Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[Breaking] Switch from rabit to the collective communicator (#8257)

Browse Source 
* Switch from rabit to the collective communicator

* fix size_t specialization

* really fix size_t

* try again

* add include

* more include

* fix lint errors

* remove rabit includes

* fix pylint error

* return dict from communicator context

* fix communicator shutdown

* fix dask test

* reset communicator mocklist

* fix distributed tests

* do not save device communicator

* fix jvm gpu tests

* add python test for federated communicator

* Update gputreeshap submodule

Co-authored-by: Hyunsu Philip Cho <chohyu01@cs.washington.edu>
This commit is contained in:
Rong Ou
2022-10-05 15:39:01 -07:00
committed by GitHub
parent e47b3a3da3
commit 668b8a0ea4
79 changed files with 805 additions and 2212 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/data/iterative_dmatrix.cc
View File

@@ -3,13 +3,11 @@
*/
#include "iterative_dmatrix.h"
#include <rabit/rabit.h>
#include <algorithm> // std::copy
#include "../collective/communicator-inl.h"
#include "../common/categorical.h" // common::IsCat
#include "../common/column_matrix.h"
#include "../common/hist_util.h" // common::HistogramCuts
#include "../tree/param.h" // FIXME(jiamingy): Find a better way to share this parameter.
#include "gradient_index.h"
#include "proxy_dmatrix.h"
@@ -140,7 +138,7 @@ void IterativeDMatrix::InitFromCPU(DataIterHandle iter_handle, float missing,
// We use do while here as the first batch is fetched in ctor
if (n_features == 0) {
n_features = num_cols();
rabit::Allreduce<rabit::op::Max>(&n_features, 1);
collective::Allreduce<collective::Operation::kMax>(&n_features, 1);
column_sizes.resize(n_features);
info_.num_col_ = n_features;
} else {
@@ -157,7 +155,7 @@ void IterativeDMatrix::InitFromCPU(DataIterHandle iter_handle, float missing,
// From here on Info() has the correct data shape
Info().num_row_ = accumulated_rows;
Info().num_nonzero_ = nnz;
rabit::Allreduce<rabit::op::Max>(&info_.num_col_, 1);
collective::Allreduce<collective::Operation::kMax>(&info_.num_col_, 1);
CHECK(std::none_of(column_sizes.cbegin(), column_sizes.cend(), [&](auto f) {
return f > accumulated_rows;
})) << "Something went wrong during iteration.";