Model Serialization (pickle)¶

This notebook shows how to save and reload a trained DeepGBoost model using Python's standard pickle module.

We cover:

1. Training a regressor and a classifier
2. Saving models to disk with pickle.dump
3. Loading them back with pickle.load
4. Verifying that predictions are identical before and after serialization

In [1]:

# If running outside of the repo, install with:
# pip install deepgboost
#
# If running from the repo root:
# pip install -e '.[dev]'
import deepgboost

print("DeepGBoost version:", deepgboost.__version__)

# If running outside of the repo, install with: # pip install deepgboost # # If running from the repo root: # pip install -e '.[dev]' import deepgboost print("DeepGBoost version:", deepgboost.__version__)

DeepGBoost version: 0.1.0

In [2]:

import pickle
import tempfile
from pathlib import Path

import numpy as np
from sklearn.datasets import load_diabetes, load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.metrics import r2_score, accuracy_score

from deepgboost import DeepGBoostRegressor, DeepGBoostClassifier

import pickle import tempfile from pathlib import Path import numpy as np from sklearn.datasets import load_diabetes, load_breast_cancer from sklearn.model_selection import train_test_split from sklearn.metrics import r2_score, accuracy_score from deepgboost import DeepGBoostRegressor, DeepGBoostClassifier

---

1. Regression¶

In [3]:

X, y = load_diabetes(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

reg = DeepGBoostRegressor(
    n_trees=10,
    n_layers=15,
    max_depth=4,
    learning_rate=0.1,
    random_state=42,
)
reg.fit(X_train, y_train)

preds_before = reg.predict(X_test)
print(f"R² before saving: {r2_score(y_test, preds_before):.4f}")

X, y = load_diabetes(return_X_y=True) X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42 ) reg = DeepGBoostRegressor( n_trees=10, n_layers=15, max_depth=4, learning_rate=0.1, random_state=42, ) reg.fit(X_train, y_train) preds_before = reg.predict(X_test) print(f"R² before saving: {r2_score(y_test, preds_before):.4f}")

R² before saving: 0.4645

Save to disk¶

In [4]:

model_path = Path(tempfile.gettempdir()) / "deepgboost_regressor.pkl"

with open(model_path, "wb") as f:
    pickle.dump(reg, f)

print(f"Model saved to: {model_path}")
print(f"File size: {model_path.stat().st_size / 1024:.1f} KB")

model_path = Path(tempfile.gettempdir()) / "deepgboost_regressor.pkl" with open(model_path, "wb") as f: pickle.dump(reg, f) print(f"Model saved to: {model_path}") print(f"File size: {model_path.stat().st_size / 1024:.1f} KB")

Model saved to: /tmp/deepgboost_regressor.pkl
File size: 347.3 KB

Load from disk and verify predictions¶

In [5]:

with open(model_path, "rb") as f:
    reg_loaded = pickle.load(f)

preds_after = reg_loaded.predict(X_test)
print(f"R² after loading: {r2_score(y_test, preds_after):.4f}")
print(f"Predictions identical: {np.allclose(preds_before, preds_after)}")

with open(model_path, "rb") as f: reg_loaded = pickle.load(f) preds_after = reg_loaded.predict(X_test) print(f"R² after loading: {r2_score(y_test, preds_after):.4f}") print(f"Predictions identical: {np.allclose(preds_before, preds_after)}")

R² after loading: 0.4645
Predictions identical: True

---

2. Classification¶

In [6]:

X, y = load_breast_cancer(return_X_y=True)
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

clf = DeepGBoostClassifier(
    n_trees=5,
    n_layers=10,
    max_depth=4,
    learning_rate=0.15,
    random_state=42,
)
clf.fit(X_train, y_train)

preds_before = clf.predict(X_test)
print(f"Accuracy before saving: {accuracy_score(y_test, preds_before):.4f}")

X, y = load_breast_cancer(return_X_y=True) X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.2, random_state=42 ) clf = DeepGBoostClassifier( n_trees=5, n_layers=10, max_depth=4, learning_rate=0.15, random_state=42, ) clf.fit(X_train, y_train) preds_before = clf.predict(X_test) print(f"Accuracy before saving: {accuracy_score(y_test, preds_before):.4f}")

Accuracy before saving: 0.9649

Save to disk¶

In [7]:

model_path = Path(tempfile.gettempdir()) / "deepgboost_classifier.pkl"

with open(model_path, "wb") as f:
    pickle.dump(clf, f)

print(f"Model saved to: {model_path}")
print(f"File size: {model_path.stat().st_size / 1024:.1f} KB")

model_path = Path(tempfile.gettempdir()) / "deepgboost_classifier.pkl" with open(model_path, "wb") as f: pickle.dump(clf, f) print(f"Model saved to: {model_path}") print(f"File size: {model_path.stat().st_size / 1024:.1f} KB")

Model saved to: /tmp/deepgboost_classifier.pkl
File size: 103.0 KB

Load from disk and verify predictions¶

In [8]:

with open(model_path, "rb") as f:
    clf_loaded = pickle.load(f)

preds_after = clf_loaded.predict(X_test)
print(f"Accuracy after loading: {accuracy_score(y_test, preds_after):.4f}")
print(f"Predictions identical: {np.array_equal(preds_before, preds_after)}")

with open(model_path, "rb") as f: clf_loaded = pickle.load(f) preds_after = clf_loaded.predict(X_test) print(f"Accuracy after loading: {accuracy_score(y_test, preds_after):.4f}") print(f"Predictions identical: {np.array_equal(preds_before, preds_after)}")

Accuracy after loading: 0.9649
Predictions identical: True

---

Note: Only load pickle files from sources you trust. Pickle does not provide any security guarantees — a malicious .pkl file can execute arbitrary code when loaded.