| 65b85f5 | Julien Herzen | 13 April 2022, 19:12:33 UTC | change github-tag-action version | 13 April 2022, 19:12:33 UTC |
| a9be91c | Julien Herzen | 13 April 2022, 19:11:01 UTC | attempt at fixing unsafe repo issue | 13 April 2022, 19:11:01 UTC |
| 045901e | Julien Herzen | 13 April 2022, 15:41:36 UTC | Release 0.19.0 (#912) * bump u8darts * update readme * update CHANGELOG | 13 April 2022, 15:41:36 UTC |
| 7a6355c | gian | 13 April 2022, 14:58:13 UTC | Fix/linear quantile interpolation (#899) * quantile sampling now linearliy interpolates between fitted quantiles * Changed sampling behaviour of regression models. Now models correctly produce sample paths. Quantile sampling changed, now uses linear interpolations between values * made sample method private * moved code back from _prepare_prediction * added type hinting * fixed docstrings, removed redundant code Co-authored-by: Julien Herzen <julien@unit8.co> | 13 April 2022, 14:58:13 UTC |
| 60fcb3b | Julien Herzen | 13 April 2022, 11:14:38 UTC | Feat/improve user guide (#905) * started TimeSeries doc * progress on time series doc * update time series guide * updated timeseries user guide * small modifs * add forecasting overview * change a bit torch user guide * improvements to user guide * add FAQ * added TCN quantile image * some improvements * some improvements * link user guide from readme | 13 April 2022, 11:14:38 UTC |
| 99b649e | Julien Herzen | 13 April 2022, 11:12:37 UTC | Feat/nhits (#898) * a first version of N-HiTS * improved comments * added unit tests * Update darts/models/forecasting/nhits.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/nhits.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/nhits.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/nhits.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * improved unit tests from PR suggestions * fix linting * Update darts/models/forecasting/nhits.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 13 April 2022, 11:12:37 UTC |
| 84801df | Julien Herzen | 13 April 2022, 11:10:57 UTC | Feat/statsforecasts (#893) * added AutoARIMASF * added Croston model * add statsforecasts to requirements * add models to init file * some updates to the models * test models * small improvements * add the AutoARIMASF file * rename AutoARIMASF to StatsForecastAutoARIMA * add statsforecasts with >= requirement * add sba and tsb flavours of Croston * adap unit test * catch an issue with statsforecast sba * correct typo * correct bug 2 sigma -> 1 sigma * rephrase error message | 13 April 2022, 11:10:57 UTC |
| 2fd9d44 | Julien Herzen | 12 April 2022, 19:17:24 UTC | update readme to include Gitter (#908) | 12 April 2022, 19:17:24 UTC |
| 9a182ce | The Gitter Badger | 12 April 2022, 19:08:38 UTC | Add Gitter badge (#901) | 12 April 2022, 19:08:38 UTC |
| 76fcf53 | Julien Herzen | 08 April 2022, 15:15:43 UTC | added transfer learning example notebook (#885) * added transfer learning example notebook * small update to transfer learning notebook * list the notebook in examples list * add example to list of examples * revert change to userguide.rst * remove a few imports * fixed typos * fix typo | 08 April 2022, 15:15:43 UTC |
| efa955a | Julien Herzen | 05 April 2022, 11:35:24 UTC | Fix/ptl1.6.0 (#888) * fix epochs trained count * save PTL module and trainer using PTL checkpointing * dynamically compute right number of epochs trained * test checkpoint file existence * restore model saving in tests | 05 April 2022, 11:35:24 UTC |
| fb5a59e | Tomas Van Pottelbergh | 05 April 2022, 09:36:09 UTC | Update black to 22.3.0 (#886) | 05 April 2022, 09:36:09 UTC |
| a3066fa | Julien Herzen | 25 March 2022, 06:36:20 UTC | Fix/link gpu guide in doc (#873) * link GPU/TPU usage guide * typo | 25 March 2022, 06:36:20 UTC |
| d5dd296 | hrzn | 22 March 2022, 16:24:56 UTC | Release 0.18.0 | 22 March 2022, 16:24:56 UTC |
| 1be416f | Julien Herzen | 22 March 2022, 15:25:26 UTC | Release 0.18.0 (#867) * bump u8darts * update contribution guidelines * update changelog * separate install page * update readme | 22 March 2022, 15:25:26 UTC |
| b4bd2b4 | gian | 21 March 2022, 18:53:18 UTC | Fix/check torch import (#866) * checks if torch is available before running tests * now all imports happen after torch check * now all imports happen after torch check | 21 March 2022, 18:53:18 UTC |
