Release patch release 1.3.2

* Remove duplicated code, which fixes typo `evals_result` -> `evals_result_`.

CMakeLists.txt

@@ -1,5 +1,5 @@
 cmake_minimum_required(VERSION 3.13)
-project(xgboost LANGUAGES CXX C VERSION 1.3.0)
+project(xgboost LANGUAGES CXX C VERSION 1.3.2)
 include(cmake/Utils.cmake)
 list(APPEND CMAKE_MODULE_PATH "${xgboost_SOURCE_DIR}/cmake/modules")
 cmake_policy(SET CMP0022 NEW)

Jenkinsfile

@@ -198,10 +198,10 @@ def BuildCUDA(args) {
     """
     if (args.cuda_version == ref_cuda_ver) {
       sh """
-      ${dockerRun} ${container_type} ${docker_binary} ${docker_args} auditwheel repair --plat ${wheel_tag} python-package/dist/*.whl
+      ${dockerRun} auditwheel_x86_64 ${docker_binary} auditwheel repair --plat ${wheel_tag} python-package/dist/*.whl
       mv -v wheelhouse/*.whl python-package/dist/
       # Make sure that libgomp.so is vendored in the wheel
-      ${dockerRun} ${container_type} ${docker_binary} ${docker_args} bash -c "unzip -l python-package/dist/*.whl | grep libgomp  || exit -1"
+      ${dockerRun} auditwheel_x86_64 ${docker_binary} bash -c "unzip -l python-package/dist/*.whl | grep libgomp  || exit -1"
       """
     }
     echo 'Stashing Python wheel...'

R-package/DESCRIPTION

@@ -1,7 +1,7 @@
 Package: xgboost
 Type: Package
 Title: Extreme Gradient Boosting
-Version: 1.3.0.1
+Version: 1.3.2.1
 Date: 2020-08-28
 Authors@R: c(
   person("Tianqi", "Chen", role = c("aut"),

R-package/tests/helper_scripts/generate_models.R

@@ -2,7 +2,6 @@
 # of saved model files from XGBoost version 0.90 and 1.0.x.
 library(xgboost)
 library(Matrix)
-source('./generate_models_params.R')
 
 set.seed(0)
 metadata <- list(
@@ -53,11 +52,16 @@ generate_logistic_model <- function () {
   y <- sample(0:1, size = metadata$kRows, replace = TRUE)
   stopifnot(max(y) == 1, min(y) == 0)
 
-  data <- xgb.DMatrix(X, label = y, weight = w)
-  params <- list(tree_method = 'hist', num_parallel_tree = metadata$kForests,
-                 max_depth = metadata$kMaxDepth, objective = 'binary:logistic')
-  booster <- xgb.train(params, data, nrounds = metadata$kRounds)
-  save_booster(booster, 'logit')
+  objective <- c('binary:logistic', 'binary:logitraw')
+  name <- c('logit', 'logitraw')
+
+  for (i in seq_len(length(objective))) {
+    data <- xgb.DMatrix(X, label = y, weight = w)
+    params <- list(tree_method = 'hist', num_parallel_tree = metadata$kForests,
+                   max_depth = metadata$kMaxDepth, objective = objective[i])
+    booster <- xgb.train(params, data, nrounds = metadata$kRounds)
+    save_booster(booster, name[i])
+  }
 }
 
 generate_classification_model <- function () {

R-package/tests/testthat/test_model_compatibility.R

@@ -39,6 +39,10 @@ run_booster_check <- function (booster, name) {
     testthat::expect_equal(config$learner$learner_train_param$objective, 'multi:softmax')
     testthat::expect_equal(as.numeric(config$learner$learner_model_param$num_class),
                            metadata$kClasses)
+  } else if (name == 'logitraw') {
+    testthat::expect_equal(get_num_tree(booster), metadata$kForests * metadata$kRounds)
+    testthat::expect_equal(as.numeric(config$learner$learner_model_param$num_class), 0)
+    testthat::expect_equal(config$learner$learner_train_param$objective, 'binary:logitraw')
   } else if (name == 'logit') {
     testthat::expect_equal(get_num_tree(booster), metadata$kForests * metadata$kRounds)
     testthat::expect_equal(as.numeric(config$learner$learner_model_param$num_class), 0)

include/xgboost/base.h

@@ -55,7 +55,7 @@
 #endif  // defined(__GNUC__) && ((__GNUC__ == 4 && __GNUC_MINOR__ >= 8) || __GNUC__ > 4)
 
