[backport] [doc][R] Update link. (#8998) (#9001)

* [CI] Fix Windows wheel to be compatible with Poetry

Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>

CMakeLists.txt

@@ -1,5 +1,5 @@
-cmake_minimum_required(VERSION 3.14 FATAL_ERROR)
+cmake_minimum_required(VERSION 3.18 FATAL_ERROR)
-project(xgboost LANGUAGES CXX C VERSION 1.7.4)
+project(xgboost LANGUAGES CXX C VERSION 1.7.5)
 include(cmake/Utils.cmake)
 list(APPEND CMAKE_MODULE_PATH "${xgboost_SOURCE_DIR}/cmake/modules")
 cmake_policy(SET CMP0022 NEW)
@@ -168,9 +168,6 @@ find_package(Threads REQUIRED)
 
 if (USE_OPENMP)
   if (APPLE)
-    # Require CMake 3.16+ on Mac OSX, as previous versions of CMake had trouble locating
-    # OpenMP on Mac. See https://github.com/dmlc/xgboost/pull/5146#issuecomment-568312706
-    cmake_minimum_required(VERSION 3.16)
     find_package(OpenMP)
     if (NOT OpenMP_FOUND)
       # Try again with extra path info; required for libomp 15+ from Homebrew

R-package/CMakeLists.txt

@@ -31,7 +31,7 @@ if (USE_OPENMP)
 endif (USE_OPENMP)
 set_target_properties(
   xgboost-r PROPERTIES
-  CXX_STANDARD 14
+  CXX_STANDARD 17
   CXX_STANDARD_REQUIRED ON
   POSITION_INDEPENDENT_CODE ON)
 

R-package/DESCRIPTION

@@ -1,8 +1,8 @@
 Package: xgboost
 Type: Package
 Title: Extreme Gradient Boosting
-Version: 1.7.4.1
+Version: 1.7.5.1
-Date: 2023-02-15
+Date: 2023-03-29
 Authors@R: c(
   person("Tianqi", "Chen", role = c("aut"),
          email = "tianqi.tchen@gmail.com"),
@@ -66,5 +66,6 @@ Imports:
     methods,
     data.table (>= 1.9.6),
     jsonlite (>= 1.0),
-RoxygenNote: 7.2.2
+RoxygenNote: 7.2.3
-SystemRequirements: GNU make, C++14
+Encoding: UTF-8
+SystemRequirements: GNU make, C++17

R-package/LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2014 by Tianqi Chen and Contributors 
+Copyright (c) 2014-2023, Tianqi Chen and XBGoost Contributors
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.

R-package/R/xgb.plot.tree.R

@@ -34,7 +34,7 @@
 #' The branches that also used for missing values are marked as bold
 #' (as in "carrying extra capacity").
 #'
-#' This function uses \href{http://www.graphviz.org/}{GraphViz} as a backend of DiagrammeR.
+#' This function uses \href{https://www.graphviz.org/}{GraphViz} as a backend of DiagrammeR.
 #'
 #' @return
 #'

R-package/configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.71 for xgboost 1.7.4.
+# Generated by GNU Autoconf 2.71 for xgboost 1.7.5.
 #
 #
 # Copyright (C) 1992-1996, 1998-2017, 2020-2021 Free Software Foundation,
@@ -607,8 +607,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='xgboost'
 PACKAGE_TARNAME='xgboost'
-PACKAGE_VERSION='1.7.4'
+PACKAGE_VERSION='1.7.5'
-PACKAGE_STRING='xgboost 1.7.4'
+PACKAGE_STRING='xgboost 1.7.5'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1225,7 +1225,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures xgboost 1.7.4 to adapt to many kinds of systems.
+\`configure' configures xgboost 1.7.5 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1287,7 +1287,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of xgboost 1.7.4:";;
+     short | recursive ) echo "Configuration of xgboost 1.7.5:";;
    esac
   cat <<\_ACEOF
 
@@ -1367,7 +1367,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-xgboost configure 1.7.4
+xgboost configure 1.7.5
 generated by GNU Autoconf 2.71
 
 Copyright (C) 2021 Free Software Foundation, Inc.
@@ -1533,7 +1533,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by xgboost $as_me 1.7.4, which was
+It was created by xgboost $as_me 1.7.5, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -2096,9 +2096,9 @@ if test -z "${R_HOME}"; then
   exit 1
 fi
 
-CXX14=`"${R_HOME}/bin/R" CMD config CXX14`
+CXX17=`"${R_HOME}/bin/R" CMD config CXX17`
-CXX14STD=`"${R_HOME}/bin/R" CMD config CXX14STD`
+CXX17STD=`"${R_HOME}/bin/R" CMD config CXX17STD`
-CXX="${CXX14} ${CXX14STD}"
+CXX="${CXX17} ${CXX17STD}"
 CXXFLAGS=`"${R_HOME}/bin/R" CMD config CXXFLAGS`
 
 CC=`"${R_HOME}/bin/R" CMD config CC`
@@ -3412,7 +3412,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by xgboost $as_me 1.7.4, which was
+This file was extended by xgboost $as_me 1.7.5, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -3467,7 +3467,7 @@ ac_cs_config_escaped=`printf "%s\n" "$ac_cs_config" | sed "s/^ //; s/'/'\\\\\\\\
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config='$ac_cs_config_escaped'
 ac_cs_version="\\
-xgboost config.status 1.7.4
+xgboost config.status 1.7.5
 configured by $0, generated by GNU Autoconf 2.71,
   with options \\"\$ac_cs_config\\"
 

R-package/configure.ac

@@ -2,7 +2,7 @@
 
 AC_PREREQ(2.69)
 
-AC_INIT([xgboost],[1.7.4],[],[xgboost],[])
+AC_INIT([xgboost],[1.7.5],[],[xgboost],[])
 
 : ${R_HOME=`R RHOME`}
 if test -z "${R_HOME}"; then
@@ -10,9 +10,9 @@ if test -z "${R_HOME}"; then
   exit 1
 fi
 
-CXX14=`"${R_HOME}/bin/R" CMD config CXX14`
+CXX17=`"${R_HOME}/bin/R" CMD config CXX17`
-CXX14STD=`"${R_HOME}/bin/R" CMD config CXX14STD`
+CXX17STD=`"${R_HOME}/bin/R" CMD config CXX17STD`
-CXX="${CXX14} ${CXX14STD}"
+CXX="${CXX17} ${CXX17STD}"
 CXXFLAGS=`"${R_HOME}/bin/R" CMD config CXXFLAGS`
 
 CC=`"${R_HOME}/bin/R" CMD config CC`

R-package/man/xgb.plot.tree.Rd

@@ -67,7 +67,7 @@ The "Yes" branches are marked by the "< split_value" label.
 The branches that also used for missing values are marked as bold
 (as in "carrying extra capacity").
 
