This release improves correctness across AutoML, forecasting, and tuning. It fixes quantized integer sampling bounds, singleton integer search-space normalization, low-cost search-space processing, and independence of suggested configurations from the portfolio. It strengthens time-series evaluation with corrected cross-validation coverage, panel time indices, naive and seasonal-naive forecasts, and R² scoring. Data preparation and metrics improvements include OrdinalEncoder-based categorical encoding, preservation of missing datetimes for imputation, weighted MAPE, clearer errors for binary-only metrics on multiclass data, and support for scalar ROUGE results. It also improves training-log appends and Spark forest prediction, updates pandas compatibility and Windows ARM64 CI, and refreshes documentation and dependencies.
Thank you to all contributors for the PRs, issues, reviewing and testing.
Welcome @raashish1601, @Linxiushen, @anishmehta24, @ktz03, @student-shivamm, @sarankumar1325, and @nicoloangileri to the community!
What's Changed
- fix: qrandint/qlograndint never sample the documented inclusive upper bound by @AmirF194 in #1606
- refactor: use sklearn OrdinalEncoder as source of truth for DataTransformer categorical encoding (#1564) by @immu4989 in #1569
- build(deps): bump codecov/codecov-action from 7.0.0 to 7.1.0 in the github-actions group by @dependabot[bot] in #1609
- fix: raise a clear error for binary-only metrics on multiclass data by @raashish1601 in #1621
- build(deps): bump codecov/codecov-action from 7.1.0 to 7.1.1 in the github-actions group by @dependabot[bot] in #1617
- build(deps): bump brace-expansion from 5.0.9 to 5.0.12 in /website by @dependabot[bot] in #1618
- fix: pass sample_weight to the mape metric by @raashish1601 in #1619
- build(deps): bump shell-quote from 1.10.0 to 1.12.0 in /website by @dependabot[bot] in #1623
- build(deps): bump source-map-js from 1.2.1 to 1.2.2 in /website by @dependabot[bot] in #1624
- build(deps): bump joi from 17.13.7 to 17.13.8 in /website by @dependabot[bot] in #1625
- build(deps): bump fast-uri from 3.1.7 to 3.1.8 in /website by @dependabot[bot] in #1626
- build(deps): bump proxy-addr from 2.0.7 to 2.0.8 in /website by @dependabot[bot] in #1627
- build(deps): bump compression from 1.7.4 to 1.8.2 in /website by @dependabot[bot] in #1628
- build(deps): bump hydra-core from 1.3.4 to 1.3.7 in /test/pipeline_tuning_example by @dependabot[bot] in #1629
- fix(automl): include latest observation in time-series CV by @Linxiushen in #1616
- Keep PyArrow compatible with the test NumPy constraint by @thinkall with @Copilot in #1638
- build(deps): bump actions/setup-node from 7.0.0 to 7.1.0 in the github-actions group by @dependabot[bot] in #1635
- Fix Windows ARM64 CI installation by excluding PyArrow by @thinkall with @Copilot in #1639
- fix: keep processing the search space after a list low-cost point by @anishmehta24 in #1637
- fix: keep suggested configs independent of the portfolio by @ktz03 in #1632
- Fix ROUGE loss adaptation for scalar Evaluate results by @student-shivamm in #1633
- build(deps): bump follow-redirects from 1.16.0 to 1.16.1 in /website by @dependabot[bot] in #1645
- Fix normalization of singleton integer search spaces by @Linxiushen in #1644
- build(deps): bump fast-uri from 3.1.8 to 3.1.9 in /website by @dependabot[bot] in #1649
- Preserve missing datetimes for median imputation by @Linxiushen in #1650
- Avoid pandas 3.1 deprecations in tabular preprocessing by @Linxiushen in #1648
- Document the real default of index_col in train_test_split_pyspark by @VenishPaneliya in #1646
- docs: fix dead links in docs, tutorials and tune README by @sarankumar1325 in #1651
- fix: compute time_idx in steps for non-monthly panel data by @anishmehta24 in #1636
- fix(automl): safely append training logs by @Linxiushen in #1615
- fix: make naive and snaive forecasts match their definitions by @nicoloangileri in #1611
- build(deps): bump joi from 17.13.8 to 17.13.9 in /website by @dependabot[bot] in #1653
- fix: prevent zero-worker forest prediction in Spark tuning by @thinkall in #1657
- Fix reversed R² inputs in time-series estimator scores by @Linxiushen in #1654
- Upgrade version to 2.8.0 by @thinkall in #1658
New Contributors
- @raashish1601 made their first contribution in #1621
- @Linxiushen made their first contribution in #1616
- @anishmehta24 made their first contribution in #1637
- @ktz03 made their first contribution in #1632
- @student-shivamm made their first contribution in #1633
- @sarankumar1325 made their first contribution in #1651
- @nicoloangileri made their first contribution in #1611
Full Changelog: v2.7.0...v2.8.0