Download Latest Version v1.2.0 source code.zip (25.7 MB) Google Add to Preferred Sources
Home / v1.2.0
Name Modified Size InfoDownloads / Week
Parent folder
README.md 2026-09-22 12.2 kB
0
v1.2.0 source code.tar.gz 2026-09-22 24.2 MB
0
v1.2.0 source code.zip 2026-09-22 25.7 MB
0
Totals: 3 Items   49.9 MB 0

Feature release.

Users should also note completed deprecations and consult notes below.

Please see our changelog for a description of all changes.

Highlights

  • better memory management: smaller sktime package size, smaller estimator memory footprint (#10459, [#10892]) @faakhir30, @fkiraly
  • pandas 3 support (#11056, [#11153]) @yash-sangwan
  • New foundation model forecasters:
    • Tafsut (#11053) @aryamanDutta
    • TimesFM3 (#10976) @hasanfaesal
    • TiRex-2 (#10786) @yash-sangwan
    • TFC-T0 (#10447) @siddharth7113
  • Recursive reduction using Monte-Carlo to probabilistic tabular regressors, MCRecursiveProbaReductionForecaster (#9242) @marrov
  • CRAN fable::ARIMA forecaster (via rpy2) (#8641) @ericjb
  • interface to hypertrees-forecasting: HyperTreeARForecaster (#11076) @oberoir080
  • forecasting evaluate utility now can run benchmarks for pretrain-able forecasters (#10560) @jgyasu
  • new transformations: SavitzkyGolayTransformer, HilbertTransformer, and WaveletPacketTransformer (#10810) @ved197338
  • EventTPR, MeanDetectionOffset and FalseAlarmRate for live event detection (#11197) @yash-sangwan
  • Hierarchical/agglomerative clustering for time series (#9187) @Muhammad-Rebaal

Core interface changes

Registry and tags

  • The craft utility for deserializing estimator specifications now has a safe argument which prevents arbitrary serialized code in the estimator definition. Only a restricted syntax with estimator names in sktime is allowed.

On-board libraries

  • The tbats package is no longer maintained and has been vendored into sktime.libs. The sktime vendor is also updated for compatibility with newest numpy and scikit-learn.

Deprecations and removals

  • From version 1.3.0, forecasters will no longer universally provide "refitting " update capability by default and as a consequence no longer store all data seen so far in self._y and/or self._X. Forecasters with an available "update" mode can now be queried via the capability:update tag. To replicate the former default behaviour of "memorize all data and refit" in an upwards compatible way, users can use the UpdateRefitsEvery compositor from sktime.forecasting.stream. In order to test future post-1.3.0 behaviour, users can set the config remember_data to False, via the .set_config(remember_data=False) (prior to forecaster use).

  • In line with the forecaster update default change, from version 1.3.0, self._y and self._X will no longer store all data seen so far by default. Users who wish to retain the previous behavior in an upwards compatible way can set the config remember_data to True via the .set_config(remember_data=True) (prior to forecaster use).

  • Deprecation of the capability:global_forecasting tag is complete, it has been removed entirely. Users should use the capability:pretrain tag instead.

All Contributors

@8rulerstar, @AanchalGupta1162, @AbdulAliMamnun, @AbiramiR-27, @adan-shahid, @adity1raut, @AH64-dll, @AlejandroCoronadoN, @AlvaroBalbin, @amarjaleelbanbhan, @Anai-Guo, @andoriyaprashant, @anupamkr1708, @archittmittal, @aryamanDutta, @Ashish-Kumar-Dash, @Ayushagrawal-cse, @baremetaldevx86, @Belagum, @blackorange-mega, @Boubker10, @chala2001, @Chetansahney, @chintam-dhanush, @chrisprown902-wq, @CloseChoice, @coding-cosmos, @dannymaaz, @DebD-max, @DebojitNath, @dhairya-motta, @dhruvb2028, @direkkakkar319-ops, @dongwonmoon, @ericjb, @Faakhir30, @fathirramadhan-web, @felipebridge, @fkiraly, @fnhirwa, @gabrielemidulla, @geetu040, @godarrenw, @goyaladitya05, @gthay, @Harshavardhan-28, @hasanfaesal, @hdimer, @HuzaifaAbdulRehman, @InnoxCodes, @JamesBoardman27, @jbbqqf, @jgyasu, @jvsch, @kajal-jotwani, @Keykyrios, @KingLizard1020, @LouisDeconinck, @LunarScoop, @luziyi123448-gif, @maarcosrmz, @marrov, @mccool1010, @Me-Priyank, @MGPOCKY, @Muhammad-Rebaal, @NAME-ASHWANIYADAV, @nicolasdmolina, @Nikunjsaini07, @Nischal1425, @oberoir080, @OfficialAbhinavSingh, @Partharsid, @PiyushKumar74110, @Pragati5-DEBUG, @pruthvipatill, @Ridadata, @RobKuebler, @sabasiddique1, @Saswatsusmoy, @Sharon-study, @shivamlalakiya, @shubham5080, @siddharth7113, @Siddhartha23i, @snoopuppy582, @Spicy-source, @srupat, @sudo-muneeb, @tarundb2005-create, @TayfurYldz, @Tharun-10Dragneel, @Udayan853, @Valentino-source-dev, @ved197338, @vedantag17, @WAHIB-EL-KHADIRI, @webzuweb, @wunianze666-netizen, @YadavAkash96, @yash-sangwan, @zelihaguven

New Contributors

Full Changelog: https://github.com/sktime/sktime/compare/v1.1.0...v1.2.0

Source: README.md, updated 2026-09-22