows of the product.
Usage Monitoring Improvements
As enterprise customers of Seldon Deploy are able to mature in their MLOps capabilities they need to scale their usage of advanced models and monitoring resources. In this release, usage monitoring now provides a more granular view of the total number of models as well as advanced analytics. These features give users the ability to view resources at a namespace level as well as advanced data science monitoring components.
HuggingFace Runtime Support
We are thrilled to announce that following the release of our new HuggingFace Runtime at KubeCon 2022 we have now integrated it to our enterprise product. This new HuggingFace Runtime provides a simple interface that allows users to benefit from a broad range of pre-trained models. These models include HuggingFace hub, together with optimisations with the ONNX Runtime through the HuggingFace Optimum framework.
This new runtime allows Seldon users to leverage pre-trained models with a simple set of configuration parameters that can be provided directly as part of the model artefact, or through the manifest resource parameters. The base parameters exposed include “task-type” which supports the “text-generation”, “text-classification”, “question-answering” – amongst various others. The runtime exposes other parameters such as “optimum” for optimisations, “batch_size” for GPU parallelisation, “device” for accelerator selection, among others outlined in the documentation.
Improvements in Advanced Monitoring
The drift detection screens provide advanced visibility on the concept and data drift of deployed models in real time. In this release, we have improved the drift detection monitoring functionalities with feature-level drift scores. The drift detection monitoring dashboard can now highlight which features are drifting over time along with other metrics. This enables actionable insights and understanding from drift detector outputs.
Try Out Seldon Deploy!
With this host of exciting new features, now is a great time to get your hands on Seldon Deploy. Set up a free trial today to test the new Seldon Deploy capabilities for yourself.