Error to predict when using a model generated with --olm

[lucmarchandhsj]

Hi. I can generate model using train and evaluate using predict without problem with standard flags but if I use --olm in the train fonction, the predict gives the following error:

Error in if (ncol(newx) != p) stop(paste0("The number of variables in newx must be ", :
argument is of length zero
Calls: predict -> predict.glmnet

Any clue of what is wrong?

[lucmarchandhsj]

I ended up commenting out this line and now it's working:

model <- model[which(names(model) %in% c('beta', 'lambda', 'a0', 'offset'))]