Registry / ai-ml / hummingbird-ml

hummingbird-ml

library0.4.12pypypi✓ verified 83d ago

Hummingbird is a Microsoft-maintained library that converts trained traditional ML models (scikit-learn, LightGBM, XGBoost, PyTorch, ONNX) into tensor computations (PyTorch, TorchScript, ONNX, TVM). It enables GPU acceleration and scaling for legacy models with minimal code changes. Current version 0.4.12, release cadence is irregular (~quarterly). Requires Python >=3.8.

ai-ml

pip install hummingbird-ml

github ↗homepage ↗

INSTALL

IMPORT

SIG · HUMMINGBIRD-ML

hummingbird-ml

ai-mlpythonv0.4.12

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

convert

✓ from hummingbird.ml import convert

✗ from hummingbird import convert

The top-level module is 'hummingbird.ml', not 'hummingbird'.

Hummingbird

✓ from hummingbird.ml import Hummingbird

Convert a scikit-learn RandomForest to PyTorch backend and run inference.

from hummingbird.ml import convert from sklearn.ensemble import RandomForestClassifier import numpy as np # Train a simple model model = RandomForestClassifier(n_estimators=10) X = np.random.rand(100, 4) y = np.random.randint(0, 2, 100) model.fit(X, y) # Convert to PyTorch pytorch_model = convert(model, backend='pytorch') # Make a prediction pytorch_model.predict(X[:5])

Debug

Known issues

gotchaThe `convert` function expects a trained model object, not an untrained estimator. Calling `convert` on an unfitted model will raise an error.

fix

Ensure the model is fitted (trained) before passing to `convert`.

affects: all

gotchaWhen using `backend='pyodide'` (for WebAssembly), the converted model may have different numerical precision, leading to slight prediction differences.

fix

Test predictions for consistency; consider using `backend='onnx'` for cross-platform stability.

affects: >=0.4.0

deprecatedThe `extra_config` parameter in `convert` is deprecated and will be removed in a future version.

fix

Use `convert` without `extra_config`; customizations may be passed via model's `__init__`.

affects: >=0.4.0

breakingSupport for Python 3.6/3.7 was dropped in version 0.4.0. Using older Python will cause import errors.

fix

Upgrade to Python 3.8 or higher.

affects: >=0.4.0

Errors

Common errors & fixes

ModuleNotFoundError: No module named 'hummingbird.ml'

Package not installed or import path wrong.

fix

Run `pip install hummingbird-ml` and use `from hummingbird.ml import convert`.

AttributeError: 'RandomForestClassifier' object has no attribute 'tree_'

Model is not trained (fitted) before conversion.

fix

Call `.fit(X, y)` on the model before passing to `convert()`.

RuntimeError: Cannot convert this model type to the requested backend

The model type or backend combination is not supported.

fix

Check official docs for supported model-backend pairs. Common: scikit-learn -> pytorch or onnx.

TypeError: __init__() got an unexpected keyword argument 'extra_config'

Using deprecated `extra_config` parameter which has been removed.

fix

Remove `extra_config`; use alternative configuration if needed.

Upgrade

Version history

0.4.12latest on PyPI · released Oct 25, 2024

Audit

Dependencies

torchrequiredRequired for conversion to PyTorch backend; most common use case.

scikit-learnoptionalRequired for loading scikit-learn models; often needed for input.

lightgbmoptionalOptional: needed if converting LightGBM models.

xgboostoptionalOptional: needed if converting XGBoost models.

Agent activity

24 hits · last 30 days

node

Perplexity

Resources

githubhttps://github.com/microsoft/hummingbird ↗

homepagehttps://github.com/microsoft/hummingbird ↗