New features and cleanup from new version release #159

rvandewater · 2024-10-17T15:59:54Z

New Features

Polars backend for preprocessing (significantly faster preprocessing)
Modality selection (for doing experiments that need to show modality contributions)
Optuna optimization with visualization and checkpointing
Options to use SHAP or other explainers (improved support pending)
Introduced configuration settings for multiple classifiers including Balanced Random Forest, CatBoost, XGBoost, and RUSBClassifier.
Refactored RNN, GRU, TCN, and Transformer.
Enhanced capabilities for binary classification, regression, and dataset imputation tasks.
Added new configuration for benchmarking experiments.
Enhanced error handling in cross-validation processes.

Chores

Updated dependencies and package versions for better compatibility.

- skip training when eval is set - change readme to reflect changes to command line

…mands-flags # Conflicts: # icu_benchmarks/models/train.py

Fix (evaluate) commands flags

logging fix

Bumps [lightning](https://github.com/Lightning-AI/lightning) from 2.3.0 to 2.3.3. - [Release notes](https://github.com/Lightning-AI/lightning/releases) - [Commits](Lightning-AI/pytorch-lightning@2.3.0...2.3.3) --- updated-dependencies: - dependency-name: lightning dependency-type: direct:production ... Signed-off-by: dependabot[bot] <[email protected]>

Bump lightning from 2.3.0 to 2.3.3

Major version release with new hyperparameter optimization, polars preprocessing, modality selection, and other new features.

rvandewater and others added 30 commits August 10, 2023 15:33

Added torchmetrics

152cbf4

refactoring

a8d7eaa

removed obsolete experiment files

63c57b5

type casted float64 to float32 to make it work with MPS

5bad6a1

make reproducible argument for evaluate too

9326393

renamed needs_training to requires_backprop, removed needs_fit.

c817ff7

Documentation update

7678dd2

fix model loading for ML

cd91449

rework metric constants and added binaryfairness wrapper

cd5ee98

feature names extraction

9a09038

Predict proba for ml models

65ef59c

small changes for lgbm model

c8499cb

small changes to transform

fcf2970

make reproducible argument for evaluate too

1b423d4

fix model loading for ML

730c7e4

rename checkpoint to hp_checkpoint

8a5dfb3

remove verbose required=False

0d2613b

remove commands and add support for eval and finetune

4e9271b

- skip training when eval is set - change readme to reflect changes to command line

move eval flag to first position

ef25a33

fix lint

2fd51b0

added support for model-v1.cpkt

25adb0d

Merge remote-tracking branch 'origin/fix-commands-flags' into fix-com…

289b4da

…mands-flags # Conflicts: # icu_benchmarks/models/train.py

Pytorch lightning syntax

e7d3c59

fixed string

66661a9

adjusted the pytorch lightning bool

fe6f679

Linting

2eb8c1c

Merge pull request #133 from rvandewater/fix-commands-flags

eef73b0

Fix (evaluate) commands flags

logging fix

afe9b32

Merge pull request #134 from rvandewater/small_logging_fix

ae4aca2

logging fix

created torchmetrics dictionary

06ee062

rvandewater and others added 27 commits October 15, 2024 09:52

Compatibility with static (non temporal) datasets

125e4f0

Experiment related tuning

b307474

revert to log loss

9777418

Saving pred indicators to evaluate predictions

808eb9f

Data sanitization

34e0573

Preprocessing improvements and failsafe when using just static features

ab3ab4b

Remove cass specific files

9905abc

update recipies to 1.0

860bf3a

Merge pull request #156 from rvandewater/dependabot/pip/lightning-2.3.3

cdb5bf4

Bump lightning from 2.3.0 to 2.3.3

Major refactor

4a4ab0c

Major refactoring round 2

9547bfa

Merge branch 'development' into version_release

4ed54cf

remove cass files

5d3ee69

remove cass files

99ce855

Added polars

151d68f

More major refactoring

47c594a

Linting

d4e8845

added checks

9a4ab8e

Increase allowed complexity

5004ae4

docs

619937c

Transformer refactor to reduce code duplication

0133dbd

transformer fix

b3d7ad0

optuna integration

23d2c68

removed superfluous test step (integrated in main wrapper now)

18175d9

xgboost linting

5372971

Merge pull request #155 from rvandewater/version_release

4dd915e

Major version release with new hyperparameter optimization, polars preprocessing, modality selection, and other new features.

rvandewater requested a review from prockenschaub October 17, 2024 16:00

rvandewater merged commit 0aa632c into master Oct 21, 2024
2 checks passed

rvandewater mentioned this pull request Oct 21, 2024

Possible Parallelization Enhancement #4

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

New features and cleanup from new version release #159

New features and cleanup from new version release #159

rvandewater commented Oct 17, 2024 •

edited

Loading

New features and cleanup from new version release #159

New features and cleanup from new version release #159

Conversation

rvandewater commented Oct 17, 2024 • edited Loading

rvandewater commented Oct 17, 2024 •

edited

Loading