Skip to content

v0.3.0

Latest

Choose a tag to compare

@hmacdope hmacdope released this 19 Nov 03:35
358a208

What's Changed

  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #570
  • Make shuffle and drop_last training-only across deep-learning workflows by @smcolby in #572
  • Consolidate random seeding under random_seed (global + per-section) by @smcolby in #573
  • Free runner disk space before Docker build by @smcolby in #574
  • Set persistent_workers=True in ChemProp DataLoader to prevent fd exhaustion by @smcolby in #575
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #578
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #581
  • Migrate ReadTheDocs conda tool from mambaforge to miniforge3 by @smcolby in #583
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #585
  • Fix for boto3 dependency in Docker containers by @tlhr in #586
  • Fix UnscaleTransform identity when no scaler is provided by @smcolby in #577
  • Make shape handling consistent across package for #509 by @dwwest in #512
  • Fix ChemProp scheduler configuration, API semantics, and model defaults by @smcolby in #576
  • Add TabICL v2 architecture with regression and classification models by @smcolby in #590
  • Rename butina cutoff kwarg to dist_cutoff by @smcolby in #599
  • Add CheMeleon embedding featurizer and model predict_embedding by @smcolby in #588
  • Add PCA dimensionality reduction to the anvil pipeline by @smcolby in #592
  • Add TrainedModelFeaturizer for featurizing with a pretrained model by @smcolby in #598
  • fix: resolve accelerator to device in TabPFN model classes by @li-lizhe in #601
  • Fix GPU CI workflow by @cynthiaxu04 in #606
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #593
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #611
  • Add MonroeFeaturizer for Monroe foundation model embeddings by @smcolby in #602
  • fix: use 'auto' accelerator (not 'gpu') as default for predict methods by @li-lizhe in #604
  • Add featurizer alias for keying same-type feature blocks by @smcolby in #603
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #613
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #619
  • Exclude fit_time and score_time from cross-validation metrics by @aaronkanzer in #621
  • Add a line to the pull request template by @dwwest in #618
  • Hook post-hoc binary metrics and plots into standard evaluation framework by @smcolby in #584
  • [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #623
  • Fix loading gpu-saved tabpfn model loading by @dwwest in #616
  • Fix pairwise inference task index selection by @SergeiNikolenko in #505
  • Honour ye olde configured aggregation whence building from a foundation model by @dwwest in #624
  • Support pixi, uv, pip, and conda installs with locked environments by @smcolby in #617

New Contributors

Full Changelog: v0.2.0...v0.3.0