| 3e4394d | gian | 21 March 2022, 16:16:52 UTC | Feat/improve seasonal decompose (#862) * added STL decompose method, added tests * fixed documentation * updated tests and documentation * updated tests * now raises an error instad of a warning if method is incompatible with seasonality mode * typo Co-authored-by: Julien Herzen <julien@unit8.co> | 21 March 2022, 16:16:52 UTC |
| c43b17b | gian | 21 March 2022, 15:43:27 UTC | checks if torch is available before running tests (#864) | 21 March 2022, 15:43:27 UTC |
| 490863f | gian | 21 March 2022, 13:58:50 UTC | Fix/quantile-multiple-ts-prediction (#853) * fixed issue with probabilistic RegressionModels not working on multiple series * Apply suggestions from code review Co-authored-by: Julien Herzen <julien@unit8.co> Co-authored-by: Julien Herzen <julien@unit8.co> | 21 March 2022, 13:58:50 UTC |
| eda8f94 | Julien Herzen | 19 March 2022, 20:40:44 UTC | Feat/more losses (#845) * added some loss functions * remove M3 loss which seems bogus * add unit tests for losses * correct unit test * correct unit test * correct unit test * added NINF case * add optional denominator computation in MAPE loss * better MAPE/MAE split * simplify MAPE loss * simplify loss tests * remove a print statement * Update darts/utils/losses.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 19 March 2022, 20:40:44 UTC |
| a7abedf | Dennis Bader | 19 March 2022, 14:33:15 UTC | fixed wrong future covariates slicing with RangeIndex (#858) | 19 March 2022, 14:33:15 UTC |
| 0233756 | gian | 11 March 2022, 06:57:58 UTC | Feat/probabilistic-lgbm-linreg (#831) * moved from forked branch * added likelihoods to linear model, updated tests, refactored lgbm quantile regression * fixed docstring * fixed test case * Apply suggestions from code review Co-authored-by: Julien Herzen <julien@unit8.co> * applied suggestions from code review, adjusted test and moved _check_quantiles to utils/utils.py * removed unnecessary enumerate * inserted else statement for clarity Co-authored-by: Julien Herzen <julien@unit8.co> | 11 March 2022, 06:57:58 UTC |
| 9cb3fd7 | Julien Herzen | 10 March 2022, 10:27:29 UTC | Fix issues with tbats (#838) | 10 March 2022, 10:27:29 UTC |
| 16bece0 | Julien Herzen | 09 March 2022, 16:38:18 UTC | Feat/tbats (#816) * Base version of TBATS * Add both BATS and TBATS * import both bats and tbats * fix an issue * added unit tests for bats and tbats probabilistic * add accuracy unit tests * Remove useless lines * Improve doc * Small fix * Add BATS/TBATS to pmdarima flavour * better frequency support * Update darts/models/forecasting/tbats.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/tbats.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * some PR comments * address PR comments Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 09 March 2022, 16:38:18 UTC |
| d48f265 | amadejkocbek | 09 March 2022, 08:23:02 UTC | Feat/backtest sliding window (#775) * Drafted implementation of sliding window * Add some unit tests for sliding window * Address comments and complete docstring. * Format logging statement * Formatting * Reformetted + min train length check * Reformat min train length method Co-authored-by: Amadej KOCBEK <amadej.kocbek@firmenich.com> Co-authored-by: Julien Herzen <julien@unit8.co> | 09 March 2022, 08:23:02 UTC |
| ae47aba | Julien Herzen | 08 March 2022, 12:58:23 UTC | Feat/stochastic inputs (#833) * Use stochastic samples for training/inference of torch models * Fix typo * add a unit test | 08 March 2022, 12:58:23 UTC |
| cb80cf3 | Kha | 01 March 2022, 15:38:52 UTC | return metric score along with untrained best models & params (#822) * return metric score along with untrained best models & params * update test and metric type * fmt * gridsearch score unittest Co-authored-by: Julien Herzen <julien@unit8.co> | 01 March 2022, 15:38:52 UTC |
| 4920de0 | Julien Herzen | 01 March 2022, 12:31:56 UTC | Fix/encoders with fit from ds (#829) * Check encoders are not None before using * Add unit test to capture the issue | 01 March 2022, 12:31:56 UTC |
