Watch this Ask the Expert session to learn how R integrates with Viya.
Watch the Webinar
You will learn:
About the SWAT package and how it enables open source integration with Viya.
How to submit CAS actions using R to run in-memory analytics.
The general integration aspects of loading data into memory, analyzing distributed tables, creating predictive models, and downloading results to the client using R functionality.
The questions from the Q&A segment held at the end of the webinar are listed below and the slides from the webinar are attached.
Q&A
Is there a SAS V10 in working, or next version of SAS Viya?
SAS Viya is on continuous development. New versions come out monthly.
R and the old Base SAS?
R can be executed using PROC IML in base SAS. This presentation discussed how to use the R API with SAS Viya to submit CAS actions.
Can Python users achieve the same types of integration with SAS?
Everything I showed you today, whether it was SAS Studio, Model Studio or the APIs, you can do the exact same thing with Python. There are two additional packages Python users can leverage. One is DLP and the other is sasoptpy. These packages essentially live on top of the swap package and help you more efficiently create your deep learning or optimization models.
Can you register open-source models in SAS Model Manager?
Absolutely. You saw me register an Rda File along with supporting JSON files. You could do the exact same thing for Python for the pickle file. So you can move local pickle files and local RDS files over to SAS Model Manager and you can govern and deploy them exactly as you would if it was a SAS model. Model Manager can deploy your local Python models. You don't even have to really touch SAS at that point. You don't have to use CAS actions if you don't want to. Of course, you don't get the benefit of the scalability, but you can deploy RDS files and pickle files to say a container or wherever you want when you go ahead and deploy it from Model Manager.
How does the speed of the action sets compare to open-source algorithms?
SAS R&D has put a tremendous amount of effort into memory allocation and speed. We just had a third-party run a very robust experiment, I think it was 1500 different technical experimentations. They ran CAS actions against other open-source platforms. On average, we found that the CAS actions run 30 times faster in memory than Spark. Because it runs so much faster, we saw about an 86% cost savings in the cloud because of it. The CAS actions are extremely fast and that's why we strongly encourage people to leverage these APIs. It's really simple to pick up if you're an open source coder and you can scale your data. It's extremely advantageous to use the swat package.
Where is the source R library package pulled from? Is it pulled from local? Or from an online source?
I'm assuming you're talking about the swat package. The swat package was created by SAS. It is technically an open-source package. It's free for anyone to download. Of course, it's not useful unless you have a SAS file license, so you'll just be downloading that like you would in R with any other package. There is a nice SAS GitHub page where you can look at more details on the swat package and it also has more information on downloading that for clarification.
The question was asking about the R package from Model Manager. So, in Model Manager, it's a little bit more work. Essentially, you need to connect an R instance or executable to SAS. There needs to be some communication there. typically, your organization’s IT professional, can download packages for you on the SAS Viya platform.
... View more