Update federated secure branch.

.github/dependabot.yml

@@ -12,7 +12,7 @@ updates:
   - package-ecosystem: "maven"
     directory: "/jvm-packages/xgboost4j"
     schedule:
-      interval: "daily"
+      interval: "monthly"
   - package-ecosystem: "maven"
     directory: "/jvm-packages/xgboost4j-gpu"
     schedule:
@@ -24,7 +24,7 @@ updates:
   - package-ecosystem: "maven"
     directory: "/jvm-packages/xgboost4j-spark"
     schedule:
-      interval: "daily"
+      interval: "monthly"
   - package-ecosystem: "maven"
     directory: "/jvm-packages/xgboost4j-spark-gpu"
     schedule:

.github/workflows/i386.yml

@@ -23,7 +23,7 @@ jobs:
       with:
         submodules: 'true'
     - name: Set up Docker Buildx
-      uses: docker/setup-buildx-action@v3
+      uses: docker/setup-buildx-action@v3.4.0
       with:
         driver-opts: network=host
     - name: Build and push container

.github/workflows/main.yml

@@ -180,7 +180,7 @@ jobs:
     - uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 # v4.1.6
       with:
         submodules: 'true'
-    - uses: actions/setup-python@82c7e631bb3cdc910f68e0081d67478d79c6982d # v5.1.0
+    - uses: actions/setup-python@39cd14951b08e74b54015e9e001cdefcf80e669f # v5.1.1
       with:
         python-version: "3.8"
         architecture: 'x64'

.github/workflows/python_tests.yml

@@ -319,7 +319,7 @@ jobs:
           submodules: 'true'
 
       - name: Set up Python 3.8
-        uses: actions/setup-python@82c7e631bb3cdc910f68e0081d67478d79c6982d # v5.1.0
+        uses: actions/setup-python@39cd14951b08e74b54015e9e001cdefcf80e669f # v5.1.1
         with:
           python-version: 3.8
 

.github/workflows/r_tests.yml

@@ -84,7 +84,7 @@ jobs:
         key: ${{ runner.os }}-r-${{ matrix.config.r }}-7-${{ hashFiles('R-package/DESCRIPTION') }}
         restore-keys: ${{ runner.os }}-r-${{ matrix.config.r }}-7-${{ hashFiles('R-package/DESCRIPTION') }}
 
-    - uses: actions/setup-python@82c7e631bb3cdc910f68e0081d67478d79c6982d # v5.1.0
+    - uses: actions/setup-python@39cd14951b08e74b54015e9e001cdefcf80e669f # v5.1.1
       with:
         python-version: "3.8"
         architecture: 'x64'

.github/workflows/scorecards.yml

@@ -41,7 +41,7 @@ jobs:
       # Upload the results as artifacts (optional). Commenting out will disable uploads of run results in SARIF
       # format to the repository Actions tab.
       - name: "Upload artifact"
-        uses: actions/upload-artifact@65462800fd760344b1a7b4382951275a0abb4808 # v4.3.3
+        uses: actions/upload-artifact@0b2256b8c012f0828dc542b3febcab082c67f72b # v4.3.4
         with:
           name: SARIF file
           path: results.sarif

CMakeLists.txt

@@ -71,7 +71,6 @@ option(HIDE_CXX_SYMBOLS "Build shared library and hide all C++ symbols" OFF)
 option(KEEP_BUILD_ARTIFACTS_IN_BINARY_DIR "Output build artifacts in CMake binary dir" OFF)
 ## CUDA
 option(USE_CUDA  "Build with GPU acceleration" OFF)
-option(USE_PER_THREAD_DEFAULT_STREAM "Build with per-thread default stream" ON)
 option(USE_NCCL  "Build with NCCL to enable distributed GPU support." OFF)
 # This is specifically designed for PyPI binary release and should be disabled for most of the cases.
 option(USE_DLOPEN_NCCL "Whether to load nccl dynamically." OFF)

R-package/src/Makevars.in

@@ -113,6 +113,7 @@ OBJECTS= \
     $(PKGROOT)/src/common/charconv.o \
     $(PKGROOT)/src/common/column_matrix.o \
     $(PKGROOT)/src/common/common.o \
+    $(PKGROOT)/src/common/cuda_rt_utils.o \
     $(PKGROOT)/src/common/error_msg.o \
     $(PKGROOT)/src/common/hist_util.o \
     $(PKGROOT)/src/common/host_device_vector.o \