-This function uses \href{http://www.graphviz.org/}{GraphViz} as a backend of DiagrammeR.
+This function uses \href{https://www.graphviz.org/}{GraphViz} as a backend of DiagrammeR.
 }
 \examples{
 data(agaricus.train, package='xgboost')

R-package/src/Makevars.in

@@ -3,7 +3,7 @@ PKGROOT=../../
 ENABLE_STD_THREAD=1
 # _*_ mode: Makefile; _*_
 
-CXX_STD = CXX14
+CXX_STD = CXX17
 
 XGB_RFLAGS = -DXGBOOST_STRICT_R_MODE=1 -DDMLC_LOG_BEFORE_THROW=0\
            -DDMLC_ENABLE_STD_THREAD=$(ENABLE_STD_THREAD) -DDMLC_DISABLE_STDIN=1\

R-package/src/Makevars.win

@@ -3,7 +3,7 @@ PKGROOT=../../
 ENABLE_STD_THREAD=0
 # _*_ mode: Makefile; _*_
 
-CXX_STD = CXX14
+CXX_STD = CXX17
 
 XGB_RFLAGS = -DXGBOOST_STRICT_R_MODE=1 -DDMLC_LOG_BEFORE_THROW=0\
            -DDMLC_ENABLE_STD_THREAD=$(ENABLE_STD_THREAD) -DDMLC_DISABLE_STDIN=1\

cmake/Utils.cmake

@@ -178,17 +178,10 @@ function(xgboost_set_cuda_flags target)
       $<$<COMPILE_LANGUAGE:CUDA>:-Xcompiler=/utf-8>)
   endif (MSVC)
 
-  if (PLUGIN_RMM)
   set_target_properties(${target} PROPERTIES
     CUDA_STANDARD 17
     CUDA_STANDARD_REQUIRED ON
     CUDA_SEPARABLE_COMPILATION OFF)
-  else ()
-    set_target_properties(${target} PROPERTIES
-      CUDA_STANDARD 14
-      CUDA_STANDARD_REQUIRED ON
-      CUDA_SEPARABLE_COMPILATION OFF)
-  endif (PLUGIN_RMM)
 endfunction(xgboost_set_cuda_flags)
 
 macro(xgboost_link_nccl target)
@@ -205,17 +198,10 @@ endmacro(xgboost_link_nccl)
 
 # compile options
 macro(xgboost_target_properties target)
-  if (PLUGIN_RMM)
   set_target_properties(${target} PROPERTIES
     CXX_STANDARD 17
     CXX_STANDARD_REQUIRED ON
     POSITION_INDEPENDENT_CODE ON)
-  else ()
-    set_target_properties(${target} PROPERTIES
-      CXX_STANDARD 14
-      CXX_STANDARD_REQUIRED ON
-      POSITION_INDEPENDENT_CODE ON)
-  endif (PLUGIN_RMM)
 
   if (HIDE_CXX_SYMBOLS)
     #-- Hide all C++ symbols

demo/c-api/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.13)
+cmake_minimum_required(VERSION 3.18)
 project(xgboost-c-examples)
 
 add_subdirectory(basic)

demo/c-api/external-memory/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.13)
+cmake_minimum_required(VERSION 3.18)
 project(external-memory-demo LANGUAGES C VERSION 0.0.1)
 
 find_package(xgboost REQUIRED)

demo/c-api/inference/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.13)
+cmake_minimum_required(VERSION 3.18)
 project(inference-demo LANGUAGES C VERSION 0.0.1)
 find_package(xgboost REQUIRED)
 

doc/python/python_api.rst

@@ -173,3 +173,13 @@ PySpark API
     :members:
     :inherited-members:
     :show-inheritance:
+
+.. autoclass:: xgboost.spark.SparkXGBRanker
+    :members:
+    :inherited-members:
+    :show-inheritance:
+
+.. autoclass:: xgboost.spark.SparkXGBRankerModel
+    :members:
+    :inherited-members:
+    :show-inheritance:

doc/tutorials/c_api_tutorial.rst

@@ -45,7 +45,7 @@ Use ``find_package()`` and ``target_link_libraries()`` in your application's CMa
 
 .. code-block:: cmake
 
-    cmake_minimum_required(VERSION 3.13)
+    cmake_minimum_required(VERSION 3.18)
     project(your_project_name LANGUAGES C CXX VERSION your_project_version)
     find_package(xgboost REQUIRED)
     add_executable(your_project_name /path/to/project_file.c)

doc/tutorials/spark_estimator.rst

@@ -43,10 +43,10 @@ in spark estimator, and some parameters are replaced with pyspark specific param
 such as `weight_col`, `validation_indicator_col`, `use_gpu`, for details please see
 `SparkXGBRegressor` doc.
 
-The following code snippet shows how to train a spark xgboost regressor model,
+The following code snippet shows how to train a spark xgboost regressor model, first we
-first we need to prepare a training dataset as a spark dataframe contains
+need to prepare a training dataset as a spark dataframe contains "label" column and
-"label" column and "features" column(s), the "features" column(s) must be `pyspark.ml.linalg.Vector`
+"features" column(s), the "features" column(s) must be ``pyspark.ml.linalg.Vector`` type
-type or spark array type or a list of feature column names.
+or spark array type or a list of feature column names.
 
 
 .. code-block:: python
@@ -54,10 +54,10 @@ type or spark array type or a list of feature column names.
   xgb_regressor_model = xgb_regressor.fit(train_spark_dataframe)
 
 
-The following code snippet shows how to predict test data using a spark xgboost regressor model,
+The following code snippet shows how to predict test data using a spark xgboost regressor
-first we need to prepare a test dataset as a spark dataframe contains
+model, first we need to prepare a test dataset as a spark dataframe contains "features"
-"features" and "label" column, the "features" column must be `pyspark.ml.linalg.Vector`
+and "label" column, the "features" column must be ``pyspark.ml.linalg.Vector`` type or
-type or spark array type.
+spark array type.
 
 .. code-block:: python
 

include/xgboost/version_config.h

@@ -6,6 +6,6 @@
 
 #define XGBOOST_VER_MAJOR 1
 #define XGBOOST_VER_MINOR 7
-#define XGBOOST_VER_PATCH 4
+#define XGBOOST_VER_PATCH 5
 
 #endif  // XGBOOST_VERSION_CONFIG_H_

jvm-packages/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>ml.dmlc</groupId>
     <artifactId>xgboost-jvm_2.12</artifactId>
-    <version>1.7.4</version>
+    <version>1.7.5</version>
     <packaging>pom</packaging>
     <name>XGBoost JVM Package</name>
     <description>JVM Package for XGBoost</description>

jvm-packages/xgboost4j-example/pom.xml

@@ -6,10 +6,10 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j-example_2.12</artifactId>
-    <version>1.7.4</version>
+    <version>1.7.5</version>
     <packaging>jar</packaging>
     <build>
         <plugins>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-spark_${scala.binary.version}</artifactId>
-            <version>1.7.4</version>
+            <version>1.7.5</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>
@@ -37,7 +37,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-flink_${scala.binary.version}</artifactId>
-            <version>1.7.4</version>
+            <version>1.7.5</version>
         </dependency>
         <dependency>
             <groupId>org.apache.commons</groupId>

jvm-packages/xgboost4j-flink/pom.xml

@@ -6,10 +6,10 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j-flink_2.12</artifactId>
-    <version>1.7.4</version>
+    <version>1.7.5</version>
     <build>
         <plugins>
             <plugin>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j_${scala.binary.version}</artifactId>
-            <version>1.7.4</version>
+            <version>1.7.5</version>
         </dependency>
         <dependency>
             <groupId>org.apache.commons</groupId>

jvm-packages/xgboost4j-gpu/pom.xml

@@ -6,10 +6,10 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j-gpu_2.12</artifactId>
-    <version>1.7.4</version>
+    <version>1.7.5</version>
     <packaging>jar</packaging>
 
