Merge branch 'dmlc:release_2.1.0' into release_2.1.0

Author: razdoburdin

CMakeLists.txt

@@ -4,7 +4,7 @@ if(PLUGIN_SYCL)
   string(REPLACE " -isystem ${CONDA_PREFIX}/include" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
 endif()
 
-project(xgboost LANGUAGES CXX C VERSION 2.1.2)
+project(xgboost LANGUAGES CXX C VERSION 2.1.3)
 include(cmake/Utils.cmake)
 list(APPEND CMAKE_MODULE_PATH "${xgboost_SOURCE_DIR}/cmake/modules")
 

R-package/DESCRIPTION

@@ -1,8 +1,8 @@
 Package: xgboost
 Type: Package
 Title: Extreme Gradient Boosting
-Version: 2.1.2.1
-Date: 2024-10-23
+Version: 2.1.3.1
+Date: 2024-11-26
 Authors@R: c(
   person("Tianqi", "Chen", role = c("aut"),
          email = "[email protected]"),

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 2.1.2.
+# Generated by GNU Autoconf 2.71 for xgboost 2.1.3.
 #
 #
 # 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='2.1.2'
-PACKAGE_STRING='xgboost 2.1.2'
+PACKAGE_VERSION='2.1.3'
+PACKAGE_STRING='xgboost 2.1.3'
 PACKAGE_BUGREPORT=''
 PACKAGE_URL=''
 
@@ -1259,7 +1259,7 @@ 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 2.1.2 to adapt to many kinds of systems.
+\`configure' configures xgboost 2.1.3 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1321,7 +1321,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of xgboost 2.1.2:";;
+     short | recursive ) echo "Configuration of xgboost 2.1.3:";;
    esac
   cat <<\_ACEOF
 
@@ -1404,7 +1404,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-xgboost configure 2.1.2
+xgboost configure 2.1.3
 generated by GNU Autoconf 2.71
 
 Copyright (C) 2021 Free Software Foundation, Inc.
@@ -1603,7 +1603,7 @@ 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 2.1.2, which was
+It was created by xgboost $as_me 2.1.3, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -3709,7 +3709,7 @@ 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 2.1.2, which was
+This file was extended by xgboost $as_me 2.1.3, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -3773,7 +3773,7 @@ ac_cs_config_escaped=`printf "%s\n" "$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 2.1.2
+xgboost config.status 2.1.3
 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],[2.1.2],[],[xgboost],[])
+AC_INIT([xgboost],[2.1.3],[],[xgboost],[])
 
 : ${R_HOME=`R RHOME`}
 if test -z "${R_HOME}"; then

include/xgboost/version_config.h

@@ -6,6 +6,6 @@
 
 #define XGBOOST_VER_MAJOR 2  /* NOLINT */
 #define XGBOOST_VER_MINOR 1  /* NOLINT */
-#define XGBOOST_VER_PATCH 2  /* NOLINT */
+#define XGBOOST_VER_PATCH 3  /* NOLINT */
 
 #endif  // XGBOOST_VERSION_CONFIG_H_

jvm-packages/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>ml.dmlc</groupId>
     <artifactId>xgboost-jvm_2.12</artifactId>
-    <version>2.1.2</version>
+    <version>2.1.3</version>
     <packaging>pom</packaging>
     <name>XGBoost JVM Package</name>
     <description>JVM Package for XGBoost</description>

jvm-packages/xgboost4j-example/pom.xml

@@ -6,11 +6,11 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
     <name>xgboost4j-example</name>
     <artifactId>xgboost4j-example_2.12</artifactId>
-    <version>2.1.2</version>
+    <version>2.1.3</version>
     <packaging>jar</packaging>
     <build>
         <plugins>

jvm-packages/xgboost4j-flink/pom.xml

@@ -6,12 +6,12 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
 
     <name>xgboost4j-flink</name>
     <artifactId>xgboost4j-flink_2.12</artifactId>
-    <version>2.1.2</version>
+    <version>2.1.3</version>
     <properties>
       <flink-ml.version>2.2.0</flink-ml.version>
     </properties>

jvm-packages/xgboost4j-gpu/pom.xml

@@ -6,11 +6,11 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
     <artifactId>xgboost4j-gpu_2.12</artifactId>
     <name>xgboost4j-gpu</name>
-    <version>2.1.2</version>
+    <version>2.1.3</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>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
     <name>xgboost4j-spark-gpu</name>
     <artifactId>xgboost4j-spark-gpu_2.12</artifactId>

