v0.10.0

This release contains an important new feature: zero-shot AutoML and mete learning. It provides a new way of doing AutoML without tuning. You can now use the existing training API from lightgbm, xgboost etc. while getting the benefit of AutoML in choosing high-performance hyperparameter configurations per task. Recommended for everyone currently using lightgbm, xgboost or random forest, regardless of previous experience in AutoML. This feature also enables continuous improvement of AutoML from historical AutoML experiments.

Other changes can be found below.

What's Changed

• Typo on the webpage's Getting Started section by @cammarb in #457
• Bump follow-redirects from 1.14.7 to 1.14.8 in /website by @sonichi in #459
• Docstr update by @qingyun-wu in #460
• update regression metrics in notebooks by @sonichi in #454
• make AutoML.classes_ an array by @sonichi in #467
• Bump prismjs from 1.25.0 to 1.27.0 in /website by @sonichi in #471
• Zero-shot AutoML by @sonichi in #468
• don't init global search with points_to_evaluate unless evaluated_rewards is provided; handle callbacks in fit kwargs by @sonichi in #469

New Contributors

• @cammarb made their first contribution in #457

Full Changelog: v0.9.7...v0.10.0

v0.9.7

What's Changed

• Update Task-Oriented-AutoML.md by @vvijayalakshmi21 in #446
• Update Task-Oriented-AutoML.md by @vvijayalakshmi21 in #447
• Update Tune-User-Defined-Function.md by @vvijayalakshmi21 in #448
• corrected typo in example xgboost documentation by @MichaelMarien in #449
• bump ray version to 1.10 by @sonichi in #450
• fix a bug when using ray & update ray on aml by @sonichi in #455

New Contributors

• @vvijayalakshmi21 made their first contribution in #446

Full Changelog: v0.9.6...v0.9.7

v0.9.6

What's Changed

• reducing AutoConfig.from_pretrained by @liususan091219 in #411
• Set use_ray to True for logging to databricks by @liususan091219 in #414
• Bump nanoid from 3.1.30 to 3.2.0 in /website by @sonichi in #420
• bump version of node-fetch to 3.1.1 in website/ by @sonichi in #423
• Use Ray _BackwardsCompatibleNumpyRng if possible by @Yard1 in #421
• remove FLAML sample size from config by @sonichi in #418
• max_iter < 2 -> no search; sign in metric constraints; test and example for forecasting by @sonichi in #415
• remove redundant imports by @liususan091219 in #426
• Support time series forecasting for discrete target variable by @int-chaos in #416
• homepage update by @sonichi in #425
• fix a broken link in README.md by @m13uz in #439
• adding catch for HTTP error by @liususan091219 in #432
• Change the upper bound for "lags" hyperparameter for sklearn forecast models by @int-chaos in #437
• Gpu support for xgboost by @sonichi in #442
• data in csv by @sonichi in #430
• note about preview feature by @sonichi in #431

New Contributors

• @m13uz made their first contribution in #439

Full Changelog: v0.9.5...v0.9.6

v0.9.5

What's Changed

• fixing load best model at the end by @liususan091219 in #389
• Regression forecast debug by @int-chaos in #391
• set verbose for transformers by @liususan091219 in #392
• Logging multiple checkpoints by @liususan091219 in #394
• postcss version update by @sonichi in #385
• fixing default metric for regression + change verbosity for transformers by @liususan091219 in #397
• fix issues in logging, bug in space.py, constraint sign, and improve code coverage by @sonichi in #388
• moving intermediate_results logging from model.py to huggingface/trainer.py by @liususan091219 in #403
• Update flaml/nlp/README.md by @liususan091219 in #404
• Logo by @qingyun-wu in #399
• update browser icon by @qingyun-wu in #407
• adding logging of training loss by @liususan091219 in #406
• Bump shelljs from 0.8.4 to 0.8.5 in /website by @sonichi in #402
• Sklearn api x by @MichaelMarien in #405

New Contributors

• @MichaelMarien made their first contribution in #405

Full Changelog: v0.9.4...v0.9.5

v0.9.4

This release enables regression models for time series forecasting. It also fixes bugs in nlp tasks, such as serialization of transformer models and automatic metrics.