     <dependencies>

jvm-packages/xgboost4j-spark-gpu/pom.xml

@@ -6,7 +6,7 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j-spark-gpu_2.12</artifactId>
     <build>
@@ -24,7 +24,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-gpu_${scala.binary.version}</artifactId>
-            <version>1.7.4</version>
+            <version>1.7.5</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>

jvm-packages/xgboost4j-spark/pom.xml

@@ -6,7 +6,7 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j-spark_2.12</artifactId>
     <build>
@@ -24,7 +24,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j_${scala.binary.version}</artifactId>
-            <version>1.7.4</version>
+            <version>1.7.5</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>

jvm-packages/xgboost4j/pom.xml

@@ -6,10 +6,10 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.7.4</version>
+        <version>1.7.5</version>
     </parent>
     <artifactId>xgboost4j_2.12</artifactId>
-    <version>1.7.4</version>
+    <version>1.7.5</version>
     <packaging>jar</packaging>
 
     <dependencies>

plugin/CMakeLists.txt

@@ -15,7 +15,7 @@ if (PLUGIN_UPDATER_ONEAPI)
   target_link_libraries(oneapi_plugin PUBLIC -fsycl)
   set_target_properties(oneapi_plugin PROPERTIES
     COMPILE_FLAGS -fsycl
-    CXX_STANDARD 14
+    CXX_STANDARD 17
     CXX_STANDARD_REQUIRED ON
     POSITION_INDEPENDENT_CODE ON)
   if (USE_OPENMP)

python-package/xgboost/VERSION

@@ -1 +1 @@
-1.7.4
+1.7.5

python-package/xgboost/spark/__init__.py

@@ -1,6 +1,5 @@
 # type: ignore
-"""PySpark XGBoost integration interface
+"""PySpark XGBoost integration interface"""
-"""
 
 try:
     import pyspark
@@ -11,6 +10,7 @@ from .estimator import (
     SparkXGBClassifier,
     SparkXGBClassifierModel,
     SparkXGBRanker,
+    SparkXGBRankerModel,
     SparkXGBRegressor,
     SparkXGBRegressorModel,
 )
@@ -21,4 +21,5 @@ __all__ = [
     "SparkXGBRegressor",
     "SparkXGBRegressorModel",
     "SparkXGBRanker",
+    "SparkXGBRankerModel",
 ]

rabit/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.3)
+cmake_minimum_required(VERSION 3.18)
 
 find_package(Threads REQUIRED)
 

src/data/iterative_dmatrix.cc

@@ -110,6 +110,7 @@ void IterativeDMatrix::InitFromCPU(DataIterHandle iter_handle, float missing,
       size_t n_threads = ctx_.Threads();
       size_t n_features = column_sizes.size();
       linalg::Tensor<size_t, 2> column_sizes_tloc({n_threads, n_features}, Context::kCpuId);
+      column_sizes_tloc.Data()->Fill(0);
       auto view = column_sizes_tloc.HostView();
       common::ParallelFor(value.Size(), n_threads, common::Sched::Static(256), [&](auto i) {
         auto const& line = value.GetLine(i);

src/tree/updater_gpu_hist.cu

@@ -306,6 +306,8 @@ struct GPUHistMakerDevice {
         matrix.is_dense
     };
     dh::TemporaryArray<GPUExpandEntry> entries(2 * candidates.size());
+    // Store the feature set ptrs so they dont go out of scope before the kernel is called
+    std::vector<std::shared_ptr<HostDeviceVector<bst_feature_t>>> feature_sets;
     for (size_t i = 0; i < candidates.size(); i++) {
       auto candidate = candidates.at(i);
       int left_nidx = tree[candidate.nid].LeftChild();
@@ -314,10 +316,12 @@ struct GPUHistMakerDevice {
       nidx[i * 2 + 1] = right_nidx;
       auto left_sampled_features = column_sampler.GetFeatureSet(tree.GetDepth(left_nidx));
       left_sampled_features->SetDevice(ctx_->gpu_id);
+      feature_sets.emplace_back(left_sampled_features);
       common::Span<bst_feature_t> left_feature_set =
           interaction_constraints.Query(left_sampled_features->DeviceSpan(), left_nidx);
       auto right_sampled_features = column_sampler.GetFeatureSet(tree.GetDepth(right_nidx));
       right_sampled_features->SetDevice(ctx_->gpu_id);
+      feature_sets.emplace_back(right_sampled_features);
       common::Span<bst_feature_t> right_feature_set =
           interaction_constraints.Query(right_sampled_features->DeviceSpan(),
                                         right_nidx);

tests/buildkite/build-containers.sh

@@ -23,10 +23,15 @@ case "${container}" in
   gpu|rmm)
     BUILD_ARGS="$BUILD_ARGS --build-arg CUDA_VERSION_ARG=$CUDA_VERSION"
     BUILD_ARGS="$BUILD_ARGS --build-arg RAPIDS_VERSION_ARG=$RAPIDS_VERSION"
+    if [[ $container == "rmm" ]]
+    then
+      BUILD_ARGS="$BUILD_ARGS --build-arg NCCL_VERSION_ARG=$NCCL_VERSION"
+    fi
     ;;
 
   gpu_build_centos7|jvm_gpu_build)
     BUILD_ARGS="$BUILD_ARGS --build-arg CUDA_VERSION_ARG=$CUDA_VERSION"
+    BUILD_ARGS="$BUILD_ARGS --build-arg NCCL_VERSION_ARG=$NCCL_VERSION"
     ;;
 
   *)

tests/buildkite/build-cuda-with-rmm.sh

@@ -15,7 +15,8 @@ fi
 
 command_wrapper="tests/ci_build/ci_build.sh rmm docker --build-arg "`
                 `"CUDA_VERSION_ARG=$CUDA_VERSION --build-arg "`
-                `"RAPIDS_VERSION_ARG=$RAPIDS_VERSION"
+                `"RAPIDS_VERSION_ARG=$RAPIDS_VERSION --build-arg "`
+                `"NCCL_VERSION_ARG=$NCCL_VERSION"
 
 echo "--- Build libxgboost from the source"
 $command_wrapper tests/ci_build/build_via_cmake.sh --conda-env=gpu_test -DUSE_CUDA=ON \

tests/buildkite/build-cuda.sh

@@ -16,7 +16,8 @@ else
 fi
 
 command_wrapper="tests/ci_build/ci_build.sh gpu_build_centos7 docker --build-arg "`
-                `"CUDA_VERSION_ARG=$CUDA_VERSION"
+                `"CUDA_VERSION_ARG=$CUDA_VERSION --build-arg "`
+                `"NCCL_VERSION_ARG=$NCCL_VERSION"
 
 echo "--- Build libxgboost from the source"
 $command_wrapper tests/ci_build/prune_libnccl.sh

tests/buildkite/build-jvm-packages-gpu.sh

@@ -14,5 +14,7 @@ else
 fi
 
 tests/ci_build/ci_build.sh jvm_gpu_build nvidia-docker \
-  --build-arg CUDA_VERSION_ARG=${CUDA_VERSION} tests/ci_build/build_jvm_packages.sh \
+  --build-arg CUDA_VERSION_ARG=${CUDA_VERSION} \
+  --build-arg NCCL_VERSION_ARG=${NCCL_VERSION} \
+  tests/ci_build/build_jvm_packages.sh \
   ${SPARK_VERSION} -Duse.cuda=ON ${arch_flag}