jvm-packages/xgboost4j-spark/pom.xml

@@ -6,7 +6,7 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
     <name>xgboost4j-spark</name>
     <artifactId>xgboost4j-spark_2.12</artifactId>

jvm-packages/xgboost4j/pom.xml

@@ -6,11 +6,11 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>2.1.2</version>
+        <version>2.1.3</version>
     </parent>
     <name>xgboost4j</name>
     <artifactId>xgboost4j_2.12</artifactId>
-    <version>2.1.2</version>
+    <version>2.1.3</version>
     <packaging>jar</packaging>
 
     <dependencies>

python-package/pyproject.toml

@@ -14,7 +14,7 @@ authors = [
     { name = "Hyunsu Cho", email = "[email protected]" },
     { name = "Jiaming Yuan", email = "[email protected]" }
 ]
-version = "2.1.2"
+version = "2.1.3"
 requires-python = ">=3.8"
 license = { text = "Apache-2.0" }
 classifiers = [

python-package/xgboost/VERSION

@@ -1 +1 @@
-2.1.2
+2.1.3

python-package/xgboost/core.py

@@ -2489,6 +2489,7 @@ def inplace_predict(
             _arrow_transform,
             _is_arrow,
             _is_cudf_df,
+            _is_cudf_pandas,
             _is_cupy_alike,
             _is_list,
             _is_np_array_like,
@@ -2498,6 +2499,9 @@ def inplace_predict(
             _transform_pandas_df,
         )
 
+        if _is_cudf_pandas(data):
+            data = data._fsproxy_fast  # pylint: disable=protected-access
+
         enable_categorical = True
         if _is_arrow(data):
             data = _arrow_transform(data)

python-package/xgboost/data.py

@@ -835,6 +835,16 @@ def _is_cudf_df(data: DataType) -> bool:
     return lazy_isinstance(data, "cudf.core.dataframe", "DataFrame")
 
 
+def _is_cudf_pandas(data: DataType) -> bool:
+    """Must go before both pandas and cudf checks."""
+    return (
+        lazy_isinstance(data, "pandas.core.frame", "DataFrame")
+        or lazy_isinstance(data, "pandas.core.series", "Series")
+    ) and lazy_isinstance(
+        type(data), "cudf.pandas.fast_slow_proxy", "_FastSlowProxyMeta"
+    )
+
+
 def _get_cudf_cat_predicate() -> Callable[[Any], bool]:
     try:
         from cudf import CategoricalDtype
@@ -1187,6 +1197,8 @@ def dispatch_data_backend(
         )
     if _is_arrow(data):
         data = _arrow_transform(data)
+    if _is_cudf_pandas(data):
+        data = data._fsproxy_fast  # pylint: disable=protected-access
     if _is_pandas_series(data):
         import pandas as pd
 
@@ -1327,6 +1339,8 @@ def dispatch_meta_backend(
         return
     if _is_arrow(data):
         data = _arrow_transform(data)
+    if _is_cudf_pandas(data):
+        data = data._fsproxy_fast  # pylint: disable=protected-access
     if _is_pandas_df(data):
         _meta_from_pandas_df(data, name, dtype=dtype, handle=handle)
         return
@@ -1398,6 +1412,8 @@ def _proxy_transform(
     feature_types: Optional[FeatureTypes],
     enable_categorical: bool,
 ) -> TransformedData:
+    if _is_cudf_pandas(data):
+        data = data._fsproxy_fast  # pylint: disable=protected-access
     if _is_cudf_df(data) or _is_cudf_ser(data):
         return _transform_cudf_df(
             data, feature_names, feature_types, enable_categorical

python-package/xgboost/spark/core.py

@@ -597,6 +597,9 @@ def _get_unwrapped_vec_cols(feature_col: Column) -> List[Column]:
 )
 
 
+_MODEL_CHUNK_SIZE = 4096 * 1024
+
+
 class _SparkXGBEstimator(Estimator, _SparkXGBParams, MLReadable, MLWritable):
     _input_kwargs: Dict[str, Any]
 
