How to Log Parameters in MLflow: Simple Guide

To log parameters in MLflow, use mlflow.log_param(key, value) inside an active MLflow run. This records key-value pairs like hyperparameters so you can track them alongside your model and metrics.

Syntax

The basic syntax to log a parameter in MLflow is mlflow.log_param(key, value).

• key: A string representing the parameter name.
• value: The parameter value, usually a string, int, or float.
• This function must be called inside an active MLflow run context.

python

import mlflow

with mlflow.start_run():
    mlflow.log_param("learning_rate", 0.01)

Example

This example shows how to start an MLflow run and log multiple parameters such as learning rate and number of trees. It demonstrates how parameters are saved for experiment tracking.

python

import mlflow

with mlflow.start_run():
    mlflow.log_param("learning_rate", 0.1)
    mlflow.log_param("num_trees", 100)
    print("Parameters logged successfully.")

Output

Parameters logged successfully.

Common Pitfalls

Common mistakes when logging parameters in MLflow include:

• Calling mlflow.log_param outside an active run, which causes an error.
• Logging non-serializable objects as parameter values.
• Using the same key multiple times in one run, which overwrites the previous value.

Always ensure you start a run with mlflow.start_run() before logging.

python

import mlflow

# Wrong way: logging outside a run
try:
    mlflow.log_param("batch_size", 32)
except Exception as e:
    print(f"Error: {e}")

# Right way:
with mlflow.start_run():
    mlflow.log_param("batch_size", 32)
    print("Logged batch_size inside run.")

Output

Error: No active run found. Please start a run before logging.
Logged batch_size inside run.

Quick Reference

Function	Description	Example
mlflow.start_run()	Starts an MLflow run context	with mlflow.start_run():
mlflow.log_param(key, value)	Logs a single parameter	mlflow.log_param('lr', 0.01)
mlflow.log_params(dict)	Logs multiple parameters at once	mlflow.log_params({'lr': 0.01, 'epochs': 10})
mlflow.end_run()	Ends the active MLflow run	mlflow.end_run()

Key Takeaways

Always log parameters inside an active MLflow run using mlflow.start_run().

Use mlflow.log_param(key, value) to record each parameter clearly.

Avoid logging complex objects; stick to simple types like strings, ints, or floats.

Repeated keys overwrite previous values within the same run.

Use mlflow.log_params() to log multiple parameters efficiently.