build(deps): Bump darts from 0.44.1 to 0.45.0

[dependabot]

Bumps darts from 0.44.1 to 0.45.0.

Release notes

Sourced from darts's releases.

Darts minor 0.45.0

We are pleased to announce the release of a new Darts version.

You can find a list with all changes in the release notes.

Changelog

Sourced from darts's changelog.

0.45.0 (2026-06-19)

For users of the library:

Improved

• Improvements to ShapExplainer : #3049 by Zhihao Dai and Dennis Bader.
  • 🚀🚀 Added support for explaining any TorchForecastingModel including regular torch models (TiDEModel, ...) as well as foundation models (Chronos2, ...). It supports global and local explanations and can output SHAP values for further analysis.
  • Added method explain_single() to explain a single model forecast in detail, in addition to the existing batched method explain(). This is useful for local explanations of individual predictions with reduced computational cost.
  • Added support for explaining the forecasted likelihood parameter of probabilistic forecasts. For PyTorch models: all likelihoods are supported. For scikit-learn-like models: quantile and poisson regression are supported.
  • Method summary_plot() can now also be computed on any optional foreground series using parameters foreground_series, foreground_past_covariates, foreground_future_covariates.
  • Added a new notebook for Explainability of Forecasting Models including detailed usage examples of ShapExplainer.
  • 🔴 Renamed method force_plot_from_ts() to force_plot() to simplify.
• Improvements to forecasting models:
  • 🚀 Added new forecasting model PatchTSTFMModel : IBM's pre-trained ~260M-parameter foundational model for zero-shot forecasting. It supports univariate, multivariate, and multiple time series forecasting without training and can output deterministic or probabilistic forecasts. #3120 by Dennis Bader.
  • Custom encoders now support functions that return multiple components. Simply pass such a function via the "custom" encoder key in the add_encoders model input parameter. #3069 by Moritz Waldleben.
  • Added use_longer_projection_head to TimesFM2p5Model to enable longer non-autoregressive prediction horizons (up to 1024 steps for output_chunk_length + output_chunk_shift). #3121 by Zhihao Dai.
  • Improved the documentation for capturing model uncertainty using Monte Carlo Dropout in the forecasting overiew user guide. #3117 by Jean-Baptiste Braun.
• Improvements to TimeSeries :
  • TimeSeries.from_dataframe() now supports time columns of type pl.Date for polars.DataFrame. #3124 by Dennis Bader
• Other improvements:
  • Optuna integration's PyTorchLightningPruningCallback for hyperparameter optimization of torch models is now natively available in Darts via darts.utils.callbacks. #3114 by Jakub Chłapek.

Fixed

• Fixed several bugs in ShapExplainer including mismatched SHAP method enum values, feature naming conventions, and inconsistent instance count in explain(). #3049 by Zhihao Dai.
• Fixed an issue in TFTModel where training under mixed precision resulted in float16 overflow. #3087 by Robert Ruidisch.
• Fixed a bug in TimeSeries.quantile() where the output dtype did not match the input series dtype for dtypes float32 or float16. Now the dtype is correctly propagated. #3124 by Dennis Bader

For developers of the library:

• Used GitHub Actions to publish the documentation to GitHub Pages, replacing the previous third-party branch-based deployment method. #3107 by Zhihao Dai
• Removed optuna-integration[pytorch-lightning] from the testing dependencies. #3114 by Jakub Chłapek.

Commits

• 0069263 Bump version: 0.44.1 → 0.45.0
• 8738663 fix bump-my-version version to 1.3.0 (#3139)
• 948e313 fix bump-my-version version to 1.2.5 (#3138)
• b0dc815 release 0.45.0 (#3137)
• 9268410 fix failing unit tests from core flavor (#3136)
• a6b7204 docs(userguide): clarify mc_dropout + likelihood combination (#3117)
• 6577569 fix failing unit tests from core flavor (#3135)
• c96169f [Feature] SHAP-based explainer for torch models (#3049)
• b0ccd3a fix unit test (#3127)
• a26bf1e Feature: Custom encoders returning multiple components (#3069)
• Additional commits viewable in compare view