| 2c0453d | gsamaras | 01 March 2022, 10:02:57 UTC | GPU and TPU usage guide (#826) * GPU and TPU usage guide * GPU and TPU usage guide Co-authored-by: Julien Herzen <julien@unit8.co> | 01 March 2022, 10:02:57 UTC |
| 7ca7801 | Dennis Bader | 28 February 2022, 19:11:24 UTC | added val loss to progess bar and updated docs (#825) | 28 February 2022, 19:11:24 UTC |
| 0d267e0 | Dennis Bader | 27 February 2022, 21:22:41 UTC | removed checkpoint from prediction (#821) | 27 February 2022, 21:22:41 UTC |
| cbe49bf | Kha | 25 February 2022, 13:21:32 UTC | fix: bad file descriptor (#769) * fix: bad file descriptor When using Prophet under a multi-threading environment such as Dask, the error `[Errno 9] Bad File Descriptor` could happen. I can consistently reproduce this error while running forecasts with Dask. This implementation seems to fix the issue. I do not fully understand why this happens, because only when running Dask for a long time would this happen. My implementation is semantically equivalent to the original, but I guess avoiding to open `devnull` too many times can help. Context: I am running forecast using Prophet for about 500_000 time series, all in 1 big Dask data frames (split into 91 smaller Pandas data frames). If I don't use Dask, and instead use *joblib* to parallelize the forecasting, then this error does not happen. I think because joblib creates a separate thread for every worker, whereas Dask can have multiple threads for a worker. * revert changes Update logging.py * make log suppression optional * add unit test * provide better data for test, update model dep * restore original intention of model construction * test case to make sure Prophet model is loaded by default * reformat to black style * sort imports * underscore private variables Co-authored-by: Julien Herzen <julien@unit8.co> | 25 February 2022, 13:21:32 UTC |
| c2123c9 | Tomas Van Pottelbergh | 17 February 2022, 16:38:17 UTC | Feat/isort pyupgrade linting (#797) * Add isort and pyupgrade pre-commit hoos * Update black to stable version * Change linting in Gradle and CI * Fix gradle task * Improve gradle task and remove pyupgrade * Add commits to .git-blame-ignore-revs * Change documentation to include linting * Fix gradle linting task * Add newlines | 17 February 2022, 16:38:17 UTC |
| 8158d3e | Tomas Van Pottelbergh | 17 February 2022, 13:08:16 UTC | Fix last black and isort issues (#798) | 17 February 2022, 13:08:16 UTC |
| bbabe54 | hrzn | 17 February 2022, 12:54:59 UTC | Release 0.17.1 | 17 February 2022, 12:54:59 UTC |
| a281861 | Julien Herzen | 17 February 2022, 11:04:09 UTC | Release 0.17.1 (#807) * bump u8darts * Update changelog | 17 February 2022, 11:04:09 UTC |
| ef72f4b | Dennis Bader | 17 February 2022, 10:55:49 UTC | Fix/ptl fixes (#806) * fixed lr schedulers and added tests * added check for invalid model creation params * additional lr scheduler fix * added deprecated to docs * move torch import in tests for flavors * fix wrong parameter docs * updated torch device docs Co-authored-by: Julien Herzen <julien@unit8.co> | 17 February 2022, 10:55:49 UTC |
| 6da5957 | Vladimir Chernykh | 17 February 2022, 09:44:50 UTC | Fix wrong seasonality basis generator for NBEATS (#804) Co-authored-by: Julien Herzen <julien@unit8.co> | 17 February 2022, 09:44:50 UTC |
| 73a1645 | Julien Herzen | 17 February 2022, 08:23:01 UTC | Added KalmanForecaster to readme (#805) | 17 February 2022, 08:23:01 UTC |
| cd361e3 | adelnick | 16 February 2022, 10:48:02 UTC | Relaxing pandas requirements to version 1.0.5 (#800) * Fixes for pandas 1.0.5 * Requirements pandas>=1.0.5 | 16 February 2022, 10:48:02 UTC |
| 00cd099 | hrzn | 15 February 2022, 17:21:00 UTC | Release 0.17.0 | 15 February 2022, 17:21:00 UTC |
| 4db18e2 | Julien Herzen | 15 February 2022, 15:23:01 UTC | Fix typo in changelog | 15 February 2022, 15:23:01 UTC |
| a904e70 | Julien Herzen | 15 February 2022, 15:17:53 UTC | Release 0.17.0 (#796) * Bump u8darts * Update changelog for 0.17.0 * Added backward compatibility warnings * Typo | 15 February 2022, 15:17:53 UTC |
| 80ae7b9 | Julien Herzen | 15 February 2022, 14:13:34 UTC | Add a section about max_samples_per_ts (#795) | 15 February 2022, 14:13:34 UTC |
