New features
- FEAT: TimeXer marcopeix (1267)
- All losses compatible with all types of models (e.g. univariate/multivariate, direct/recurrent) OR appropriate protection added.
- DistributionLoss now supports the use of `quantiles` in `predict`, allowing for easy quantile retrieval for all `DistributionLosses`.
- Mixture losses (GMM, PMM and NBMM) now support learned weights for weighted mixture distribution outputs.
- Mixture losses now support the use of `quantiles` in `predict`, allowing for easy quantile retrieval.
- Improved stability of `ISQF` by adding softplus protection around some parameters instead of using `.abs`.
- Unified API for any quantile or any confidence level during predict for both point and distribution losses.
Enhancements
- [DOCS] Docstrings elephaint (1279)
- FIX: Minor bug fix in TFT and a nicer error message for fitting with the wrong val_size marcopeix (1275)
- [FIX] Adds bfloat16 support elephaint (1265)
- Recurrent models can now produce forecasts recursively or directly.
- IQLoss now gives monotonic quantiles
- MASE loss now works
Breaking Changes
- [FIX] Unify API elephaint (1023)
- RMoK uses the `revin_affine` parameter instead of `revine_affine`. This was a typo in the previous version.
- All models now inherit the `BaseModel` class. This changes how we implement new models in neuralforecast.
- Recurrent models now require an `input_size` parameter.
- `TCN` and `DRNN` are now window models, not recurrent models
- We cannot load a recurrent model from a previous version to v3.0.0
Bug Fixes
- [FIX] Multivariate models give error when predicting when n_series > batch_size elephaint (1276)
- [FIX]: Insample predictions with series of varying lengths marcopeix (1246)
Documentation
- [DOCS] Update documentation elephaint (1274)
- [DOCS] Add example of modifying the default configure_optimizers() behavior (use of ReduceLROnPlateau scheduler) JQGoh (1015)