Hands-On Automated Machine Learning

Code profiling in Python

Some applications will require your machine learning models to be performant in terms of training and scoring time. For example, a recommender engine might require you to generate recommendations in less than a second and if you have more than a second latency, profiling is one way to understand intensive operations. Code profiling will help you a lot to understand how different parts of your program are executed. Profiling stats will give metrics, such as the number of calls, the total time spent to execute a function call including/excluding calls to its sub-functions, and incremental and total memory usage.

The cProfile module in Python will help you to see time statistics for every function. Here's a small example: ...