• Classes
  • 11 - Logging

Logging

Warming up

Analyze the following code:

a = int(input())
b = int(input())

print(a / b)

Question 1

Run the code and enter the values 3 and 2. What happens?

Answer!

1.5 is printed in stdout (terminal).

Question 2

Run the code and enter the values 6 and 0. What happens?

Answer!

An exception occurs, which is displayed in the terminal.

Question 3

So the exception is also printed to stdout?

Answer

It is printed to stderr. In this case, stdout and stderr are both pointing to the terminal, but it could be different.

Run the code and check: 

import sys

# Open a file in write mode to redirect stderr
file = open("error_output.txt", "w")
sys.stderr = file

a = 5
b = 0

print(a / b)

file.close()

Now check the content of error_output.txt file.

Logging Importance

Logging refers to the practice of recording events or messages from a software application, system, or device into a file or a centralized logging system.

Logging and monitoring are fundamental pillars of DevOps principles, playing a vital role in ensuring robust ML practices. It allows developers and data scientists to track the execution flow of an ML system, identify errors, and diagnose issues.

By logging relevant information, such as intermediate results, input data, and model predictions, it becomes easier to understand the behavior of the ML system and pinpoint the source of any problems.

In today's complex production infrastructures, which often involve multiple models deployed across multiple servers, effectively monitoring a live system, with or without machine learning components, entails collecting and aggregating data about its states. This highlights the increasing importance of having a robust logging system in place.

Important!

Every successful machine learning project incorporates a data feedback loop, where valuable information from the production environment is fed back to the model prototyping environment to enhance its performance.

Advance to the next topic. Let's explore some logging options!

References