Run predict() on a loaded model

Once the model is loaded, call .predict(X) to get predictions. This is the core of inference benchmarking.

script.py v3.1

import joblib
import numpy as np
from sklearn.linear_model import LinearRegression
from sklearn.datasets import make_regression
from sklearn.model_selection import train_test_split

# Train + save a model (simulating what a user would upload)
X, y = make_regression(n_samples=200, n_features=4, noise=10, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
model = LinearRegression().fit(X_train, y_train)
joblib.dump(model, 'model.pkl')

# ── BENCHMARK ENGINE: load the model + run predictions ──
loaded_model = joblib.load('model.pkl')
print(f'Loaded model: {type(loaded_model).__name__}')

# predict() takes a 2D array — even for a single sample
# Shape: (n_samples, n_features)
print(f'X_test shape: {X_test.shape}')  # (40, 4) — 40 samples, 4 features

# Run predictions on the WHOLE test set
y_pred = loaded_model.predict(X_test)
print(f'y_pred shape: {y_pred.shape}')  # (40,) — 40 predictions
print(f'First 5 predictions: {y_pred[:5]}')
print(f'First 5 actual:      {y_test[:5]}')

# Predict on a SINGLE new sample (must be 2D!)
new_sample = np.array([[1.5, -0.5, 2.0, 0.8]])  # shape (1, 4)
single_pred = loaded_model.predict(new_sample)
print(f'\nSingle sample prediction: {single_pred[0]:.4f}')

# COMMON BUG: passing a 1D array — sklearn will crash or give wrong results
try:
    bad_pred = loaded_model.predict([1.5, -0.5, 2.0, 0.8])  # 1D!
except Exception as e:
    print(f'\n1D array error: {type(e).__name__}: {e}')

Python output click Run to execute

Click "Run" to execute the code on the left.
The output will appear here.

🧠 How it works

model.predict(X) is the core of inference. X must be 2D: (n_samples, n_features). Even for 1 sample: shape (1, 4), not (4,). Returns a 1D array: (n_samples,) — one prediction per input row. For classification: predict() returns class labels (0, 1, 2...). predict_proba() returns probabilities (n_samples, n_classes). The COMMON BUG at the bottom shows what happens with a 1D array — sklearn either crashes or reshapes it wrong.

✏️ Try these changes

Change test_size from 0.2 to 0.5
→ bigger test set → more predictions
Replace LinearRegression with RandomForestRegressor(n_estimators=50)
→ different model, same predict() call
Change new_sample values
→ different prediction
Add predict_proba (need a classifier): use make_classification instead of make_regression
→ returns probabilities instead of values

← Previous lab

Load a .pkl model file (joblib + pickle)

Next lab →

Measure training time (time.perf_counter)