tests/buildkite/build-win64-gpu.ps1

@@ -12,10 +12,10 @@ if ( $is_release_branch -eq 0 ) {
 }
 mkdir build
 cd build
-cmake .. -G"Visual Studio 15 2017 Win64" -DUSE_CUDA=ON -DCMAKE_VERBOSE_MAKEFILE=ON `
+cmake .. -G"Visual Studio 17 2022" -A x64 -DUSE_CUDA=ON -DCMAKE_VERBOSE_MAKEFILE=ON `
-  -DGOOGLE_TEST=ON -DUSE_DMLC_GTEST=ON -DCMAKE_UNITY_BUILD=ON ${arch_flag}
+  -DGOOGLE_TEST=ON -DUSE_DMLC_GTEST=ON ${arch_flag}
 $msbuild = -join @(
-  "C:\\Program Files (x86)\\Microsoft Visual Studio\\2017\\Community\\MSBuild\\15.0"
+  "C:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\MSBuild\\Current"
   "\\Bin\\MSBuild.exe"
 )
 & $msbuild xgboost.sln /m /p:Configuration=Release /nodeReuse:false

tests/buildkite/conftest.sh

@@ -22,9 +22,10 @@ function set_buildkite_env_vars_in_container {
 
 set -x
 
-CUDA_VERSION=11.0.3
+CUDA_VERSION=11.8.0
-RAPIDS_VERSION=22.10
+NCCL_VERSION=2.16.5-1
-SPARK_VERSION=3.0.1
+RAPIDS_VERSION=23.02
+SPARK_VERSION=3.1.1
 JDK_VERSION=8
 
 if [[ -z ${BUILDKITE:-} ]]

tests/buildkite/deploy-jvm-packages.sh

@@ -9,5 +9,6 @@ then
   echo "--- Deploy JVM packages to xgboost-maven-repo S3 repo"
   tests/ci_build/ci_build.sh jvm_gpu_build docker \
     --build-arg CUDA_VERSION_ARG=${CUDA_VERSION} \
+    --build-arg NCCL_VERSION_ARG=${NCCL_VERSION} \
     tests/ci_build/deploy_jvm_packages.sh ${SPARK_VERSION}
 fi

tests/buildkite/infrastructure/aws-stack-creator/create_stack.py

@@ -2,12 +2,16 @@ import argparse
 import copy
 import os
 import re
+import sys
 
 import boto3
 import botocore
 from metadata import AMI_ID, COMMON_STACK_PARAMS, STACK_PARAMS
 
 current_dir = os.path.dirname(__file__)
+sys.path.append(os.path.join(current_dir, ".."))
+
+from common_blocks.utils import create_or_update_stack, wait
 
 TEMPLATE_URL = "https://s3.amazonaws.com/buildkite-aws-stack/latest/aws-stack.yml"
 
@@ -68,72 +72,7 @@ def get_full_stack_id(stack_id):
     return f"buildkite-{stack_id}-autoscaling-group"
 
 
-def stack_exists(args, *, stack_name):
+def create_agent_iam_policy(args, *, client):
-    client = boto3.client("cloudformation", region_name=args.aws_region)
-    waiter = client.get_waiter("stack_exists")
-    try:
-        waiter.wait(StackName=stack_name, WaiterConfig={"MaxAttempts": 1})
-        return True
-    except botocore.exceptions.WaiterError as e:
-        return False
-
-
-def create_or_update_stack(
-    args, *, stack_name, template_url=None, template_body=None, params=None
-):
-    kwargs = {
-        "StackName": stack_name,
-        "Capabilities": [
-            "CAPABILITY_IAM",
-            "CAPABILITY_NAMED_IAM",
-            "CAPABILITY_AUTO_EXPAND",
-        ],
-    }
-    if template_url:
-        kwargs["TemplateURL"] = template_url
-    if template_body:
-        kwargs["TemplateBody"] = template_body
-    if params:
-        kwargs["Parameters"] = params
-
-    client = boto3.client("cloudformation", region_name=args.aws_region)
-
-    if stack_exists(args, stack_name=stack_name):
-        print(f"Stack {stack_name} already exists. Updating...")
-        try:
-            response = client.update_stack(**kwargs)
-            return {"StackName": stack_name, "Action": "update"}
-        except botocore.exceptions.ClientError as e:
-            if e.response["Error"]["Code"] == "ValidationError" and re.search(
-                "No updates are to be performed", e.response["Error"]["Message"]
-            ):
-                print(f"No update was made to {stack_name}")
-                return {"StackName": stack_name, "Action": "noop"}
-            else:
-                raise e
-    else:
-        kwargs.update({"OnFailure": "ROLLBACK", "EnableTerminationProtection": False})
-        response = client.create_stack(**kwargs)
-        return {"StackName": stack_name, "Action": "create"}
-
-
-def wait(promise):
-    client = boto3.client("cloudformation", region_name=args.aws_region)
-    stack_name = promise["StackName"]
-    print(f"Waiting for {stack_name}...")
-    if promise["Action"] == "create":
-        waiter = client.get_waiter("stack_create_complete")
-        waiter.wait(StackName=stack_name)
-        print(f"Finished creating stack {stack_name}")
-    elif promise["Action"] == "update":
-        waiter = client.get_waiter("stack_update_complete")
-        waiter.wait(StackName=stack_name)
-        print(f"Finished updating stack {stack_name}")
-    elif promise["Action"] != "noop":
-        raise ValueError(f"Invalid promise {promise}")
-
-
-def create_agent_iam_policy(args):
     policy_stack_name = "buildkite-agent-iam-policy"
     print(f"Creating stack {policy_stack_name} for agent IAM policy...")
     with open(
@@ -142,9 +81,9 @@ def create_agent_iam_policy(args):
     ) as f:
         policy_template = f.read()
     promise = create_or_update_stack(
-        args, stack_name=policy_stack_name, template_body=policy_template
+        args, client=client, stack_name=policy_stack_name, template_body=policy_template
     )
-    wait(promise)
+    wait(promise, client=client)
 
     cf = boto3.resource("cloudformation", region_name=args.aws_region)
     policy = cf.StackResource(policy_stack_name, "BuildkiteAgentManagedPolicy")
@@ -152,10 +91,10 @@ def create_agent_iam_policy(args):
 
 
 def main(args):
-    agent_iam_policy = create_agent_iam_policy(args)
-
     client = boto3.client("cloudformation", region_name=args.aws_region)
 
+    agent_iam_policy = create_agent_iam_policy(args, client=client)
+
     promises = []
 
     for stack_id in AMI_ID:
@@ -167,13 +106,17 @@ def main(args):
         )
 
         promise = create_or_update_stack(
-            args, stack_name=stack_id_full, template_url=TEMPLATE_URL, params=params
+            args,
+            client=client,
+            stack_name=stack_id_full,
+            template_url=TEMPLATE_URL,
+            params=params,
         )
         promises.append(promise)
         print(f"CI stack {stack_id_full} is in progress in the background")
 
     for promise in promises:
-        wait(promise)
+        wait(promise, client=client)
 
 
 if __name__ == "__main__":

