Loading
MindXchange
MindXchange
MindXchange

Quipoin Menu

Learn • Practice • Grow

Tutorial Exercises MCQs Interview Q & A Tests
machine-learning / Model Persistence
tutorial
VideosDiscussion

Model Persistence

After training a model, you want to save it to disk and load it later without retraining. This is called model persistence. The two most common methods are `pickle` and `joblib`.

Using `joblib` (Recommended for scikit‑learn)

import joblib

# Save model
joblib.dump(model, 'model.joblib')

# Load model
loaded_model = joblib.load('model.joblib')

# Predict with loaded model
predictions = loaded_model.predict(X_test)

Using `pickle` (Python built‑in)

import pickle

# Save model
with open('model.pkl', 'wb') as f:
    pickle.dump(model, f)

# Load model
with open('model.pkl', 'rb') as f:
    loaded_model = pickle.load(f)

Why Save Models?

  • Avoid retraining (time‑consuming).
  • Deploy models in production applications.
  • Share models with team members.
  • Keep a baseline for future comparison.

Best Practices

  • Save preprocessors (scalers, encoders) together with the model.
  • Use `joblib` for large NumPy arrays (more efficient).
  • Version control model files? Usually no (large files). Use DVC or cloud storage.


Two Minute Drill
  • Save models with `joblib.dump()` or `pickle.dump()`.
  • Load with `joblib.load()` or `pickle.load()`.
  • `joblib` is preferred for scikit‑learn models.
  • Save preprocessors together with the model.

Need more clarification?

Drop us an email at career@quipoinfotech.com