 #if defined(__GNUC__) && ((__GNUC__ == 4 && __GNUC_MINOR__ >= 8) || __GNUC__ > 4) && \
-    !defined(__CUDACC__)
+    !defined(__CUDACC__) && !defined(__sun) && !defined(sun)
 #include <parallel/algorithm>
 #define XGBOOST_PARALLEL_SORT(X, Y, Z) __gnu_parallel::sort((X), (Y), (Z))
 #define XGBOOST_PARALLEL_STABLE_SORT(X, Y, Z) \

include/xgboost/version_config.h

@@ -6,6 +6,6 @@
 
 #define XGBOOST_VER_MAJOR 1
 #define XGBOOST_VER_MINOR 3
-#define XGBOOST_VER_PATCH 0
+#define XGBOOST_VER_PATCH 2
 
 #endif  // XGBOOST_VERSION_CONFIG_H_

jvm-packages/dev/change_version.sh

@@ -34,9 +34,9 @@ TO_VERSION=$2
 sed_i() {
   perl -p -000 -e "$1" "$2" > "$2.tmp" && mv "$2.tmp" "$2"
 }
-   
+
 export -f sed_i
- 
+
 BASEDIR=$(dirname $0)/..
 find "$BASEDIR" -name 'pom.xml' -not -path '*target*' -print \
   -exec bash -c \

jvm-packages/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>ml.dmlc</groupId>
     <artifactId>xgboost-jvm_2.12</artifactId>
-    <version>1.3.0</version>
+    <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</version>
     </parent>
     <artifactId>xgboost4j-example_2.12</artifactId>
-    <version>1.3.0</version>
+    <version>1.3.2</version>
     <packaging>jar</packaging>
     <build>
         <plugins>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-spark_${scala.binary.version}</artifactId>
-            <version>1.3.0</version>
+            <version>1.3.2</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>
@@ -37,7 +37,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-flink_${scala.binary.version}</artifactId>
-            <version>1.3.0</version>
+            <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</version>
     </parent>
     <artifactId>xgboost4j-flink_2.12</artifactId>
-    <version>1.3.0</version>
+    <version>1.3.2</version>
     <build>
         <plugins>
             <plugin>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j_${scala.binary.version}</artifactId>
-            <version>1.3.0</version>
+            <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</version>
     </parent>
     <artifactId>xgboost4j-gpu_2.12</artifactId>
-    <version>1.3.0</version>
+    <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</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>1.3.0</version>
+            <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</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>1.3.0</version>
+            <version>1.3.2</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>1.3.0</version>
+        <version>1.3.2</version>
     </parent>
     <artifactId>xgboost4j_2.12</artifactId>
-    <version>1.3.0</version>
+    <version>1.3.2</version>
     <packaging>jar</packaging>
 
     <dependencies>

python-package/xgboost/VERSION

@@ -1 +1 @@
-1.3.0
+1.3.2

python-package/xgboost/callback.py

@@ -456,6 +456,7 @@ class LearningRateScheduler(TrainingCallback):
 
     def after_iteration(self, model, epoch, evals_log):
         model.set_param('learning_rate', self.learning_rates(epoch))
+        return False
 
 
 # pylint: disable=too-many-instance-attributes
@@ -565,7 +566,7 @@ class EarlyStopping(TrainingCallback):
     def after_training(self, model: Booster):
         try:
             if self.save_best:
-                model = model[: int(model.attr('best_iteration'))]
+                model = model[: int(model.attr('best_iteration')) + 1]
         except XGBoostError as e:
             raise XGBoostError('`save_best` is not applicable to current booster') from e
         return model
@@ -621,7 +622,7 @@ class EvaluationMonitor(TrainingCallback):
                     msg += self._fmt_metric(data, metric_name, score, stdv)
             msg += '\n'
 
-            if (epoch % self.period) != 0 or self.period == 1:
+            if (epoch % self.period) == 0 or self.period == 1:
                 rabit.tracker_print(msg)
                 self._latest = None
             else:
@@ -677,6 +678,7 @@ class TrainingCheckPoint(TrainingCallback):
                 else:
                     model.save_model(path)
         self._epoch += 1
+        return False
 
 
 class LegacyCallbacks:

python-package/xgboost/core.py

@@ -1,11 +1,12 @@
 # coding: utf-8
 # pylint: disable=too-many-arguments, too-many-branches, invalid-name
-# pylint: disable=too-many-lines, too-many-locals
+# pylint: disable=too-many-lines, too-many-locals, no-self-use
 """Core XGBoost Library."""
 import collections
 # pylint: disable=no-name-in-module,import-error
 from collections.abc import Mapping
 # pylint: enable=no-name-in-module,import-error
+from typing import Dict, Union, List
 import ctypes
 import os
 import re
@@ -1012,6 +1013,7 @@ class Booster(object):
         _check_call(_LIB.XGBoosterCreate(dmats, c_bst_ulong(len(cache)),
                                          ctypes.byref(self.handle)))
         params = params or {}
+        params = self._configure_metrics(params.copy())
         if isinstance(params, list):
             params.append(('validate_parameters', True))
         else:
@@ -1041,6 +1043,17 @@ class Booster(object):
         else:
             raise TypeError('Unknown type:', model_file)
 