tests/buildkite/infrastructure/aws-stack-creator/metadata.py

@@ -1,27 +1,27 @@
 AMI_ID = {
     # Managed by XGBoost team
     "linux-amd64-gpu": {
-        "us-west-2": "ami-00ed92bd37f77bc33",
+        "us-west-2": "ami-094271bed4788ddb5",
     },
     "linux-amd64-mgpu": {
-        "us-west-2": "ami-00ed92bd37f77bc33",
+        "us-west-2": "ami-094271bed4788ddb5",
     },
     "windows-gpu": {
-        "us-west-2": "ami-0a1a2ea551a07ad5f",
+        "us-west-2": "ami-0839681594a1d7627",
     },
     "windows-cpu": {
-        "us-west-2": "ami-0a1a2ea551a07ad5f",
+        "us-west-2": "ami-0839681594a1d7627",
     },
     # Managed by BuildKite
     # from https://s3.amazonaws.com/buildkite-aws-stack/latest/aws-stack.yml
     "linux-amd64-cpu": {
-        "us-west-2": "ami-075d4c25d5f0c17c1",
+        "us-west-2": "ami-00f2127550cf03658",
     },
     "pipeline-loader": {
-        "us-west-2": "ami-075d4c25d5f0c17c1",
+        "us-west-2": "ami-00f2127550cf03658",
     },
     "linux-arm64-cpu": {
-        "us-west-2": "ami-0952c6fb6db9a9891",
+        "us-west-2": "ami-0c5789068f4a2d1b5",
     },
 }
 

tests/buildkite/infrastructure/common_blocks/utils.py

@@ -0,0 +1,97 @@
+import re
+
+import boto3
+import botocore
+
+
+def stack_exists(args, *, stack_name):
+    client = boto3.client("cloudformation", region_name=args.aws_region)
+    waiter = client.get_waiter("stack_exists")
+    try:
+        waiter.wait(StackName=stack_name, WaiterConfig={"MaxAttempts": 1})
+        return True
+    except botocore.exceptions.WaiterError as e:
+        return False
+
+
+def create_or_update_stack(
+    args, *, client, stack_name, template_url=None, template_body=None, params=None
+):
+    kwargs = {
+        "StackName": stack_name,
+        "Capabilities": [
+            "CAPABILITY_IAM",
+            "CAPABILITY_NAMED_IAM",
+            "CAPABILITY_AUTO_EXPAND",
+        ],
+    }
+    if template_url:
+        kwargs["TemplateURL"] = template_url
+    if template_body:
+        kwargs["TemplateBody"] = template_body
+    if params:
+        kwargs["Parameters"] = params
+
+    if stack_exists(args, stack_name=stack_name):
+        print(f"Stack {stack_name} already exists. Updating...")
+        try:
+            response = client.update_stack(**kwargs)
+            return {"StackName": stack_name, "Action": "update"}
+        except botocore.exceptions.ClientError as e:
+            if e.response["Error"]["Code"] == "ValidationError" and re.search(
+                "No updates are to be performed", e.response["Error"]["Message"]
+            ):
+                print(f"No update was made to {stack_name}")
+                return {"StackName": stack_name, "Action": "noop"}
+            else:
+                raise e
+    else:
+        kwargs.update({"OnFailure": "ROLLBACK", "EnableTerminationProtection": False})
+        response = client.create_stack(**kwargs)
+        return {"StackName": stack_name, "Action": "create"}
+
+
+def replace_stack(
+    args, *, client, stack_name, template_url=None, template_body=None, params=None
+):
+    """Delete an existing stack and create a new stack with identical name"""
+
+    if not stack_exists(args, stack_name=stack_name):
+        raise ValueError(f"Stack {stack_name} does not exist")
+    r = client.delete_stack(StackName=stack_name)
+    delete_waiter = client.get_waiter("stack_delete_complete")
+    delete_waiter.wait(StackName=stack_name)
+
+    kwargs = {
+        "StackName": stack_name,
+        "Capabilities": [
+            "CAPABILITY_IAM",
+            "CAPABILITY_NAMED_IAM",
+            "CAPABILITY_AUTO_EXPAND",
+        ],
+        "OnFailure": "ROLLBACK",
+        "EnableTerminationProtection": False,
+    }
+    if template_url:
+        kwargs["TemplateURL"] = template_url
+    if template_body:
+        kwargs["TemplateBody"] = template_body
+    if params:
+        kwargs["Parameters"] = params
+    response = client.create_stack(**kwargs)
+    return {"StackName": stack_name, "Action": "create"}
+
+
+def wait(promise, *, client):
+    stack_name = promise["StackName"]
+    print(f"Waiting for {stack_name}...")
+    if promise["Action"] == "create":
+        waiter = client.get_waiter("stack_create_complete")
+        waiter.wait(StackName=stack_name)
+        print(f"Finished creating stack {stack_name}")
+    elif promise["Action"] == "update":
+        waiter = client.get_waiter("stack_update_complete")
+        waiter.wait(StackName=stack_name)
+        print(f"Finished updating stack {stack_name}")
+    elif promise["Action"] != "noop":
+        raise ValueError(f"Invalid promise {promise}")

tests/buildkite/infrastructure/worker-image-pipeline/create_worker_image_pipelines.py

@@ -2,6 +2,7 @@ import argparse
 import copy
 import json
 import os
+import sys
 from urllib.request import urlopen
 
 import boto3
@@ -9,6 +10,9 @@ import cfn_flip
 from metadata import IMAGE_PARAMS
 
 current_dir = os.path.dirname(__file__)
+sys.path.append(os.path.join(current_dir, ".."))
+
+from common_blocks.utils import replace_stack, wait
 
 BUILDKITE_CF_TEMPLATE_URL = (
     "https://s3.amazonaws.com/buildkite-aws-stack/latest/aws-stack.yml"
@@ -47,6 +51,9 @@ def main(args):
 
     ami_mapping = get_ami_mapping()
 
+    client = boto3.client("cloudformation", region_name=args.aws_region)
+    promises = []
+
     for stack_id in IMAGE_PARAMS:
         stack_id_full = get_full_stack_id(stack_id)
         print(f"Creating EC2 image builder stack {stack_id_full}...")
@@ -55,28 +62,20 @@ def main(args):
             stack_id=stack_id, aws_region=args.aws_region, ami_mapping=ami_mapping
         )
 
-        client = boto3.client("cloudformation", region_name=args.aws_region)
+        promise = replace_stack(
-        response = client.create_stack(
+            args,
-            StackName=stack_id_full,
+            client=client,
-            TemplateBody=ec2_image_pipeline_template,
+            stack_name=stack_id_full,
-            Capabilities=[
+            template_body=ec2_image_pipeline_template,
-                "CAPABILITY_IAM",
+            params=params,
-                "CAPABILITY_NAMED_IAM",
-                "CAPABILITY_AUTO_EXPAND",
-            ],
-            OnFailure="ROLLBACK",
-            EnableTerminationProtection=False,
-            Parameters=params,
         )
+        promises.append(promise)
         print(
             f"EC2 image builder stack {stack_id_full} is in progress in the background"
         )
 
-    for stack_id in IMAGE_PARAMS:
+    for promise in promises:
-        stack_id_full = get_full_stack_id(stack_id)
+        wait(promise, client=client)
-        waiter = client.get_waiter("stack_create_complete")
-        waiter.wait(StackName=stack_id_full)
-        print(f"EC2 image builder stack {stack_id_full} is now finished.")
 
 
 if __name__ == "__main__":

tests/buildkite/infrastructure/worker-image-pipeline/ec2-image-builder-pipeline-template.yml

