Multiclass Training - WARNING: amalgamation/../src/learner.cc:541

Hi,

I've been using this package for over a year now and I really like it. Earlier in the year, I made an update and everything still worked fine until the following update I made about a week ago. Calling the "train_model" function below, on the features (data) and errors given in the files at the end of this message, makes it unable to make predictions (using the function stats::predict below) and I now obtain this warning message:

*
WARNING: amalgamation/../src/learner.cc:541:
Parameters: { silent } might not be used.

This may not be accurate due to some parameters are only used in language bindings but
passed down to XGBoost core. Or some parameters are not used but slip through this
verification. Please open an issue if you find above cases.
*

I use these functions in a multiclassification problem using the classification errors and the features as input. If the issue is due to a misuse of your functions, I'd appreciate your corrections to my code.

Thank you, in advance, for your kind help :)

# Train Function
train_model <- function (data, errors) {
dtrain <- xgboost::xgb.DMatrix(data)
attr(dtrain, "errors") <- errors

param <- list(max_depth=14, eta=0.575188, nthread = 3, silent=1,
objective=error_softmax_obj,
num_class=ncol(errors),
subsample=0.9161483,
colsample_bytree=0.7670739, eval_metric = "merror"
)

return(xgboost::xgb.train(param, dtrain, 94))
}

# Prediction line
pred <- stats::predict(xgb_returned_model, new_data, outputmargin = TRUE, reshape=TRUE)

# Folder containing sample of features and errors
[issue_xgboost.zip](https://github.com/dmlc/xgboost/files/6305841/issue_xgboost.zip)