A current article got here out with an up to date listing of vital elements for MLOps and LLMOps. And whereas this listing could appear lengthy, studying by the capabilities and elements, I noticed that SAS Viya already covers a lot of the required performance. Organizations can have a hodgepodge of instruments that they need to string collectively themselves or an analytics platform that already works together with your favourite container registry.
SAS Viya is an end-to-end analytics platform with instruments for:
Accessing, managing, and getting ready knowledge
Visualizing knowledge, creating reviews, and constructing dashboards
Coaching machine studying fashions pipelines utilizing no-code, low-code, SAS code, Python code, or R code
Managing, deploying, and monitoring your fashions.
The whole lot you want is within the SAS Viya foundational providing. Thus, I wish to suggest that the final word must-have for MLOps is SAS Viya.
Model Management
Mannequin objects are versioned inside SAS Viya’s mannequin administration device, together with mannequin code, metadata, properties, and extra. Moreover, every model of a mannequin could be rating examined, deployed, and monitored inside SAS Viya. No must rollback or copy your rating code to work with a earlier model of your mannequin. Moreover, historical past is tracked with timestamps and usernames, so who accountable.
Orchestrator
SAS Viya comes with a workflow device that lets you specify your ideally suited MLOps processes and even implement them! We just lately used the workflow device to create a reliable AI lifecycle workflow for embedding suggestions from the AI Danger Administration Framework throughout the analytics lifecycle. However in the event you want different instruments, no worries. SAS Viya options open APIs that can be utilized to construct orchestration flows utilizing your tooling of selection.
Mannequin Registry
The mannequin registry on SAS Viya organizes fashions into tasks based mostly on use case, not language. Tasks share key info, like function and coaching knowledge, permitting customers to match fashions side-by-side to find out the very best match for manufacturing. Fashions are saved with sturdy metadata, together with efficiency on coaching knowledge, enter variables, output variables, modeler, creation date, modified knowledge, created by, modified by, modeling device, and I can preserve going! However stayed tuned for an thrilling replace to our mannequin registry. Mannequin playing cards will quickly be accessible to mechanically visualize mannequin well being and well-being.
Compute
SAS Viya has compute to assist SAS, Python, and R mannequin coaching and execution by SAS Cloud Analytic Providers (CAS). CAS gives the run-time surroundings for knowledge administration and analytics on SAS Viya. Thus, you’ll hear CAS referenced as a spot to retailer knowledge, but in addition as an analytical engine. From the mannequin pipelining device on SAS Viya, you possibly can create mannequin coaching pipelines consisting of out-of-the-box nodes, SAS code, Python code, and R code, which are in contrast side-by-side to search out essentially the most correct mannequin.
Serving
SAS Viya options varied deployment locations so fashions could be executed in the very best kind for every use case. For instance, with SAS Viya you possibly can:
Host Python and SAS fashions as a REST-API endpoints on SAS Viya for ad-hoc transaction scoring
Construct containers for SAS, Python, and R fashions with the required dependencies and a minimal footprint and publish to registries in Docker, Azure, AWS, and GCP
Rating fashions on knowledge tables in batch, which can be utilized in course of flows, scheduled, or visualized in a dashboard
Add enterprise logic and orchestrate amongst a number of fashions inside a choice circulation.
Monitoring
Mannequin efficiency and knowledge drift are additionally monitored inside SAS Viya’s mannequin administration device. The efficiency monitoring reviews mechanically generate a wide range of metrics and charts, relying on the kind of mannequin being monitored. These reviews can be personalized so as to add new charts or take away charts you don’t want.
Experiment Monitoring
The mannequin pipeline device on SAS Viya makes experimentation simple! The pipelining device lets you string knowledge preprocessing steps with a number of fashions and ensembling to construct a mannequin experiment in only a few clicks. To work even quicker, you possibly can leverage the out-of-box templates, create and share your individual templates, or use Automated Machine Studying.
This device makes an information scientist’s job a lot simpler by mechanically splitting the info into coaching, testing, and validation, mechanically calculating evaluation metrics throughout all three units, and offering hyperparameter autotuning. Information scientists can view the outcomes of every node, every pipeline, or throughout pipelines to focus their consideration on the fashions with essentially the most potential. As soon as they’re pleased with their fashions, they are often added to the mannequin repository in two clicks.
Accountable AI
SAS leads in Accountable AI. SAS takes a multi-prong strategy for serving to organizations construct extra reliable AI programs, by participation in a number of AI boards, advisory providers, embedded options in SAS Viya, and extra. Simply to call just a few options, SAS Viya assist info privateness classifications for knowledge, lineage, bias evaluation, bias mitigation, Reliable AI Lifecycle workflow, and shortly, the mannequin card! To dive deeper, you possibly can discover the opposite SAS Reliable AI blogs.
Human within the Loop
Protecting people within the loop is straightforward on SAS Viya. Utilizing the workflow device, you possibly can add system duties to inform or electronic mail customers or teams at any time when one thing wants their consideration. Customers can be prompted to take actions by assigned person duties. And with an abundance of visualizations summarizing key info all through SAS Viya, customers can rapidly view standing and know when to behave.
Conclusion
SAS Viya is the most efficient analytics platform. Inside just a few hours, you possibly can add an information set, discover and visualize the info to construct out your understanding, practice high performing machine studying fashions, and deploy the champion mannequin to make higher selections and enhance outcomes. As a substitute of spending your time stringing collectively a bunch of elements, spend time constructing extra correct fashions. Mitigate threat of error, scale back vulnerabilities, and reduce cognitive load brought about from gluing collectively instruments and make analytics simpler with SAS Viya.
Don’t consider me? Effectively, there may be nonetheless time to register for the SAS Hackathon. Contributors within the SAS Hackathon convey an issue, construct a staff, discover knowledge, and have a month to construct their answer. Most significantly, SAS gives entry to SAS Viya, coaching, and mentoring without cost. However you must register earlier than August thirtieth!