| 51c480a | Dennis Bader | 15 February 2022, 14:13:12 UTC | fixed pytorch lightning import for flavor tests (#794) | 15 February 2022, 14:13:12 UTC |
| dda4d9c | Dennis Bader | 15 February 2022, 13:43:16 UTC | Feat/pytorch lightning (#702) * first draft proposal for pytorch lightning integration * further cleanup * removed unnecessary file * fix for multiple-TS * moved prediction timeseries generation back to TorchForecastingModel * support for custom trainer in fit() * removed unused methods from TorchForecastingModel * checkpoint loading now correctly resumes training * rewrote TorchForecastingModel * rewrote TFTModel * rewrote rnn models * rewrote nbeats models * rewrote tcn model * rewrote transformer model * removed unused import * resolve failing tests part 1 * resolve failing tests part 2 * adapted the way how model parameters are saved * moved TFTModel predict method into TorchForecastingModel subclass * further simplification of model calls * integrated ProbabilisticTorchForecastingModel into PLForecastingModule * integrated _produce_predict_output into PLForecastingModule * reintegrated original random state handling * removed unused pl random state wrapper function * use OrderedDict for savety in model parameter extraction * made TFM and PLFM paramater extraction generic * added types for variables in TFM init * made predictions deterministic for same fit predict process for non-loaded models * fix flake8 issues * fix flake8 issues part 2 * added pytorch-lightning to torch requirements * fixed loading models with wrong precision * fixed is_probabilistic() * fixed failing tests for epoch count tracker * removed input/output_chunk_length from TorchForecastingModel __init__ * unit tests save models to temp dir * added documentation for ModelMeta * apply suggestions from PR review part 1 * deprecated `torch_device_str` * updated optimizer docs * updated retrain warning * made PLMixedCovariatesModule more generic * added docs * added PTL trainer unit tests * update model docs * fixed broken url in TFM and covariates userguide * removed input/output chunk length from PL modules * relaxed pytorch-lightning requirement * isort * isort part 2 | 15 February 2022, 13:43:16 UTC |
| da84f07 | Julien Herzen | 15 February 2022, 12:45:50 UTC | Fix/int index warnings (#777) * Started removing Int64Index * some progress on introducing RangeIndex everywhere * Handle creation from a DF with integer column * adapt TimeSeries to support RangeIndex only * Keep RangeIndex only in the whole codebase * Update darts/timeseries.py Co-authored-by: Dustin Brunner <92083143+brunnedu@users.noreply.github.com> * Removed commented lines Co-authored-by: Dustin Brunner <92083143+brunnedu@users.noreply.github.com> | 15 February 2022, 12:45:50 UTC |
| d8e056e | Tomas Van Pottelbergh | 15 February 2022, 11:11:34 UTC | Fix/restore datasets hashes (#792) * Restore datasets hashes * Restore newline in monthly-milk.csv * Restore newlines in monthly-milk-incomplete.csv * Restore newlines in monthly-milk.csv Co-authored-by: Julien Herzen <julien@unit8.co> | 15 February 2022, 11:11:34 UTC |
| 46c6249 | Julien Herzen | 15 February 2022, 10:08:56 UTC | Restore datasets hashes (#791) | 15 February 2022, 10:08:56 UTC |
| 38cc671 | Greg DeVos | 13 February 2022, 17:56:05 UTC | Fix/isort and pyupgrade issues (#765) * changes from isort etc * fixing import errors from isort * Clean up dev.txt * Clean up release.txt Co-authored-by: Greg DeVos <gregory.devos@ngc.com> Co-authored-by: Tomas Van Pottelbergh <85937396+tomasvanpottelbergh@users.noreply.github.com> Co-authored-by: Julien Herzen <julien@unit8.co> | 13 February 2022, 17:56:05 UTC |
| 12059b0 | gian | 11 February 2022, 15:20:11 UTC | formatting (#784) | 11 February 2022, 15:20:11 UTC |
| b51b3af | gian | 10 February 2022, 16:01:08 UTC | Fix/simple statistics (#773) * Removed dependencies on old TimeSeries.mean() method * Expanded functionality of mean, median, sum, max, min to work over all axes. Removed dependency on pandas from TimeSeries.quantile_timeseries() method. * Added unit tests for simple statistics methods * Added documentation for simple statistics methods | 10 February 2022, 16:01:08 UTC |
| 508d60a | Julien Herzen | 08 February 2022, 10:08:47 UTC | update video link to pydata 2021 presentation (#771) | 08 February 2022, 10:08:47 UTC |
