As we enter the holiday season, we are reminded of the importance of coming together and spending time with those we care about. Our families and friends may come from different backgrounds and have different preferences, but during the holidays we come together and put aside our differences.
Much like families and friends gathering during the holidays, in analytics and model building data scientists and analysts may come from different backgrounds and have different preferences when it comes to model building. What becomes important here is that these users have a gathering place where they can collaborate and work together. SAS Model Manager in SAS Viya provides a centralized repository where data scientists can register models that they have built using different languages and tools.
For data scientists developing models in Python, there exists a handy package that makes it easier to work with SAS Model Manager from a Python interface. The sasctl package simplifies the model management process by simplifying the process of importing Python models into the common model repository.
To work with SAS Model Manager from our Python interface, we must first establish a session.
from sasctl import Session
with Session(host, username, password):
#Here is where we can specify if we want to register our model. We will show this in the next step.
Once we have developed our model, we can register our python model with only a few lines of code.
Register a pure Python model in Model Manager:
from sasctl import Session, register_model
from sklearn import linear_model as lm
with Session(host, authinfo=<authinfo file>):
model = lm.LogisticRegression()
register_model(model, 'Sklearn Model', 'My Project')
When we have successfully registered our Python models in SAS Model Manager, we can easily compare our python model with our other models registered in the same project, as well as govern, deploy, and monitor our models.
Happy holidays and happy modeling ❄️ 🎄 ☃️
If you are interested in learning more, additional information can be found at the links below:
