This release improves compatibility with the latest ecosystem packages, including scikit-learn 1.7+, Transformers 4.26+/5.0+, pandas 3.0, and recent setuptools changes. It strengthens packaging by unblocking editable installs and migrating metadata to PEP 621, fixes CatBoost and Optuna-related test issues, and adds timestamp generation for time series predictions without test data. It also includes website dependency updates, CI/deployment maintenance, and documentation updates for LightGBM quantile regression and supported Python versions.

Welcome OpenByte (@OpenByteDev) to the community!

What's Changed

• Bump lodash from 4.17.21 to 4.17.23 in /website by Dependabot (@dependabot)[bot] in #1509
• Only run save dependencies on main branch by Li Jiang (@thinkall) in #1510
• Fix sklearn 1.7+ compatibility: BaseEstimator type detection for ensemble by @Copilot in #1512
• ci: skip pre-commit workflow on main by Li Jiang (@thinkall) in #1513
• Fix transformers API compatibility: support v4.26+ and v5.0+ with version-aware parameter selection by @Copilot in #1514
• Bump webpack from 5.94.0 to 5.105.0 in /website by Dependabot (@dependabot)[bot] in #1515
• Pin setuptools version<82 to fix pkg_resources not found error by Li Jiang (@thinkall) in #1516
• Fix CatBoost.get_params() TypeError in test_regression by @Copilot in #1520
• Bump minimatch from 3.0.5 to 10.2.1 in /website by Dependabot (@dependabot)[bot] in #1518
• Bump minimatch from 10.2.1 to 10.2.3 in /website by Dependabot (@dependabot)[bot] in #1521
• Fix test_no_optuna reinstalling optuna at wrong version by Li Jiang (@thinkall) in #1522
• Bump yaml from 1.10.2 to 1.10.3 in /website by Dependabot (@dependabot)[bot] in #1523
• Bump picomatch from 2.3.1 to 2.3.2 in /website by Dependabot (@dependabot)[bot] in #1524
• Bump brace-expansion from 5.0.3 to 5.0.5 in /website by Dependabot (@dependabot)[bot] in #1525
• Bump lodash from 4.17.23 to 4.18.1 in /website by Dependabot (@dependabot)[bot] in #1528
• Fix pandas 3.0 compatibility: StringDtype, datetime resolution, deprecated APIs, pyspark.pandas import by @Copilot in #1527
• Bump follow-redirects from 1.15.6 to 1.16.0 in /website by Dependabot (@dependabot)[bot] in #1529
• Fix invalid key in deploy-website.yml by OpenByte (@OpenByteDev) in #1532
• Bump postcss from 8.4.31 to 8.5.10 in /website by Dependabot (@dependabot)[bot] in #1533
• Bump version to 2.6.0 by Li Jiang (@thinkall) in #1534
• Document LightGBM quantile regression with FLAML by @Copilot in #1535
• Fix pyproject.toml: add build-system section to unblock editable installs by @Copilot in #1531
• Fix editable install failure: migrate packaging metadata to PEP 621 (pyproject.toml) by @Copilot in #1538
• Generate timestamps for time series predictions without test data by @Copilot in #1536
• Revise Python support version in README by Li Jiang (@thinkall) in #1539

New Contributors

• OpenByte (@OpenByteDev) made their first contribution in #1532

Full Changelog: v2.5.0...v2.6.0

This release adds Python 3.13 support, fixes several bugs in search, configuration handling, and estimator preprocessing, and improves overall robustness. It also includes major documentation updates covering metrics, estimators, best practices, and hierarchical search spaces. Most of the work was contributed by @Copilot.