R-package/src/Makevars.win

@@ -113,6 +113,7 @@ OBJECTS= \
     $(PKGROOT)/src/common/charconv.o \
     $(PKGROOT)/src/common/column_matrix.o \
     $(PKGROOT)/src/common/common.o \
+    $(PKGROOT)/src/common/cuda_rt_utils.o \
     $(PKGROOT)/src/common/error_msg.o \
     $(PKGROOT)/src/common/hist_util.o \
     $(PKGROOT)/src/common/host_device_vector.o \

README.md

@@ -4,8 +4,8 @@
 [![Build Status](https://badge.buildkite.com/aca47f40a32735c00a8550540c5eeff6a4c1d246a580cae9b0.svg?branch=master)](https://buildkite.com/xgboost/xgboost-ci)
 [![XGBoost-CI](https://github.com/dmlc/xgboost/workflows/XGBoost-CI/badge.svg?branch=master)](https://github.com/dmlc/xgboost/actions)
 [![Documentation Status](https://readthedocs.org/projects/xgboost/badge/?version=latest)](https://xgboost.readthedocs.org)
-[![GitHub license](http://dmlc.github.io/img/apache2.svg)](./LICENSE)
-[![CRAN Status Badge](http://www.r-pkg.org/badges/version/xgboost)](http://cran.r-project.org/web/packages/xgboost)
+[![GitHub license](https://dmlc.github.io/img/apache2.svg)](./LICENSE)
+[![CRAN Status Badge](https://www.r-pkg.org/badges/version/xgboost)](https://cran.r-project.org/web/packages/xgboost)
 [![PyPI version](https://badge.fury.io/py/xgboost.svg)](https://pypi.python.org/pypi/xgboost/)
 [![Conda version](https://img.shields.io/conda/vn/conda-forge/py-xgboost.svg)](https://anaconda.org/conda-forge/py-xgboost)
 [![Optuna](https://img.shields.io/badge/Optuna-integrated-blue)](https://optuna.org)
@@ -35,7 +35,7 @@ Checkout the [Community Page](https://xgboost.ai/community).
 
 Reference
 ---------
-- Tianqi Chen and Carlos Guestrin. [XGBoost: A Scalable Tree Boosting System](http://arxiv.org/abs/1603.02754). In 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, 2016
+- Tianqi Chen and Carlos Guestrin. [XGBoost: A Scalable Tree Boosting System](https://arxiv.org/abs/1603.02754). In 22nd SIGKDD Conference on Knowledge Discovery and Data Mining, 2016
 - XGBoost originates from research project at University of Washington.
 
 Sponsors

cmake/Utils.cmake

@@ -80,12 +80,8 @@ function(xgboost_set_cuda_flags target)
     $<$<COMPILE_LANGUAGE:CUDA>:--expt-extended-lambda>
     $<$<COMPILE_LANGUAGE:CUDA>:--expt-relaxed-constexpr>
     $<$<COMPILE_LANGUAGE:CUDA>:-Xcompiler=${OpenMP_CXX_FLAGS}>
-    $<$<COMPILE_LANGUAGE:CUDA>:-Xfatbin=-compress-all>)
-
-  if(USE_PER_THREAD_DEFAULT_STREAM)
-    target_compile_options(${target} PRIVATE
-            $<$<COMPILE_LANGUAGE:CUDA>:--default-stream per-thread>)
-  endif()
+    $<$<COMPILE_LANGUAGE:CUDA>:-Xfatbin=-compress-all>
+    $<$<COMPILE_LANGUAGE:CUDA>:--default-stream per-thread>)
 
   if(FORCE_COLORED_OUTPUT)
     if(FORCE_COLORED_OUTPUT AND (CMAKE_GENERATOR STREQUAL "Ninja") AND

dev/release-artifacts.py

@@ -2,6 +2,7 @@
 
 tqdm, sh are required to run this script.
 """
+
 import argparse
 import os
 import shutil
@@ -106,6 +107,15 @@ def make_pysrc_wheel(
     if not os.path.exists(dist):
         os.mkdir(dist)
 
+    # Apply patch to remove NCCL dependency
+    # Save the original content of pyproject.toml so that we can restore it later
+    with DirectoryExcursion(ROOT):
+        with open("python-package/pyproject.toml", "r") as f:
+            orig_pyproj_lines = f.read()
+        with open("tests/buildkite/remove_nccl_dep.patch", "r") as f:
+            patch_lines = f.read()
+        subprocess.run(["patch", "-p0"], input=patch_lines, text=True)
+
     with DirectoryExcursion(os.path.join(ROOT, "python-package")):
         subprocess.check_call(["python", "-m", "build", "--sdist"])
         if rc is not None:
@@ -117,6 +127,10 @@ def make_pysrc_wheel(
         target = os.path.join(dist, name)
         shutil.move(src, target)
 
