Applying our learning
Let’s use what we have learned to productionalize our models.
Technical requirements
Here are the technical requirements needed to complete the hands-on examples in this chapter:
- On-demand features require the use of DBR ML 13.1 or higher.
- RAG and CV parts require DBR ML 14.2 and higher.
- Python UDFs are created and governed in UC; hence, Unity Catalog must be enabled for the workspace – no shared clusters.
- The Streaming Transactions project uses
scikit-learn==1.4.0rc1. The notebooks that need it install it. - The Streaming Transactions project, again, performs better with parallel compute. We’ll use the multi-node cluster from Chapter 5. See Figure 7.6 for the multi-node CPU configuration:
Figure 7.6 – Multi-node CPU cluster configuration (on AWS)
Project – Favorita Sales forecasting
In this chapter, we discussed using managed MLflow and the UC Model Registry...