@@ -1091,25 +1094,27 @@ def _train_booster(
             context.barrier()
 
             if context.partitionId() == 0:
-                yield pd.DataFrame(
-                    data={
-                        "config": [booster.save_config()],
-                        "booster": [booster.save_raw("json").decode("utf-8")],
-                    }
-                )
+                config = booster.save_config()
+                yield pd.DataFrame({"data": [config]})
+                booster_json = booster.save_raw("json").decode("utf-8")
+
+                for offset in range(0, len(booster_json), _MODEL_CHUNK_SIZE):
+                    booster_chunk = booster_json[offset : offset + _MODEL_CHUNK_SIZE]
+                    yield pd.DataFrame({"data": [booster_chunk]})
 
         def _run_job() -> Tuple[str, str]:
             rdd = (
                 dataset.mapInPandas(
                     _train_booster,  # type: ignore
-                    schema="config string, booster string",
+                    schema="data string",
                 )
                 .rdd.barrier()
                 .mapPartitions(lambda x: x)
             )
             rdd_with_resource = self._try_stage_level_scheduling(rdd)
-            ret = rdd_with_resource.collect()[0]
-            return ret[0], ret[1]
+            ret = rdd_with_resource.collect()
+            data = [v[0] for v in ret]
+            return data[0], "".join(data[1:])
 
         get_logger(_LOG_TAG).info(
             "Running xgboost-%s on %s workers with"
@@ -1690,7 +1695,12 @@ def saveImpl(self, path: str) -> None:
         _SparkXGBSharedReadWrite.saveMetadata(self.instance, path, self.sc, self.logger)
         model_save_path = os.path.join(path, "model")
         booster = xgb_model.get_booster().save_raw("json").decode("utf-8")
-        _get_spark_session().sparkContext.parallelize([booster], 1).saveAsTextFile(
+        booster_chunks = []
+
+        for offset in range(0, len(booster), _MODEL_CHUNK_SIZE):
+            booster_chunks.append(booster[offset : offset + _MODEL_CHUNK_SIZE])
+
+        _get_spark_session().sparkContext.parallelize(booster_chunks, 1).saveAsTextFile(
             model_save_path
         )
 
@@ -1721,8 +1731,8 @@ def load(self, path: str) -> "_SparkXGBModel":
         )
         model_load_path = os.path.join(path, "model")
 
-        ser_xgb_model = (
-            _get_spark_session().sparkContext.textFile(model_load_path).collect()[0]
+        ser_xgb_model = "".join(
+            _get_spark_session().sparkContext.textFile(model_load_path).collect()
         )
 
         def create_xgb_model() -> "XGBModel":

src/common/random.h

@@ -230,7 +230,7 @@ class ColumnSampler {
 };
 
 inline auto MakeColumnSampler(Context const* ctx) {
-  std::uint32_t seed = common::GlobalRandomEngine()();
+  std::uint32_t seed = common::GlobalRandom()();
   auto rc = collective::Broadcast(ctx, linalg::MakeVec(&seed, 1), 0);
   collective::SafeColl(rc);
   auto cs = std::make_shared<common::ColumnSampler>(seed);

src/tree/updater_gpu_hist.cu

@@ -867,12 +867,7 @@ class GPUHistMaker : public TreeUpdater {
     CHECK_GE(ctx_->Ordinal(), 0) << "Must have at least one device";
     info_ = &dmat->Info();
 
-    // Synchronise the column sampling seed
-    uint32_t column_sampling_seed = common::GlobalRandom()();
-    auto rc = collective::Broadcast(
-        ctx_, linalg::MakeVec(&column_sampling_seed, sizeof(column_sampling_seed)), 0);
-    SafeColl(rc);
-    this->column_sampler_ = std::make_shared<common::ColumnSampler>(column_sampling_seed);
+    this->column_sampler_ = common::MakeColumnSampler(ctx_);
 
     auto batch_param = BatchParam{param->max_bin, TrainParam::DftSparseThreshold()};
     dh::safe_cuda(cudaSetDevice(ctx_->Ordinal()));
@@ -1012,8 +1007,7 @@ class GPUGlobalApproxMaker : public TreeUpdater {
 
     monitor_.Start(__func__);
     CHECK(ctx_->IsCUDA()) << error::InvalidCUDAOrdinal();
-    uint32_t column_sampling_seed = common::GlobalRandom()();
-    this->column_sampler_ = std::make_shared<common::ColumnSampler>(column_sampling_seed);
+    this->column_sampler_ = common::MakeColumnSampler(ctx_);
 
     p_last_fmat_ = p_fmat;
     initialised_ = true;