For enterprises, the ability to put models into production at scale is critical to ensure business value is generated. The process which defines the model lifecycle, from when a model is being developed, to production is ModelOps.    The ModelOps process is critical. Different models serve different use-cases and what production looks like varies, at the same time, there needs to be a consistent framework such that production can be repeatable, and trusted.   In this webinar you will learn:   To develop mission critical models using SAS and open-source How both SAS & open-source models are managed in a centralised model repository How SAS Viya supports model deployment through a variety of different production endpoints   Watch the Webinar   Questions and Answers:   What cloud(s) does SAS Viya Support? SAS Viya supports deployment on cloud providers like Azure, AWS & GCP as well as other on-prem options like open-source Kubernetes or on OpenShift. For a more detail, refer to the documentation: https://go.documentation.sas.com/doc/en/itopscdc/v_045/itopssr/n098rczq46ffjfn1xbgfzahytnmx.htm   Can Viya Integrate with CI/CD? I know I did show in my deployment part, for instance, a Git repository and integrating with Git. I know YJ actually showcased SAS Studio flows. And one of the things about SAS Studio flows is you can actually integrate it with your Git repository. So, I know a lot of organizations are leveraging Git for a lot of their CICD, their version management. So having that is really, really important and we can certainly integrate with that. So, it would be really pretty interesting to hear from people on the call post this session what kind of services are they leveraging for CI/CD and we can answer that on a case-by-case basis.   What Python packages are supported on Viya? The way it works is that we still use the open source package management tools. Things like pip and condo Viya is still using underneath. So the answer to that question is almost like how long is a piece of string. It's like whatever packages you want to configure you can use. So what is required from BYO is that a Python environment is configured. You can either bring your own or when you set up BYO for the first time, there are some applications that actually help you set up your own Python environment and then all that is done is in ensuring that that Python environment can be mounted to the BYO environment. So things like you can see paths and host paths and all of that sort of stuff or whatever's on that environment that has been configured using these open source tools, it will be there in on Viya.   Can you build your own custom model interpretability reports in Viya in addition to using the out of box in model manager? Yeah, great question and certainly I know that YJ actually showed this in the demo. So instead of just going into Model Manager and leveraging those out-of-the-box reports, you can actually go into Visual Analytics, you get access to all of the underlying data and you can build custom model degradation reports, model interpretability reports. It's very flexible in terms of how you can actually leverage that and what you can do. So that's certainly possible.   What is the significant benefit using Python in SAS, instead of using SAS only? I think one benefit is that it's around more efficient sort of model workflows. The pattern that we've been noticing talking to a lot of organisations and clients is that now a lot of organizations have multidisciplinary teams in terms of programming languages. So you have SAS users working with open source users because a lot of the processes traditionally at organizations have been run in SAS, for example, a lot of the data processing, the data management and then you also have a lot of people using open source coming from universities as well as from a more innovative or innovation standpoint, because all of the packages are being developed in open source and it's hard to match the speed of the open source software model.   So by bringing these two worlds together, we essentially are saying to organizations, you can keep doing what has been really successful for you, but you can also then use what's great out in the community. And you're not limited by who you're hiring because it doesn't matter whether they have Python skills, R skills, SAS skills, whatever, they can all join in.   So it's all about you create the best model, workflow, the best pipeline for your organization, and everyone can use the skill of choice.   Must the data be in a CAS library? You can add in anything if you want. So a lot of the capability we've shown is using CAS because we just want to show the latest and greatest stuff.   If you are coming from the perspective of you are a SAS 9 user, there's still the SAS 9 Compute engine that is here on fire as well. As a user, you don't really have to explicitly go OK, I'm using compute. OK, I'm using CAS. The platform is smart enough to determine what you're doing and your code will just run. It's just that if you want some additional benefits of maybe the speed, and of course all of this UI functionality also runs in CAS. That's when you go to CAS. So from the perspective of, let's say I'm a SAS user and just want my code to run, it will run. You just have to do things maybe like file paths that might not exist on one server or another, but your code will run as is.   If you want to speed it up, use CAS for all of the stuff that's happened in the UI, all the great things that we've shown that's happening under the hood. You don't really have to think about it.   It's also important to consider the deployment part of the story. So obviously when you're discovering, testing, building your models and you're building it on the SAS platform, it's going to need to be in CAS. That's just how it is. Now when you deploy it, ideally what you want to do is you want to actually then think about deploying that close to where the data is. So a lot of what I talked about around embedded processing, things like event stream processing that really considers that you're actually now taking that production model and you're pushing it close to where the data source is.   In those instances, you don't need a CAS engine to do that. You could be actually just leveraging the database where the data source is sitting in. Now obviously that's going to depend on your organization and where you want to deploy it. Thus the answer to the question is, you don't just have to use the CAS library, but there is a performance advantage to doing so.   This presentation was answer to many questions our site raised. We hope to utilise containers and GPU etc. with existing SAS code. It seems we have to purchase the Viya license, however there is no trial whatsoever. How to commit the among of investment without even trial at all?  There is actually a free online trial available here - Experience SAS Viya for yourself ... View more