Skip to content

etna 2.10.0

Latest
Latest
Compare
Choose a tag to compare
@d-a-bunin d-a-bunin released this 09 Jan 14:46
· 9 commits to master since this release
2.10.0
297a66f
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.

Highlights

  • Add foundational models: ChornosModel, ChronosBoltModel, TimesFMModel
  • Add ability to handle missing values by adding parameter missing_mode into metrics
  • Optimize memory usage in TFTNativeModel, DeepStateModel, DeepARNativeModel
  • Fix working with NaN target in MeanEncoderTransform
  • Fix target leakage in MeanSegmentEncoderTransform
  • Add handling scikit-learn version >= 1.4 in OneHotEncoderTransform and HierarchicalClustering

Full changelog

Added

  • Add load_dataset to public API (#484)
  • Add example of using custom pipeline pools in Auto (#504)
  • Add MetricWithMissingHandling base class, change signature of etna.metrics.Metric to return None values (#514)
  • Add ChronosModel (#511)
  • Add ChronosBoltModel (#511)
  • Add usage example of ChronosModel and ChronosBoltModel in 202-NN_examples notebook (#511)
  • Add TimesFMModel (#544)
  • Add usage example of TimesFMModel in 202-NN_examples notebook (#544)
  • Add MissingCounter metric (#520)

Changed

  • Add docstring warning about handling non-regressors (including target) to children of WindowStatisticsTransform (#474)
  • Add parameter missing_mode into MSE metric (#515)
  • Add parameter missing_mode into MAE metric (#523)
  • Add parameter missing_mode into MAPE and SMAPE metrics (#524)
  • Add parameter missing_mode into Sign, WAPE and MaxDeviation metrics (#530)
  • Add parameter missing_mode into Coverage and Width metrics (#541)
  • Update aggregate_metrics_df to work with None values (#522)
  • Rework validation of FoldMask to not fail on tail nans (#536)
  • Add parameter missing_mode into R2 and MedAE metrics (#537)
  • Update analysis.forecast.plots.plot_metric_per_segment to handle None from metrics (#540)
  • Add parameter missing_mode into RMSE and MSLE metrics (#542)
  • Update analysis.forecast.plots.metric_per_segment_distribution_plot to handle None from metrics (#543)
  • Add example on using custom params_to_tune in Tune (#547)

Fixed

  • Fix working with embedding_sizes in 202-NN_examples notebook (#489)
  • Disallow dropping target in TSDataset.drop_features (#491)
  • Optimize memory usage in TFTNativeModel by eliminating copying during making samples (#494)
  • Optimize memory usage in DeepStateModel and DeepARNativeModel by eliminating copying during making samples (#499)
  • Fix working with NaN target in MeanEncoderTransform (#492)
  • Fix target leakage in MeanSegmentEncoderTransform (#503)
  • Add handling scikit-learn version >= 1.4 in OneHotEncoderTransform and HierarchicalClustering (#529)
Assets 2
Loading