+    def _configure_metrics(self, params: Union[Dict, List]) -> Union[Dict, List]:
+        if isinstance(params, dict) and 'eval_metric' in params \
+           and isinstance(params['eval_metric'], list):
+            params = dict((k, v) for k, v in params.items())
+            eval_metrics = params['eval_metric']
+            params.pop("eval_metric", None)
+            params = list(params.items())
+            for eval_metric in eval_metrics:
+                params += [('eval_metric', eval_metric)]
+        return params
+
     def __del__(self):
         if hasattr(self, 'handle') and self.handle is not None:
             _check_call(_LIB.XGBoosterFree(self.handle))

python-package/xgboost/dask.py

@@ -1210,10 +1210,10 @@ class DaskXGBClassifier(DaskScikitLearnBase, XGBClassifierBase):
                                 early_stopping_rounds=early_stopping_rounds,
                                 verbose=verbose)
 
-    async def _predict_proba_async(self, data, output_margin=False,
+    async def _predict_proba_async(self, X, output_margin=False,
                                    base_margin=None):
         test_dmatrix = await DaskDMatrix(
-            client=self.client, data=data, base_margin=base_margin,
+            client=self.client, data=X, base_margin=base_margin,
             missing=self.missing
         )
         pred_probs = await predict(client=self.client,
@@ -1223,11 +1223,11 @@ class DaskXGBClassifier(DaskScikitLearnBase, XGBClassifierBase):
         return pred_probs
 
     # pylint: disable=arguments-differ,missing-docstring
-    def predict_proba(self, data, output_margin=False, base_margin=None):
+    def predict_proba(self, X, output_margin=False, base_margin=None):
         _assert_dask_support()
         return self.client.sync(
             self._predict_proba_async,
-            data,
+            X=X,
             output_margin=output_margin,
             base_margin=base_margin
         )

python-package/xgboost/sklearn.py

@@ -4,6 +4,7 @@
 import copy
 import warnings
 import json
+from typing import Optional
 import numpy as np
 from .core import Booster, DMatrix, XGBoostError, _deprecate_positional_args
 from .training import train
@@ -494,6 +495,13 @@ class XGBModel(XGBModelBase):
         # Delete the attribute after load
         self.get_booster().set_attr(scikit_learn=None)
 
+    def _set_evaluation_result(self, evals_result: Optional[dict]) -> None:
+        if evals_result:
+            for val in evals_result.items():
+                evals_result_key = list(val[1].keys())[0]
+                evals_result[val[0]][evals_result_key] = val[1][evals_result_key]
+            self.evals_result_ = evals_result
+
     @_deprecate_positional_args
     def fit(self, X, y, *, sample_weight=None, base_margin=None,
             eval_set=None, eval_metric=None, early_stopping_rounds=None,
@@ -565,13 +573,6 @@ class XGBModel(XGBModelBase):
 
         """
         self.n_features_in_ = X.shape[1]
-
-        train_dmatrix = DMatrix(data=X, label=y, weight=sample_weight,
-                                base_margin=base_margin,
-                                missing=self.missing,
-                                nthread=self.n_jobs)
-        train_dmatrix.set_info(feature_weights=feature_weights)
-
         evals_result = {}
 
         train_dmatrix, evals = self._wrap_evaluation_matrices(
@@ -601,12 +602,7 @@ class XGBModel(XGBModelBase):
                               verbose_eval=verbose, xgb_model=xgb_model,
                               callbacks=callbacks)
 
-        if evals_result:
-            for val in evals_result.items():
-                evals_result_key = list(val[1].keys())[0]
-                evals_result[val[0]][evals_result_key] = val[1][
-                    evals_result_key]
-            self.evals_result_ = evals_result
+        self._set_evaluation_result(evals_result)
 
         if early_stopping_rounds is not None:
             self.best_score = self._Booster.best_score
@@ -841,14 +837,18 @@ class XGBClassifier(XGBModel, XGBClassifierBase):
             self.classes_ = cp.unique(y.values)
             self.n_classes_ = len(self.classes_)
             can_use_label_encoder = False
-            if not cp.array_equal(self.classes_, cp.arange(self.n_classes_)):
+            expected_classes = cp.arange(self.n_classes_)
+            if (self.classes_.shape != expected_classes.shape or
+                    not (self.classes_ == expected_classes).all()):
                 raise ValueError(label_encoding_check_error)
         elif _is_cupy_array(y):
             import cupy as cp  # pylint: disable=E0401
             self.classes_ = cp.unique(y)
             self.n_classes_ = len(self.classes_)
             can_use_label_encoder = False
-            if not cp.array_equal(self.classes_, cp.arange(self.n_classes_)):
+            expected_classes = cp.arange(self.n_classes_)
+            if (self.classes_.shape != expected_classes.shape or
+                    not (self.classes_ == expected_classes).all()):
                 raise ValueError(label_encoding_check_error)
         else:
             self.classes_ = np.unique(y)
@@ -915,12 +915,7 @@ class XGBClassifier(XGBModel, XGBClassifierBase):
                               callbacks=callbacks)
 
         self.objective = xgb_options["objective"]
-        if evals_result:
-            for val in evals_result.items():
-                evals_result_key = list(val[1].keys())[0]
-                evals_result[val[0]][
-                    evals_result_key] = val[1][evals_result_key]
-            self.evals_result_ = evals_result
+        self._set_evaluation_result(evals_result)
 
         if early_stopping_rounds is not None:
             self.best_score = self._Booster.best_score
@@ -991,10 +986,9 @@ class XGBClassifier(XGBModel, XGBClassifierBase):
             return self._le.inverse_transform(column_indexes)
         return column_indexes
 
-    def predict_proba(self, data, ntree_limit=None, validate_features=False,
+    def predict_proba(self, X, ntree_limit=None, validate_features=False,
                       base_margin=None):
-        """
-        Predict the probability of each `data` example being of a given class.
+        """ Predict the probability of each `X` example being of a given class.
 
         .. note:: This function is not thread safe
 
@@ -1004,21 +998,22 @@ class XGBClassifier(XGBModel, XGBClassifierBase):
 
         Parameters
         ----------
-        data : array_like
+        X : array_like
             Feature matrix.
         ntree_limit : int
-            Limit number of trees in the prediction; defaults to best_ntree_limit if defined
-            (i.e. it has been trained with early stopping), otherwise 0 (use all trees).
+            Limit number of trees in the prediction; defaults to best_ntree_limit if
+            defined (i.e. it has been trained with early stopping), otherwise 0 (use all
+            trees).
         validate_features : bool
-            When this is True, validate that the Booster's and data's feature_names are identical.
-            Otherwise, it is assumed that the feature_names are the same.
+            When this is True, validate that the Booster's and data's feature_names are
+            identical.  Otherwise, it is assumed that the feature_names are the same.
 
         Returns
         -------
         prediction : numpy array
             a numpy array with the probability of each data example being of a given class.
         """
-        test_dmatrix = DMatrix(data, base_margin=base_margin,
+        test_dmatrix = DMatrix(X, base_margin=base_margin,
                                missing=self.missing, nthread=self.n_jobs)
         if ntree_limit is None:
             ntree_limit = getattr(self, "best_ntree_limit", 0)
@@ -1324,12 +1319,7 @@ class XGBRanker(XGBModel):
 
         self.objective = params["objective"]
 
-        if evals_result:
-            for val in evals_result.items():
-                evals_result_key = list(val[1].keys())[0]
-                evals_result[val[0]][evals_result_key] = val[1][evals_result_key]
-            self.evals_result = evals_result
-
+        self._set_evaluation_result(evals_result)
         if early_stopping_rounds is not None:
             self.best_score = self._Booster.best_score
             self.best_iteration = self._Booster.best_iteration

python-package/xgboost/training.py

@@ -4,6 +4,7 @@
 """Training Library containing training routines."""
 import warnings
 import copy
+import json
 
 import numpy as np
 from .core import Booster, XGBoostError
@@ -40,18 +41,6 @@ def _is_new_callback(callbacks):
                for c in callbacks) or not callbacks
 
 
-def _configure_metrics(params):
-    if isinstance(params, dict) and 'eval_metric' in params \
-       and isinstance(params['eval_metric'], list):
-        params = dict((k, v) for k, v in params.items())
-        eval_metrics = params['eval_metric']
-        params.pop("eval_metric", None)
-        params = list(params.items())
-        for eval_metric in eval_metrics:
-            params += [('eval_metric', eval_metric)]
-    return params
-
-
 def _train_internal(params, dtrain,
                     num_boost_round=10, evals=(),
                     obj=None, feval=None,
@@ -61,7 +50,6 @@ def _train_internal(params, dtrain,
     """internal training function"""
     callbacks = [] if callbacks is None else copy.copy(callbacks)
     evals = list(evals)
-    params = _configure_metrics(params.copy())
 
     bst = Booster(params, [dtrain] + [d[0] for d in evals])
     nboost = 0
@@ -136,7 +124,28 @@ def _train_internal(params, dtrain,
         bst.best_iteration = int(bst.attr('best_iteration'))
     else:
         bst.best_iteration = nboost - 1
-    bst.best_ntree_limit = (bst.best_iteration + 1) * num_parallel_tree
+
+    config = json.loads(bst.save_config())
+    booster = config['learner']['gradient_booster']['name']
+    if booster == 'gblinear':
+        num_parallel_tree = 0
+    elif booster == 'dart':
+        num_parallel_tree = int(
+            config['learner']['gradient_booster']['gbtree']['gbtree_train_param'][
+                'num_parallel_tree'
+            ]
+        )
+    elif booster == 'gbtree':
+        num_parallel_tree = int(
+            config['learner']['gradient_booster']['gbtree_train_param'][
+                'num_parallel_tree']
+        )
+    else:
+        raise ValueError(f'Unknown booster: {booster}')
+    num_groups = int(config['learner']['learner_model_param']['num_class'])
+    num_groups = 1 if num_groups == 0 else num_groups
+    bst.best_ntree_limit = (bst.best_iteration + 1) * num_parallel_tree * num_groups
+
     # Copy to serialise and unserialise booster to reset state and free
     # training memory
     return bst.copy()

rabit/include/rabit/internal/socket.h

@@ -25,6 +25,10 @@
 #include <sys/socket.h>
 #include <sys/ioctl.h>
 
+#if defined(__sun) || defined(sun)
+#include <sys/sockio.h>
+#endif  // defined(__sun) || defined(sun)
+
 #endif  // defined(_WIN32)
 
 #include <string>

src/objective/regression_loss.h

@@ -162,6 +162,9 @@ struct LogisticRaw : public LogisticRegression {
     predt = common::Sigmoid(predt);
     return std::max(predt * (T(1.0f) - predt), eps);
   }
+  static bst_float ProbToMargin(bst_float base_score) {
+    return base_score;
+  }
   static const char* DefaultEvalMetric() { return "auc"; }
 
   static const char* Name() { return "binary:logitraw"; }

tests/ci_build/Dockerfile.auditwheel_x86_64

@@ -0,0 +1,15 @@
+FROM quay.io/pypa/manylinux2010_x86_64
+
+# Install lightweight sudo (not bound to TTY)
+ENV GOSU_VERSION 1.10
+RUN set -ex; \
+    curl -o /usr/local/bin/gosu -L "https://github.com/tianon/gosu/releases/download/$GOSU_VERSION/gosu-amd64" && \
+    chmod +x /usr/local/bin/gosu && \
+    gosu nobody true
+
+# Default entry-point to use if running locally
+# It will preserve attributes of created files
+COPY entrypoint.sh /scripts/
+
+WORKDIR /workspace
+ENTRYPOINT ["/scripts/entrypoint.sh"]

tests/ci_build/conda_env/win64_test.yml

@@ -9,7 +9,6 @@ dependencies:
 - scikit-learn
 - pandas
 - pytest
-- python-graphviz
 - boto3
 - hypothesis
 - jsonschema
@@ -17,3 +16,4 @@ dependencies:
 - pip:
   - cupy-cuda101
   - modin[all]
+  - graphviz

tests/python-gpu/test_gpu_with_dask.py

@@ -5,8 +5,10 @@ import numpy as np
 import asyncio
 import xgboost
 import subprocess
+import hypothesis
 from hypothesis import given, strategies, settings, note
 from hypothesis._settings import duration
+from hypothesis import HealthCheck
 from test_gpu_updaters import parameter_strategy
 
 if sys.platform.startswith("win"):
@@ -19,6 +21,11 @@ from test_with_dask import _get_client_workers  # noqa
 from test_with_dask import generate_array     # noqa
 import testing as tm                          # noqa
 
+if hasattr(HealthCheck, 'function_scoped_fixture'):
+    suppress = [HealthCheck.function_scoped_fixture]
+else:
+    suppress = hypothesis.utils.conventions.not_set
+
 
 try:
     import dask.dataframe as dd
@@ -161,19 +168,24 @@ class TestDistributedGPU:
             run_with_dask_dataframe(dxgb.DaskDMatrix, client)
             run_with_dask_dataframe(dxgb.DaskDeviceQuantileDMatrix, client)
 
-    @given(params=parameter_strategy, num_rounds=strategies.integers(1, 20),
-           dataset=tm.dataset_strategy)
-    @settings(deadline=duration(seconds=120))
+    @given(
+        params=parameter_strategy,
+        num_rounds=strategies.integers(1, 20),
+        dataset=tm.dataset_strategy,
+    )
+    @settings(deadline=duration(seconds=120), suppress_health_check=suppress)
     @pytest.mark.skipif(**tm.no_dask())
     @pytest.mark.skipif(**tm.no_dask_cuda())
-    @pytest.mark.parametrize('local_cuda_cluster', [{'n_workers': 2}], indirect=['local_cuda_cluster'])
+    @pytest.mark.parametrize(
+        "local_cuda_cluster", [{"n_workers": 2}], indirect=["local_cuda_cluster"]
+    )
     @pytest.mark.mgpu
     def test_gpu_hist(self, params, num_rounds, dataset, local_cuda_cluster):
         with Client(local_cuda_cluster) as client:
-            run_gpu_hist(params, num_rounds, dataset, dxgb.DaskDMatrix,
-                         client)
-            run_gpu_hist(params, num_rounds, dataset,
-                         dxgb.DaskDeviceQuantileDMatrix, client)
+            run_gpu_hist(params, num_rounds, dataset, dxgb.DaskDMatrix, client)
+            run_gpu_hist(
+                params, num_rounds, dataset, dxgb.DaskDeviceQuantileDMatrix, client
+            )
 
     @pytest.mark.skipif(**tm.no_cupy())
     @pytest.mark.skipif(**tm.no_dask())

tests/python/generate_models.py

@@ -64,22 +64,24 @@ def generate_logistic_model():
     y = np.random.randint(0, 2, size=kRows)
     assert y.max() == 1 and y.min() == 0
 
-    data = xgboost.DMatrix(X, label=y, weight=w)
-    booster = xgboost.train({'tree_method': 'hist',
-                             'num_parallel_tree': kForests,
-                             'max_depth': kMaxDepth,
-                             'objective': 'binary:logistic'},
-                            num_boost_round=kRounds, dtrain=data)
-    booster.save_model(booster_bin('logit'))
-    booster.save_model(booster_json('logit'))
+    for objective, name in [('binary:logistic', 'logit'), ('binary:logitraw', 'logitraw')]:
+        data = xgboost.DMatrix(X, label=y, weight=w)
+        booster = xgboost.train({'tree_method': 'hist',
+                                 'num_parallel_tree': kForests,
+                                 'max_depth': kMaxDepth,
+                                 'objective': objective},
+                                num_boost_round=kRounds, dtrain=data)
+        booster.save_model(booster_bin(name))
+        booster.save_model(booster_json(name))
 
-    reg = xgboost.XGBClassifier(tree_method='hist',
-                                num_parallel_tree=kForests,
-                                max_depth=kMaxDepth,
-                                n_estimators=kRounds)
-    reg.fit(X, y, w)
-    reg.save_model(skl_bin('logit'))
-    reg.save_model(skl_json('logit'))
+        reg = xgboost.XGBClassifier(tree_method='hist',
+                                    num_parallel_tree=kForests,
+                                    max_depth=kMaxDepth,
+                                    n_estimators=kRounds,
+                                    objective=objective)
+        reg.fit(X, y, w)
+        reg.save_model(skl_bin(name))
+        reg.save_model(skl_json(name))
 
 
 def generate_classification_model():

tests/python/test_basic.py

@@ -57,6 +57,25 @@ class TestBasic:
             # assert they are the same
             assert np.sum(np.abs(preds2 - preds)) == 0
 
+    def test_metric_config(self):
+        # Make sure that the metric configuration happens in booster so the
+        # string `['error', 'auc']` doesn't get passed down to core.
+        dtrain = xgb.DMatrix(dpath + 'agaricus.txt.train')
+        dtest = xgb.DMatrix(dpath + 'agaricus.txt.test')
+        param = {'max_depth': 2, 'eta': 1, 'verbosity': 0,
+                 'objective': 'binary:logistic', 'eval_metric': ['error', 'auc']}
+        watchlist = [(dtest, 'eval'), (dtrain, 'train')]
+        num_round = 2
+        booster = xgb.train(param, dtrain, num_round, watchlist)
+        predt_0 = booster.predict(dtrain)
+        with tempfile.TemporaryDirectory() as tmpdir:
+            path = os.path.join(tmpdir, 'model.json')
+            booster.save_model(path)
+
+            booster = xgb.Booster(params=param, model_file=path)
+            predt_1 = booster.predict(dtrain)
+            np.testing.assert_allclose(predt_0, predt_1)
+
     def test_record_results(self):
         dtrain = xgb.DMatrix(dpath + 'agaricus.txt.train')
         dtest = xgb.DMatrix(dpath + 'agaricus.txt.test')
@@ -124,8 +143,8 @@ class TestBasic:
 
         dump2 = bst.get_dump(with_stats=True)
         assert dump2[0].count('\n') == 3, 'Expected 1 root and 2 leaves - 3 lines in dump.'
-        assert (dump2[0].find('\n') > dump1[0].find('\n'),
-                'Expected more info when with_stats=True is given.')
+        msg = 'Expected more info when with_stats=True is given.'
+        assert dump2[0].find('\n') > dump1[0].find('\n'), msg
 
         dump3 = bst.get_dump(dump_format="json")
         dump3j = json.loads(dump3[0])
@@ -248,13 +267,11 @@ class TestBasicPathLike:
         assert binary_path.exists()
         Path.unlink(binary_path)
 
-
     def test_Booster_init_invalid_path(self):
         """An invalid model_file path should raise XGBoostError."""
         with pytest.raises(xgb.core.XGBoostError):
             xgb.Booster(model_file=Path("invalidpath"))
 
-
     def test_Booster_save_and_load(self):
         """Saving and loading model files from paths."""
         save_path = Path("saveload.model")

tests/python/test_callback.py

@@ -33,15 +33,18 @@ class TestCallbacks:
                       verbose_eval=verbose_eval)
             output: str = out.getvalue().strip()
 
-        pos = 0
-        msg = 'Train-error'
-        for i in range(rounds // int(verbose_eval)):
-            pos = output.find('Train-error', pos)
-            assert pos != -1
-            pos += len(msg)
-
-        assert output.find('Train-error', pos) == -1
-
+        if int(verbose_eval) == 1:
+            # Should print each iteration info
+            assert len(output.split('\n')) == rounds
+        elif int(verbose_eval) > rounds:
+            # Should print first and latest iteration info
+            assert len(output.split('\n')) == 2
+        else:
+            # Should print info by each period additionaly to first and latest iteration
+            num_periods = rounds // int(verbose_eval)
+            # Extra information is required for latest iteration
+            is_extra_info_required = num_periods * int(verbose_eval) < (rounds - 1) 
+            assert len(output.split('\n')) == 1 + num_periods + int(is_extra_info_required)
 
     def test_evaluation_monitor(self):
         D_train = xgb.DMatrix(self.X_train, self.y_train)
@@ -57,8 +60,10 @@ class TestCallbacks:
         assert len(evals_result['Train']['error']) == rounds
         assert len(evals_result['Valid']['error']) == rounds
 
-        self.run_evaluation_monitor(D_train, D_valid, rounds, 2)
         self.run_evaluation_monitor(D_train, D_valid, rounds, True)
+        self.run_evaluation_monitor(D_train, D_valid, rounds, 2)
+        self.run_evaluation_monitor(D_train, D_valid, rounds, 4)
+        self.run_evaluation_monitor(D_train, D_valid, rounds, rounds + 1)        
 
     def test_early_stopping(self):
         D_train = xgb.DMatrix(self.X_train, self.y_train)
@@ -148,7 +153,7 @@ class TestCallbacks:
                 eval_metric=tm.eval_error_metric, callbacks=[early_stop])
         booster = cls.get_booster()
         dump = booster.get_dump(dump_format='json')
-        assert len(dump) == booster.best_iteration
+        assert len(dump) == booster.best_iteration + 1
 
         early_stop = xgb.callback.EarlyStopping(rounds=early_stopping_rounds,
                                                 save_best=True)

tests/python/test_model_compatibility.py

@@ -24,6 +24,10 @@ def run_booster_check(booster, name):
             config['learner']['learner_model_param']['base_score']) == 0.5
         assert config['learner']['learner_train_param'][
             'objective'] == 'multi:softmax'
+    elif name.find('logitraw') != -1:
+        assert len(booster.get_dump()) == gm.kForests * gm.kRounds
+        assert config['learner']['learner_model_param']['num_class'] == str(0)
+        assert config['learner']['learner_train_param']['objective'] == 'binary:logitraw'
     elif name.find('logit') != -1:
         assert len(booster.get_dump()) == gm.kForests * gm.kRounds
         assert config['learner']['learner_model_param']['num_class'] == str(0)
@@ -77,6 +81,13 @@ def run_scikit_model_check(name, path):
         assert config['learner']['learner_train_param'][
             'objective'] == 'rank:ndcg'
         run_model_param_check(config)
+    elif name.find('logitraw') != -1:
+        logit = xgboost.XGBClassifier()
+        logit.load_model(path)
+        assert (len(logit.get_booster().get_dump()) ==
+                gm.kRounds * gm.kForests)
+        config = json.loads(logit.get_booster().save_config())
+        assert config['learner']['learner_train_param']['objective'] == 'binary:logitraw'
     elif name.find('logit') != -1:
         logit = xgboost.XGBClassifier()
         logit.load_model(path)

tests/python/test_training_continuation.py

@@ -123,13 +123,13 @@ class TestTrainingContinuation:
         gbdt_05 = xgb.train(xgb_params_03, dtrain_5class,
                             num_boost_round=7)
         assert gbdt_05.best_ntree_limit == (
-            gbdt_05.best_iteration + 1) * self.num_parallel_tree
+            gbdt_05.best_iteration + 1) * self.num_parallel_tree * 5
         gbdt_05 = xgb.train(xgb_params_03,
                             dtrain_5class,
                             num_boost_round=3,
                             xgb_model=gbdt_05)
         assert gbdt_05.best_ntree_limit == (
-            gbdt_05.best_iteration + 1) * self.num_parallel_tree
+            gbdt_05.best_iteration + 1) * self.num_parallel_tree * 5
 
         res1 = gbdt_05.predict(dtrain_5class)
         res2 = gbdt_05.predict(dtrain_5class,

tests/python/test_with_dask.py

@@ -8,7 +8,8 @@ import asyncio
 from sklearn.datasets import make_classification
 import os
 import subprocess
-from hypothesis import given, settings, note
+import hypothesis
+from hypothesis import given, settings, note, HealthCheck
 from test_updaters import hist_parameter_strategy, exact_parameter_strategy
 
 if sys.platform.startswith("win"):
@@ -17,6 +18,12 @@ if tm.no_dask()['condition']:
     pytest.skip(msg=tm.no_dask()['reason'], allow_module_level=True)
 
 
+if hasattr(HealthCheck, 'function_scoped_fixture'):
+    suppress = [HealthCheck.function_scoped_fixture]
+else:
+    suppress = hypothesis.utils.conventions.not_set
+
+
 try:
     from distributed import LocalCluster, Client, get_client
     from distributed.utils_test import client, loop, cluster_fixture
@@ -668,14 +675,14 @@ class TestWithDask:
 
     @given(params=hist_parameter_strategy,
            dataset=tm.dataset_strategy)
-    @settings(deadline=None)
+    @settings(deadline=None, suppress_health_check=suppress)
     def test_hist(self, params, dataset, client):
         num_rounds = 30
         self.run_updater_test(client, params, num_rounds, dataset, 'hist')
 
     @given(params=exact_parameter_strategy,
            dataset=tm.dataset_strategy)
-    @settings(deadline=None)
+    @settings(deadline=None, suppress_health_check=suppress)
     def test_approx(self, client, params, dataset):
         num_rounds = 30
         self.run_updater_test(client, params, num_rounds, dataset, 'approx')
@@ -795,7 +802,6 @@ class TestDaskCallbacks:
                 merged = xgb.dask._get_workers_from_data(train, evals=[(valid, 'Valid')])
                 assert len(merged) == 2
 
-
     def test_data_initialization(self):
         '''Assert each worker has the correct amount of data, and DMatrix initialization doesn't
         generate unnecessary copies of data.

tests/python/test_with_sklearn.py

@@ -78,6 +78,34 @@ def test_multiclass_classification():
         check_pred(preds4, labels, output_margin=False)
 
 
+def test_best_ntree_limit():
+    from sklearn.datasets import load_iris
+
+    X, y = load_iris(return_X_y=True)
+
+    def train(booster, forest):
+        rounds = 4
+        cls = xgb.XGBClassifier(
+            n_estimators=rounds, num_parallel_tree=forest, booster=booster
+        ).fit(
+            X, y, eval_set=[(X, y)], early_stopping_rounds=3
+        )
+
+        if forest:
+            assert cls.best_ntree_limit == rounds * forest * cls.n_classes_
+        else:
+            assert cls.best_ntree_limit == 0
+
+        # best_ntree_limit is used by default, assert that under gblinear it's
+        # automatically ignored due to being 0.
+        cls.predict(X)
+
+    num_parallel_tree = 4
+    train('gbtree', num_parallel_tree)
+    train('dart', num_parallel_tree)
+    train('gblinear', None)
+
+
 def test_ranking():
     # generate random data
     x_train = np.random.rand(1000, 10)
@@ -94,6 +122,8 @@ def test_ranking():
     model = xgb.sklearn.XGBRanker(**params)
     model.fit(x_train, y_train, group=train_group,
               eval_set=[(x_valid, y_valid)], eval_group=[valid_group])
+    assert model.evals_result()
+
     pred = model.predict(x_test)
 
     train_data = xgb.DMatrix(x_train, y_train)