| 021a9a9 | Julien Herzen | 07 February 2022, 09:12:41 UTC | Feat/kalman forecaster (#743) * avoid copying dataframe in regression models * improve formating in KF file * make Kalman filter work on missing values * generate forecasts with correct time axis name * improve docstring of Kalman filter * added Kalman Forecaster * added KalmanForecaster to unit tests * custom kalman step function for missing values * Fix the nfoursid version * Update test * Added a test case * Rely on new nfoursid step() function * Update version of nfoursid * Remove unused import * Removed unnecessary lines | 07 February 2022, 09:12:41 UTC |
| 0e5b5ad | Dustin Brunner | 05 February 2022, 17:55:14 UTC | Feat/output chunk length regression model (#761) * fix multicollinearity in regression ensemble model tests causing exploding coefficients * reset seed to intial value * add output_chunk_length parameter to regression model * add output_chunk_length to fit method of regressionmodel * add check if model support multi output regression natively * remove _shift_matrices test * update the LightGBMModel * update linear regression model * update random forest regression model * update LightGBMModel docstring * use dict for lags in regressionmodel and adjust all models and tests accordingly * reformat regression_ensemble_model using pre-commit * reformat test_regression_models with pre-commit * shorten comment line length * remove unused import to pass flake8 * reformat with black * reformat with black * update docstring of _create_lagged_data * reformat using black * improve error message when unable to build any samples to fit and when input_dim doesn't match * return self at the end of fit() in regressionmodel * remove numpydoc type hints and add n_jobs_multioutput_wrapper parameter to fit() * add comments | 05 February 2022, 17:55:14 UTC |
| c2d91e0 | Greg DeVos | 04 February 2022, 09:22:23 UTC | Fix/added unique ids to model_name (#760) * added unique ids to model_name * removed pid from the folder name Co-authored-by: Greg DeVos <gregory.devos@ngc.com> Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 04 February 2022, 09:22:23 UTC |
| ec6ccf8 | Julien Herzen | 01 February 2022, 11:54:16 UTC | Feat/return estimator (#741) * Return estimator by forecasting models * Return KF in fit() * Added a unit test * Merge unit test with an existing one * Fix typo | 01 February 2022, 11:54:16 UTC |
| 50c7391 | Dustin Brunner | 01 February 2022, 10:48:59 UTC | add favicon (#763) | 01 February 2022, 10:48:59 UTC |
| ba4e81c | Tomas Van Pottelbergh | 31 January 2022, 18:01:29 UTC | Add workaround for broken pip-tools (#762) | 31 January 2022, 18:01:29 UTC |
| 151ba96 | John Ramsey | 24 January 2022, 19:12:45 UTC | allow for ETS with seasonal=None (#729) * allow for ETS with seasonal=None * Added SeasonalityModel for ETS seasonal arg, clarified docs * formatted with black * Update darts/models/forecasting/exponential_smoothing.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/exponential_smoothing.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/exponential_smoothing.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/models/forecasting/exponential_smoothing.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * oxford comma * Update darts/models/forecasting/exponential_smoothing.py * Apply black reformatting * fix too long lines Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 24 January 2022, 19:12:45 UTC |
| 841a42e | hrzn | 24 January 2022, 16:05:06 UTC | Release 0.16.1 | 24 January 2022, 16:05:06 UTC |
| 0990a6e | Julien Herzen | 24 January 2022, 15:13:07 UTC | Update changelog and u8darts version (#753) | 24 January 2022, 15:13:07 UTC |
| bed6e52 | Julien Herzen | 24 January 2022, 15:01:58 UTC | Remove broken type annotation (#752) * Remove broken type annotation * Solve issues with calling pandas mean() | 24 January 2022, 15:01:58 UTC |
| d94428a | Tomas Van Pottelbergh | 21 January 2022, 15:27:36 UTC | Feat/set up flake8 (#749) * Modify flake8 config for Black compatibility * Add flake8 pre-commit * Pin flake8 requirement * Add flake8 Github action * Modify test command to include linting * Ignore flake8 issues fix in git blame * Mention flake8 pre-commit linting | 21 January 2022, 15:27:36 UTC |