What's Changed

• Fix Best Practices not shown by Li Jiang (@thinkall) in #1483
• Fix isinstance usage issues by Li Jiang (@thinkall) in #1488
• Add support to python 3.13 by Li Jiang (@thinkall) in #1486
• Add objective parameter to LGBMEstimator search space by @Copilot in #1474
• Fix BlendSearch OptunaSearch warning for non-hierarchical spaces with Ray Tune domains by @Copilot in #1477
• Fix: Preserve FLAML_sample_size in best_config_per_estimator by @Copilot in #1475
• Bump version to 2.5.0 by Li Jiang (@thinkall) in #1492
• Add comprehensive metric documentation and URL reference to AutoML docstrings by @Copilot in #1471
• Expand docs to include all flamlized estimators by @Copilot in #1472
• Fix indents in Task-Oriented-AutoML.md by Li Jiang (@thinkall) in #1493
• Fix nested dictionary merge in SearchThread losing sampled hyperparameters by @Copilot in #1494
• Add GitHub Copilot instructions for FLAML repository by @Copilot in #1502
• Fix eval_set preprocessing for XGBoost estimators with categorical features by @Copilot in #1470
• Document missing value handling behavior for AutoML estimators by @Copilot in #1473
• Document that final_estimator parameters in ensemble are not auto-tuned by @Copilot in #1499
• Add validation and clear error messages for custom_metric parameter by @Copilot in #1500
• Use scientific notation for best error in logger output by @Copilot in #1498
• Add configurable label overlap handling for classification holdout strategy by @Copilot in #1491
• Clarify period parameter and automatic label lagging in time series forecasting by @Copilot in #1495
• Expose task-level and estimator-level preprocessors as public API by @Copilot in #1497
• Add hierarchical search space documentation by @Copilot in #1496
• Update Copilot Instructions by Li Jiang (@thinkall) in #1505

Full Changelog: v2.4.1...v2.5.0

This release includes comprehensive documentation updates to reflect recent changes across the project, adds a new best practices page, introduces support for pickling the entire AutoML instance, and fixes several bugs.

What's Changed

• Update readme, enable all python versions for macos tests by Li Jiang (@thinkall) in #1478
• Add timeout for tests and remove macos test envs by Li Jiang (@thinkall) in #1479
• Support pickling the whole AutoML instance, Sync Fabric till 0d4ab16f by Li Jiang (@thinkall) in #1481
• Update documents, Bump version to 2.4.1, Sync Fabric till 088cfb98 by Li Jiang (@thinkall) in #1482

Full Changelog: v2.4.0...v2.4.1

Python 3.12 Support Added

This release introduces official support for Python 3.12 and includes several bug fixes and enhancements that improve stability, performance, and overall user experience.

Thank you Azamatkhan Arifkhanov (@Sm1Ling), Keita Onabuta (@konabuta) and all the other contributors for the PRs, issues, reviewing and testing.

Welcome Azamatkhan Arifkhanov (@Sm1Ling) and Keita Onabuta (@konabuta) to the community!

What's Changed

• Fix OSError: [Errno 24] Too many open files: 'nul' by Azamatkhan Arifkhanov (@Sm1Ling) in #1455
• Bump version to 2.3.7 by Li Jiang (@thinkall) in #1457
• update loc second argument to column by Keita Onabuta (@konabuta) in #1458
• Bump algoliasearch-helper from 3.11.1 to 3.26.0 in /website by Dependabot (@dependabot)[bot] in #1461
• Update readme, bump version to 2.4.0, fix CI errors by Li Jiang (@thinkall) in #1466
• Add support to Python 3.12, Sync Fabric till dc382961 by Li Jiang (@thinkall) in #1467
• Fix log_training_metric causing IndexError for time series models by @Copilot in #1469

New Contributors

• Azamatkhan Arifkhanov (@Sm1Ling) made their first contribution in #1455
• Keita Onabuta (@konabuta) made their first contribution in #1458
• @Copilot made their first contribution in #1469

Full Changelog: v2.3.6...v2.4.0

This release addresses a bug and enhances the documentation to provide a smoother, more user-friendly experience.

What's Changed

• Improve FAQ docs by Li Jiang (@thinkall) in #1448
• Fix TypeError of customized kfold method which needs 'y' by Li Jiang (@thinkall) in #1450
• Bump version to 2.3.6 by Li Jiang (@thinkall) in #1451
• Bump brace-expansion from 1.1.11 to 1.1.12 in /website by Dependabot (@dependabot)[bot] in #1453

Full Changelog: v2.3.5...v2.3.6

This release includes several bug fixes and enhancements aimed at improving overall user experience. Key updates include:

• Improved MLflow Logging: MLflow logging is now partially asynchronous, significantly boosting logging performance.

