FLAML
MicrosoftMicrosoft's fast AutoML and hyperparameter-tuning library. Given training data and a task type, its AutoML class searches across learners such as LightGBM, XGBoost, CatBoost, random forests, linear models and time-series models and their hyperparameters within a time budget, using cost-aware search methods, and can stack the results. Its tune module runs the same search over any user function.
Openness
5 high confidence- license
- MIT(OSI)
- source
- public(github.com/microsoft/FLAML)
- core features withheld
- no — Microsoft Fabric's AutoML runs it as a separate product
MIT-licensed and built from the public repository. Microsoft uses it inside the AutoML feature of its Fabric data platform, which is a separate paid product; nothing is held back from the library.
- https://pypi.org/pypi/flaml/json recorded 2026-09-27
PyPI record for flaml 2.7.0, uploaded 2026-09-18, license MIT, Homepage github.com/microsoft/FLAML.
- https://raw.githubusercontent.com/microsoft/FLAML/main/LICENSE recorded 2026-09-27
LICENSE is the "MIT License", "Copyright (c) Microsoft Corporation."
- https://raw.githubusercontent.com/microsoft/FLAML/main/README.md recorded 2026-09-27
README: "FLAML supports AutoML and Hyperparameter Tuning in Microsoft Fabric Data Science."; no paid edition of the library mentioned.
- https://ungh.cc/repos/microsoft/FLAML/files/main recorded 2026-09-27
Repository tree of 463 paths with no ee, enterprise, commercial or pro directory.
Adoption
3 high confidenceMeasured on monthly PyPI downloads of the flaml package.
- https://pypistats.org/api/packages/flaml/recent recorded 2026-09-27
last_month 324,343 downloads of flaml
Capability
5 medium confidenceGiven only the training data and the task, FLAML chooses the learner and its settings itself, which is the automation AutoGluon provides. Unlike AutoGluon it bundles no pretrained tabular model, and its search is tuned for low compute cost.
- https://raw.githubusercontent.com/microsoft/FLAML/main/website/docs/Use-Cases/Task-Oriented-AutoML.md recorded 2026-09-27
Task-Oriented AutoML: "flaml.AutoML is a class for task-oriented AutoML ... The minimal inputs from users are the training data and the task type."; "Two optional inputs are time_budget and max_iter for searching models and hyperparameters."; built-in estimators lgbm, xgboost, xgb_limitdepth, rf, extra_tree, histgb, lrl1, lrl2, catboost, kneighbor, prophet, arima, sarimax, holt-winters, transformer, tft, tcn.
Verified 2026-09-27