| 96125f8 | Tomas Van Pottelbergh | 21 January 2022, 14:23:44 UTC | Fix/flake8 issues (#748) * Fix W605 * Fix W291 * Fix F401 * Fix F541 * Fix F841 (partially) * Fix E713 * Fix E731 * Fix E714 * Fix F841 * Fix E501 * Fix E501 | 21 January 2022, 14:23:44 UTC |
| 074be2a | Dennis Bader | 18 January 2022, 17:41:25 UTC | fix/TFTModel_flask (#745) | 18 January 2022, 17:41:25 UTC |
| 334b0d8 | Dustin Brunner | 18 January 2022, 15:11:00 UTC | Fix/revert numpydoc docstrings (#744) * fix/update statistics docstrings * fix/update timeseries generation docstrings | 18 January 2022, 15:11:00 UTC |
| ddc84a2 | Dennis Bader | 15 January 2022, 19:13:25 UTC | fixed non contiguous error when using lstm_layers > 1 on gpu (#740) | 15 January 2022, 19:13:25 UTC |
| adde52a | Dennis Bader | 14 January 2022, 09:27:47 UTC | fixed timedelta issue in longest_contiguous_slice (#725) * fixed timedelta issue in longest_contiguous_slice * made slice size calculation more comprehensible | 14 January 2022, 09:27:47 UTC |
| 3882a0c | hrzn | 13 January 2022, 17:23:03 UTC | Release 0.16.0 | 13 January 2022, 17:23:03 UTC |
| b27a8ea | Julien Herzen | 13 January 2022, 16:00:47 UTC | Release 0.16.0 (#738) * Update changelog * Bump u8darts | 13 January 2022, 16:00:47 UTC |
| a5b1d9c | Julien Herzen | 13 January 2022, 15:42:16 UTC | Feat/quickstart (#731) * Started updating quickstart * Update quickstart * add covariates section to quickstart * Add probabilistic forecasting to quickstart * Update quickstart * Update quickstart * Update quickstart * Small cosmetics improvements * Cosmetic improvements * add encoders placeholder * Slight improvements to quickstart * Small cosmetic improvements to quickstart * formatting * formatting * formatting * Small improvement to quickstart * Address PR comments * Cosmetics on quickstart | 13 January 2022, 15:42:16 UTC |
| 2f3be1a | Dustin Brunner | 13 January 2022, 13:50:38 UTC | Fix/optimize regression model fit (#736) * optimize computation of X and y matrices for fitting of RegressionModel and update unittests * remove unused import * modify unittests to pass with new preprocessing method * reformat with black | 13 January 2022, 13:50:38 UTC |
| 8c139fc | Julien Herzen | 13 January 2022, 08:56:54 UTC | Add brunnedu to code owners | 13 January 2022, 08:56:54 UTC |
| ddc91b9 | Tomas Van Pottelbergh | 12 January 2022, 09:34:44 UTC | Feat/set up black (#732) * Add jupyter variant to black requirement * Add black pre-commit hook * Add instructions for code formatting * Add linting with Black to CI * Add badge for Black * Add git blame ignore file for Black migration Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 12 January 2022, 09:34:44 UTC |
| 3416fc7 | Tomas Van Pottelbergh | 12 January 2022, 08:02:57 UTC | Feat/format using black (#733) * Apply black and fix some strings * Replace inconsistent quote characters * Apply black to example notebooks | 12 January 2022, 08:02:57 UTC |
| 4a97140 | Dennis Bader | 09 January 2022, 20:25:04 UTC | added check for untrained models for copying of model parameters (#728) * added check for untrained models for copying of model parameters * added test and changed position of checks due to failing tests | 09 January 2022, 20:25:04 UTC |
| 8fa7dbf | Tomas Van Pottelbergh | 06 January 2022, 08:55:24 UTC | Feat/kalman with fitting (#692) * First proof of concept * Clean up and add checks * Remove support for multiple timeseries * Add tests and fixes * Add test given KF and fixes * Add docstrings * Add and adapt example * Improve covariates (error) handling * Remove unused imports Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Improve docstring line length Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Fix typos * Remove filterpy from requirements * Remove unnecessary parentheses * Add num_samples to docstring * Add num_block_rows to docstring * Clarify covariance formula * Update to nfoursid with relaxed requirements * Bump nfoursid requirement Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 06 January 2022, 08:55:24 UTC |