@@ -58,7 +58,7 @@ Resources:
   BootstrapComponent:
     Type: AWS::ImageBuilder::Component
     Properties:
-      Name: !Sub "${AWS::StackName}-bootstrap-component"
+      Name: !Join ["-", [!Ref AWS::StackName, "bootstrap-component", !Select [2, !Split ['/', !Ref AWS::StackId]]]]
       Platform: !Ref InstanceOperatingSystem
       Version: "1.0.0"
       Description: Execute a bootstrap script.
@@ -67,7 +67,7 @@ Resources:
   Recipe:
     Type: AWS::ImageBuilder::ImageRecipe
     Properties:
-      Name: !Sub "${AWS::StackName}-image"
+      Name: !Join ["-", [!Ref AWS::StackName, "image", !Select [2, !Split ['/', !Ref AWS::StackId]]]]
       Components:
         - ComponentArn: !Ref BootstrapComponent
       ParentImage: !Ref BaseImageId
@@ -83,7 +83,7 @@ Resources:
   Infrastructure:
     Type: AWS::ImageBuilder::InfrastructureConfiguration
     Properties:
-      Name: !Sub "${AWS::StackName}-image-pipeline-infrastructure"
+      Name: !Join ["-", [!Ref AWS::StackName, "image-pipeline-infrastructure", !Select [2, !Split ['/', !Ref AWS::StackId]]]]
       InstanceProfileName: !Ref InstanceProfile
       InstanceTypes:
         - !Ref InstanceType
@@ -93,7 +93,7 @@ Resources:
   Distribution:
     Type: AWS::ImageBuilder::DistributionConfiguration
     Properties:
-      Name: !Sub "${AWS::StackName}-image-pipeline-distribution-config"
+      Name: !Join ["-", [!Ref AWS::StackName, "image-pipeline-distribution-config", !Select [2, !Split ['/', !Ref AWS::StackId]]]]
       Distributions:
         - Region: !Ref AWS::Region
           AmiDistributionConfiguration: {}
@@ -102,7 +102,7 @@ Resources:
   Pipeline:
     Type: AWS::ImageBuilder::ImagePipeline
     Properties:
-      Name: !Sub "${AWS::StackName}-image-pipeline"
+      Name: !Join ["-", [!Ref AWS::StackName, "image-pipeline", !Select [2, !Split ['/', !Ref AWS::StackId]]]]
       DistributionConfigurationArn: !Ref Distribution
       ImageRecipeArn: !Ref Recipe
       InfrastructureConfigurationArn: !Ref Infrastructure

tests/buildkite/infrastructure/worker-image-pipeline/metadata.py

@@ -13,6 +13,6 @@ IMAGE_PARAMS = {
         "BootstrapScript": "windows-gpu-bootstrap.yml",
         "InstanceType": "g4dn.2xlarge",
         "InstanceOperatingSystem": "Windows",
-        "VolumeSize": "80",  # in GiBs
+        "VolumeSize": "120",  # in GiBs
     },
 }

tests/buildkite/infrastructure/worker-image-pipeline/windows-gpu-bootstrap.yml

@@ -15,9 +15,9 @@ phases:
               choco --version
               choco feature enable -n=allowGlobalConfirmation
 
-              # CMake 3.18
+              # CMake 3.25
-              Write-Host '>>> Installing CMake 3.18...'
+              Write-Host '>>> Installing CMake 3.25...'
-              choco install cmake --version 3.18.0 --installargs "ADD_CMAKE_TO_PATH=System"
+              choco install cmake --version 3.25.2 --installargs "ADD_CMAKE_TO_PATH=System"
               if ($LASTEXITCODE -ne 0) { throw "Last command failed" }
 
               # Notepad++
@@ -45,18 +45,18 @@ phases:
               choco install graphviz
               if ($LASTEXITCODE -ne 0) { throw "Last command failed" }
 
-              # Install Visual Studio Community 2017 (15.9)
+              # Install Visual Studio 2022 Community
-              Write-Host '>>> Installing Visual Studio 2017 Community (15.9)...'
+              Write-Host '>>> Installing Visual Studio 2022 Community...'
-              choco install visualstudio2017community --version 15.9.23.0 `
+              choco install visualstudio2022community `
                   --params "--wait --passive --norestart"
               if ($LASTEXITCODE -ne 0) { throw "Last command failed" }
-              choco install visualstudio2017-workload-nativedesktop --params `
+              choco install visualstudio2022-workload-nativedesktop --params `
                   "--wait --passive --norestart --includeOptional"
               if ($LASTEXITCODE -ne 0) { throw "Last command failed" }
 
-              # Install CUDA 11.0
+              # Install CUDA 11.8
-              Write-Host '>>> Installing CUDA 11.0...'
+              Write-Host '>>> Installing CUDA 11.8...'
-              choco install cuda --version 11.0.3
+              choco install cuda --version=11.8.0.52206
               if ($LASTEXITCODE -ne 0) { throw "Last command failed" }
 
               # Install Python packages

tests/buildkite/pipeline.yml

@@ -22,11 +22,11 @@ steps:
       queue: linux-amd64-cpu
   - wait
   #### -------- BUILD --------
-  - label: ":console: Run clang-tidy"
+  # - label: ":console: Run clang-tidy"
-    command: "tests/buildkite/run-clang-tidy.sh"
+  #   command: "tests/buildkite/run-clang-tidy.sh"
-    key: run-clang-tidy
+  #   key: run-clang-tidy
-    agents:
+  #  agents:
-      queue: linux-amd64-cpu
+  #    queue: linux-amd64-cpu
   - wait
   - label: ":console: Build CPU"
     command: "tests/buildkite/build-cpu.sh"

tests/buildkite/test-cpp-gpu.sh

@@ -20,4 +20,5 @@ tests/ci_build/ci_build.sh gpu nvidia-docker \
 # tests/ci_build/ci_build.sh rmm nvidia-docker \
 #   --build-arg CUDA_VERSION_ARG=$CUDA_VERSION \
 #   --build-arg RAPIDS_VERSION_ARG=$RAPIDS_VERSION bash -c \
+#   --build-arg NCCL_VERSION_ARG=$NCCL_VERSION bash -c \
 #   "source activate gpu_test && build/testxgboost --use-rmm-pool"

tests/ci_build/Dockerfile.clang_tidy

@@ -15,8 +15,8 @@ RUN \
     add-apt-repository -u 'deb http://apt.llvm.org/bionic/ llvm-toolchain-bionic-11 main' && \
     apt-get update && \
     apt-get install -y llvm-11 clang-tidy-11 clang-11 && \
-    wget -nv -nc https://cmake.org/files/v3.14/cmake-3.14.0-Linux-x86_64.sh --no-check-certificate && \
+    wget -nv -nc https://cmake.org/files/v3.18/cmake-3.18.0-Linux-x86_64.sh --no-check-certificate && \
-    bash cmake-3.14.0-Linux-x86_64.sh --skip-license --prefix=/usr
+    bash cmake-3.18.0-Linux-x86_64.sh --skip-license --prefix=/usr
 
