Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[refactor] Refactor __init__ of abstract evaluator
[refactor] Collect shared variables in NamedTuples [fix] Copy the budget passed to the evaluator params [refactor] Add cross validation result manager for separate management [refactor] Separate pipeline classes from abstract evaluator [refactor] Refactor tae.py [refactor] Increase the safety level of pipeline config [test] Fix test_evaluation.py [test] Fix test_abstract_evaluator.py 1 -- 3 [test] Add default pipeline config [test] Modify queue.empty in a safer way [test] Fix test_api.py [test] Fix test_train_evaluator.py [refactor] Refactor test_api.py before adding new tests [refactor] Refactor test_tabular_xxx [fix] Find the error in test_tabular_xxx Since pipeline is updated after the evaluations and the previous code updated self.pipeline in the predict method, dummy class only needs to override this method. However, the new code does it separately, so I override get_pipeline method so that we can reproduce the same results. [fix] Fix the shape issue in regression and add bug comment in a test [refactor] Use keyword args to avoid unexpected bugs [fix] Fix the ground truth of test_cv Since we changed the weighting strategy for the cross validation in the validation phase so that we weight performance from each model proportionally to the size of each VALIDATION split. I needed to change the answer. Note that the previous was weighting the performance proportionally to the TRAINING splits for both training and validation phases. [fix] Change qsize --> Empty since qsize might not be reliable [refactor] Add cost for crash in autoPyTorchMetrics [test] Remove self.pipeline since this is a duplication of self.pipelines [fix] Fix attribute errors caused by the last change in curve extraction [fix] Fix the issue when taking num_classes from regression task [fix] Deactivate the save of cv model in the case of holdout
- Loading branch information