• New Utilities:

  • Added get_random_dataframe for generating random test data.

  • Added auto_convert_dtypes_spark and auto_convert_dtypes_pandas for automatic data type conversion.

  These utilities are now available in the flaml.automl.data module.

• Python 3.8 Deprecation: Support for Python 3.8 is officially deprecated starting with this release.

Thank you SkBlaz (@SkBlaz), Stickic-cyber, hexiang-x, murunlin and all the other contributors for the PRs, issues, reviewing and testing.

Welcome SkBlaz (@SkBlaz), Stickic-cyber, hexiang-x and murunlin to the community!

What's Changed

• Bump version to 2.3.5 by Li Jiang (@thinkall) in #1409
• Upgrade setuptools to fix pipeline errors in python 3.8 and 3.9 by Li Jiang (@thinkall) in #1410
• Improved error handling in case no scikit present by SkBlaz (@SkBlaz) in #1402
• Bump @babel/runtime-corejs3 from 7.20.1 to 7.26.10 in /website by Dependabot (@dependabot) in #1412
• Bump prismjs from 1.29.0 to 1.30.0 in /website by Dependabot (@dependabot) in #1411
• Bump @babel/runtime from 7.20.1 to 7.26.10 in /website by Dependabot (@dependabot) in #1414
• Bump @babel/helpers from 7.20.1 to 7.26.10 in /website by Dependabot (@dependabot) in #1413
• Fix issue with "list index out of range" when max_iter=1 by Stickic-cyber in #1419
• Numpy 2.x is not supported yet. by Li Jiang (@thinkall) in #1424
• Revert "Numpy 2.x is not supported yet." by Li Jiang (@thinkall) in #1426
• Bump http-proxy-middleware from 2.0.7 to 2.0.9 in /website by Dependabot (@dependabot) in #1425
• fix:When use_spark = True and mlflow_logging = True are set, an error is reported when logging the best model: 'NoneType' object has no attribute 'save' bug Something isn't working by hexiang-x in #1432
• fix: best_model_for_estimator returns inconsistent feature_importances_ compared to automl.model by murunlin in #1429
• Sync Fabric till 2cd1c3d by Li Jiang (@thinkall) in #1433
• Add the detailed info for parameter 'verbose' by murunlin in #1435
• Sync Fabric till deb6dbe by Li Jiang (@thinkall) in #1438
• Upgrade python 3.8 to 3.10 in github actions by Li Jiang (@thinkall) in #1440
• Fix CD actions by Li Jiang (@thinkall) in #1441
• Fix typo, add quotes to python-version by Li Jiang (@thinkall) in #1442
• Install wheel and setuptools by Li Jiang (@thinkall) in #1443

New Contributors

• SkBlaz (@SkBlaz) made their first contribution in #1402
• Stickic-cyber made their first contribution in #1419
• hexiang-x made their first contribution in #1432
• murunlin made their first contribution in #1429

Full Changelog: v2.3.4...v2.3.5

This release fixes some bugs and improves user experience.

Thank you Will Charles (@drwillcharles) , Daniel Grindrod (@dannycg1996) and all the other contributors for the PRs, issues, reviewing and testing.

Welcome Will Charles (@drwillcharles) to the community!

What's Changed

• Bump version to 2.3.4 by Li Jiang (@thinkall) in #1389
• Fix test hang issue by Li Jiang (@thinkall) in #1406
• Changed tune.report import for ray>=2 by Will Charles (@drwillcharles) in #1392
• Bump serialize-javascript from 6.0.0 to 6.0.2 in /website by Dependabot (@dependabot) in #1407
• fix: Fixed bug where group folds and sample weights couldn't be used together by Daniel Grindrod (@dannycg1996) in #1405
• Fix params pop error by Li Jiang (@thinkall) in #1408

New Contributors

• Will Charles (@drwillcharles) made their first contribution in #1392

Full Changelog: v2.3.3...v2.3.4

This release fixes some bugs and improves user experience.

