OHDSI Home | Forums | Wiki | Github

PatientLevelPrediction Package Error

Developers
#1

I am running into the following error when training a model with Patient Level Prediction R package. I am training a model using standard features and custom attributes defined cohort attributes tables in CDM database.

Limiting covariate data took: 42.5452976147334 mins
Train Set:
Fold 1 1271393 patients with 21393 outcomes - Fold 2 1271392 patients with 21392 outcomes - Fold 3 1271392 patients with 21392 outcomes
398763 covariates in train data
Test Set:
1271392 patients with 21392 outcomes
Starting data sampling
Applying sameData
No sampling - returning same data
Finished data sampling
Train Set:
Fold 1 1271393 patients with 21393 outcomes - Fold 2 1271392 patients with 21392 outcomes - Fold 3 1271392 patients with 21392 outcomes
398763 covariates in train data
Test Set:
1271392 patients with 21392 outcomes
Starting Feature Engineering
Applying sameData
No sampling - returning same data
Done Feature Engineering
Train Set:
Fold 1 1271393 patients with 21393 outcomes - Fold 2 1271392 patients with 21392 outcomes - Fold 3 1271392 patients with 21392 outcomes
398763 covariates in train data
Test Set:
1271392 patients with 21392 outcomes
Removing 1 redundant covariates
Removing 380592 infrequent covariates
Normalizing covariates
Error: database or disk is full

Error in runPlp(plpData = plpData, outcomeId = xxx, analysisId = “abcdsds”, :
train data NULL after preprocessing
Execution halted

|2023-11-16 20:21:45|[Main thread]|ERROR|PatientLevelPrediction|3|Error: database or disk is full
|2023-11-16 20:21:45|[Main thread]|FATAL|PatientLevelPrediction|runPlp|train data NULL after preprocessing|

This error is being generated internally within the PLP package. Can you let me know where is this data being stored? .

Thanks
Mahi

#2

There are two places internal data are stored:

  1. The result artifacts are stored in the saveDirectory specified by the user.
  2. Intermediate objects are temporarily stored in the Andromeda temp folder. You can specify where that is by setting the option, like:
    options(andromedaTempFolder = "d:/andromedaTemp")
t