self is localhost, we are sending the signal to the error handling thread for it to close.

* @brief Compute the sparse peer set for a given rank: ring neighbors union binomial tree * neighbors (rooted at rank 0). */

NOLINT

customization part synchronize the best solution of each node

Synchronize feature type in case of empty DMatrix

* Train a regression model and save it into JSON model file. */

Compute the path from `src` to rank 0 through the binomial tree (excluding 0).

try to prune off current leaf

! * \brief update the split entry, replace it if e is better * \param e candidate split solution * \return whether the proposed split is bett

! * \brief update the split entry, replace it if e is better * \param e candidate split solution * \return whether the proposed split is bett

Update position biased for unbiased click data

! \brief update queue expand add in new leaves */

* \brief Updates the gradient vector with respect to a change in weight. * * \param fidx The feature index. * \param group_idx Zero-based inde

update the solution candidate

! * \brief Whether should we use onehot encoding for categorical data. */