From 790da458e7575affb53a5dab5750d7940d1d4945 Mon Sep 17 00:00:00 2001
From: EvanChong <remixzyj@gmail.com>
Date: Wed, 29 Nov 2017 11:19:12 -0800
Subject: [PATCH] Sync number of features after loaded matrix in different
 workers. (#2722)

---
 src/data/data.cc | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/src/data/data.cc b/src/data/data.cc
index 7a092f459..d7887a363 100644
--- a/src/data/data.cc
+++ b/src/data/data.cc
@@ -210,6 +210,10 @@ DMatrix* DMatrix::Load(const std::string& uri,
     LOG(CONSOLE) << dmat->info().num_row << 'x' << dmat->info().num_col << " matrix with "
                  << dmat->info().num_nonzero << " entries loaded from " << uri;
   }
+  /* sync up number of features after matrix loaded.
+   * partitioned data will fail the train/val validation check 
+   * since partitioned data not knowing the real number of features. */
+  rabit::Allreduce<rabit::op::Max>(&dmat->info().num_col, 1);
   // backward compatiblity code.
   if (!load_row_split) {
     MetaInfo& info = dmat->info();