Tsai

- support to MPS backend. Both the MPS accelerator and the PyTorch backend are still experimental. As such, not all operations are currently supported.
- compatibility with torch 2.2
- ability to pass arch_config to multimodal models

New Features

- added Hydra and HydraMultiRocket archs ([800](https://github.com/timeseriesAI/tsai/issues/800))

Bugs Squashed

- UCR Dataset download link has been updated ([827](https://github.com/timeseriesAI/tsai/issues/827))

- mWDNPlus now supports multidimensional outputs ([802](https://github.com/timeseriesAI/tsai/issues/802))

- Fixed import issues with demo code in Readme.MD ([798](https://github.com/timeseriesAI/tsai/issues/798))

New Features

- added functionality to support inputs with static/ observed (time-dependent) features

- added functionality to support inputs with categorical/ continuous features

- added functionality to apply patches to time series models

- Added `MultiRocket`/ `MultiRocketPlus` architectures

- added `TSSelfDropout` ([790](https://github.com/timeseriesAI/tsai/issues/790))

- added `get_feat_idxs` to calculate multimodal indices ([789](https://github.com/timeseriesAI/tsai/issues/789))

- remaining features assigned to o_cont_idxs by default ([788](https://github.com/timeseriesAI/tsai/issues/788))

- added patch encoder to `MultiInputWrapper` ([787](https://github.com/timeseriesAI/tsai/issues/787))

- added `TSTargetEncoder` transform ([769](https://github.com/timeseriesAI/tsai/issues/769))

- added `TSRobustScaler` to tfm pipelines ([763](https://github.com/timeseriesAI/tsai/issues/763))

- added new tfms - `TSDropIfTrueCols` and ApplyFunc ([760](https://github.com/timeseriesAI/tsai/issues/760))

- tensor slices in different devices when using `TensorSplitter` ([799](https://github.com/timeseriesAI/tsai/issues/799))

Bugs Squashed

- mixed augmentations (`MixUp1d`, `CutMix1d`,..) are not updating labels ([791](https://github.com/timeseriesAI/tsai/issues/791))

- `get_UCR_data` function fails due to changed download link ([785](https://github.com/timeseriesAI/tsai/issues/785))

- error when using `TSSelectColumns` due to pandas df slicing ([762](https://github.com/timeseriesAI/tsai/issues/762))

- short arrays create issues when running `get_usable_idxs` ([761](https://github.com/timeseriesAI/tsai/issues/761))

- `get_X_pred` creates different probablities when using numpy array or torch tensor ([754](https://github.com/timeseriesAI/tsai/issues/754))

- `partial_n` is applied to all datasets by default ([748](https://github.com/timeseriesAI/tsai/issues/748))

- `get_best_dls_params` function still prints output when the verbose parameter is set to false ([737](https://github.com/timeseriesAI/tsai/issues/737))

- using xresnet for vision classification raises an error ([728](https://github.com/timeseriesAI/tsai/issues/728))

New Features

- added optional activation to get_X_preds ([715](https://github.com/timeseriesAI/tsai/issues/715))

- added external vocab option to dls ([705](https://github.com/timeseriesAI/tsai/issues/705))

- allow classification outputs with n dimensions ([704](https://github.com/timeseriesAI/tsai/issues/704))

- added get_sweep_config to wandb module ([687](https://github.com/timeseriesAI/tsai/issues/687))

- added functionality to run pipeline sweeps ([686](https://github.com/timeseriesAI/tsai/issues/686))

- added seed to learners to make training reproducible ([685](https://github.com/timeseriesAI/tsai/issues/685))

- added functionality to filter df for required forecasting dates ([679](https://github.com/timeseriesAI/tsai/issues/679))

- added option to train model on train only ([671](https://github.com/timeseriesAI/tsai/issues/671))

Bugs Squashed

- access all available dataloaders in dls ([724](https://github.com/timeseriesAI/tsai/issues/724))

- make all models ending in Plus work with ndim classification targets ([719](https://github.com/timeseriesAI/tsai/issues/719))

