MLOS Versions Save

What's Changed

• Add tests for mixed numerical type conversions of tunable config data from storage by @bpkroth in https://github.com/microsoft/MLOS/pull/670
• Fixtures context fixup by @bpkroth in https://github.com/microsoft/MLOS/pull/673
• Address upcoming changes to default GITHUB_TOKEN permissions by @bpkroth in https://github.com/microsoft/MLOS/pull/674
• Return status and actual timestamp from all environments by @motus in https://github.com/microsoft/MLOS/pull/671
• Handle some mypy errors and increase test coverage by @bpkroth in https://github.com/microsoft/MLOS/pull/680
• New pytest version dependency tweaks by @bpkroth in https://github.com/microsoft/MLOS/pull/681
• Address pandas warnings about future pyarrow requirements by @bpkroth in https://github.com/microsoft/MLOS/pull/682
• Address pandas timestamp warnings by @bpkroth in https://github.com/microsoft/MLOS/pull/683
• Scheduling queue: Allow submitting trials with future timestamps and use watermarks to retrieve configs and results by @motus in https://github.com/microsoft/MLOS/pull/676
• use named constants to define columns containing IDs and generic values by @motus in https://github.com/microsoft/MLOS/pull/678
• Allow use of Azure VM Service for only remote exec. by @bpkroth in https://github.com/microsoft/MLOS/pull/675
• Tweaks for anchored stdout parsing by @bpkroth in https://github.com/microsoft/MLOS/pull/685
• Allow no tunable configs for benchmarking by @bpkroth in https://github.com/microsoft/MLOS/pull/686
• Add tests for config data retrieval for experiments with no tunables by @bpkroth in https://github.com/microsoft/MLOS/pull/689
• Initial work to separate the scheduler and the optimizer (no async yet) by @motus in https://github.com/microsoft/MLOS/pull/684
• Bump version: 0.5.0 → 0.5.1 to mark a stable version before the big changes (async Environment etc.) by @motus in https://github.com/microsoft/MLOS/pull/695

Full Changelog: https://github.com/microsoft/MLOS/compare/v0.5.0...v0.5.1

What's Changed

• Reorg doc deploy step for new GITHUB_TOKEN permissions requirements by @bpkroth in https://github.com/microsoft/MLOS/pull/659
• Omit tests from docs by @bpkroth in https://github.com/microsoft/MLOS/pull/660
• Add weights for tunables with categorical or special numerical values by @motus in https://github.com/microsoft/MLOS/pull/646
• Add log scale and quantization to the tunables by @motus in https://github.com/microsoft/MLOS/pull/662
• Fix mixed numeric datatypes for optimizers by @ephoris in https://github.com/microsoft/MLOS/pull/667
• Add support for ConfigSpace distributions by @motus in https://github.com/microsoft/MLOS/pull/663
• Bump version: 0.4.2 → 0.5.0 by @motus in https://github.com/microsoft/MLOS/pull/672

New Contributors

• @ephoris made their first contribution in https://github.com/microsoft/MLOS/pull/667

Full Changelog: https://github.com/microsoft/MLOS/compare/v0.4.2...v0.5.0

What's Changed

• README documentation updates by @bpkroth in https://github.com/microsoft/MLOS/pull/647
• Remove TunableGroups from Storage classes and add TrialData tests by @bpkroth in https://github.com/microsoft/MLOS/pull/644
• Refactor mlos_bench.storage and add TunableConfigTrialGroup property for TrialData and ExperimentData by @bpkroth in https://github.com/microsoft/MLOS/pull/648
• Add default_tunable_config_id property to ExperimentData by @bpkroth in https://github.com/microsoft/MLOS/pull/650
• Remove mistakenly added columns by @bpkroth in https://github.com/microsoft/MLOS/pull/652
• Fixes to max_iterations calculation when trial_config_repeat_count is given. by @bpkroth in https://github.com/microsoft/MLOS/pull/651
• Address some pandas FuturesWarnings by @bpkroth in https://github.com/microsoft/MLOS/pull/653
• Improve mlos-viz for multiple repeats of a config and add tests by @bpkroth in https://github.com/microsoft/MLOS/pull/633
• Workaround pytest version bump incompatibility with pytest-lazy-fixtures plugin by @bpkroth in https://github.com/microsoft/MLOS/pull/656
• Add demo video link to README by @bpkroth in https://github.com/microsoft/MLOS/pull/657
• Bump version: 0.4.1 → 0.4.2 by @bpkroth in https://github.com/microsoft/MLOS/pull/658

Full Changelog: https://github.com/microsoft/MLOS/compare/v0.3.2...v0.4.2

What's Changed

• Rename and improve tunable vs config API and documentation by @bpkroth in https://github.com/microsoft/MLOS/pull/629
• CI tweak: use "release" event instead of "tag push" for pypi publishing by @bpkroth in https://github.com/microsoft/MLOS/pull/631
• Add support for storing the optimization targets and direction of an experiment by @bpkroth in https://github.com/microsoft/MLOS/pull/628
• Introduce basic experiment visualization module mlos_viz via dabl by @bpkroth in https://github.com/microsoft/MLOS/pull/624
• Github Action syntax fixup by @bpkroth in https://github.com/microsoft/MLOS/pull/632
• Add labels to the optimizer trends plot by @bpkroth in https://github.com/microsoft/MLOS/pull/634
• Add more info on const_args vs. tunable_params; start the DEVNOTES by @motus in https://github.com/microsoft/MLOS/pull/630
• Bump actions/cache from 3 to 4 in /.github/workflows by @dependabot in https://github.com/microsoft/MLOS/pull/637
• Support for special tunable values outside of the range by @motus in https://github.com/microsoft/MLOS/pull/617
• Tell pylint to ignore FIXME comments in our code. by @bpkroth in https://github.com/microsoft/MLOS/pull/639
• Add unit tests for MlosCoreoptimizer._to_df() functionality; use special values in tunables by @motus in https://github.com/microsoft/MLOS/pull/638
• Fixup a vscode config file. by @bpkroth in https://github.com/microsoft/MLOS/pull/643
• Disallow ! characters in tunable parameter names. by @bpkroth in https://github.com/microsoft/MLOS/pull/641
• Add basic support for repeating a trial config. by @bpkroth in https://github.com/microsoft/MLOS/pull/642
• PyPi publish makefile rules fixup by @bpkroth in https://github.com/microsoft/MLOS/pull/640

Full Changelog: https://github.com/microsoft/MLOS/compare/v0.2.5...v0.3.2

Adds

• NuGet packaging for C# codegen and external agent components
• Some cmake integration patterns for external project use of shared memory codegen pieces
• Anonymous shared memory support for Linux