Logging artifacts and models in MLOps - Commands & Configuration

Introduction

When you train machine learning models, you want to save the results and files so you can use them later or share them. Logging artifacts and models means saving these files in a safe place automatically during training.

When you want to save your trained model to use it later for predictions.

When you want to keep track of files like plots or data samples created during training.

When you want to compare different model versions by saving each one separately.

When you want to share your model and related files with your team easily.

When you want to keep a history of your experiments and their outputs.

Commands

This command installs MLflow, a tool that helps you log models and artifacts easily.

Terminal

pip install mlflow

Expected Output

Collecting mlflow

  Downloading mlflow-2.7.0-py3-none-any.whl (18.7 MB)

Installing collected packages: mlflow

Successfully installed mlflow-2.7.0

This runs a Python script that trains a model and logs the model and an artifact file using MLflow.

Terminal

python log_model.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 12:00:01 INFO mlflow.tracking.fluent: Logging model and artifact

Model and artifact logged successfully

Key Concept

If you remember nothing else from this pattern, remember: logging saves your model and files automatically so you never lose your work.

Code Example

MLOps

import mlflow
import mlflow.sklearn
from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
import joblib

# Load data
iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target, random_state=42)

# Train model
model = RandomForestClassifier(n_estimators=10, random_state=42)
model.fit(X_train, y_train)

# Save a sample artifact
joblib.dump(model, 'model.joblib')

# Start MLflow run
with mlflow.start_run():
    # Log model
    mlflow.sklearn.log_model(model, 'random_forest_model')
    # Log artifact file
    mlflow.log_artifact('model.joblib')

print('Model and artifact logged successfully')

Common Mistakes

Not calling mlflow.start_run() before logging artifacts or models

MLflow needs a run context to save logs; without it, logging commands fail silently or cause errors.

Always wrap your logging code inside mlflow.start_run() as a context manager.

Logging model files without specifying the correct path

If the path is wrong, MLflow cannot find the files to save, so nothing gets logged.

Make sure the file paths you pass to mlflow.log_artifact() or mlflow.sklearn.log_model() are correct and files exist.

Summary

Install MLflow to enable logging of models and artifacts.

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

Log your trained model and any files you want to keep using MLflow functions.

This process helps you save and track your machine learning work automatically.