What's Changed

• citation file by @sonichi in #364
• Fix several issues for nlp tasks by @sonichi in #380
• serialize TransformerEstimator by @sonichi in #381
• Time series forecasting with sklearn regressors by @int-chaos in #362
• fixing auto metric bug by @liususan091219 in #387

Full Changelog: v0.9.3...v0.9.4

v0.9.3

What's Changed

• Finish the Multiple Choice Classification by @oberonbot in #367
• logging by @sonichi in #371
• adding token classification by @liususan091219 and @siddheshshaji in #376

New Contributors

• @oberonbot and @siddheshshaji made their first contribution in #367

Full Changelog: v0.9.2...v0.9.3

v0.9.2

New Features:

• New task: text summarization
• Reproducibility of hyperparameter search sequence
• Run flaml in azureml + ray

What's Changed

• url update for doc edit by @sonichi in #345
• Adding the NLP task summarization by @liususan091219 @XinZofStevens @GideonWu0105 in #346
• reproducibility for random sampling by @sonichi in #349
• doc update by @sonichi in #352
• azureml + ray by @sonichi in #344
• Fixing the bug in custom metric by @liususan091219 in #356
• Simplify lgbm example by @ruizhuanguw in #358
• fixing custom metric by @liususan091219 in #357
• Example by @sonichi in #359

New Contributors

• @ruizhuanguw @XinZofStevens @GideonWu0105 made their first contribution in #358

Full Changelog: v0.9.1...v0.9.2

v0.9.1

This release contains several feature improvements and bug fixes. For example,

• support for custom data splitter.
• evaluation_function can receive incumbent result in local search and perform domain-specific early stopping by comparing with the incumbent result. As long as the comparison result (better or worse) is known, the evaluation can be stopped.
• support and automate huggingface metrics.
• use cfo in tune.run if bs is not installed.
• fixed a bug in modifying n_estimators to satisfy constraints.
• new documentation website.

What's Changed

• Update flaml_pytorch_cifar10.ipynb by @sonichi in #328
• adding HF metrics by @liususan091219 in #335
• train at least one iter when not trained by @sonichi in #336
• use cfo in tune.run if bs is not installed by @sonichi in #334
• Makes the evaluation_function could receive the incumbent best result as input in Tune by @Shao-kun-Zhang in #339
• support for customized splitters by @wuchihsu in #333
• Deploy a new doc website by @sonichi, @qingyun-wu and @Shao-kun-Zhang in #338
• version update by @sonichi in #341

New Contributors

• @Shao-kun-Zhang made their first contribution in #339

Full Changelog: v0.9.0...v0.9.1

v0.9.0

1. Revise flaml.tune API

• Add a “scheduler” argument (a user can choose from “flaml”, “asha” or a customized scheduler)
• Rename "prune_attr" to "resource_attr"
• Rename “training_function” to “evaluation_function”
• Remove the “report_intermediate_result” argument (covered by “scheduler” instead)
• Add tests for the supported schedulers
• Re-run notebooks that use schedulers

2. Add save_best_config() to save best config in a json file

What's Changed

• add save_best_config() by @sonichi in #324
• tune api for schedulers by @qingyun-wu in #322
• add init.py in nlp by @sonichi in #325
• rename training_function by @qingyun-wu in #327

Full Changelog: v0.8.2...v0.9.0

v0.8.2

What's Changed

• include default value in rf search space by @sonichi in #317
• adding TODOs for NLP module, so students can implement other tasks easier by @liususan091219 in #321
• pred_time_limit clarification and logging by @sonichi in #319
• bug fix in confg2params by @sonichi in #323

Full Changelog: v0.8.1...v0.8.2