# What changed in flaml

> 26 changes in flaml 2 from its official release notes, upgrade guides and advisories: 5 new APIs and options. Each one cited. Sources checked 2026-10-19. HTML: https://unstale.sh/pypi/flaml/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## flaml 2

### New APIs and options

- Adds an optional resampler keyword argument for per-fold class-imbalance handling. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Generates timestamps for time series predictions when test data is not provided. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Adds configurable label overlap handling for the classification holdout strategy. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- The LGBMEstimator search space adds the objective parameter. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- Task-level and estimator-level preprocessors are exposed as a public API. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)

### Fixes and other changes

- Preserves anonymous tuning metrics whose value is zero. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Categorical codes are fixed at fit time, and unseen values trigger a warning. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Float sampling now honors unit quantization. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Handles mixed-type conditional parameters. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Forwards free_mem_ratio from compute_estimator to the evaluation. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Fixes crashes and misaligned sample weights when preparing data with non-shuffled or group holdout splits. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- SGD, logistic regression, random forest, and XGBoost estimators now honor random_seed for reproducible results. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Supports recent scikit-learn changes, including LRL1/LRL2 deprecations and fitted-state detection when reloading MLflow autologged models. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Naive forecasters support statsmodels 0.15, and future forecast horizons are corrected. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Trial averages are computed from per-metric report counts. (2.7.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.7.0)
- Fixes compatibility with pandas 3.0, including StringDtype, datetime resolution, deprecated APIs, and pyspark.pandas imports. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Adds a build-system section and migrates packaging metadata to PEP 621 to unblock editable installs. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Fixes BaseEstimator type detection for ensembles to support scikit-learn 1.7+. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Pins setuptools below version 82 to prevent a pkg_resources not found error. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Selects parameters according to the Transformers version to support v4.26+ and v5.0+. (2.6.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.6.0)
- Preserves FLAML_sample_size in best_config_per_estimator. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- Fixes an OptunaSearch warning for non-hierarchical spaces with Ray Tune domains. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- Adds validation and clear error messages for the custom_metric parameter. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- Fixes eval_set preprocessing for XGBoost estimators when categorical features are present. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- FLAML adds support for Python 3.13. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)
- Fixes nested dictionary merging that could lose sampled hyperparameters. (2.5.0) — [release notes](https://github.com/microsoft/FLAML/releases/tag/v2.5.0)