 # Set default clang-tidy version
 RUN \

tests/ci_build/Dockerfile.cpu

@@ -12,8 +12,8 @@ RUN \
     apt-get update && \
     apt-get install -y tar unzip wget git build-essential doxygen graphviz llvm libasan2 libidn11 ninja-build gcc-8 g++-8 openjdk-8-jdk-headless && \
     # CMake
-    wget -nv -nc https://cmake.org/files/v3.14/cmake-3.14.0-Linux-x86_64.sh --no-check-certificate && \
+    wget -nv -nc https://cmake.org/files/v3.18/cmake-3.18.0-Linux-x86_64.sh --no-check-certificate && \
-    bash cmake-3.14.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
+    bash cmake-3.18.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
     # Python
     wget -nv https://github.com/conda-forge/miniforge/releases/latest/download/Mambaforge-Linux-x86_64.sh && \
     bash Mambaforge-Linux-x86_64.sh -b -p /opt/python

tests/ci_build/Dockerfile.gpu

@@ -22,10 +22,10 @@ ENV PATH=/opt/python/bin:$PATH
 RUN \
     conda install -c conda-forge mamba && \
     mamba create -n gpu_test -c rapidsai-nightly -c rapidsai -c nvidia -c conda-forge -c defaults \
-        python=3.9 cudf=$RAPIDS_VERSION_ARG* rmm=$RAPIDS_VERSION_ARG* cudatoolkit=$CUDA_VERSION_ARG \
+        python=3.10 cudf=$RAPIDS_VERSION_ARG* rmm=$RAPIDS_VERSION_ARG* cudatoolkit=$CUDA_VERSION_ARG \
         dask dask-cuda=$RAPIDS_VERSION_ARG* dask-cudf=$RAPIDS_VERSION_ARG* cupy \
         numpy pytest pytest-timeout scipy scikit-learn pandas matplotlib wheel python-kubernetes urllib3 graphviz hypothesis \
-        pyspark cloudpickle cuda-python=11.7.0 && \
+        pyspark cloudpickle cuda-python && \
     mamba clean --all && \
     conda run --no-capture-output -n gpu_test pip install buildkite-test-collector
 

tests/ci_build/Dockerfile.gpu_build_centos7

@@ -1,6 +1,7 @@
 ARG CUDA_VERSION_ARG
 FROM nvidia/cuda:$CUDA_VERSION_ARG-devel-centos7
 ARG CUDA_VERSION_ARG
+ARG NCCL_VERSION_ARG
 
 # Install all basic requirements
 RUN \
@@ -21,7 +22,7 @@ RUN \
 # NCCL2 (License: https://docs.nvidia.com/deeplearning/sdk/nccl-sla/index.html)
 RUN \
     export CUDA_SHORT=`echo $CUDA_VERSION_ARG | grep -o -E '[0-9]+\.[0-9]'` && \
-    export NCCL_VERSION=2.13.4-1 && \
+    export NCCL_VERSION=$NCCL_VERSION_ARG && \
     wget -nv -nc https://developer.download.nvidia.com/compute/machine-learning/repos/rhel7/x86_64/nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm && \
     rpm -i nvidia-machine-learning-repo-rhel7-1.0.0-1.x86_64.rpm && \
     yum -y update && \

tests/ci_build/Dockerfile.gpu_build_r_centos7

@@ -36,8 +36,8 @@ RUN \
     bash Miniconda3.sh -b -p /opt/python && \
     /opt/python/bin/python -m pip install auditwheel awscli && \
     # CMake
-    wget -nv -nc https://cmake.org/files/v3.14/cmake-3.14.0-Linux-x86_64.sh --no-check-certificate && \
+    wget -nv -nc https://cmake.org/files/v3.18/cmake-3.18.0-Linux-x86_64.sh --no-check-certificate && \
-    bash cmake-3.14.0-Linux-x86_64.sh --skip-license --prefix=/usr
+    bash cmake-3.18.0-Linux-x86_64.sh --skip-license --prefix=/usr
 
 ENV GOSU_VERSION 1.10
 

tests/ci_build/Dockerfile.jvm

@@ -12,8 +12,8 @@ RUN \
     wget -nv -nc -O Miniconda3.sh https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && \
     bash Miniconda3.sh -b -p /opt/python && \
     # CMake
-    wget -nv -nc https://cmake.org/files/v3.14/cmake-3.14.0-Linux-x86_64.sh --no-check-certificate && \
+    wget -nv -nc https://cmake.org/files/v3.18/cmake-3.18.0-Linux-x86_64.sh --no-check-certificate && \
-    bash cmake-3.14.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
+    bash cmake-3.18.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
     # Maven
     wget -nv -nc https://archive.apache.org/dist/maven/maven-3/3.6.1/binaries/apache-maven-3.6.1-bin.tar.gz && \
     tar xvf apache-maven-3.6.1-bin.tar.gz -C /opt && \

tests/ci_build/Dockerfile.jvm_gpu_build

@@ -1,6 +1,7 @@
 ARG CUDA_VERSION_ARG
 FROM nvidia/cuda:$CUDA_VERSION_ARG-devel-centos7
 ARG CUDA_VERSION_ARG
+ARG NCCL_VERSION_ARG
 
 # Install all basic requirements
 RUN \
@@ -14,8 +15,8 @@ RUN \
     wget -nv -nc -O Miniconda3.sh https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh && \
     bash Miniconda3.sh -b -p /opt/python && \
     # CMake
-    wget -nv -nc https://cmake.org/files/v3.14/cmake-3.14.0-Linux-x86_64.sh --no-check-certificate && \
+    wget -nv -nc https://cmake.org/files/v3.18/cmake-3.18.0-Linux-x86_64.sh --no-check-certificate && \
-    bash cmake-3.14.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
+    bash cmake-3.18.0-Linux-x86_64.sh --skip-license --prefix=/usr && \
     # Maven
     wget -nv -nc https://archive.apache.org/dist/maven/maven-3/3.6.1/binaries/apache-maven-3.6.1-bin.tar.gz && \
     tar xvf apache-maven-3.6.1-bin.tar.gz -C /opt && \
@@ -24,7 +25,7 @@ RUN \
 # NCCL2 (License: https://docs.nvidia.com/deeplearning/sdk/nccl-sla/index.html)
 RUN \
     export CUDA_SHORT=`echo $CUDA_VERSION_ARG | grep -o -E '[0-9]+\.[0-9]'` && \
-    export NCCL_VERSION=2.13.4-1 && \
+    export NCCL_VERSION=$NCCL_VERSION_ARG && \
     yum-config-manager --add-repo http://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo && \
     yum -y update && \
     yum install -y libnccl-${NCCL_VERSION}+cuda${CUDA_SHORT} libnccl-devel-${NCCL_VERSION}+cuda${CUDA_SHORT} libnccl-static-${NCCL_VERSION}+cuda${CUDA_SHORT}

tests/ci_build/Dockerfile.rmm

@@ -1,7 +1,8 @@
 ARG CUDA_VERSION_ARG
-FROM nvidia/cuda:$CUDA_VERSION_ARG-devel-ubuntu18.04
+FROM nvidia/cuda:$CUDA_VERSION_ARG-devel-ubuntu20.04
 ARG CUDA_VERSION_ARG
 ARG RAPIDS_VERSION_ARG
+ARG NCCL_VERSION_ARG
 
 # Environment
 ENV DEBIAN_FRONTEND noninteractive
@@ -19,7 +20,7 @@ RUN \
 # NCCL2 (License: https://docs.nvidia.com/deeplearning/sdk/nccl-sla/index.html)
 RUN \
     export CUDA_SHORT=`echo $CUDA_VERSION_ARG | grep -o -E '[0-9]+\.[0-9]'` && \
-    export NCCL_VERSION=2.13.4-1 && \
+    export NCCL_VERSION=$NCCL_VERSION_ARG && \
     apt-get update && \
     apt-get install -y --allow-downgrades --allow-change-held-packages libnccl2=${NCCL_VERSION}+cuda${CUDA_SHORT} libnccl-dev=${NCCL_VERSION}+cuda${CUDA_SHORT}
 
@@ -29,7 +30,7 @@ ENV PATH=/opt/python/bin:$PATH
 RUN \
     conda install -c conda-forge mamba && \
     mamba create -n gpu_test -c rapidsai-nightly -c rapidsai -c nvidia -c conda-forge -c defaults \
-        python=3.9 rmm=$RAPIDS_VERSION_ARG* cudatoolkit=$CUDA_VERSION_ARG cmake && \
+        python=3.10 rmm=$RAPIDS_VERSION_ARG* cudatoolkit=$CUDA_VERSION_ARG cmake && \
     mamba clean --all
 