Thank you kernelmethod (@kernelmethod) , EgorKraevTransferwise , Daniel Grindrod (@dannycg1996) and all the other contributors for the PRs, issues, reviewing and testing.

Welcome kernelmethod (@kernelmethod) to the community!

What's Changed

• Fix logging issues that can arise when importing flaml by kernelmethod (@kernelmethod) in #1377
• Bump version to 2.3.3 by Li Jiang (@thinkall) in #1378
• Bump cross-spawn from 7.0.3 to 7.0.6 in /website by Dependabot (@dependabot) in #1379
• Pass cost_attr and cost_budget from flaml.tune.run() to the search algo by EgorKraevTransferwise in #1382
• fix: Fixed bug with catboost and groups by Daniel Grindrod (@dannycg1996) in #1383
• Bump nanoid from 3.3.6 to 3.3.8 in /website by Dependabot (@dependabot) in #1387
• fix: KeyError no longer occurs when using groupfolds for regression tasks. by Daniel Grindrod (@dannycg1996) in #1385

New Contributors

• kernelmethod (@kernelmethod) made their first contribution in #1377

Full Changelog: v2.3.2...v2.3.3

This release improves the reproducibility of a bunch of models. Huge thanks to Daniel Grindrod (@dannycg1996) !

Thank you Daniel Grindrod (@dannycg1996) , smty2018, @Programmer-RD-AI and all the other contributors for the PRs, issues, reviewing and testing.

What's Changed

• fix: Cross validation process isn't always run to completion by Daniel Grindrod (@dannycg1996) in #1360
• test: Adding tests to verify model reproducibility by Daniel Grindrod (@dannycg1996) in #1362
• fix: FLAML catboost metrics arent reproducible by Daniel Grindrod (@dannycg1996) in #1364
• Included that retrain_full = True does not include the user provided validation data in the docs. #1228 by smty2018 in #1245
• Remove outdated comment by Li Jiang (@thinkall) in #1366
• Bump http-proxy-middleware from 2.0.6 to 2.0.7 in /website by Dependabot (@dependabot) in #1370
• Bump version to 2.3.2 by Li Jiang (@thinkall) in #1372
• Fix data transform issue, spark log_loss metric compute error and json dumps TypeError (Sync Fabric till 3c545e6) by Li Jiang (@thinkall) in #1371
• fix: Ensure Reproducibility of ElasticNetEstimator by Daniel Grindrod (@dannycg1996) in #1374
• fix: LinearSVC results now reproducible by Daniel Grindrod (@dannycg1996) in #1376
• Add documentation for automl.model.estimator usage by @Programmer-RD-AI in #1311
• Flaml: fix lgbm reproducibility by Daniel Grindrod (@dannycg1996) in #1369

New Contributors

• Daniel Grindrod (@dannycg1996) made their first contribution in #1360
• smty2018 made their first contribution in #1245

Full Changelog: v2.3.1...v2.3.2

Hotfix release to resolve two urgent issues.

Thank you Ian Hellen (@ianhelle) and Harutaka Kawamura (@harupy) for reporting them. Thank you Harutaka Kawamura (@harupy) and leafy-lee for reviewing PRs.

What's Changed

• Add contributors wall by Li Jiang (@thinkall) in #1341
• Add info of tutorial automl 2024 by Li Jiang (@thinkall) in #1344
• Bump micromatch from 4.0.5 to 4.0.8 in /website by Dependabot (@dependabot) in #1343
• Bump webpack from 5.76.1 to 5.94.0 in /website by Dependabot (@dependabot) in #1342
• Remove temporary pickle files by Li Jiang (@thinkall) in #1354
• Support logger_formatter without automl dependencies by Li Jiang (@thinkall) in #1356
• Bump express from 4.19.2 to 4.21.0 in /website by Dependabot (@dependabot) in #1357

Full Changelog: v2.3.0...v2.3.1