A DLPy sequential model is saved using model.save_to_astore(path='models/', name='fraud_dlpy_v1'). What is the purpose of saving to ASTORE format?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — saving to ASTORE gives you a SAS-native binary model you can score with PROC ASTORE, SAS Micro Analytic Service, or SAS Model Manager without Python or CAS. That's the point: deployment inside SAS, not compression or export.
Full Explanation
The correct purpose is that saving to ASTORE enables the DLPy model to be scored in SAS environments without a Python or CAS dependency, through PROC ASTORE, SAS Micro Analytic Service, or SAS Model Manager. ASTORE, short for Analytic Store, encapsulates the model in a SAS-native binary format, making it the primary SAS deployment format for DLPy models. The claim that ASTORE compresses model weights for a smaller disk footprint compared to HDF5 is wrong; compression is not the purpose. The claim that ASTORE enables transfer learning by preserving only convolutional layer weights is wrong; ASTORE saves the full model, not just convolutional layers. The claim that ASTORE is required to export DLPy models to ONNX format for deployment outside SAS is wrong; ONNX export uses a different workflow.