 ENV GOSU_VERSION 1.10

tests/ci_build/build_r_pkg_with_cuda_win64.sh

@@ -15,7 +15,7 @@ mv xgboost/ xgboost_rpack/
 
 mkdir build
 cd build
-cmake .. -G"Visual Studio 15 2017 Win64" -DUSE_CUDA=ON -DR_LIB=ON -DLIBR_HOME="c:\\Program Files\\R\\R-3.6.3"
+cmake .. -G"Visual Studio 17 2022" -A x64 -DUSE_CUDA=ON -DR_LIB=ON -DLIBR_HOME="c:\\Program Files\\R\\R-3.6.3"
 cmake --build . --config Release --parallel
 cd ..
 

tests/ci_build/insert_vcomp140.py

@@ -1,19 +1,102 @@
-import sys
+import argparse
-import re
+import base64
-import zipfile
 import glob
+import hashlib
+import os
+import pathlib
+import re
+import shutil
+import tempfile
 
-if len(sys.argv) != 2:
+VCOMP140_PATH = "C:\\Windows\\System32\\vcomp140.dll"
-    print('Usage: {} [wheel]'.format(sys.argv[0]))
-    sys.exit(1)
 
-vcomp140_path = 'C:\\Windows\\System32\\vcomp140.dll'
 
-for wheel_path in sorted(glob.glob(sys.argv[1])):
+def get_sha256sum(path):
-    m = re.search(r'xgboost-(.*)-py3', wheel_path)
+    return (
-    assert m, f'wheel_path = {wheel_path}'
+        base64.urlsafe_b64encode(hashlib.sha256(open(path, "rb").read()).digest())
+        .decode("latin1")
+        .rstrip("=")
+    )
+
+
+def update_record(*, wheel_content_dir, xgboost_version):
+    vcomp140_size = os.path.getsize(VCOMP140_PATH)
+    vcomp140_hash = get_sha256sum(VCOMP140_PATH)
+
+    record_path = wheel_content_dir / pathlib.Path(
+        f"xgboost-{xgboost_version}.dist-info/RECORD"
+    )
+    with open(record_path, "r") as f:
+        record_content = f.read()
+    record_content += f"xgboost-{xgboost_version}.data/data/xgboost/vcomp140.dll,"
+    record_content += f"sha256={vcomp140_hash},{vcomp140_size}\n"
+    with open(record_path, "w") as f:
+        f.write(record_content)
+
+
+def main(args):
+    candidates = list(sorted(glob.glob(args.wheel_path)))
+    for wheel_path in candidates:
+        print(f"Processing wheel {wheel_path}")
+        m = re.search(r"xgboost-(.*)\+.*-py3", wheel_path)
+        if not m:
+            raise ValueError(f"Wheel {wheel_path} has unexpected name")
         version = m.group(1)
-
+        print(f"  Detected version for {wheel_path}: {version}")
         print(f"  Inserting vcomp140.dll into {wheel_path}...")
-    with zipfile.ZipFile(wheel_path, 'a') as f:
+        with tempfile.TemporaryDirectory() as tempdir:
-        f.write(vcomp140_path, 'xgboost-{}.data/data/xgboost/vcomp140.dll'.format(version))
+            wheel_content_dir = pathlib.Path(tempdir) / "wheel_content"
+            print(f"    Extract {wheel_path} into {wheel_content_dir}")
+            shutil.unpack_archive(
+                wheel_path, extract_dir=wheel_content_dir, format="zip"
+            )
+            data_dir = wheel_content_dir / pathlib.Path(
+                f"xgboost-{version}.data/data/xgboost"
+            )
+            data_dir.mkdir(parents=True, exist_ok=True)
+
+            print(f"    Copy {VCOMP140_PATH} -> {data_dir}")
+            shutil.copy(VCOMP140_PATH, data_dir)
+
+            print(f"    Update RECORD")
+            update_record(wheel_content_dir=wheel_content_dir, xgboost_version=version)
+
+            print(f"    Content of {wheel_content_dir}:")
+            for e in sorted(wheel_content_dir.rglob("*")):
+                if e.is_file():
+                    r = e.relative_to(wheel_content_dir)
+                    print(f"      {r}")
+
+            print(f"    Create new wheel...")
+            new_wheel_tmp_path = pathlib.Path(tempdir) / "new_wheel"
+            shutil.make_archive(
+                str(new_wheel_tmp_path.resolve()),
+                format="zip",
+                root_dir=wheel_content_dir,
+            )
+            new_wheel_tmp_path = new_wheel_tmp_path.resolve().with_suffix(".zip")
+            new_wheel_tmp_path = new_wheel_tmp_path.rename(
+                new_wheel_tmp_path.with_suffix(".whl")
+            )
+            print(f"    Created new wheel {new_wheel_tmp_path}")
+
+            # Rename the old wheel with suffix .bak
+            # The new wheel takes the name of the old wheel
+            wheel_path_obj = pathlib.Path(wheel_path).resolve()
+            backup_path = wheel_path_obj.with_suffix(".whl.bak")
+            print(f"    Rename {wheel_path_obj} -> {backup_path}")
+            wheel_path_obj.replace(backup_path)
+            print(f"    Rename {new_wheel_tmp_path} -> {wheel_path_obj}")
+            new_wheel_tmp_path.replace(wheel_path_obj)
+
+            shutil.rmtree(wheel_content_dir)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "wheel_path", type=str, help="Path to wheel (wildcard permitted)"
+    )
+    args = parser.parse_args()
+
+    main(args)

tests/python-gpu/test_gpu_prediction.py

@@ -216,6 +216,7 @@ class TestGPUPredict:
     def test_inplace_predict_cupy(self):
         self.run_inplace_predict_cupy(0)
 
+    @pytest.mark.xfail
     @pytest.mark.skipif(**tm.no_cupy())
     @pytest.mark.mgpu
     def test_inplace_predict_cupy_specified_device(self):

tests/python-gpu/test_gpu_updaters.py

@@ -215,3 +215,34 @@ class TestGPUUpdaters:
     @pytest.mark.parametrize("weighted", [True, False])
     def test_adaptive(self, weighted) -> None:
         self.cputest.run_adaptive("gpu_hist", weighted)
+
+    @pytest.mark.skipif(**tm.no_pandas())
+    def test_issue8824(self):
+        # column sampling by node crashes because shared pointers go out of scope
+        import pandas as pd
+
+        data = pd.DataFrame(np.random.rand(1024, 8))
+        data.columns = "x" + data.columns.astype(str)
+        features = data.columns
+        data["y"] = data.sum(axis=1) < 4
+        dtrain = xgb.DMatrix(data[features], label=data["y"])
+        model = xgb.train(
+            dtrain=dtrain,
+            params={
+                "max_depth": 5,
+                "learning_rate": 0.05,
+                "objective": "binary:logistic",
+                "tree_method": "gpu_hist",
+                "colsample_bytree": 0.5,
+                "colsample_bylevel": 0.5,
+                "colsample_bynode": 0.5,  # Causes issues
+                "reg_alpha": 0.05,
+                "reg_lambda": 0.005,
+                "seed": 66,
+                "subsample": 0.5,
+                "gamma": 0.2,
+                "predictor": "auto",
+                "eval_metric": "auc",
+            },
+            num_boost_round=150,
+        )