| ed88209 | Julien Herzen | 05 January 2022, 20:35:30 UTC | Feat/refactor doc (#720) * renamed intro notebook into quickstart * renamed examples/ into notebooks/ * moved doc/ --> docs/userguide/ * adapt Makefile to include more targets * use build-all-docs target from gradle * add quickstart and userguide to doc index * Fix toctree issues with numpydoc * removing display of inheritance * small changes to README * small adaptations in transformers docstrings * update docstring for transformers * fix typo * Add some examples to transformers * Improve docstrings of datasets and metrics * Some more docstrings improvements * Improved docstrings of models * more docstrings improvements * Improve docstrings of TimeSeries * Small documentation improvements * Rename notebooks folder * Change refs to notebooks * Renamed notebooks * Fix link typo * Simplify build.gradle * correct typo in intro notebook name * correct intro notebook name * Correct pointers to example notebooks * Fix table * Clean links in covariates.md * Solve a weird issue with Prophet tests | 05 January 2022, 20:35:30 UTC |
| 60121e4 | Francesco Lässig | 29 December 2021, 15:24:51 UTC | fix(TorchForecastingModel): added support for ReduceLROnPlateau scheduler with training loss (#711) | 29 December 2021, 15:24:51 UTC |
| 86443d0 | Francesco Lässig | 29 December 2021, 13:52:38 UTC | feat(README): added a link to the prob forecasting blog post (#712) | 29 December 2021, 13:52:38 UTC |
| 145a30f | hrzn | 24 December 2021, 11:15:07 UTC | Release 0.15.0 | 24 December 2021, 11:15:07 UTC |
| e47373b | Julien Herzen | 24 December 2021, 10:31:58 UTC | Release 0.15.0 (#706) * bump u8darts * update changelog for 0.15.0 | 24 December 2021, 10:31:58 UTC |
| dcfc81d | Julien Herzen | 24 December 2021, 10:13:34 UTC | Update README.md | 24 December 2021, 10:13:34 UTC |
| dad1591 | Julien Herzen | 23 December 2021, 14:00:57 UTC | Change some relative imports into absolute (#705) * Change some relative imports into absolute * Fix an import issue | 23 December 2021, 14:00:57 UTC |
| cd05435 | Julien Herzen | 23 December 2021, 08:31:50 UTC | Adding codecov badge (#704) * try adding codecov to develop workflow * add codecov badge to readme * try specifying coverage file * remove tests from coverage computation * produce coverage report in xml * add source to coveragerc file * Add codecov to merge workflow * remove coverage.sh | 23 December 2021, 08:31:50 UTC |
| 299139c | dumjax | 22 December 2021, 11:01:53 UTC | granger causality init (#695) * granger causality init * small adjustments * link to statsmodels * stationarity tests and warning in granger * slice_intersect in granger * test freq for granger causality * add some unittests * Update darts/utils/statistics.py Co-authored-by: Julien Herzen <julien.herzen@unit8.co> * add unittests for stationarity * small add * merge master 2 Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 22 December 2021, 11:01:53 UTC |
| 5fba889 | Julien Herzen | 21 December 2021, 19:06:01 UTC | Feat/improve tests (#700) * absolute imports & black style in tests * Correct issues * Some linting * submodules for tests * Correct issues * Remove some warnings, linting * slightly speed up a test * Use pytest, compute coverage in one go | 21 December 2021, 19:06:01 UTC |
| 6604c3c | Julien Herzen | 19 December 2021, 20:30:10 UTC | Fix/failing notebooks (#698) * fix a bug in TCN model * fix and speedup some notebooks * fix and slightly speedup some notebooks | 19 December 2021, 20:30:10 UTC |
| 055d327 | Dennis Bader | 18 December 2021, 09:23:51 UTC | Fix/example notebooks (#682) * added TFT and DTW example notebooks to the docs example page * updated TFTModel example with new add_encoders * updates to bug issue template and github actions examples tests * make examples doc page less verbose * added all notebook examples to tests on merge Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 18 December 2021, 09:23:51 UTC |
| 1b9dad6 | Dennis Bader | 17 December 2021, 23:52:02 UTC | Docs/cov update2 (#691) * added covariates guide * added images * updated docs * split into two guides - covariates and global forecasting models * additional changes * final adaptions for PR * Apply suggestions from code review Co-authored-by: Julien Herzen <julien.herzen@unit8.co> * applied suggestions from code review p2 * last changes * renamed folder to doc * updated guide paths in doc/ Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 17 December 2021, 23:52:02 UTC |
