Closed lhjohn closed 1 year ago
The training cache allows us to resume training at the last unsolved hyperparameter combination. However, when we try to resume training from a completed cache it will return NA for the index to return from and throw an exception: https://github.com/OHDSI/DeepPatientLevelPrediction/blob/2aba7580f8410d623dfb6f012ce6fed95bdf417f/R/TrainingCache-class.R#L75
NA
Instead, in the case of a full cache this for loop in gridCvDeep could be skipped to directly train the final model: https://github.com/OHDSI/DeepPatientLevelPrediction/blob/2aba7580f8410d623dfb6f012ce6fed95bdf417f/R/Estimator.R#L313
gridCvDeep
Fixed in develop
The training cache allows us to resume training at the last unsolved hyperparameter combination. However, when we try to resume training from a completed cache it will return
NA
for the index to return from and throw an exception: https://github.com/OHDSI/DeepPatientLevelPrediction/blob/2aba7580f8410d623dfb6f012ce6fed95bdf417f/R/TrainingCache-class.R#L75Instead, in the case of a full cache this for loop in
gridCvDeep
could be skipped to directly train the final model: https://github.com/OHDSI/DeepPatientLevelPrediction/blob/2aba7580f8410d623dfb6f012ce6fed95bdf417f/R/Estimator.R#L313