Is there any way to implement Pearson Correlation as loss function on XGBoost?

Question

I\'m trying to implement Pearson Correlation for XGBoost, I wrote the gradient and the hessian code, however the training does not start because the first prediction vector