| 9933d29 | Julien Herzen | 14 December 2021, 10:53:17 UTC | Feat/extend map to stochastic (#688) * remove a print * generalise TimeSeries.map() * Add a unit test for invert mapping stochastic series * fix a test * computationally more efficient mapping | 14 December 2021, 10:53:17 UTC |
| 8077cfc | Julien Herzen | 13 December 2021, 21:21:53 UTC | Fix/min train length on torch models (#686) * return correct min series length in torch models * unit test for residuals on global models * Update darts/tests/test_global_forecasting_models.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> * Update darts/tests/test_global_forecasting_models.py Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> Co-authored-by: Dennis Bader <dennis.bader@gmx.ch> | 13 December 2021, 21:21:53 UTC |
| 35cb36e | Dennis Bader | 10 December 2021, 08:41:19 UTC | Feat/future covariate slicing (#685) * future covariates are now sliced when used with local forecasting models * removed some blank lines * applied suggestions for PR review. Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 10 December 2021, 08:41:19 UTC |
| ab04f9e | Francesco Lässig | 09 December 2021, 17:29:25 UTC | feat(ForecastingModel): added call to fit method when model is untrained (#681) Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 09 December 2021, 17:29:25 UTC |
| 2050541 | Julien Herzen | 09 December 2021, 15:33:32 UTC | Update CODEOWNERS Add @dennisbader to code owners. | 09 December 2021, 15:33:32 UTC |
| 8fa3bfa | Dennis Bader | 09 December 2021, 14:36:07 UTC | fixed encoder tests crashing with non-torch flavors (#683) | 09 December 2021, 14:36:07 UTC |
| 4355210 | Francesco Lässig | 09 December 2021, 13:59:21 UTC | feat(ExponentialSmoothing): added model parameters to __str__ method (#680) Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 09 December 2021, 13:59:21 UTC |
| c475af3 | Dennis Bader | 09 December 2021, 10:07:16 UTC | refactored `_get_batch_prediction()` for PastCovariatesTorchModel and added unit tests (#674) Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 09 December 2021, 10:07:16 UTC |
| c2f5a49 | Dennis Bader | 09 December 2021, 06:49:09 UTC | Feat/add encoders (#565) * data loader performance improvements through leveraging index * replaced isin with >/< operators * integrated index caching to reduce comp. cost * small fix * added support for integer-indexed loader * cleaning up for PR * added index generator classes for encoders * added cyclic encoder * non-generic encoding with cyclic encoder yields identical results * setup sequential encocer class * added encode_train() to EncoderSequence * EncoderSequence now fully operable * added index generator tests * added tests and docs * additional tests * added datetime_attribute encoders and updated TFTModel * cleaned up encoders.py and encoders_base.py * non-lazy case works * lazy encoders can now be passed through Datasets * added covariate types to training datasets * refactored inference datasets and added lazy encoding * replaced past and future inference datasets with generic one * fixed inference dataset according to errors from tests * refactored SequenceEncoder to act on `àdd_encoders` directly * removed lazy encoding * refactored generic inference dataset * added absolute and relative integer index position encoders * moved initialize_encoders() into fit() so that training, validation and inference all use the same encoder objects * updated docs for position integer index encoding * integer index encoders now take only the train target series as a reference point * added support for custom callable encoders * added support for global scaler * splitted transform into past and future parts * updated docs for PR * Apply suggestions from code review Co-authored-by: Julien Herzen <julien.herzen@unit8.co> * applied suggestions from PR review part 2 * added encoder support to all TorchForecastingModels * fixed some indents Co-authored-by: Julien Herzen <julien.herzen@unit8.co> | 09 December 2021, 06:49:09 UTC |