+    with DirectoryExcursion(ROOT):
+        with open("python-package/pyproject.toml", "w") as f:
+            print(orig_pyproj_lines, file=f, end="")
+
 
 def download_py_packages(
     branch: str, major: int, minor: int, commit_hash: str, outdir: str

doc/install.rst

@@ -76,6 +76,19 @@ Capabilities of binary wheels for each platform:
 | Windows             | |tick|  |  |cross|             |
 +---------------------+---------+----------------------+
 
+Minimal installation (CPU-only)
+*******************************
+The default installation with ``pip`` will install the full XGBoost package, including the support for the GPU algorithms and federated learning.
+
+You may choose to reduce the size of the installed package and save the disk space, by opting to install ``xgboost-cpu`` instead:
+
+.. code-block:: bash
+
+  pip install xgboost-cpu
+
+The ``xgboost-cpu`` variant will have drastically smaller disk footprint, but does not provide some features, such as the GPU algorithms and
+federated learning.
+
 Conda
 *****
 

include/xgboost/base.h

@@ -7,6 +7,8 @@
 #define XGBOOST_BASE_H_
 
 #include <dmlc/omp.h>  // for omp_uint, omp_ulong
+// Put the windefs here to guard as many files as possible.
+#include <xgboost/windefs.h>
 
 #include <cstdint>  // for int32_t, uint64_t, int16_t
 #include <ostream>  // for ostream

include/xgboost/collective/poll_utils.h

@@ -4,13 +4,14 @@
  * \author Tianqi Chen
  */
 #pragma once
-#include "xgboost/collective/result.h"
-#include "xgboost/collective/socket.h"
+#include <xgboost/collective/result.h>
+#include <xgboost/collective/socket.h>
 
 #if defined(_WIN32)
+#include <xgboost/windefs.h>
+// Socket API
 #include <winsock2.h>
 #include <ws2tcpip.h>
-
 #else
 
 #include <arpa/inet.h>

include/xgboost/collective/socket.h

@@ -1,12 +1,8 @@
 /**
- * Copyright (c) 2022-2024, XGBoost Contributors
+ * Copyright 2022-2024, XGBoost Contributors
  */
 #pragma once
 
-#if !defined(NOMINMAX) && defined(_WIN32)
-#define NOMINMAX
-#endif                   // !defined(NOMINMAX)
-
 #include <cerrno>        // errno, EINTR, EBADF
 #include <climits>       // HOST_NAME_MAX
 #include <cstddef>       // std::size_t
@@ -18,18 +14,12 @@
 
 #if defined(__linux__)
 #include <sys/ioctl.h>  // for TIOCOUTQ, FIONREAD
-#endif  // defined(__linux__)
-
-#if !defined(xgboost_IS_MINGW)
-
-#if defined(__MINGW32__)
-#define xgboost_IS_MINGW 1
-#endif  // defined(__MINGW32__)
-
-#endif  // xgboost_IS_MINGW
+#endif                  // defined(__linux__)
 
 #if defined(_WIN32)
-
+// Guard the include.
+#include <xgboost/windefs.h>
+// Socket API
 #include <winsock2.h>
 #include <ws2tcpip.h>
 
@@ -41,9 +31,9 @@ using in_port_t = std::uint16_t;
 
 #if !defined(xgboost_IS_MINGW)
 using ssize_t = int;
-#endif                    // !xgboost_IS_MINGW()
+#endif  // !xgboost_IS_MINGW()
 
-#else                     // UNIX
+#else  // UNIX
 
 #include <arpa/inet.h>    // inet_ntop
 #include <fcntl.h>        // fcntl, F_GETFL, O_NONBLOCK
@@ -839,7 +829,3 @@ Result INetNToP(H const &host, std::string *p_out) {
 }  // namespace xgboost
 
 #undef xgboost_CHECK_SYS_CALL
-
-#if defined(xgboost_IS_MINGW)
-#undef xgboost_IS_MINGW
-#endif

include/xgboost/windefs.h

@@ -0,0 +1,33 @@
+/**
+ * Copyright 2024, XGBoost Contributors
+ *
+ * @brief Macro for Windows.
+ */
+#pragma once
+
+#if !defined(xgboost_IS_WIN)
+
+#if defined(_MSC_VER) || defined(__MINGW32__)
+#define xgboost_IS_WIN 1
+#endif  // defined(_MSC_VER) || defined(__MINGW32__)
+
+#endif  // !defined(xgboost_IS_WIN)
+
+#if defined(xgboost_IS_WIN)
+
+#if !defined(NOMINMAX)
+#define NOMINMAX
+#endif  // !defined(NOMINMAX)
+
+// A macro used inside `windows.h` to avoid conflicts with `winsock2.h`
+#define WIN32_LEAN_AND_MEAN
+
+#if !defined(xgboost_IS_MINGW)
+
+#if defined(__MINGW32__)
+#define xgboost_IS_MINGW 1
+#endif  // defined(__MINGW32__)
+
+#endif  // xgboost_IS_MINGW
+
+#endif  // defined(xgboost_IS_WIN)

