v3.2.1

New features

• [FEAT] Categorical feature support for distributed workflows @marcopeix (#1581)
• [FEAT] Support categorical features for explainability @marcopeix (#1575)
• [FEAT] Support categorical features for simulation paths @marcopeix (#1574)
• [FEAT] Support categorical features for multivariate and recurrent models @marcopeix (#1573)

Bug fixes

• [FIX] Schaake shuffle @elephaint (#1586)
• [FIX] Initialize static scalers in distributed prediction workers @ShreyanshGoyal (#1579)
• [FIX] Support DBFS paths in distributed partition listing @vikas-prasad-cx (#1576)
• [FIX] Set default gpu=1 for Ray @marcopeix (#1572)
• [FIX] Fix loss computation in multi-device training @marcopeix (#1571)
• [FIX] Fix AutoHINT ray serialization issue @marcopeix (#1570)
• [FIX] Use per-series static features in predict_insample @Vedant-Agarwal (#1568)

v3.2.0

New features

• [FEAT] Support categorical exog for univariate models @marcopeix (#1552)
• [FEAT] Add FreDF loss combining time-domain MSE and frequency-domain MAE @DresdenGman (#1563)
• [FEAT] Add GELU as a grn_activation option in TFT @autokaiai (#1560)
• [FEAT] Remove cpus and gpus in favor of ray_options @marcopeix (#1544)

Bugfixes

• [FIX] Multi-device training for multivariate models @marcopeix (#1557)
• [FIX] Translate input_size_multiplier in user-supplied Auto configs @Vedant-Agarwal (#1553)
• [FIX] Further fixes for CV with no df @marcopeix (#1548)
• [FIX] Allow CV without passing a df @marcopeix (#1545)
• [FIX] Prevent a second optimization study when using conformal predictions @marcopeix (#1539)
• [FIX] Raise error when missing params in auto config @marcopeix (#1540)

Documentation

• [DOC] Add xLSTM to docs @marcopeix (#1543)
• [DOC] Add tutorial to convert nf models to onnx @marcopeix (#1542)

v3.1.9

New features

• [FEAT] Add SOFTSSharp model @hljubic (#1516)
• [FEAT] Add ray scheduler and refactor optuna and ray kwargs @marcopeix (#1531)
• [FEAT] Transfer learning with cross-validation @marcopeix (#1530)
• [FEAT] Add Optuna create_study_kwargs @marcopeix (#1528)
• [FEAT] Add study kwargs for Optuna @marcopeix (#1527)
• [FEAT] Can pass run_config with Ray @marcopeix (#1525)

Documentation

• [DOC] Updates for v.3.1.9 @marcopeix (#1533)
• [DOC] Fix image paths @deven367 (#1517)

Bug fixes

• [FIX] Get rid of Lightning tips and deprecation messages @elephaint (#1532)
• [FIX] ResetEarlyStopping state in cross-validation when refit=True @marcopeix (#1529)
• [FIX] Upgrade idna to fix vulnerability @marcopeix (#1526)
• [FIX] Return self so we don't lose HINT wrapper and preds are reconciled @marcopeix (#1524)

v3.1.8

Changes

• [CHORE] version bump for release @deven367 (#1514)
• [FIX] Patch for pytorch-lightning @deven367 (#1513)
• [CHORE] missed fixes @deven367 (#1505)

Dependencies

• Bump the ci-dependencies group across 1 directory with 3 updates @dependabot[bot] (#1504)
• Bump gitpython from 3.1.45 to 3.1.47 @dependabot[bot] (#1511)
• Bump ray from 2.54.0 to 2.55.0 @dependabot[bot] (#1509)
• Bump mako from 1.3.10 to 1.3.11 @dependabot[bot] (#1508)
• Bump pytest from 8.4.2 to 9.0.3 @dependabot[bot] (#1507)
• Bump pillow from 12.1.1 to 12.2.0 @dependabot[bot] (#1506)

v3.1.7

Features

• [FEAT] Add sample_weight capability @marcopeix (#1501)
• [FEAT] Pass a separate validation dataframe @marcopeix (#1498)
• [FEAT] Time budget for Auto models @marcopeix (#1493)

Bug fixes

• [FIX] Add scipy runtime dependency @chandraveshchaudhari (#1496)
• [FIX] Raise error for short series @marcopeix (#1492)

Documentation

• [DOC] Fix remaining broken links @deven367 (#1499)
• [DOC] Fix predictive maintenance tutorial @marcopeix (#1495)

v3.1.6

New Features

• [FEAT] Cache trainer for faster inference @marcopeix (#1478)
• [FEAT] Choose monitor for validation @marcopeix (#1488)
• [FEAT] Simulation paths @elephaint (#1483)

Bugfixes

• [chore] fix links @deven367 (#1476)
• [FIX] Error when saving model with custom callbacks @marcopeix (#1487)
• [FIX] Refit scalers for transfer learning @marcopeix (#1486)
• [FIX] Spark to optional-dependencies @nasaul (#1481)
• [FIX] Use the common MLP as decoder in DeepAR @marcopeix (#1472)
• [FIX] Deprecate the losses.numpy module @marcopeix (#1471)
• [FIX] Mismtached quantiles in loss and valid_loss @marcopeix (#1470)
• [FIX] Add XLinear to MODEL_FILENAME_DICT for save/load support @ph3on1x (#1469)

v3.1.5

New features

• [FEAT] Explainability for multivariate models @marcopeix (#1464)
• [FEAT] Add XLinear @marcopeix (#1445)

Fixes

• [FIX] Add protection against duplicated level and quantiles in distribution losses @marcopeix (#1448)
• [FIX] Inverse scaling with MASE loss @marcopeix (#1447)
• [FIX] Exogenous support in TimeXer @marcopeix (#1444)
• [FIX] Correct min_samples formula in conformal prediction for step_size @W057 (#1461)
• [FIX] Fix save() to prevent FileExistsError with DDP @W057 (#1460)

Documentation

• [DOC] update title @deven367 (#1459)

v3.1.4

Changes

• [CHORE] Version bump @marcopeix (#1437)
• [CHORE]: fix requirements to build whl @deven367 (#1436)

v3.1.3

Features

• [FEAT] Support normalization for static exogenous features @JuditHalperin (#1406)
• [FEAT] Add support for historical and static exogenous in TimeXer @marcopeix (#1404)

Bug fixes

• [FIX] Prevent inverse normalization of quantile validation loss @marcopeix (#1432)
• [FIX] Fix broken links @deven367 (#1433)
• [FIX] Fix type of n_windows in cross_validation @Riklia (#1378)
• [FIX] Apply windows batch size @marcopeix @elephaint (#1429)

Documentation

• Updates contributing for building documentation @nasaul (#1430)
• Improve grammar and readability in documentation @rogeliomj (#1423)
• Add explanability to the sidebar @deven367 (#1428)
• Fixed missing import @ngupta23 (#1425)
• Migrate to docs.json @deven367 (#1422)
• Lazydocs to mkdocstrings @deven367 (#1399)
• Fix table render @deven367 (#1409)

Changes

• [CHORE] Drop Python 3.9, add Python 3.13 @marcopeix (#1434)

Many thanks to @nasaul for his work on the CI! We usually don't list these PRs in the release notes, but thanks to his work, tests run better for all contributors.

v3.1.2

Features

• [FEAT] Any horizon explanation @marcopeix (#1393)

Bug fixes

• [FIX] Remove torch fix @elephaint (#1392)