Skip to content

CHANGELOG.md

Latest commit

 

History

History
220 lines (178 loc) · 7.38 KB

CHANGELOG.md

File metadata and controls

220 lines (178 loc) · 7.38 KB

CHANGELOG

Change log for the codebase. Initialised from the developments following version V0.11.3

[v1.3.0] - 2025-08-01

Added

  • Added: test case for variance threshold
  • Added: `.env' file to set some required env vars
  • Added: extra & missing cicd packages

Changed

  • changed: dockerfile to only install main dependencies
  • changed: streamlined imports
  • changed: updated as many packages as possible
  • Changed: updated pre-commit-config

Fixed

  • fix: omic path parsing bug
  • fix: plotting bug arising from api change
  • fix: correct file permissions
  • fix: cicd to build image and push to the ghcr.io
  • fix: shuffle bug

Security

  • security: updated packages to resolve dependabot alerts

v1.2.0 - 2025-05-09

Changed

  • Changed logger to also send to stdout and converted more print statements to loggers.
  • Upgraded:
    • Eli5
    • scikit-learn (also resolves dependabot alert)
    • xgboost
    • shap
  • Changed env manager to be poetry based

v1.1.1 - 2025-03-04

Fixed

  • Bug fixed: type issue in feature selection

V1.1.0 - 2025-01-30

Added

  • added pydantic env
  • created initial pydantic models:
    • DataModel
    • PredictionModel
    • PlottingModel
    • omics:
      • TabularModel
      • MetabolomicModel
      • GeneExpressionModel
      • MicrobiomeModel
    • FeatureSelectionModel, AutoModel & MethodModel
    • Models:
      • AutoKerasModel
      • AutoLgbmModel
      • AutoXgboostModel
    • MlModel
    • ConfigModel
  • added batch mode processing
  • Added: new flag to skip ML standardizing in MlModel
  • Added: created load_transformed_data from pre-existing code
  • Added: new option DataModel.data_type "R2G" for datasets are Ready to Go to be trained on

Changed

  • upgraded:

    • scikit-learn to 1.2.2
    • keras to 2.14.0
    • lightgbm to 3.3.5
    • numpy to 1.23.5
    • optuna to 3.6.*
    • tensorflow to 2.14.0
    • xgboost to 1.7.6

  • moved dev packages to requirements_dev.txt

  • upgraded tensorflow and keras to be able to create env on arm64

  • changed DEV_MANUAL.md to capture venv changes

  • Changed: simplified load_data

  • Changed: extracted save_transformed_data from pre-existing code

  • Changed: extracted out ml preprocessing into function learn_ml_preprocessing and apply_ml_preprocessing

Fixed

  • Fixed: deprecation warnings

Removed

  • removed auto-sklearn

V1.0.1 - 2024-03-11

Changed

  • cleaned up docstring comment
  • removed copyright notices from example json and data
  • Readme updated

Removed

  • Removed config jsons that do not have paired data files

Fixed

  • copy-paste corrections
  • redundant set casting

Security

  • switched os.mkdir for os.makedirs

V1.0.0 - 2024-01-22

  • Introduction of pre-commit hooks (Ruff & Blacks)
  • Linting of code
  • Creation of feature selection mode for AO
  • max_features kwarg added for feature selection
  • moved source code into an src folder & modified dockerfile
  • created utils subfolder
  • created models subfolder
  • added in structure.md within ./src to explain the structure format adopted
  • renamed mode scripts to have a mode_ pre-fix
  • added extra assertion in mode checks
  • removed redundant check_config, check_keys, load_params_json, activate
  • refactored plot_graphs to remove unnessicary plot_dict var
  • extracted load funcs from utils into own file
  • extracted save funcs from utils into own file
  • extracted util funcs from mode_plotting into respective files
  • extracted plotting funcs from mode_plotting into own files
  • removed exit() calls
  • bugfixes
  • extracted save funcs from models into utils.save
  • extracted load funcs from data_processing into utils.load
  • streamlined init
  • removed globals() call
  • extracted funcs from data_processing into appropriate locations, and file fully dispersed
  • changed call of parsing FS entry to correct place
  • separated problem specific plots into own files
  • extracted shap and permutation importance plots into their own subfolders given their importance
  • consolidate metric functions into one location
  • consolidate model defs into one location
  • custom model clean up
  • shifted repeated save_model code from TabAuto children into TabAuto
  • shifted repeated load_model code from TabAuto children into TabAuto, except for FixedKeras & AutoKeras being keras based they load in a different was so have left their re-implementation of load model
  • shifted repeated fit code from TabAuto children into TabAuto
  • shifted repeated init code from TabAuto children into TabAuto
  • generalised repeated _define_model code and shifted from TabAuto children into TabAuto
  • removed repeated set_params from TabAuto
  • removed unnessicary data being passed to plot_reg, plot_clf, perm_imp and most of plot_both functions
  • removed unnessicary data being passed to plots_shap functions and separated out problem specific logic into separate function.
  • extract repeated getting model_path code to function
  • streamlined custom model setup
  • TabAuto Class merged with CustomModel Class
  • Deleted autogulon src code, as was depreciated ages ago
  • Custom SKLearnModel code only ever instansiated in auto mode for AutoSklearn so removed redundant code
  • added test coverage config
  • tests, type hinting and docstring created for:
    • metrics.define_scorers
    • utils.ml.class_balancing
    • utils.ml.data_split.std_split & strat_split
    • plotting.plot_utils
  • reduce ussage of full config dict in:
    • R_replacement.py
  • added probabilities for predictions when doing classification problems
  • code comment clean up
  • multiline strings causing extra space fixed
  • added pytest marks to distinguish binary classiification problems
  • microbiome pre-processing bugfix
  • lgbm warning re: suggest_loguniform & suggest_uniform resolved
  • AutoKeras bugfix re: batch_size
  • tests_mode grabbing the wrong file at times when lots present
  • Arm fixes for previous security fix
  • test_modes bugfix
  • name change within code base & repo (from Auto-Omics to AutoXAI4Omics)
  • provided example data & config users can run
  • corrected test_model_outputs
  • added skip conditions for test_omic_datasets
  • bug fix of xticks for bar & box plots
  • added statement about supported python version to Readme and dev manual
  • License file created and headers added to files
  • fixed commenting of license in requirements.txt
  • .pre-commit-config.yaml cleaned
  • tests requiring a container marked and fixture added to build the container
  • CONTRIBUTING.md added and info added to DEV_MANUAL.md
  • DEV_MANUAL.md updated
  • Detect secrets added
  • Upgraded python base image from 3.9.14 to 3.9.18 for additional security fixes