Truncated error message of XGBoost when tuning

[Siegfried]

I failed to tune my XGBoost regression model. And I'm having problems locating the error because the error messages were truncated like this:

❯  Generating a set of 3 initial parameter results
x Fold1: preprocessor 1/1, model 1/3: Error in xgboost::xgb.DMatrix(x, label = y,...
x Fold1: preprocessor 1/1, model 2/3: Error in xgboost::xgb.DMatrix(x, label = y,...
x Fold1: preprocessor 1/1, model 3/3: Error in xgboost::xgb.DMatrix(x, label = y,...
x Fold2: preprocessor 1/1, model 1/3: Error in xgboost::xgb.DMatrix(x, label = y,...
x Fold2: preprocessor 1/1, model 2/3: Error in xgboost::xgb.DMatrix(x, label = y,...
x Fold2: preprocessor 1/1, model 3/3: Error in xgboost::xgb.DMatrix(x, label = y,...
✓ Initialization complete

How can I get the complete messages? Thanks.

[arthur.t]

Sorry I don't have a more complete answer, but perhaps you're tuning a grid of hyperparameters? Maybe if you run the function in a simpler way (only one value of parameters) then you'll see the full message.

[Max]

A reproducible example would help.

I think that you are using tune::tune_grid() for this output. If so, check the .notes column for the full message.

[hnagaty]

I use something like the below code to be able to print the full error message. Ofcourse, you will need to modify it to your code.
I'm not sure if there is a more neat solution or not. I remember I once saw a Github issue suggesting a convenience function to pull the notes. I searched for it now but couldn't find it, may be my memory is playing tricks on me

pois_res %>% # this is the o/p object of tune_grid()
  chuck(".notes",  1,  1) %>%
  print()

[Siegfried]

I was using tune::tune_bayes(). .notes only works if the results are returned, if I'm correct. However, there were times no result returned.