Logging

Loguru is a library that simplifies logging and has useful features.

Log levels:

• debug: Detailed information useful during development

• info: High level operation of the code

• warning: Might be an issue

• error: Problem that needs attention

• critical: Fatal error

loguru

from loguru import logger

def fit_model(model, params):
    logger.debug(f"Using params: {params}")
    logger.info(f'Starting training.')

    if params['learning_rate'] is None:
        logger.warning("learning_rate param not specified, using 3e-4.")

    if hasattr(model, 'fit'):
        model.fit()
    else:
        logger.error("The model provided does not have a fit() method.")

Logging to a File

By default, Loguru will output to the console (stdout). The logger.add() method is use to save logs to a file.

logger.add("file.log"): Save log data to “file.log” path. Parameters of note:

• retention pamameter controls how long we keep logs. E.g. retention="1 HOUR"

• rotation: Rotates log file when it exceeds a certain size. E.g. rotation="10 MB"

Function Decoratoring:

@logger.catch allows us to automatically log the function’s entry, exit, and exceptions. Parameters:

• reraise: If set to True, the decorator will re-raise the exception after logging it. This allows the program to continue handling the error as it normally would.

• level: Specify the log level (e.g., “DEBUG”, “ERROR”, etc.) at which the exception should be logged.

• onerror: Provide a fallback function to execute when an exception occurs.

from loguru import logger

@logger.catch(onerror=lambda _: print("An error occurred, but we are moving on."))
def run_error():
    return 1 / 0

# run_error()  # This will raise ZeroDivisionError and log it