Model Persistence

Save and load trained OpenBoost models.

Basic Save/Load

import openboost as ob

# Train
model = ob.GradientBoosting(n_trees=100)
model.fit(X_train, y_train)

# Save
model.save('my_model.joblib')

# Load
loaded_model = ob.GradientBoosting.load('my_model.joblib')
predictions = loaded_model.predict(X_test)

Supported Models

All models support save/load:

• GradientBoosting
• MultiClassGradientBoosting
• DART
• OpenBoostGAM
• NaturalBoostNormal, NaturalBoostGamma, etc.
• LinearLeafGBDT

Using joblib/pickle Directly

import joblib

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

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

Or with pickle:

import pickle

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

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

Cross-Platform Loading

Models trained on GPU can be loaded on CPU machines:

# Train on GPU
model = ob.GradientBoosting(n_trees=100)
model.fit(X_train, y_train)  # Using CUDA
model.save('model.joblib')

# Load on CPU machine
loaded = ob.GradientBoosting.load('model.joblib')
predictions = loaded.predict(X_test)  # Works on CPU

Versioning

Include version information when saving:

import openboost as ob

metadata = {
    'openboost_version': ob.__version__,
    'model_type': 'GradientBoosting',
    'training_date': '2026-01-20',
}

joblib.dump({'model': model, 'metadata': metadata}, 'model_with_meta.joblib')

Model Checkpointing

Save best models during training:

from openboost import ModelCheckpoint

model.fit(
    X_train, y_train,
    eval_set=[(X_val, y_val)],
    callbacks=[
        ModelCheckpoint(
            filepath='checkpoints/best_model.joblib',
            save_best_only=True,
        )
    ],
)

File Formats

Format	Extension	Pros	Cons
joblib	.joblib	Fast, compressed	Python only
pickle	.pkl	Standard library	Slower, larger

Recommended: Use joblib for best performance.