AutoXGBoostClassifier sometimes rejects "f" metric

[eafpres]

I am running the following:

xgboost_results <- AutoXGBoostClassifier(data = cbind(train_x_data, act = train_y_data[, 1]), ValidationData = cbind(val_x_data, act = val_y_data[, 1])[1:(ceiling(nrow(val_x_data) / 2)), ], TestData = cbind(val_x_data, act = val_y_data[, 1])[(ceiling(nrow(val_x_data) / 2) + 1):nrow(val_x_data), ], TargetColumnName = "act", FeatureColNames = seq(1, ncol(train_x_data)), Trees = 25, GridTune = TRUE, MaxModelsInGrid = 15, grid_eval_metric = "f", TreeMethod = "hist", ModelID = "xgboost_rev6", NThreads = 256)

and get this error:
Error in metric %chin% c("auc", "tpr", "tnr", "prbe", "f", "odds") :
object 'metric' not found

Changing the metric to "auc" it runs. However, since "f" is in the allowed list, why is this occurring?

Also, I'm "sure" I ran it with "f" before, but now I cannot reproduce how.

[AdrianAntico]

I just had to make a small change. In the AutoXGBoostClassifier() function, the grid_eval_metric is used to determine which of the models built in the grid search is the winner. Based on the error output, "metric" was being used instead of "grid_eval_metric" like the rest of the function was using. I fixed that and you should be good to go once you re-install.

Thanks for letting me know!

[AdrianAntico]

Turns out that AutoCatBoostClassifier() had the same issue. Just fixed that one too.

[eafpres]

I will reinstall and verify; if OK I will close this. Thank you for taking care of this so quickly.

[eafpres]

Confirmed it now works.