Installation¶

Stable release¶

Install from PyPI (recommended):

pip install lazypredict

Install from conda-forge:

conda install -c conda-forge lazypredict

Core dependencies¶

The base install includes: scikit-learn, pandas, numpy, tqdm, joblib, click.

Optional extras¶

Lazy Predict has several optional dependency groups. Install them with pip extras:

Boosting libraries¶

XGBoost, LightGBM, and CatBoost (all support GPU acceleration):

pip install lazypredict[boost]

On macOS, XGBoost and LightGBM also need the OpenMP runtime:

brew install libomp

Without it, lazypredict warns that the library could not be imported and skips its models.

Categorical encoders¶

Target and binary encoding (categorical_encoder="target" / "binary") via category-encoders:

pip install lazypredict[encoders]

Time series forecasting¶

Statistical models (Holt-Winters, SARIMAX, Theta):

pip install lazypredict[timeseries]

Add deep learning models (LSTM, GRU via PyTorch):

pip install lazypredict[timeseries,deeplearning]

Add the Google TimesFM foundation model (installs timesfm[torch]):

pip install lazypredict[timeseries,foundation]

MLflow experiment tracking¶

pip install lazypredict[mlflow]

Activate tracking by setting an environment variable before running:

import os
os.environ["MLFLOW_TRACKING_URI"] = "sqlite:///mlflow.db"

Hyperparameter tuning¶

pip install lazypredict[tune]      # Optuna backend
pip install lazypredict[flaml]     # FLAML backend

The flaml extra installs flaml[automl], which the FLAML backend needs. Its dependency constraints (at the time of writing, xgboost<3) apply to your environment as well.

Enable with tune=True in any Lazy estimator:

clf = LazyClassifier(tune=True, tune_top_k=5, tune_trials=50)

Visualization¶

pip install lazypredict[viz]

Required for LazyForecaster.plot_results() and all plotting utilities.

Explainability¶

pip install lazypredict[explain]     # SHAP values
pip install lazypredict[interpret]   # InterpretML (EBM)

Distributed / Spark¶

pip install lazypredict[spark]       # PySpark MLlib

All optional dependencies¶

pip install lazypredict[all]

GPU acceleration¶

GPU acceleration requires an NVIDIA GPU with CUDA support:

pip install lazypredict[boost]   # XGBoost, LightGBM, CatBoost (all support GPU)
pip install cuml-cu12            # cuML (RAPIDS) GPU-native sklearn replacements

Enable with use_gpu=True:

clf = LazyClassifier(use_gpu=True)
models, predictions = clf.fit(X_train, X_test, y_train, y_test)

Supported GPU backends:

Backend	Package	Parameter
XGBoost	`xgboost`	`device="cuda"`
LightGBM	`lightgbm`	`device="gpu"`
CatBoost	`catboost`	`task_type="GPU"`
cuML	`cuml-cu12`	GPU-native (no extra params)
LSTM / GRU	`torch` with CUDA	CUDA device
TimesFM	`timesfm`	CUDA device

All GPU backends fall back to CPU automatically if CUDA is not available.

Intel CPU acceleration¶

For improved performance on Intel CPUs:

pip install scikit-learn-intelex

Patch scikit-learn before importing lazypredict, so that the accelerated estimators are the ones lazypredict collects:

from sklearnex import patch_sklearn
patch_sklearn()

from lazypredict.Supervised import LazyClassifier

Changed in 0.4.0

Earlier versions called patch_sklearn() automatically on import lazypredict, which changed scikit-learn for the whole Python process. The patch is now opt-in.

Development install¶

To set up a local development environment:

git clone https://github.com/shankarpandala/lazypredict
cd lazypredict
pip install -e ".[all,dev]"

Run the tests:

pytest tests/

Installing from source (archive)¶

Download and install without git:

curl -OJL https://github.com/shankarpandala/lazypredict/tarball/master
pip install lazypredict-*.tar.gz