Python developer need for algorithm task

Adjust the XGBoost (that uses quantile sketch and gradients histograms) ML algorithm so that the training\building of each tree of the XGBoost would be based on the sum of the gradient and hessian histograms that are collected from different parties. The parties can be considered as two computers that hold the data (so the data is partitioned between two computers or entities). Every party calculates the gradient and hessian histogram for all features for a given node of the tree, then we sum the histograms of each feature and give it back to the parties to divide the data according to the summed histograms. This process repeats for each node of the tree.
please check the attached file