- make all models ending in Plus work with ndim work with ndim regression/ forecasting targets ([718](https://github.com/timeseriesAI/tsai/issues/718))

- added MiniRocket to get_arch ([717](https://github.com/timeseriesAI/tsai/issues/717))

- fixed issue with get_arch missing new models ([709](https://github.com/timeseriesAI/tsai/issues/709))

- valid_metrics causes an error when using TSLearners ([708](https://github.com/timeseriesAI/tsai/issues/708))

- valid_metrics are not shown when an array is passed within splits ([707](https://github.com/timeseriesAI/tsai/issues/707))

- TSDatasets w/o tfms and inplace=False creates new X ([695](https://github.com/timeseriesAI/tsai/issues/695))

- Prediction and True Values Swapped in plot_forecast (utils.py) ([690](https://github.com/timeseriesAI/tsai/issues/690))

- MiniRocket incompatible with latest scikit-learn version ([677](https://github.com/timeseriesAI/tsai/issues/677))

- Df2xy causing incorrect splits ([666](https://github.com/timeseriesAI/tsai/issues/666))

- Feature Importance & Step Importance Not working ([647](https://github.com/timeseriesAI/tsai/issues/647))

- multi-horizon forecasting ([591](https://github.com/timeseriesAI/tsai/issues/591))

- Issues saving models with TSMetaDataset Dataloader ([317](https://github.com/timeseriesAI/tsai/issues/317))

Breaking Changes

- removed default transforms from TSClassifier, TSRegressor and TSForecaster ([665](https://github.com/timeseriesAI/tsai/issues/665))

New Features

- add option to pass an instantiated model to TSLearners ([650](https://github.com/timeseriesAI/tsai/issues/650))

- Added PatchTST model to tsai ([638](https://github.com/timeseriesAI/tsai/issues/638))

- Added new long-term time series forecasting tutorial notebook

Bugs Squashed

- Undefined variable ([662](https://github.com/timeseriesAI/tsai/issues/662))

- Multivariate Regression and Forecasting basic tutorials throw an error ([629](https://github.com/timeseriesAI/tsai/issues/629))

- TypeError: __init__() got an unexpected keyword argument 'custom_head' ([597](https://github.com/timeseriesAI/tsai/issues/597))

- Issues with TSMultiLabelClassification ([533](https://github.com/timeseriesAI/tsai/issues/533))

- Incompatible errors or missing functions in 'tutorial_nbs' notebooks, please fix. ([447](https://github.com/timeseriesAI/tsai/issues/447))

- Saving models with TSUnwindowedDataset Dataloaders: AttributeError: 'TSUnwindowedDataset' object has no attribute 'new_empty' ([215](https://github.com/timeseriesAI/tsai/issues/215))

New Features

- compatibility with Pytorch 1.13 ([619](https://github.com/timeseriesAI/tsai/issues/619))

- added sel_vars to get_robustscale_params ([610](https://github.com/timeseriesAI/tsai/issues/610))

- added sel_steps to TSRandom2Value ([607](https://github.com/timeseriesAI/tsai/issues/607))

- new walk forward cross-validation in tsai ([582](https://github.com/timeseriesAI/tsai/issues/582))

Bugs Squashed

- fixed issue when printing an empty dataset wo transforms NoTfmLists ([622](https://github.com/timeseriesAI/tsai/issues/622))

- fixed minor issue in get_robustscaler params with sel_vars ([615](https://github.com/timeseriesAI/tsai/issues/615))

- fixed issue when using tsai in dev with VSCode ([614](https://github.com/timeseriesAI/tsai/issues/614))

- issue when using lists as sel_vars and sel_steps in TSRandom2Value ([612](https://github.com/timeseriesAI/tsai/issues/612))

- fixed issue with feature_importance and step_importance when using metrics ([609](https://github.com/timeseriesAI/tsai/issues/609))

- renamed data processing tfms feature_idxs as sel_vars for consistency ([608](https://github.com/timeseriesAI/tsai/issues/608))

- fixed issue when importing 'GatedTabTransformer' ([536](https://github.com/timeseriesAI/tsai/issues/536))