Model saving, loading, and prediction: Worked Example — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)

Model Saving, Loading, and Prediction: A Worked Example for Accelerated Data Science

In the context of MLOps practices for the NVIDIA-Certified Associate: Accelerated Data Science certification, managing models efficiently is crucial. This includes saving trained models, loading them for inference, and performing predictions in production or testing environments. Below is a detailed, step-by-step worked example demonstrating these tasks using a realistic scenario.

Scenario

You have trained a GPU-accelerated machine learning model using scikit-learn on a dataset to predict customer churn. After training, you want to save the model, load it later for inference, and run predictions on new customer data.

Step 1: Train the Model

Assume you have already trained a model. For illustration:

from sklearn.ensemble import RandomForestClassifier from sklearn.datasets import make_classification

X, y = make_classification(n_samples=1000, n_features=20, random_state=42) model = RandomForestClassifier(n_estimators=100, random_state=42) model.fit(X, y)

This model is trained on synthetic data representing customer features.

Step 2: Save the Model

Use joblib or pickle to serialize the trained model. Joblib is preferred for large numpy arrays common in models.

import joblib

model_filename = 'customer_churn_model.joblib' joblib.dump(model, model_filename)

This step creates a file customer_churn_model.joblib that stores the model state.

Step 3: Load the Model

When you need to perform predictions later, load the model from disk:

loaded_model = joblib.load(model_filename)

This restores the model to the exact state it was when saved.

Step 4: Prepare New Data for Prediction

Prepare new customer data in the same feature format as training data. For example:

import numpy as np

new_customer = np.array([[0.5, -1.2, 0.3, 0.7, -0.8, 1.1, 0.0, 0.2, -0.5, 1.3, 0.4, -0.6, 0.9, -1.0, 0.1, 0.3, -0.7, 0.8, 0.6, -0.2]])

Step 5: Predict Using the Loaded Model

Run prediction on the new data:

prediction = loaded_model.predict(new_customer) print(f'Predicted churn class: {prediction[0]}')

This outputs the predicted class (e.g., 0 for no churn, 1 for churn).

Step 6: Save and Load Model with GPU-Accelerated Frameworks (Optional)

If using GPU-accelerated libraries such as cuML from RAPIDS, the saving and loading process is similar but may require specific serialization methods. For example, cuML models can be saved using pickle or their own save_model methods.

Summary

• Saving: Serialize the trained model to disk using joblib or pickle.
• Loading: Deserialize the model when needed for inference.
• Prediction: Use the loaded model to predict on new data formatted consistently with training data.

This workflow ensures reproducibility and efficient deployment of GPU-accelerated machine learning models, a foundational MLOps practice for the NVIDIA certification.

More in this topic

Benchmarking workflows and selecting hardware: Worked Example — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring production models for drift: Quick Reference — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Benchmarking workflows and selecting hardware: Common Mistakes — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Model saving, loading, and prediction — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Model saving, loading, and prediction: Common Mistakes — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Model saving, loading, and prediction: Quick Reference — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Introductory MLOps Practices — NVIDIA-Certified Associate: Accelerated Data ScienceBenchmarking workflows and selecting hardware: Practice Questions — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Experiment tracking with MLflow and Weights & Biases — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Model saving, loading, and prediction: Practice Questions — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring production models for drift: Practice Questions — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring and optimizing ML pipelines: Practice Questions — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring and optimizing ML pipelines: Common Mistakes — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring and optimizing ML pipelines: Worked Example — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring and optimizing ML pipelines: Quick Reference — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring production models for drift: Common Mistakes — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring and optimizing ML pipelines — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring production models for drift — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Benchmarking workflows and selecting hardware — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Monitoring production models for drift: Worked Example — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)Benchmarking workflows and selecting hardware: Quick Reference — Introductory MLOps Practices (NVIDIA-Certified Associate: Accelerated Data Science)