Logging parameters and metrics in MLOps - Commands & Configuration

Introduction

When you train a machine learning model, you want to keep track of what settings you used and how well the model performed. Logging parameters and metrics helps you remember these details so you can compare different runs and pick the best model.

When you want to save the learning rate and number of training steps used in a model run.

When you want to record the accuracy or loss of a model after training to see how well it did.

When you want to compare different model versions by their performance numbers.

When you want to keep a history of experiments to avoid repeating the same mistakes.

When you want to share your model results with teammates in a clear, organized way.

Commands

This command installs MLflow, a tool that helps you log parameters and metrics easily during machine learning experiments.

Terminal

pip install mlflow

Expected Output

Collecting mlflow

  Downloading mlflow-2.6.0-py3-none-any.whl (18.3 MB)

Installing collected packages: mlflow

Successfully installed mlflow-2.6.0

This runs a Python script that logs parameters and metrics using MLflow. It shows how to save settings and results for a model training run.

Terminal

python log_params_metrics.py

Expected Output

2024/06/01 12:00:00 INFO mlflow.tracking.fluent: Experiment with name 'Default' does not exist. Creating a new experiment.

2024/06/01 12:00:00 INFO mlflow.tracking.fluent: Experiment created with ID '1'

2024/06/01 12:00:01 INFO mlflow.tracking.fluent: Run with ID '1234567890abcdef' started

2024/06/01 12:00:01 INFO mlflow.tracking.fluent: Logged parameter: learning_rate = 0.01

2024/06/01 12:00:01 INFO mlflow.tracking.fluent: Logged parameter: epochs = 10

2024/06/01 12:00:01 INFO mlflow.tracking.fluent: Logged metric: accuracy = 0.85

2024/06/01 12:00:01 INFO mlflow.tracking.fluent: Run with ID '1234567890abcdef' ended

Key Concept

If you remember nothing else from this pattern, remember: logging parameters and metrics lets you track what you tried and how well it worked so you can improve your models.

Code Example

MLOps

import mlflow

with mlflow.start_run():
    mlflow.log_param("learning_rate", 0.01)
    mlflow.log_param("epochs", 10)
    mlflow.log_metric("accuracy", 0.85)
print("Parameters and metrics logged successfully.")

Common Mistakes

Not starting an MLflow run before logging parameters and metrics.

Without starting a run, MLflow does not know where to save the data, so logging calls fail or data is lost.

Always use mlflow.start_run() as a context manager before logging parameters and metrics.

Logging parameters or metrics with wrong data types, like passing a list instead of a number or string.

MLflow expects parameters as strings or numbers and metrics as numbers; wrong types cause errors.

Convert parameters to strings and metrics to floats or ints before logging.

Summary

Install MLflow to enable logging of parameters and metrics.

Use mlflow.start_run() to begin a logging session.

Log parameters with mlflow.log_param() and metrics with mlflow.log_metric() inside the run.

This process helps track experiment settings and results for better model comparison.