jvm-packages/pom.xml

@@ -37,17 +37,17 @@
         <junit.version>4.13.2</junit.version>
         <spark.version>3.5.1</spark.version>
         <spark.version.gpu>3.5.1</spark.version.gpu>
-        <fasterxml.jackson.version>2.15.2</fasterxml.jackson.version>
+        <fasterxml.jackson.version>2.17.2</fasterxml.jackson.version>
         <scala.version>2.12.18</scala.version>
         <scala.binary.version>2.12</scala.binary.version>
         <hadoop.version>3.4.0</hadoop.version>
         <maven.wagon.http.retryHandler.count>5</maven.wagon.http.retryHandler.count>
         <log.capi.invocation>OFF</log.capi.invocation>
         <use.cuda>OFF</use.cuda>
-        <cudf.version>24.04.0</cudf.version>
-        <spark.rapids.version>24.04.1</spark.rapids.version>
+        <cudf.version>24.06.0</cudf.version>
+        <spark.rapids.version>24.06.0</spark.rapids.version>
         <cudf.classifier>cuda12</cudf.classifier>
-        <scalatest.version>3.2.18</scalatest.version>
+        <scalatest.version>3.2.19</scalatest.version>
         <scala-collection-compat.version>2.12.0</scala-collection-compat.version>
         <skip.native.build>false</skip.native.build>
 

python-package/pyproject.toml

@@ -7,13 +7,13 @@ build-backend = "packager.pep517"
 
 [project]
 name = "xgboost"
-version = "2.2.0-dev"
+description = "XGBoost Python Package"
+readme = { file = "README.rst", content-type = "text/x-rst" }
 authors = [
     { name = "Hyunsu Cho", email = "chohyu01@cs.washington.edu" },
     { name = "Jiaming Yuan", email = "jm.yuan@outlook.com" }
 ]
-description = "XGBoost Python Package"
-readme = { file = "README.rst", content-type = "text/x-rst" }
+version = "2.2.0-dev"
 requires-python = ">=3.8"
 license = { text = "Apache-2.0" }
 classifiers = [

python-package/xgboost/testing/__init__.py

@@ -45,6 +45,7 @@
     get_cancer,
     get_digits,
     get_sparse,
+    make_batches,
     memory,
 )
 
@@ -161,7 +162,16 @@ def no_cudf() -> PytestSkip:
 
 
 def no_cupy() -> PytestSkip:
-    return no_mod("cupy")
+    skip_cupy = no_mod("cupy")
+    if not skip_cupy["condition"] and system() == "Windows":
+        import cupy as cp
+
+        # Cupy might run into issue on Windows due to missing compiler
+        try:
+            cp.array([1, 2, 3]).sum()
+        except Exception:  # pylint: disable=broad-except
+            skip_cupy["condition"] = True
+    return skip_cupy
 
 
 def no_dask_cudf() -> PytestSkip:
@@ -248,35 +258,6 @@ def as_arrays(
         return X, y, w
 
 
-def make_batches(  # pylint: disable=too-many-arguments,too-many-locals
-    n_samples_per_batch: int,
-    n_features: int,
-    n_batches: int,
-    use_cupy: bool = False,
-    *,
-    vary_size: bool = False,
-    random_state: int = 1994,
-) -> Tuple[List[np.ndarray], List[np.ndarray], List[np.ndarray]]:
-    X = []
-    y = []
-    w = []
-    if use_cupy:
-        import cupy
-
-        rng = cupy.random.RandomState(random_state)
-    else:
-        rng = np.random.RandomState(random_state)
-    for i in range(n_batches):
-        n_samples = n_samples_per_batch + i * 10 if vary_size else n_samples_per_batch
-        _X = rng.randn(n_samples, n_features)
-        _y = rng.randn(n_samples)
-        _w = rng.uniform(low=0, high=1, size=n_samples)
-        X.append(_X)
-        y.append(_y)
-        w.append(_w)
-    return X, y, w
-
-
 def make_regression(
     n_samples: int, n_features: int, use_cupy: bool
 ) -> Tuple[ArrayLike, ArrayLike, ArrayLike]: