[R] release 1.5.0.2 (#7452)

* [R] release 1.5.0.2

* Add cmake list to r build ignore.

.github/workflows/python_tests.yml

@@ -45,59 +45,13 @@ jobs:
         cd ..
         python -c 'import xgboost'
 
-  python-tests-on-win:
+  python-tests:
     name: Test XGBoost Python package on ${{ matrix.config.os }}
     runs-on: ${{ matrix.config.os }}
     strategy:
       matrix:
         config:
           - {os: windows-2016, python-version: '3.8'}
-
-    steps:
-    - uses: actions/checkout@v2
-      with:
-        submodules: 'true'
-
-    - uses: conda-incubator/setup-miniconda@v2
-      with:
-        auto-update-conda: true
-        python-version: ${{ matrix.config.python-version }}
-        activate-environment: win64_env
-        environment-file: tests/ci_build/conda_env/win64_cpu_test.yml
-
-    - name: Display Conda env
-      shell: bash -l {0}
-      run: |
-        conda info
-        conda list
-
-    - name: Build XGBoost on Windows
-      shell: bash -l {0}
-      run: |
-        mkdir build_msvc
-        cd build_msvc
-        cmake .. -G"Visual Studio 15 2017" -DCMAKE_CONFIGURATION_TYPES="Release" -A x64 -DGOOGLE_TEST=ON  -DUSE_DMLC_GTEST=ON
-        cmake --build . --config Release --parallel $(nproc)
-
-    - name: Install Python package
-      shell: bash -l {0}
-      run: |
-        cd python-package
-        python --version
-        python setup.py bdist_wheel --universal
-        pip install ./dist/*.whl
-
-    - name: Test Python package
-      shell: bash -l {0}
-      run: |
-        pytest -s -v ./tests/python
-
-  python-tests-on-macos:
-    name: Test XGBoost Python package on ${{ matrix.config.os }}
-    runs-on: ${{ matrix.config.os }}
-    strategy:
-      matrix:
-        config:
           - {os: macos-10.15, python-version "3.8" }
 
     steps:
@@ -109,8 +63,8 @@ jobs:
       with:
         auto-update-conda: true
         python-version: ${{ matrix.config.python-version }}
-        activate-environment: macos_test
-        environment-file: tests/ci_build/conda_env/macos_cpu_test.yml
+        activate-environment: win64_test
+        environment-file: tests/ci_build/conda_env/win64_cpu_test.yml
 
     - name: Display Conda env
       shell: bash -l {0}
@@ -118,7 +72,17 @@ jobs:
         conda info
         conda list
 
+    - name: Build XGBoost on Windows
+      shell: bash -l {0}
+      if: matrix.config.os == 'windows-2016'
+      run: |
+        mkdir build_msvc
+        cd build_msvc
+        cmake .. -G"Visual Studio 15 2017" -DCMAKE_CONFIGURATION_TYPES="Release" -A x64 -DGOOGLE_TEST=ON  -DUSE_DMLC_GTEST=ON
+        cmake --build . --config Release --parallel $(nproc)
+
     - name: Build XGBoost on macos
+      if: matrix.config.os == 'macos-10.15'
       run: |
         wget https://raw.githubusercontent.com/Homebrew/homebrew-core/679923b4eb48a8dc7ecc1f05d06063cd79b3fc00/Formula/libomp.rb -O $(find $(brew --repository) -name libomp.rb)
         brew install ninja libomp
@@ -144,19 +108,16 @@ jobs:
 
     - name: Rename Python wheel
       shell: bash -l {0}
+      if: matrix.config.os == 'macos-10.15'
       run: |
         TAG=macosx_10_15_x86_64.macosx_11_0_x86_64.macosx_12_0_x86_64
         python tests/ci_build/rename_whl.py python-package/dist/*.whl ${{ github.sha }} ${TAG}
 
-    - name: Extract branch name
-      shell: bash
-      run: echo "##[set-output name=branch;]$(echo ${GITHUB_REF#refs/heads/})"
-      id: extract_branch
-      if: github.ref == 'refs/heads/master' || contains(github.ref, 'refs/heads/release_')
-
     - name: Upload Python wheel
       shell: bash -l {0}
-      if: github.ref == 'refs/heads/master' || contains(github.ref, 'refs/heads/release_')
+      if: |
+        (github.ref == 'refs/heads/master' || contains(github.ref, 'refs/heads/release_')) &&
+        matrix.os == 'macos-latest'
       run: |
         python -m awscli s3 cp python-package/dist/*.whl s3://xgboost-nightly-builds/${{ steps.extract_branch.outputs.branch }}/ --acl public-read
       env:

CMakeLists.txt

@@ -1,5 +1,5 @@
 cmake_minimum_required(VERSION 3.14 FATAL_ERROR)
-project(xgboost LANGUAGES CXX C VERSION 1.5.2)
+project(xgboost LANGUAGES CXX C VERSION 1.5.0)
 include(cmake/Utils.cmake)
 list(APPEND CMAKE_MODULE_PATH "${xgboost_SOURCE_DIR}/cmake/modules")
 cmake_policy(SET CMP0022 NEW)
@@ -135,10 +135,6 @@ if (USE_CUDA)
   set(GEN_CODE "")
   format_gencode_flags("${GPU_COMPUTE_VER}" GEN_CODE)
   add_subdirectory(${PROJECT_SOURCE_DIR}/gputreeshap)
-
-  if ((${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 11.4) AND (NOT BUILD_WITH_CUDA_CUB))
-    message(SEND_ERROR "`BUILD_WITH_CUDA_CUB` should be set to `ON` for CUDA >= 11.4")
-  endif ()
 endif (USE_CUDA)
 
 if (FORCE_COLORED_OUTPUT AND (CMAKE_GENERATOR STREQUAL "Ninja") AND

R-package/.Rbuildignore

@@ -4,3 +4,4 @@
 ^.*\.Rproj$
 ^\.Rproj\.user$
 README.md
+CMakeLists.txt

R-package/DESCRIPTION

@@ -1,8 +1,8 @@
 Package: xgboost
 Type: Package
 Title: Extreme Gradient Boosting
-Version: 1.5.2.1
-Date: 2022-1-17
+Version: 1.5.0.2
+Date: 2021-11-19
 Authors@R: c(
   person("Tianqi", "Chen", role = c("aut"),
          email = "tianqi.tchen@gmail.com"),
@@ -26,11 +26,9 @@ Authors@R: c(
   person("Min", "Lin", role = c("aut")),
   person("Yifeng", "Geng", role = c("aut")),
   person("Yutian", "Li", role = c("aut")),
-  person("Jiaming", "Yuan", role = c("aut")),
   person("XGBoost contributors", role = c("cph"),
          comment = "base XGBoost implementation")
   )
-Maintainer: Jiaming Yuan <jm.yuan@outlook.com>
 Description: Extreme Gradient Boosting, which is an efficient implementation
     of the gradient boosting framework from Chen & Guestrin (2016) <doi:10.1145/2939672.2939785>.
     This package is its R interface. The package includes efficient linear

R-package/R/xgb.Booster.R

@@ -435,8 +435,7 @@ predict.xgb.Booster <- function(object, newdata, missing = NA, outputmargin = FA
       lapply(seq_len(n_groups), function(g) arr[g, , ])
     } else {
       ## remove the first axis (group)
-      dn <- dimnames(arr)
-      matrix(arr[1, , ], nrow = dim(arr)[2], ncol = dim(arr)[3], dimnames = c(dn[2], dn[3]))
+      as.matrix(arr[1, , ])
     }
   } else if (predinteraction) {
     ## Predict interaction
@@ -448,8 +447,7 @@ predict.xgb.Booster <- function(object, newdata, missing = NA, outputmargin = FA
       lapply(seq_len(n_groups), function(g) arr[g, , , ])
     } else {
       ## remove the first axis (group)
-      arr <- arr[1, , , , drop = FALSE]
-      array(arr, dim = dim(arr)[2:4], dimnames(arr)[2:4])
+      arr[1, , , ]
     }
   } else {
     ## Normal prediction

R-package/R/xgb.create.features.R

@@ -18,7 +18,7 @@
 #'
 #' International Workshop on Data Mining for Online Advertising (ADKDD) - August 24, 2014
 #'
-#' \url{https://research.facebook.com/publications/practical-lessons-from-predicting-clicks-on-ads-at-facebook/}.
+#' \url{https://research.fb.com/publications/practical-lessons-from-predicting-clicks-on-ads-at-facebook/}.
 #'
 #' Extract explaining the method:
 #'

R-package/R/xgb.dump.R

@@ -6,6 +6,8 @@
 #' @param fname the name of the text file where to save the model text dump.
 #'        If not provided or set to \code{NULL}, the model is returned as a \code{character} vector.
 #' @param fmap feature map file representing feature types.
+#'        Detailed description could be found at
+#'        \url{https://github.com/dmlc/xgboost/wiki/Binary-Classification#dump-model}.
 #'        See demo/ for walkthrough example in R, and
 #'        \url{https://github.com/dmlc/xgboost/blob/master/demo/data/featmap.txt}
 #'        for example Format.

R-package/man/xgb.create.features.Rd

@@ -29,7 +29,7 @@ Joaquin Quinonero Candela)}
 
 International Workshop on Data Mining for Online Advertising (ADKDD) - August 24, 2014
 
-\url{https://research.facebook.com/publications/practical-lessons-from-predicting-clicks-on-ads-at-facebook/}.
+\url{https://research.fb.com/publications/practical-lessons-from-predicting-clicks-on-ads-at-facebook/}.
 
 Extract explaining the method:
 

R-package/man/xgb.dump.Rd

@@ -20,6 +20,8 @@ xgb.dump(
 If not provided or set to \code{NULL}, the model is returned as a \code{character} vector.}
 
 \item{fmap}{feature map file representing feature types.
+Detailed description could be found at
+\url{https://github.com/dmlc/xgboost/wiki/Binary-Classification#dump-model}.
 See demo/ for walkthrough example in R, and
 \url{https://github.com/dmlc/xgboost/blob/master/demo/data/featmap.txt}
 for example Format.}

R-package/tests/testthat/test_glm.R

@@ -46,31 +46,3 @@ test_that("gblinear works", {
   expect_equal(dim(h), c(n, ncol(dtrain) + 1))
   expect_s4_class(h, "dgCMatrix")
 })
-
-test_that("gblinear early stopping works", {
-  data(agaricus.train, package = 'xgboost')
-  data(agaricus.test, package = 'xgboost')
-  dtrain <- xgb.DMatrix(agaricus.train$data, label = agaricus.train$label)
-  dtest <- xgb.DMatrix(agaricus.test$data, label = agaricus.test$label)
-
-  param <- list(
-    objective = "binary:logistic", eval_metric = "error", booster = "gblinear",
-    nthread = 2, eta = 0.8, alpha = 0.0001, lambda = 0.0001,
-    updater = "coord_descent"
-  )
-
-  es_round <- 1
-  n <- 10
-  booster <- xgb.train(
-    param, dtrain, n, list(eval = dtest, train = dtrain), early_stopping_rounds = es_round
-  )
-  expect_equal(booster$best_iteration, 5)
-  predt_es <- predict(booster, dtrain)
-
-  n <- booster$best_iteration + es_round
-  booster <- xgb.train(
-    param, dtrain, n, list(eval = dtest, train = dtrain), early_stopping_rounds = es_round
-  )
-  predt <- predict(booster, dtrain)
-  expect_equal(predt_es, predt)
-})

R-package/tests/testthat/test_interactions.R

@@ -157,28 +157,3 @@ test_that("multiclass feature interactions work", {
   # sums WRT columns must be close to feature contributions
   expect_lt(max(abs(apply(intr, c(1, 2, 3), sum) - aperm(cont, c(3, 1, 2)))), 0.00001)
 })
-
-
-test_that("SHAP single sample works", {
-  train <- agaricus.train
-  test <- agaricus.test
-  booster <- xgboost(
-    data = train$data,
-    label = train$label,
-    max_depth = 2,
-    nrounds = 4,
-    objective = "binary:logistic",
-  )
-
-  predt <- predict(
-    booster,
-    newdata = train$data[1, , drop = FALSE], predcontrib = TRUE
-  )
-  expect_equal(dim(predt), c(1, dim(train$data)[2] + 1))
-
-  predt <- predict(
-    booster,
-    newdata = train$data[1, , drop = FALSE], predinteraction = TRUE
-  )
-  expect_equal(dim(predt), c(1, dim(train$data)[2] + 1, dim(train$data)[2] + 1))
-})

R-package/vignettes/discoverYourData.Rmd

@@ -138,7 +138,7 @@ levels(df[,Treatment])
 
 Next step, we will transform the categorical data to dummy variables.
 Several encoding methods exist, e.g., [one-hot encoding](https://en.wikipedia.org/wiki/One-hot) is a common approach.
-We will use the [dummy contrast coding](https://stats.oarc.ucla.edu/r/library/r-library-contrast-coding-systems-for-categorical-variables/) which is popular because it produces "full rank" encoding (also see [this blog post by Max Kuhn](http://appliedpredictivemodeling.com/blog/2013/10/23/the-basics-of-encoding-categorical-data-for-predictive-models)).
+We will use the [dummy contrast coding](https://stats.idre.ucla.edu/r/library/r-library-contrast-coding-systems-for-categorical-variables/) which is popular because it produces "full rank" encoding (also see [this blog post by Max Kuhn](http://appliedpredictivemodeling.com/blog/2013/10/23/the-basics-of-encoding-categorical-data-for-predictive-models)).
 
 The purpose is to transform each value of each *categorical* feature into a *binary* feature `{0, 1}`.
 

doc/build.rst

@@ -148,8 +148,7 @@ From the command line on Linux starting from the XGBoost directory:
 
   mkdir build
   cd build
-  # For CUDA toolkit >= 11.4, `BUILD_WITH_CUDA_CUB` is required.
-  cmake .. -DUSE_CUDA=ON -DBUILD_WITH_CUDA_CUB=ON
+  cmake .. -DUSE_CUDA=ON
   make -j4
 
 .. note:: Specifying compute capability

doc/gpu/index.rst

@@ -95,13 +95,13 @@ XGBoost makes use of `GPUTreeShap <https://github.com/rapidsai/gputreeshap>`_ as
   shap_interaction_values = model.predict(dtrain, pred_interactions=True)
 
 See examples `here
-<https://github.com/dmlc/xgboost/tree/master/demo/gpu_acceleration>`__.
+<https://github.com/dmlc/xgboost/tree/master/demo/gpu_acceleration>`_.
 
 Multi-node Multi-GPU Training
 =============================
 XGBoost supports fully distributed GPU training using `Dask <https://dask.org/>`_. For
 getting started see our tutorial :doc:`/tutorials/dask` and worked examples `here
-<https://github.com/dmlc/xgboost/tree/master/demo/dask>`__, also Python documentation
+<https://github.com/dmlc/xgboost/tree/master/demo/dask>`_, also Python documentation
 :ref:`dask_api` for complete reference.
 
 
@@ -238,7 +238,7 @@ Working memory is allocated inside the algorithm proportional to the number of r
 
 The quantile finding algorithm also uses some amount of working device memory. It is able to operate in batches, but is not currently well optimised for sparse data.
 
-If you are getting out-of-memory errors on a big dataset, try the :doc:`external memory version </tutorials/external_memory>`.
+If you are getting out-of-memory errors on a big dataset, try the `external memory version <../tutorials/external_memory.html>`_.
 
 Developer notes
 ===============

doc/jvm/xgboost4j_spark_tutorial.rst

@@ -79,7 +79,7 @@ The first thing in data transformation is to load the dataset as Spark's structu
     StructField("class", StringType, true)))
   val rawInput = spark.read.schema(schema).csv("input_path")
 
-At the first line, we create a instance of `SparkSession <https://spark.apache.org/docs/latest/sql-getting-started.html#starting-point-sparksession>`_ which is the entry of any Spark program working with DataFrame. The ``schema`` variable defines the schema of DataFrame wrapping Iris data. With this explicitly set schema, we can define the columns' name as well as their types; otherwise the column name would be the default ones derived by Spark, such as ``_col0``, etc. Finally, we can use Spark's built-in csv reader to load Iris csv file as a DataFrame named ``rawInput``.
+At the first line, we create a instance of `SparkSession <http://spark.apache.org/docs/latest/sql-programming-guide.html#starting-point-sparksession>`_ which is the entry of any Spark program working with DataFrame. The ``schema`` variable defines the schema of DataFrame wrapping Iris data. With this explicitly set schema, we can define the columns' name as well as their types; otherwise the column name would be the default ones derived by Spark, such as ``_col0``, etc. Finally, we can use Spark's built-in csv reader to load Iris csv file as a DataFrame named ``rawInput``.
 
 Spark also contains many built-in readers for other format. The latest version of Spark supports CSV, JSON, Parquet, and LIBSVM.
 
@@ -130,7 +130,7 @@ labels. A DataFrame like this (containing vector-represented features and numeri
 Dealing with missing values
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
-XGBoost supports missing values by default (`as desribed here <https://xgboost.readthedocs.io/en/latest/faq.html#how-to-deal-with-missing-values>`_).
+XGBoost supports missing values by default (`as desribed here <https://xgboost.readthedocs.io/en/latest/faq.html#how-to-deal-with-missing-value>`_).
 If given a SparseVector, XGBoost will treat any values absent from the SparseVector as missing. You are also able to
 specify to XGBoost to treat a specific value in your Dataset as if it was a missing value. By default XGBoost will treat NaN as the value representing missing.
 
@@ -369,7 +369,7 @@ Then we can load this model with single node Python XGBoost:
 
   When interacting with other language bindings, XGBoost also supports saving-models-to and loading-models-from file systems other than the local one. You can use HDFS and S3 by prefixing the path with ``hdfs://`` and ``s3://`` respectively. However, for this capability, you must do **one** of the following:
 
-  1. Build XGBoost4J-Spark with the steps described in :ref:`here <install_jvm_packages>`, but turning `USE_HDFS <https://github.com/dmlc/xgboost/blob/e939192978a0c152ad7b49b744630e99d54cffa8/jvm-packages/create_jni.py#L18>`_ (or USE_S3, etc. in the same place) switch on. With this approach, you can reuse the above code example by replacing "nativeModelPath" with a HDFS path.
+  1. Build XGBoost4J-Spark with the steps described in `here <https://xgboost.readthedocs.io/en/latest/jvm/index.html#installation-from-source>`_, but turning `USE_HDFS <https://github.com/dmlc/xgboost/blob/e939192978a0c152ad7b49b744630e99d54cffa8/jvm-packages/create_jni.py#L18>`_ (or USE_S3, etc. in the same place) switch on. With this approach, you can reuse the above code example by replacing "nativeModelPath" with a HDFS path.
 
      - However, if you build with USE_HDFS, etc. you have to ensure that the involved shared object file, e.g. libhdfs.so, is put in the LIBRARY_PATH of your cluster. To avoid the complicated cluster environment configuration, choose the other option.
 

doc/parameter.rst

@@ -366,8 +366,8 @@ Specify the learning task and the corresponding learning objective. The objectiv
   - ``rank:pairwise``: Use LambdaMART to perform pairwise ranking where the pairwise loss is minimized
   - ``rank:ndcg``: Use LambdaMART to perform list-wise ranking where `Normalized Discounted Cumulative Gain (NDCG) <http://en.wikipedia.org/wiki/NDCG>`_ is maximized
   - ``rank:map``: Use LambdaMART to perform list-wise ranking where `Mean Average Precision (MAP) <http://en.wikipedia.org/wiki/Mean_average_precision#Mean_average_precision>`_ is maximized
-  - ``reg:gamma``: gamma regression with log-link. Output is a mean of gamma distribution. It might be useful, e.g., for modeling insurance claims severity, or for any outcome that might be `gamma-distributed <https://en.wikipedia.org/wiki/Gamma_distribution#Occurrence_and_applications>`_.
-  - ``reg:tweedie``: Tweedie regression with log-link. It might be useful, e.g., for modeling total loss in insurance, or for any outcome that might be `Tweedie-distributed <https://en.wikipedia.org/wiki/Tweedie_distribution#Occurrence_and_applications>`_.
+  - ``reg:gamma``: gamma regression with log-link. Output is a mean of gamma distribution. It might be useful, e.g., for modeling insurance claims severity, or for any outcome that might be `gamma-distributed <https://en.wikipedia.org/wiki/Gamma_distribution#Applications>`_.
+  - ``reg:tweedie``: Tweedie regression with log-link. It might be useful, e.g., for modeling total loss in insurance, or for any outcome that might be `Tweedie-distributed <https://en.wikipedia.org/wiki/Tweedie_distribution#Applications>`_.
 
 * ``base_score`` [default=0.5]
 
@@ -390,7 +390,7 @@ Specify the learning task and the corresponding learning objective. The objectiv
     - ``error@t``: a different than 0.5 binary classification threshold value could be specified by providing a numerical value through 't'.
     - ``merror``: Multiclass classification error rate. It is calculated as ``#(wrong cases)/#(all cases)``.
     - ``mlogloss``: `Multiclass logloss <http://scikit-learn.org/stable/modules/generated/sklearn.metrics.log_loss.html>`_.
-    - ``auc``: `Receiver Operating Characteristic Area under the Curve <https://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_the_curve>`_.
+    - ``auc``: `Receiver Operating Characteristic Area under the Curve <http://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_curve>`_.
       Available for classification and learning-to-rank tasks.
 
       - When used with binary classification, the objective should be ``binary:logistic`` or similar functions that work on probability.

doc/tutorials/kubernetes.rst

@@ -11,7 +11,7 @@ In order to run a XGBoost job in a Kubernetes cluster, perform the following ste
 
 1. Install XGBoost Operator on the Kubernetes cluster.
 
-   a. XGBoost Operator is designed to manage the scheduling and monitoring of XGBoost jobs. Follow `this installation guide <https://github.com/kubeflow/xgboost-operator#install-xgboost-operator>`_ to install XGBoost Operator.
+   a. XGBoost Operator is designed to manage the scheduling and monitoring of XGBoost jobs. Follow `this installation guide <https://github.com/kubeflow/xgboost-operator#installing-xgboost-operator>`_ to install XGBoost Operator.
 
 2. Write application code that will be executed by the XGBoost Operator.
 

doc/tutorials/saving_model.rst

@@ -227,15 +227,15 @@ XGBoost has a function called ``dump_model`` in Booster object, which lets you t
 the model in a readable format like ``text``, ``json`` or ``dot`` (graphviz).  The primary
 use case for it is for model interpretation or visualization, and is not supposed to be
 loaded back to XGBoost.  The JSON version has a `schema
-<https://github.com/dmlc/xgboost/blob/master/doc/dump.schema>`__.  See next section for
+<https://github.com/dmlc/xgboost/blob/master/doc/dump.schema>`_.  See next section for
 more info.
 
 ***********
 JSON Schema
 ***********
 
-Another important feature of JSON format is a documented `schema
-<https://json-schema.org/>`__, based on which one can easily reuse the output model from
+Another important feature of JSON format is a documented `Schema
+<https://json-schema.org/>`_, based on which one can easily reuse the output model from
 XGBoost.  Here is the initial draft of JSON schema for the output model (not
 serialization, which will not be stable as noted above).  It's subject to change due to
 the beta status.  For an example of parsing XGBoost tree model, see ``/demo/json-model``.

include/xgboost/version_config.h

@@ -6,6 +6,6 @@
 
 #define XGBOOST_VER_MAJOR 1
 #define XGBOOST_VER_MINOR 5
-#define XGBOOST_VER_PATCH 2
+#define XGBOOST_VER_PATCH 0
 
 #endif  // XGBOOST_VERSION_CONFIG_H_

jvm-packages/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>ml.dmlc</groupId>
     <artifactId>xgboost-jvm_2.12</artifactId>
-    <version>1.5.2</version>
+    <version>1.5.0</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>1.5.2</version>
+        <version>1.5.0</version>
     </parent>
     <artifactId>xgboost4j-example_2.12</artifactId>
-    <version>1.5.2</version>
+    <version>1.5.0</version>
     <packaging>jar</packaging>
     <build>
         <plugins>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-spark_${scala.binary.version}</artifactId>
-            <version>1.5.2</version>
+            <version>1.5.0</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>
@@ -37,7 +37,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j-flink_${scala.binary.version}</artifactId>
-            <version>1.5.2</version>
+            <version>1.5.0</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>1.5.2</version>
+        <version>1.5.0</version>
     </parent>
     <artifactId>xgboost4j-flink_2.12</artifactId>
-    <version>1.5.2</version>
+    <version>1.5.0</version>
     <build>
         <plugins>
             <plugin>
@@ -26,7 +26,7 @@
         <dependency>
             <groupId>ml.dmlc</groupId>
             <artifactId>xgboost4j_${scala.binary.version}</artifactId>
-            <version>1.5.2</version>
+            <version>1.5.0</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>1.5.2</version>
+        <version>1.5.0</version>
     </parent>
     <artifactId>xgboost4j-gpu_2.12</artifactId>
-    <version>1.5.2</version>
+    <version>1.5.0</version>
     <packaging>jar</packaging>
 
     <properties>

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

@@ -6,7 +6,7 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.5.2</version>
+        <version>1.5.0</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>1.5.2</version>
+            <version>1.5.0</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>1.5.2</version>
+        <version>1.5.0</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>1.5.2</version>
+            <version>1.5.0</version>
         </dependency>
         <dependency>
             <groupId>org.apache.spark</groupId>

jvm-packages/xgboost4j-spark/src/main/scala/ml/dmlc/xgboost4j/scala/spark/params/DefaultXGBoostParamsWriter.scala

@@ -17,13 +17,11 @@
 package ml.dmlc.xgboost4j.scala.spark.params
 
 import org.apache.hadoop.fs.Path
-
 import org.apache.spark.SparkContext
 import org.apache.spark.ml.param.{ParamPair, Params}
+import org.json4s.JsonDSL._
 import org.json4s.jackson.JsonMethods._
-import org.json4s.{JArray, JBool, JDouble, JField, JInt, JNothing, JObject, JString, JValue}
-
-import JsonDSLXGBoost._
+import org.json4s.{JObject, _}
 
 // This originates from apache-spark DefaultPramsWriter copy paste
 private[spark] object DefaultXGBoostParamsWriter {
@@ -89,62 +87,3 @@ private[spark] object DefaultXGBoostParamsWriter {
     metadataJson
   }
 }
-
-// Fix json4s bin-incompatible issue.
-// This originates from org.json4s.JsonDSL of 3.6.6
-object JsonDSLXGBoost {
-
-  implicit def seq2jvalue[A](s: Iterable[A])(implicit ev: A => JValue): JArray =
-    JArray(s.toList.map(ev))
-
-  implicit def map2jvalue[A](m: Map[String, A])(implicit ev: A => JValue): JObject =
-    JObject(m.toList.map { case (k, v) => JField(k, ev(v)) })
-
-  implicit def option2jvalue[A](opt: Option[A])(implicit ev: A => JValue): JValue = opt match {
-    case Some(x) => ev(x)
-    case None => JNothing
-  }
-
-  implicit def short2jvalue(x: Short): JValue = JInt(x)
-  implicit def byte2jvalue(x: Byte): JValue = JInt(x)
-  implicit def char2jvalue(x: Char): JValue = JInt(x)
-  implicit def int2jvalue(x: Int): JValue = JInt(x)
-  implicit def long2jvalue(x: Long): JValue = JInt(x)
-  implicit def bigint2jvalue(x: BigInt): JValue = JInt(x)
-  implicit def double2jvalue(x: Double): JValue = JDouble(x)
-  implicit def float2jvalue(x: Float): JValue = JDouble(x.toDouble)
-  implicit def bigdecimal2jvalue(x: BigDecimal): JValue = JDouble(x.doubleValue)
-  implicit def boolean2jvalue(x: Boolean): JValue = JBool(x)
-  implicit def string2jvalue(x: String): JValue = JString(x)
-
-  implicit def symbol2jvalue(x: Symbol): JString = JString(x.name)
-  implicit def pair2jvalue[A](t: (String, A))(implicit ev: A => JValue): JObject =
-    JObject(List(JField(t._1, ev(t._2))))
-  implicit def list2jvalue(l: List[JField]): JObject = JObject(l)
-  implicit def jobject2assoc(o: JObject): JsonListAssoc = new JsonListAssoc(o.obj)
-  implicit def pair2Assoc[A](t: (String, A))(implicit ev: A => JValue): JsonAssoc[A] =
-    new JsonAssoc(t)
-}
-
-final class JsonAssoc[A](private val left: (String, A)) extends AnyVal {
-  def ~[B](right: (String, B))(implicit ev1: A => JValue, ev2: B => JValue): JObject = {
-    val l: JValue = ev1(left._2)
-    val r: JValue = ev2(right._2)
-    JObject(JField(left._1, l) :: JField(right._1, r) :: Nil)
-  }
-
-  def ~(right: JObject)(implicit ev: A => JValue): JObject = {
-    val l: JValue = ev(left._2)
-    JObject(JField(left._1, l) :: right.obj)
-  }
-  def ~~[B](right: (String, B))(implicit ev1: A => JValue, ev2: B => JValue): JObject =
-    this.~(right)
-  def ~~(right: JObject)(implicit ev: A => JValue): JObject = this.~(right)
-}
-
-final class JsonListAssoc(private val left: List[JField]) extends AnyVal {
-  def ~(right: (String, JValue)): JObject = JObject(left ::: List(JField(right._1, right._2)))
-  def ~(right: JObject): JObject = JObject(left ::: right.obj)
-  def ~~(right: (String, JValue)): JObject = this.~(right)
-  def ~~(right: JObject): JObject = this.~(right)
-}

jvm-packages/xgboost4j/pom.xml

@@ -6,10 +6,10 @@
     <parent>
         <groupId>ml.dmlc</groupId>
         <artifactId>xgboost-jvm_2.12</artifactId>
-        <version>1.5.2</version>
+        <version>1.5.0</version>
     </parent>
     <artifactId>xgboost4j_2.12</artifactId>
-    <version>1.5.2</version>
+    <version>1.5.0</version>
     <packaging>jar</packaging>
 
     <dependencies>

python-package/setup.py

@@ -4,7 +4,6 @@ import shutil
 import subprocess
 import logging
 import distutils
-from typing import Optional, List
 import sys
 from platform import system
 from setuptools import setup, find_packages, Extension
@@ -37,7 +36,7 @@ NEED_CLEAN_FILE = set()
 BUILD_TEMP_DIR = None
 
 
-def lib_name() -> str:
+def lib_name():
     '''Return platform dependent shared object name.'''
     if system() == 'Linux' or system().upper().endswith('BSD'):
         name = 'libxgboost.so'
@@ -48,13 +47,13 @@ def lib_name() -> str:
     return name
 
 
-def copy_tree(src_dir: str, target_dir: str) -> None:
+def copy_tree(src_dir, target_dir):
     '''Copy source tree into build directory.'''
-    def clean_copy_tree(src: str, dst: str) -> None:
+    def clean_copy_tree(src, dst):
         distutils.dir_util.copy_tree(src, dst)
         NEED_CLEAN_TREE.add(os.path.abspath(dst))
 
-    def clean_copy_file(src: str, dst: str) -> None:
+    def clean_copy_file(src, dst):
         distutils.file_util.copy_file(src, dst)
         NEED_CLEAN_FILE.add(os.path.abspath(dst))
 
@@ -78,7 +77,7 @@ def copy_tree(src_dir: str, target_dir: str) -> None:
     clean_copy_file(lic, os.path.join(target_dir, 'LICENSE'))
 
 
-def clean_up() -> None:
+def clean_up():
     '''Removed copied files.'''
     for path in NEED_CLEAN_TREE:
         shutil.rmtree(path)
@@ -88,7 +87,7 @@ def clean_up() -> None:
 
 class CMakeExtension(Extension):  # pylint: disable=too-few-public-methods
     '''Wrapper for extension'''
-    def __init__(self, name: str) -> None:
+    def __init__(self, name):
         super().__init__(name=name, sources=[])
 
 
@@ -98,14 +97,7 @@ class BuildExt(build_ext.build_ext):  # pylint: disable=too-many-ancestors
     logger = logging.getLogger('XGBoost build_ext')
 
     # pylint: disable=too-many-arguments,no-self-use
-    def build(
-        self,
-        src_dir: str,
-        build_dir: str,
-        generator: str,
-        build_tool: Optional[str] = None,
-        use_omp: int = 1,
-    ) -> None:
+    def build(self, src_dir, build_dir, generator, build_tool=None, use_omp=1):
         '''Build the core library with CMake.'''
         cmake_cmd = ['cmake', src_dir, generator]
 
@@ -124,14 +116,13 @@ class BuildExt(build_ext.build_ext):  # pylint: disable=too-many-ancestors
 
         if system() != 'Windows':
             nproc = os.cpu_count()
-            assert build_tool is not None
             subprocess.check_call([build_tool, '-j' + str(nproc)],
                                   cwd=build_dir)
         else:
             subprocess.check_call(['cmake', '--build', '.',
                                    '--config', 'Release'], cwd=build_dir)
 
-    def build_cmake_extension(self) -> None:
+    def build_cmake_extension(self):
         '''Configure and build using CMake'''
         if USER_OPTIONS['use-system-libxgboost'][2]:
             self.logger.info('Using system libxgboost.')
@@ -183,14 +174,14 @@ class BuildExt(build_ext.build_ext):  # pylint: disable=too-many-ancestors
                 self.logger.warning('Disabling OpenMP support.')
                 self.build(src_dir, build_dir, gen, build_tool, use_omp=0)
 
-    def build_extension(self, ext: Extension) -> None:
+    def build_extension(self, ext):
         '''Override the method for dispatching.'''
         if isinstance(ext, CMakeExtension):
             self.build_cmake_extension()
         else:
             super().build_extension(ext)
 
-    def copy_extensions_to_source(self) -> None:
+    def copy_extensions_to_source(self):
         '''Dummy override.  Invoked during editable installation.  Our binary
         should available in `lib`.
 
@@ -205,7 +196,7 @@ class Sdist(sdist.sdist):       # pylint: disable=too-many-ancestors
     '''Copy c++ source into Python directory.'''
     logger = logging.getLogger('xgboost sdist')
 
-    def run(self) -> None:
+    def run(self):
         copy_tree(os.path.join(CURRENT_DIR, os.path.pardir),
                   os.path.join(CURRENT_DIR, 'xgboost'))
         libxgboost = os.path.join(
@@ -222,7 +213,7 @@ class InstallLib(install_lib.install_lib):
     '''Copy shared object into installation directory.'''
     logger = logging.getLogger('xgboost install_lib')
 
-    def install(self) -> List[str]:
+    def install(self):
         outfiles = super().install()
 
         if USER_OPTIONS['use-system-libxgboost'][2] != 0:
@@ -264,7 +255,7 @@ class Install(install.install):  # pylint: disable=too-many-instance-attributes
     user_options = install.install.user_options + list(
         (k, v[0], v[1]) for k, v in USER_OPTIONS.items())
 
-    def initialize_options(self) -> None:
+    def initialize_options(self):
         super().initialize_options()
         self.use_openmp = 1
         self.use_cuda = 0
@@ -280,7 +271,7 @@ class Install(install.install):  # pylint: disable=too-many-instance-attributes
 
         self.use_system_libxgboost = 0
 
-    def run(self) -> None:
+    def run(self):
         # setuptools will configure the options according to user supplied command line
         # arguments, then here we propagate them into `USER_OPTIONS` for visibility to
         # other sub-commands like `build_ext`.
@@ -350,9 +341,7 @@ if __name__ == '__main__':
                        'Programming Language :: Python :: 3',
                        'Programming Language :: Python :: 3.6',
                        'Programming Language :: Python :: 3.7',
-                       'Programming Language :: Python :: 3.8',
-                       'Programming Language :: Python :: 3.9',
-                       'Programming Language :: Python :: 3.10'],
+                       'Programming Language :: Python :: 3.8'],
           python_requires='>=3.6',
           url='https://github.com/dmlc/xgboost')
 

python-package/xgboost/VERSION

@@ -1 +1 @@
-1.5.2
+1.5.0

python-package/xgboost/core.py

@@ -229,7 +229,7 @@ def _numpy2ctypes_type(dtype):
     }
     if np.intc is not np.int32:  # Windows
         _NUMPY_TO_CTYPES_MAPPING[np.intc] = _NUMPY_TO_CTYPES_MAPPING[np.int32]
-    if dtype not in _NUMPY_TO_CTYPES_MAPPING:
+    if dtype not in _NUMPY_TO_CTYPES_MAPPING.keys():
         raise TypeError(
             f"Supported types: {_NUMPY_TO_CTYPES_MAPPING.keys()}, got: {dtype}"
         )
@@ -266,7 +266,7 @@ def ctypes2cupy(cptr, length, dtype):
     from cupy.cuda.memory import UnownedMemory
 
     CUPY_TO_CTYPES_MAPPING = {cupy.float32: ctypes.c_float, cupy.uint32: ctypes.c_uint}
-    if dtype not in CUPY_TO_CTYPES_MAPPING:
+    if dtype not in CUPY_TO_CTYPES_MAPPING.keys():
         raise RuntimeError(f"Supported types: {CUPY_TO_CTYPES_MAPPING.keys()}")
     addr = ctypes.cast(cptr, ctypes.c_void_p).value
     # pylint: disable=c-extension-no-member,no-member
@@ -386,7 +386,7 @@ class DataIter:  # pylint: disable=too-many-instance-attributes
             raise exc  # pylint: disable=raising-bad-type
 
     def __del__(self) -> None:
-        assert self._temporary_data is None
+        assert self._temporary_data is None, self._temporary_data
         assert self._exception is None
 
     def _reset_wrapper(self, this: None) -> None:  # pylint: disable=unused-argument
@@ -410,19 +410,19 @@ class DataIter:  # pylint: disable=too-many-instance-attributes
             feature_names: Optional[List[str]] = None,
             feature_types: Optional[List[str]] = None,
             **kwargs: Any,
-        ) -> None:
+        ):
             from .data import dispatch_proxy_set_data
             from .data import _proxy_transform
 
-            new, cat_codes, feature_names, feature_types = _proxy_transform(
+            transformed, feature_names, feature_types = _proxy_transform(
                 data,
                 feature_names,
                 feature_types,
                 self._enable_categorical,
             )
             # Stage the data, meta info are copied inside C++ MetaInfo.
-            self._temporary_data = (new, cat_codes)
-            dispatch_proxy_set_data(self.proxy, new, cat_codes, self._allow_host)
+            self._temporary_data = transformed
+            dispatch_proxy_set_data(self.proxy, transformed, self._allow_host)
             self.proxy.set_info(
                 feature_names=feature_names,
                 feature_types=feature_types,
@@ -1103,7 +1103,7 @@ class _ProxyDMatrix(DMatrix):
         self.handle = ctypes.c_void_p()
         _check_call(_LIB.XGProxyDMatrixCreate(ctypes.byref(self.handle)))
 
-    def _set_data_from_cuda_interface(self, data) -> None:
+    def _set_data_from_cuda_interface(self, data):
         """Set data from CUDA array interface."""
         interface = data.__cuda_array_interface__
         interface_str = bytes(json.dumps(interface, indent=2), "utf-8")
@@ -1111,11 +1111,11 @@ class _ProxyDMatrix(DMatrix):
             _LIB.XGProxyDMatrixSetDataCudaArrayInterface(self.handle, interface_str)
         )
 
-    def _set_data_from_cuda_columnar(self, data, cat_codes: list) -> None:
+    def _set_data_from_cuda_columnar(self, data):
         """Set data from CUDA columnar format."""
         from .data import _cudf_array_interfaces
 
-        interfaces_str = _cudf_array_interfaces(data, cat_codes)
+        _, interfaces_str = _cudf_array_interfaces(data)
         _check_call(_LIB.XGProxyDMatrixSetDataCudaColumnar(self.handle, interfaces_str))
 
     def _set_data_from_array(self, data: np.ndarray):
@@ -1805,7 +1805,7 @@ class Booster(object):
         .. note::
 
             See `Prediction
-            <https://xgboost.readthedocs.io/en/latest/prediction.html>`_
+            <https://xgboost.readthedocs.io/en/latest/tutorials/prediction.html>`_
             for issues like thread safety and a summary of outputs from this function.
 
         Parameters
@@ -1986,6 +1986,13 @@ class Booster(object):
         preds = ctypes.POINTER(ctypes.c_float)()
 
         # once caching is supported, we can pass id(data) as cache id.
+        try:
+            import pandas as pd
+
+            if isinstance(data, pd.DataFrame):
+                data = data.values
+        except ImportError:
+            pass
         args = {
             "type": 0,
             "training": False,
@@ -2020,20 +2027,7 @@ class Booster(object):
                     f"got {data.shape[1]}"
                 )
 
-        from .data import _is_pandas_df, _transform_pandas_df
         from .data import _array_interface
-        if (
-            _is_pandas_df(data)
-            or lazy_isinstance(data, "cudf.core.dataframe", "DataFrame")
-        ):
-            ft = self.feature_types
-            if ft is None:
-                enable_categorical = False
-            else:
-                enable_categorical = any(f == "c" for f in ft)
-        if _is_pandas_df(data):
-            data, _, _ = _transform_pandas_df(data, enable_categorical)
-
         if isinstance(data, np.ndarray):
             from .data import _ensure_np_dtype
             data, _ = _ensure_np_dtype(data, data.dtype)
@@ -2086,11 +2080,9 @@ class Booster(object):
             )
             return _prediction_output(shape, dims, preds, True)
         if lazy_isinstance(data, "cudf.core.dataframe", "DataFrame"):
-            from .data import _cudf_array_interfaces, _transform_cudf_df
-            data, cat_codes, _, _ = _transform_cudf_df(
-                data, None, None, enable_categorical
-            )
-            interfaces_str = _cudf_array_interfaces(data, cat_codes)
+            from .data import _cudf_array_interfaces
+
+            _, interfaces_str = _cudf_array_interfaces(data)
             _check_call(
                 _LIB.XGBoosterPredictFromCudaColumnar(
                     self.handle,

python-package/xgboost/dask.py

@@ -1606,9 +1606,8 @@ class DaskScikitLearnBase(XGBModel):
         should use `worker_client' instead of default client.
 
         """
-
+        asynchronous = getattr(self, "_asynchronous", False)
         if self._client is None:
-            asynchronous = getattr(self, "_asynchronous", False)
             try:
                 distributed.get_worker()
                 in_worker = True
@@ -1621,7 +1620,7 @@ class DaskScikitLearnBase(XGBModel):
                         return ret
                     return ret
 
-        return self.client.sync(func, **kwargs, asynchronous=self.client.asynchronous)
+        return self.client.sync(func, **kwargs, asynchronous=asynchronous)
 
 
 @xgboost_model_doc(
@@ -1834,7 +1833,7 @@ class DaskXGBClassifier(DaskScikitLearnBase, XGBClassifierBase):
         vstack = update_wrapper(
             partial(da.vstack, allow_unknown_chunksizes=True), da.vstack
         )
-        return _cls_predict_proba(getattr(self, "n_classes_", 0), predts, vstack)
+        return _cls_predict_proba(getattr(self, "n_classes_", None), predts, vstack)
 
     # pylint: disable=missing-function-docstring
     def predict_proba(

python-package/xgboost/data.py

@@ -1,4 +1,4 @@
-# pylint: disable=too-many-arguments, too-many-branches, too-many-lines
+# pylint: disable=too-many-arguments, too-many-branches
 # pylint: disable=too-many-return-statements, import-error
 '''Data dispatching for DMatrix.'''
 import ctypes
@@ -12,7 +12,7 @@ import numpy as np
 from .core import c_array, _LIB, _check_call, c_str
 from .core import _cuda_array_interface
 from .core import DataIter, _ProxyDMatrix, DMatrix
-from .compat import lazy_isinstance, DataFrame
+from .compat import lazy_isinstance
 
 c_bst_ulong = ctypes.c_uint64   # pylint: disable=invalid-name
 
@@ -217,48 +217,36 @@ _pandas_dtype_mapper = {
 }
 
 
-def _invalid_dataframe_dtype(data) -> None:
-    # pandas series has `dtypes` but it's just a single object
-    # cudf series doesn't have `dtypes`.
-    if hasattr(data, "dtypes") and hasattr(data.dtypes, "__iter__"):
-        bad_fields = [
-            str(data.columns[i])
-            for i, dtype in enumerate(data.dtypes)
-            if dtype.name not in _pandas_dtype_mapper
-        ]
-        err = " Invalid columns:" + ", ".join(bad_fields)
-    else:
-        err = ""
-
-    msg = """DataFrame.dtypes for data must be int, float, bool or category.  When
-categorical type is supplied, DMatrix parameter `enable_categorical` must
-be set to `True`.""" + err
-    raise ValueError(msg)
-
-
 def _transform_pandas_df(
-    data: DataFrame,
+    data,
     enable_categorical: bool,
     feature_names: Optional[List[str]] = None,
     feature_types: Optional[List[str]] = None,
-    meta: Optional[str] = None,
-    meta_type: Optional[str] = None,
-) -> Tuple[np.ndarray, Optional[List[str]], Optional[List[str]]]:
+    meta=None,
+    meta_type=None,
+):
     import pandas as pd
     from pandas.api.types import is_sparse, is_categorical_dtype
 
-    if not all(
-        dtype.name in _pandas_dtype_mapper
-        or is_sparse(dtype)
-        or (is_categorical_dtype(dtype) and enable_categorical)
-        for dtype in data.dtypes
-    ):
-        _invalid_dataframe_dtype(data)
+    if not all(dtype.name in _pandas_dtype_mapper or is_sparse(dtype) or
+               (is_categorical_dtype(dtype) and enable_categorical)
+               for dtype in data.dtypes):
+        bad_fields = [
+            str(data.columns[i]) for i, dtype in enumerate(data.dtypes)
+            if dtype.name not in _pandas_dtype_mapper
+        ]
+
+        msg = """DataFrame.dtypes for data must be int, float, bool or category.  When
+                categorical type is supplied, DMatrix parameter `enable_categorical` must
+                be set to `True`."""
+        raise ValueError(msg + ', '.join(bad_fields))
 
     # handle feature names
     if feature_names is None and meta is None:
         if isinstance(data.columns, pd.MultiIndex):
-            feature_names = [" ".join([str(x) for x in i]) for i in data.columns]
+            feature_names = [
+                ' '.join([str(x) for x in i]) for i in data.columns
+            ]
         elif isinstance(data.columns, (pd.Int64Index, pd.RangeIndex)):
             feature_names = list(map(str, data.columns))
         else:
@@ -275,24 +263,21 @@ def _transform_pandas_df(
             else:
                 feature_types.append(_pandas_dtype_mapper[dtype.name])
 
-    # handle category codes.
+    # handle categorical codes.
     transformed = pd.DataFrame()
     if enable_categorical:
         for i, dtype in enumerate(data.dtypes):
             if is_categorical_dtype(dtype):
-                # pandas uses -1 as default missing value for categorical data
-                transformed[data.columns[i]] = (
-                    data[data.columns[i]]
-                    .cat.codes.astype(np.float32)
-                    .replace(-1.0, np.NaN)
-                )
+                transformed[data.columns[i]] = data[data.columns[i]].cat.codes
             else:
                 transformed[data.columns[i]] = data[data.columns[i]]
     else:
         transformed = data
 
     if meta and len(data.columns) > 1:
-        raise ValueError(f"DataFrame for {meta} cannot have multiple columns")
+        raise ValueError(
+            f"DataFrame for {meta} cannot have multiple columns"
+        )
 
     dtype = meta_type if meta_type else np.float32
     arr = transformed.values
@@ -302,7 +287,7 @@ def _transform_pandas_df(
 
 
 def _from_pandas_df(
-    data: DataFrame,
+    data,
     enable_categorical: bool,
     missing,
     nthread,
@@ -315,7 +300,6 @@ def _from_pandas_df(
                              feature_types)
 
 
-
 def _is_pandas_series(data):
     try:
         import pandas as pd
@@ -334,26 +318,13 @@ def _is_modin_series(data):
 
 def _from_pandas_series(
     data,
-    missing: float,
-    nthread: int,
-    enable_categorical: bool,
+    missing,
+    nthread,
     feature_names: Optional[List[str]],
     feature_types: Optional[List[str]],
 ):
-    from pandas.api.types import is_categorical_dtype
-
-    if (data.dtype.name not in _pandas_dtype_mapper) and not (
-        is_categorical_dtype(data.dtype) and enable_categorical
-    ):
-        _invalid_dataframe_dtype(data)
-    if enable_categorical and is_categorical_dtype(data.dtype):
-        data = data.cat.codes
     return _from_numpy_array(
-        data.values.reshape(data.shape[0], 1).astype("float"),
-        missing,
-        nthread,
-        feature_names,
-        feature_types,
+        data.values.astype("float"), missing, nthread, feature_names, feature_types
     )
 
 
@@ -457,7 +428,7 @@ def _is_cudf_df(data):
     return hasattr(cudf, 'DataFrame') and isinstance(data, cudf.DataFrame)
 
 
-def _cudf_array_interfaces(data, cat_codes: list) -> bytes:
+def _cudf_array_interfaces(data) -> Tuple[list, bytes]:
     """Extract CuDF __cuda_array_interface__.  This is special as it returns a new list of
     data and a list of array interfaces.  The data is list of categorical codes that
     caller can safely ignore, but have to keep their reference alive until usage of array
@@ -469,27 +440,23 @@ def _cudf_array_interfaces(data, cat_codes: list) -> bytes:
     except ImportError:
         from cudf.utils.dtypes import is_categorical_dtype
 
+    cat_codes = []
     interfaces = []
     if _is_cudf_ser(data):
-        if is_categorical_dtype(data.dtype):
-            interface = cat_codes[0].__cuda_array_interface__
-        else:
-            interface = data.__cuda_array_interface__
-        if "mask" in interface:
-            interface["mask"] = interface["mask"].__cuda_array_interface__
-        interfaces.append(interface)
+        interfaces.append(data.__cuda_array_interface__)
     else:
-        for i, col in enumerate(data):
+        for col in data:
             if is_categorical_dtype(data[col].dtype):
-                codes = cat_codes[i]
+                codes = data[col].cat.codes
                 interface = codes.__cuda_array_interface__
+                cat_codes.append(codes)
             else:
                 interface = data[col].__cuda_array_interface__
             if "mask" in interface:
                 interface["mask"] = interface["mask"].__cuda_array_interface__
             interfaces.append(interface)
     interfaces_str = bytes(json.dumps(interfaces, indent=2), "utf-8")
-    return interfaces_str
+    return cat_codes, interfaces_str
 
 
 def _transform_cudf_df(
@@ -503,57 +470,25 @@ def _transform_cudf_df(
     except ImportError:
         from cudf.utils.dtypes import is_categorical_dtype
 
-    if _is_cudf_ser(data):
-        dtypes = [data.dtype]
-    else:
-        dtypes = data.dtypes
-
-    if not all(
-        dtype.name in _pandas_dtype_mapper
-        or (is_categorical_dtype(dtype) and enable_categorical)
-        for dtype in dtypes
-    ):
-        _invalid_dataframe_dtype(data)
-
-    # handle feature names
     if feature_names is None:
         if _is_cudf_ser(data):
             feature_names = [data.name]
         elif lazy_isinstance(data.columns, "cudf.core.multiindex", "MultiIndex"):
             feature_names = [" ".join([str(x) for x in i]) for i in data.columns]
-        elif (
-            lazy_isinstance(data.columns, "cudf.core.index", "RangeIndex")
-            or lazy_isinstance(data.columns, "cudf.core.index", "Int64Index")
-            # Unique to cuDF, no equivalence in pandas 1.3.3
-            or lazy_isinstance(data.columns, "cudf.core.index", "Int32Index")
-        ):
-            feature_names = list(map(str, data.columns))
         else:
             feature_names = data.columns.format()
-
-    # handle feature types
     if feature_types is None:
         feature_types = []
+        if _is_cudf_ser(data):
+            dtypes = [data.dtype]
+        else:
+            dtypes = data.dtypes
         for dtype in dtypes:
             if is_categorical_dtype(dtype) and enable_categorical:
                 feature_types.append(CAT_T)
             else:
                 feature_types.append(_pandas_dtype_mapper[dtype.name])
-
-    # handle categorical data
-    cat_codes = []
-    if _is_cudf_ser(data):
-        # unlike pandas, cuDF uses NA for missing data.
-        if is_categorical_dtype(data.dtype) and enable_categorical:
-            codes = data.cat.codes
-            cat_codes.append(codes)
-    else:
-        for col in data:
-            if is_categorical_dtype(data[col].dtype) and enable_categorical:
-                codes = data[col].cat.codes
-                cat_codes.append(codes)
-
-    return data, cat_codes, feature_names, feature_types
+    return data, feature_names, feature_types
 
 
 def _from_cudf_df(
@@ -564,10 +499,10 @@ def _from_cudf_df(
     feature_types: Optional[List[str]],
     enable_categorical: bool,
 ) -> Tuple[ctypes.c_void_p, Any, Any]:
-    data, cat_codes, feature_names, feature_types = _transform_cudf_df(
+    data, feature_names, feature_types = _transform_cudf_df(
         data, feature_names, feature_types, enable_categorical
     )
-    interfaces_str = _cudf_array_interfaces(data, cat_codes)
+    _, interfaces_str = _cudf_array_interfaces(data)
     handle = ctypes.c_void_p()
     config = bytes(json.dumps({"missing": missing, "nthread": nthread}), "utf-8")
     _check_call(
@@ -772,9 +707,8 @@ def dispatch_data_backend(
         return _from_pandas_df(data, enable_categorical, missing, threads,
                                feature_names, feature_types)
     if _is_pandas_series(data):
-        return _from_pandas_series(
-            data, missing, threads, enable_categorical, feature_names, feature_types
-        )
+        return _from_pandas_series(data, missing, threads, feature_names,
+                                   feature_types)
     if _is_cudf_df(data) or _is_cudf_ser(data):
         return _from_cudf_df(
             data, missing, threads, feature_names, feature_types, enable_categorical
@@ -798,9 +732,8 @@ def dispatch_data_backend(
         return _from_pandas_df(data, enable_categorical, missing, threads,
                                feature_names, feature_types)
     if _is_modin_series(data):
-        return _from_pandas_series(
-            data, missing, threads, enable_categorical, feature_names, feature_types
-        )
+        return _from_pandas_series(data, missing, threads, feature_names,
+                                   feature_types)
     if _has_array_protocol(data):
         array = np.asarray(data)
         return _from_numpy_array(array, missing, threads, feature_names, feature_types)
@@ -814,7 +747,7 @@ def dispatch_data_backend(
 
 def _to_data_type(dtype: str, name: str):
     dtype_map = {'float32': 1, 'float64': 2, 'uint32': 3, 'uint64': 4}
-    if dtype not in dtype_map:
+    if dtype not in dtype_map.keys():
         raise TypeError(
             f'Expecting float32, float64, uint32, uint64, got {dtype} ' +
             f'for {name}.')
@@ -933,7 +866,8 @@ def dispatch_meta_backend(matrix: DMatrix, data, name: str, dtype: str = None):
         _meta_from_dt(data, name, dtype, handle)
         return
     if _is_modin_df(data):
-        data, _, _ = _transform_pandas_df(data, False, meta=name, meta_type=dtype)
+        data, _, _ = _transform_pandas_df(
+            data, False, meta=name, meta_type=dtype)
         _meta_from_numpy(data, name, dtype, handle)
         return
     if _is_modin_series(data):
@@ -983,38 +917,30 @@ def _proxy_transform(
         )
     if _is_cupy_array(data):
         data = _transform_cupy_array(data)
-        return data, None, feature_names, feature_types
+        return data, feature_names, feature_types
     if _is_dlpack(data):
-        return _transform_dlpack(data), None, feature_names, feature_types
+        return _transform_dlpack(data), feature_names, feature_types
     if _is_numpy_array(data):
-        return data, None, feature_names, feature_types
+        return data, feature_names, feature_types
     if _is_scipy_csr(data):
-        return data, None, feature_names, feature_types
+        return data, feature_names, feature_types
     if _is_pandas_df(data):
         arr, feature_names, feature_types = _transform_pandas_df(
             data, enable_categorical, feature_names, feature_types
         )
-        return arr, None, feature_names, feature_types
+        return arr, feature_names, feature_types
     raise TypeError("Value type is not supported for data iterator:" + str(type(data)))
 
 
-def dispatch_proxy_set_data(
-    proxy: _ProxyDMatrix,
-    data: Any,
-    cat_codes: Optional[list],
-    allow_host: bool,
-) -> None:
+def dispatch_proxy_set_data(proxy: _ProxyDMatrix, data: Any, allow_host: bool) -> None:
     """Dispatch for DeviceQuantileDMatrix."""
     if not _is_cudf_ser(data) and not _is_pandas_series(data):
         _check_data_shape(data)
-
     if _is_cudf_df(data):
-        # pylint: disable=W0212
-        proxy._set_data_from_cuda_columnar(data, cat_codes)
+        proxy._set_data_from_cuda_columnar(data)  # pylint: disable=W0212
         return
     if _is_cudf_ser(data):
-        # pylint: disable=W0212
-        proxy._set_data_from_cuda_columnar(data, cat_codes)
+        proxy._set_data_from_cuda_columnar(data)  # pylint: disable=W0212
         return
     if _is_cupy_array(data):
         proxy._set_data_from_cuda_interface(data)  # pylint: disable=W0212

python-package/xgboost/sklearn.py

@@ -1354,7 +1354,9 @@ class XGBClassifier(XGBModel, XGBClassifierBase):
             iteration_range=iteration_range
         )
         # If model is loaded from a raw booster there's no `n_classes_`
-        return _cls_predict_proba(getattr(self, "n_classes_", 0), class_probs, np.vstack)
+        return _cls_predict_proba(
+            getattr(self, "n_classes_", None), class_probs, np.vstack
+        )
 
     def evals_result(self) -> TrainingCallback.EvalsLog:
         """Return the evaluation results.

python-package/xgboost/tracker.py

@@ -144,7 +144,7 @@ class RabitTracker(object):
     """
 
     def __init__(
-        self, hostIP, nslave, port=9091, port_end=9999, use_logger: bool = False
+        self, hostIP, nslave, port=9091, port_end=9999, use_logger: bool = True
     ) -> None:
         """A Python implementation of RABIT tracker.
 
@@ -384,17 +384,16 @@ def start_rabit_tracker(args):
     ----------
     args: arguments to start the rabit tracker.
     """
-    envs = {"DMLC_NUM_WORKER": args.num_workers, "DMLC_NUM_SERVER": args.num_servers}
-    rabit = RabitTracker(
-        hostIP=get_host_ip(args.host_ip), nslave=args.num_workers, use_logger=True
-    )
+    envs = {'DMLC_NUM_WORKER': args.num_workers,
+            'DMLC_NUM_SERVER': args.num_servers}
+    rabit = RabitTracker(hostIP=get_host_ip(args.host_ip), nslave=args.num_workers)
     envs.update(rabit.slave_envs())
     rabit.start(args.num_workers)
-    sys.stdout.write("DMLC_TRACKER_ENV_START\n")
+    sys.stdout.write('DMLC_TRACKER_ENV_START\n')
     # simply write configuration to stdout
     for k, v in envs.items():
         sys.stdout.write(f"{k}={v}\n")
-    sys.stdout.write("DMLC_TRACKER_ENV_END\n")
+    sys.stdout.write('DMLC_TRACKER_ENV_END\n')
     sys.stdout.flush()
     rabit.join()
 

src/common/bitfield.h

@@ -58,15 +58,14 @@ __forceinline__ __device__ BitFieldAtomicType AtomicAnd(BitFieldAtomicType* addr
 template <typename VT, typename Direction, bool IsConst = false>
 struct BitFieldContainer {
   using value_type = std::conditional_t<IsConst, VT const, VT>;  // NOLINT
-  using index_type = size_t;                                     // NOLINT
-  using pointer = value_type*;                                   // NOLINT
+  using pointer = value_type*;  // NOLINT
 
-  static index_type constexpr kValueSize = sizeof(value_type) * 8;
-  static index_type constexpr kOne = 1;  // force correct type.
+  static value_type constexpr kValueSize = sizeof(value_type) * 8;
+  static value_type constexpr kOne = 1;  // force correct type.
 
   struct Pos {
-    index_type int_pos{0};
-    index_type bit_pos{0};
+    std::remove_const_t<value_type> int_pos {0};
+    std::remove_const_t<value_type> bit_pos {0};
   };
 
  private:
@@ -74,13 +73,13 @@ struct BitFieldContainer {
   static_assert(!std::is_signed<VT>::value, "Must use unsiged type as underlying storage.");
 
  public:
-  XGBOOST_DEVICE static Pos ToBitPos(index_type pos) {
+  XGBOOST_DEVICE static Pos ToBitPos(value_type pos) {
     Pos pos_v;
     if (pos == 0) {
       return pos_v;
     }
-    pos_v.int_pos = pos / kValueSize;
-    pos_v.bit_pos = pos % kValueSize;
+    pos_v.int_pos =  pos / kValueSize;
+    pos_v.bit_pos =  pos % kValueSize;
     return pos_v;
   }
 
@@ -97,7 +96,7 @@ struct BitFieldContainer {
   /*\brief Compute the size of needed memory allocation.  The returned value is in terms
    *       of number of elements with `BitFieldContainer::value_type'.
    */
-  XGBOOST_DEVICE static size_t ComputeStorageSize(index_type size) {
+  XGBOOST_DEVICE static size_t ComputeStorageSize(size_t size) {
     return common::DivRoundUp(size, kValueSize);
   }
 #if defined(__CUDA_ARCH__)
@@ -139,14 +138,14 @@ struct BitFieldContainer {
 #endif  // defined(__CUDA_ARCH__)
 
 #if defined(__CUDA_ARCH__)
-  __device__ auto Set(index_type pos) {
+  __device__ auto Set(value_type pos) {
     Pos pos_v = Direction::Shift(ToBitPos(pos));
     value_type& value = bits_[pos_v.int_pos];
     value_type set_bit = kOne << pos_v.bit_pos;
     using Type = typename dh::detail::AtomicDispatcher<sizeof(value_type)>::Type;
     atomicOr(reinterpret_cast<Type *>(&value), set_bit);
   }
-  __device__ void Clear(index_type pos) {
+  __device__ void Clear(value_type pos) {
     Pos pos_v = Direction::Shift(ToBitPos(pos));
     value_type& value = bits_[pos_v.int_pos];
     value_type clear_bit = ~(kOne << pos_v.bit_pos);
@@ -154,13 +153,13 @@ struct BitFieldContainer {
     atomicAnd(reinterpret_cast<Type *>(&value), clear_bit);
   }
 #else
-  void Set(index_type pos) {
+  void Set(value_type pos) {
     Pos pos_v = Direction::Shift(ToBitPos(pos));
     value_type& value = bits_[pos_v.int_pos];
     value_type set_bit = kOne << pos_v.bit_pos;
     value |= set_bit;
   }
-  void Clear(index_type pos) {
+  void Clear(value_type pos) {
     Pos pos_v = Direction::Shift(ToBitPos(pos));
     value_type& value = bits_[pos_v.int_pos];
     value_type clear_bit = ~(kOne << pos_v.bit_pos);
@@ -176,7 +175,7 @@ struct BitFieldContainer {
     value_type result = test_bit & value;
     return static_cast<bool>(result);
   }
-  XGBOOST_DEVICE bool Check(index_type pos) const {
+  XGBOOST_DEVICE bool Check(value_type pos) const {
     Pos pos_v = ToBitPos(pos);
     return Check(pos_v);
   }

src/common/categorical.h

@@ -1,5 +1,5 @@
 /*!
- * Copyright 2020-2021 by XGBoost Contributors
+ * Copyright 2020 by XGBoost Contributors
  * \file categorical.h
  */
 #ifndef XGBOOST_COMMON_CATEGORICAL_H_
@@ -42,11 +42,6 @@ inline XGBOOST_DEVICE bool Decision(common::Span<uint32_t const> cats, bst_cat_t
   return !s_cats.Check(cat);
 }
 
-inline void CheckCat(bst_cat_t cat) {
-  CHECK_GE(cat, 0) << "Invalid categorical value detected.  Categorical value "
-                      "should be non-negative.";
-}
-
 struct IsCatOp {
   XGBOOST_DEVICE bool operator()(FeatureType ft) {
     return ft == FeatureType::kCategorical;

src/common/device_helpers.cuh

@@ -711,12 +711,6 @@ constexpr std::pair<int, int> CUDAVersion() {
 constexpr std::pair<int32_t, int32_t> ThrustVersion() {
   return std::make_pair(THRUST_MAJOR_VERSION, THRUST_MINOR_VERSION);
 }
-// Whether do we have thrust 1.x with x >= minor
-template <int32_t minor>
-constexpr bool HasThrustMinorVer() {
-  return (ThrustVersion().first == 1 && ThrustVersion().second >= minor) ||
-         ThrustVersion().first > 1;
-}
 
 namespace detail {
 template <typename T>
@@ -731,8 +725,10 @@ class TypedDiscard : public thrust::discard_iterator<T> {
 
 template <typename T>
 using TypedDiscard =
-    std::conditional_t<HasThrustMinorVer<12>(), detail::TypedDiscardCTK114<T>,
-                       detail::TypedDiscard<T>>;
+    std::conditional_t<((ThrustVersion().first == 1 &&
+                         ThrustVersion().second >= 12) ||
+                        ThrustVersion().first > 1),
+                       detail::TypedDiscardCTK114<T>, detail::TypedDiscard<T>>;
 
 /**
  * \class AllReducer
@@ -1446,39 +1442,24 @@ void ArgSort(xgboost::common::Span<U> keys, xgboost::common::Span<IdxT> sorted_i
 namespace detail {
 // Wrapper around cub sort for easier `descending` sort.
 template <bool descending, typename KeyT, typename ValueT,
-          typename BeginOffsetIteratorT, typename EndOffsetIteratorT>
+          typename OffsetIteratorT>
 void DeviceSegmentedRadixSortPair(
     void *d_temp_storage, size_t &temp_storage_bytes, const KeyT *d_keys_in, // NOLINT
     KeyT *d_keys_out, const ValueT *d_values_in, ValueT *d_values_out,
-    size_t num_items, size_t num_segments, BeginOffsetIteratorT d_begin_offsets,
-    EndOffsetIteratorT d_end_offsets, int begin_bit = 0,
+    size_t num_items, size_t num_segments, OffsetIteratorT d_begin_offsets,
+    OffsetIteratorT d_end_offsets, int begin_bit = 0,
     int end_bit = sizeof(KeyT) * 8) {
   cub::DoubleBuffer<KeyT> d_keys(const_cast<KeyT *>(d_keys_in), d_keys_out);
   cub::DoubleBuffer<ValueT> d_values(const_cast<ValueT *>(d_values_in),
                                      d_values_out);
-  // In old version of cub, num_items in dispatch is also int32_t, no way to change.
-  using OffsetT =
-      std::conditional_t<BuildWithCUDACub() && HasThrustMinorVer<13>(), size_t,
-                         int32_t>;
-  CHECK_LE(num_items, std::numeric_limits<OffsetT>::max());
-  // For Thrust >= 1.12 or CUDA >= 11.4, we require system cub installation
-
-#if (THRUST_MAJOR_VERSION == 1 && THRUST_MINOR_VERSION >= 13) || THRUST_MAJOR_VERSION > 1
+  using OffsetT = int32_t;  // num items in dispatch is also int32_t, no way to change.
+  CHECK_LE(num_items, std::numeric_limits<int32_t>::max());
   safe_cuda((cub::DispatchSegmentedRadixSort<
-             descending, KeyT, ValueT, BeginOffsetIteratorT, EndOffsetIteratorT,
+             descending, KeyT, ValueT, OffsetIteratorT,
              OffsetT>::Dispatch(d_temp_storage, temp_storage_bytes, d_keys,
                                 d_values, num_items, num_segments,
                                 d_begin_offsets, d_end_offsets, begin_bit,
                                 end_bit, false, nullptr, false)));
-#else
-  safe_cuda((cub::DispatchSegmentedRadixSort<
-             descending, KeyT, ValueT, BeginOffsetIteratorT,
-             OffsetT>::Dispatch(d_temp_storage, temp_storage_bytes, d_keys,
-                                d_values, num_items, num_segments,
-                                d_begin_offsets, d_end_offsets, begin_bit,
-                                end_bit, false, nullptr, false)));
-#endif
-
 }
 }  // namespace detail
 

src/common/hist_util.cu

@@ -133,7 +133,6 @@ void RemoveDuplicatedCategories(
     int32_t device, MetaInfo const &info, Span<bst_row_t> d_cuts_ptr,
     dh::device_vector<Entry> *p_sorted_entries,
     dh::caching_device_vector<size_t> *p_column_sizes_scan) {
-  info.feature_types.SetDevice(device);
   auto d_feature_types = info.feature_types.ConstDeviceSpan();
   CHECK(!d_feature_types.empty());
   auto &column_sizes_scan = *p_column_sizes_scan;

src/common/hist_util.cuh

@@ -124,11 +124,6 @@ void MakeEntriesFromAdapter(AdapterBatch const& batch, BatchIter batch_iter,
 
 void SortByWeight(dh::device_vector<float>* weights,
                   dh::device_vector<Entry>* sorted_entries);
-
-void RemoveDuplicatedCategories(
-    int32_t device, MetaInfo const &info, Span<bst_row_t> d_cuts_ptr,
-    dh::device_vector<Entry> *p_sorted_entries,
-    dh::caching_device_vector<size_t> *p_column_sizes_scan);
 }  // namespace detail
 
 // Compute sketch on DMatrix.
@@ -137,10 +132,9 @@ HistogramCuts DeviceSketch(int device, DMatrix* dmat, int max_bins,
                            size_t sketch_batch_num_elements = 0);
 
 template <typename AdapterBatch>
-void ProcessSlidingWindow(AdapterBatch const &batch, MetaInfo const &info,
-                          int device, size_t columns, size_t begin, size_t end,
-                          float missing, SketchContainer *sketch_container,
-                          int num_cuts) {
+void ProcessSlidingWindow(AdapterBatch const& batch, int device, size_t columns,
+                          size_t begin, size_t end, float missing,
+                          SketchContainer* sketch_container, int num_cuts) {
   // Copy current subset of valid elements into temporary storage and sort
   dh::device_vector<Entry> sorted_entries;
   dh::caching_device_vector<size_t> column_sizes_scan;
@@ -148,7 +142,6 @@ void ProcessSlidingWindow(AdapterBatch const &batch, MetaInfo const &info,
       thrust::make_counting_iterator(0llu),
       [=] __device__(size_t idx) { return batch.GetElement(idx); });
   HostDeviceVector<SketchContainer::OffsetT> cuts_ptr;
-  cuts_ptr.SetDevice(device);
   detail::MakeEntriesFromAdapter(batch, batch_iter, {begin, end}, missing,
                                  columns, num_cuts, device,
                                  &cuts_ptr,
@@ -158,14 +151,8 @@ void ProcessSlidingWindow(AdapterBatch const &batch, MetaInfo const &info,
   thrust::sort(thrust::cuda::par(alloc), sorted_entries.begin(),
                sorted_entries.end(), detail::EntryCompareOp());
 
-  if (sketch_container->HasCategorical()) {
-    auto d_cuts_ptr = cuts_ptr.DeviceSpan();
-    detail::RemoveDuplicatedCategories(device, info, d_cuts_ptr,
-                                       &sorted_entries, &column_sizes_scan);
-  }
-
+  auto const& h_cuts_ptr = cuts_ptr.ConstHostVector();
   auto d_cuts_ptr = cuts_ptr.DeviceSpan();
-  auto const &h_cuts_ptr = cuts_ptr.HostVector();
   // Extract the cuts from all columns concurrently
   sketch_container->Push(dh::ToSpan(sorted_entries),
                          dh::ToSpan(column_sizes_scan), d_cuts_ptr,
@@ -235,12 +222,6 @@ void ProcessWeightedSlidingWindow(Batch batch, MetaInfo const& info,
 
   detail::SortByWeight(&temp_weights, &sorted_entries);
 
-  if (sketch_container->HasCategorical()) {
-    auto d_cuts_ptr = cuts_ptr.DeviceSpan();
-    detail::RemoveDuplicatedCategories(device, info, d_cuts_ptr,
-                                       &sorted_entries, &column_sizes_scan);
-  }
-
   auto const& h_cuts_ptr = cuts_ptr.ConstHostVector();
   auto d_cuts_ptr = cuts_ptr.DeviceSpan();
 
@@ -293,8 +274,8 @@ void AdapterDeviceSketch(Batch batch, int num_bins,
         device, num_cuts_per_feature, false);
     for (auto begin = 0ull; begin < batch.Size(); begin += sketch_batch_num_elements) {
       size_t end = std::min(batch.Size(), size_t(begin + sketch_batch_num_elements));
-      ProcessSlidingWindow(batch, info, device, num_cols, begin, end, missing,
-                           sketch_container, num_cuts_per_feature);
+      ProcessSlidingWindow(batch, device, num_cols,
+                           begin, end, missing, sketch_container, num_cuts_per_feature);
     }
   }
 }

src/data/adapter.h

@@ -21,7 +21,6 @@
 
 #include "array_interface.h"
 #include "../c_api/c_api_error.h"
-#include "../common/math.h"
 
 namespace xgboost {
 namespace data {
@@ -81,24 +80,6 @@ struct COOTuple {
   float value{0};
 };
 
-struct IsValidFunctor {
-  float missing;
-
-  XGBOOST_DEVICE explicit IsValidFunctor(float missing) : missing(missing) {}
-
-  XGBOOST_DEVICE bool operator()(float value) const {
-    return !(common::CheckNAN(value) || value == missing);
-  }
-
-  XGBOOST_DEVICE bool operator()(const data::COOTuple& e) const {
-    return !(common::CheckNAN(e.value) || e.value == missing);
-  }
-
-  XGBOOST_DEVICE bool operator()(const Entry& e) const {
-    return !(common::CheckNAN(e.fvalue) || e.fvalue == missing);
-  }
-};
-
 namespace detail {
 
 /**

src/data/data.cc

@@ -987,19 +987,18 @@ uint64_t SparsePage::Push(const AdapterBatchT& batch, float missing, int nthread
 
   // Second pass over batch, placing elements in correct position
 
-  auto is_valid = data::IsValidFunctor{missing};
 #pragma omp parallel num_threads(nthread)
   {
     exec.Run([&]() {
       int tid = omp_get_thread_num();
-      size_t begin = tid * thread_size;
-      size_t end = tid != (nthread - 1) ? (tid + 1) * thread_size : batch_size;
+      size_t begin = tid*thread_size;
+      size_t end = tid != (nthread-1) ? (tid+1)*thread_size : batch_size;
       for (size_t i = begin; i < end; ++i) {
         auto line = batch.GetLine(i);
         for (auto j = 0ull; j < line.Size(); j++) {
           auto element = line.GetElement(j);
           const size_t key = (element.row_idx - base_rowid);
-          if (is_valid(element)) {
+          if (!common::CheckNAN(element.value) && element.value != missing) {
             builder.Push(key, Entry(element.column_idx, element.value), tid);
           }
         }

src/data/device_adapter.cuh

@@ -15,6 +15,29 @@
 namespace xgboost {
 namespace data {
 
+struct IsValidFunctor : public thrust::unary_function<Entry, bool> {
+  float missing;
+
+  XGBOOST_DEVICE explicit IsValidFunctor(float missing) : missing(missing) {}
+
+  __device__ bool operator()(float value) const {
+    return !(common::CheckNAN(value) || value == missing);
+  }
+
+  __device__ bool operator()(const data::COOTuple& e) const {
+    if (common::CheckNAN(e.value) || e.value == missing) {
+      return false;
+    }
+    return true;
+  }
+  __device__ bool operator()(const Entry& e) const {
+    if (common::CheckNAN(e.fvalue) || e.fvalue == missing) {
+      return false;
+    }
+    return true;
+  }
+};
+
 class CudfAdapterBatch : public detail::NoMetaInfo {
   friend class CudfAdapter;
 

src/data/iterative_device_dmatrix.cu

@@ -152,7 +152,6 @@ void IterativeDeviceDMatrix::Initialize(DataIterHandle iter_handle, float missin
 
   if (batches == 1) {
     this->info_ = std::move(proxy->Info());
-    this->info_.num_nonzero_ = nnz;
     CHECK_EQ(proxy->Info().labels_.Size(), 0);
   }
 

src/gbm/gblinear.cc

@@ -62,8 +62,9 @@ struct GBLinearTrainParam : public XGBoostParameter<GBLinearTrainParam> {
   }
 };
 
-void LinearCheckLayer(unsigned layer_begin) {
+void LinearCheckLayer(unsigned layer_begin, unsigned layer_end) {
   CHECK_EQ(layer_begin, 0) << "Linear booster does not support prediction range.";
+  CHECK_EQ(layer_end, 0)   << "Linear booster does not support prediction range.";
 }
 
 /*!
@@ -151,7 +152,7 @@ class GBLinear : public GradientBooster {
   void PredictBatch(DMatrix *p_fmat, PredictionCacheEntry *predts,
                     bool training, unsigned layer_begin, unsigned layer_end) override {
     monitor_.Start("PredictBatch");
-    LinearCheckLayer(layer_begin);
+    LinearCheckLayer(layer_begin, layer_end);
     auto* out_preds = &predts->predictions;
     this->PredictBatchInternal(p_fmat, &out_preds->HostVector());
     monitor_.Stop("PredictBatch");
@@ -160,7 +161,7 @@ class GBLinear : public GradientBooster {
   void PredictInstance(const SparsePage::Inst &inst,
                        std::vector<bst_float> *out_preds,
                        unsigned layer_begin, unsigned layer_end) override {
-    LinearCheckLayer(layer_begin);
+    LinearCheckLayer(layer_begin, layer_end);
     const int ngroup = model_.learner_model_param->num_output_group;
     for (int gid = 0; gid < ngroup; ++gid) {
       this->Pred(inst, dmlc::BeginPtr(*out_preds), gid,
@@ -176,8 +177,8 @@ class GBLinear : public GradientBooster {
                            HostDeviceVector<bst_float>* out_contribs,
                            unsigned layer_begin, unsigned layer_end, bool, int, unsigned) override {
     model_.LazyInitModel();
-    LinearCheckLayer(layer_begin);
-    const auto &base_margin = p_fmat->Info().base_margin_.ConstHostVector();
+    LinearCheckLayer(layer_begin, layer_end);
+    const auto& base_margin = p_fmat->Info().base_margin_.ConstHostVector();
     const int ngroup = model_.learner_model_param->num_output_group;
     const size_t ncolumns = model_.learner_model_param->num_feature + 1;
     // allocate space for (#features + bias) times #groups times #rows
@@ -213,7 +214,7 @@ class GBLinear : public GradientBooster {
   void PredictInteractionContributions(DMatrix* p_fmat,
                                        HostDeviceVector<bst_float>* out_contribs,
                                        unsigned layer_begin, unsigned layer_end, bool) override {
-    LinearCheckLayer(layer_begin);
+    LinearCheckLayer(layer_begin, layer_end);
     std::vector<bst_float>& contribs = out_contribs->HostVector();
 
     // linear models have no interaction effects

src/gbm/gblinear_model.cc

@@ -18,7 +18,6 @@ void GBLinearModel::SaveModel(Json* p_out) const {
     j_weights[i] = weight[i];
   }
   out["weights"] = std::move(j_weights);
-  out["boosted_rounds"] = Json{this->num_boosted_rounds};
 }
 
 void GBLinearModel::LoadModel(Json const& in) {
@@ -28,13 +27,6 @@ void GBLinearModel::LoadModel(Json const& in) {
   for (size_t i = 0; i < n_weights; ++i) {
     weight[i] = get<Number const>(j_weights[i]);
   }
-  auto const& obj = get<Object const>(in);
-  auto boosted_rounds = obj.find("boosted_rounds");
-  if (boosted_rounds != obj.cend()) {
-    this->num_boosted_rounds = get<Integer const>(boosted_rounds->second);
-  } else {
-    this->num_boosted_rounds = 0;
-  }
 }
 
 DMLC_REGISTER_PARAMETER(DeprecatedGBLinearModelParam);

src/gbm/gbtree.h

@@ -273,7 +273,6 @@ class GBTree : public GradientBooster {
     uint32_t tree_begin, tree_end;
     std::tie(tree_begin, tree_end) =
         detail::LayerToTree(model_, tparam_, layer_begin, layer_end);
-    CHECK_LE(tree_end, model_.trees.size()) << "Invalid number of trees.";
     std::vector<Predictor const *> predictors{
       cpu_predictor_.get(),
 #if defined(XGBOOST_USE_CUDA)

src/tree/updater_gpu_hist.cu

@@ -585,7 +585,6 @@ struct GPUHistMakerDevice {
       CHECK_LT(candidate.split.fvalue, std::numeric_limits<bst_cat_t>::max())
           << "Categorical feature value too large.";
       auto cat = common::AsCat(candidate.split.fvalue);
-      common::CheckCat(cat);
       std::vector<uint32_t> split_cats(LBitField32::ComputeStorageSize(std::max(cat+1, 1)), 0);
       LBitField32 cats_bits(split_cats);
       cats_bits.Set(cat);

tests/ci_build/conda_env/macos_cpu_test.yml

@@ -1,4 +1,4 @@
-name: macos_test
+name: cpu_test
 channels:
 - conda-forge
 dependencies:

tests/cpp/common/test_bitfield.cc

@@ -38,14 +38,6 @@ TEST(BitField, Check) {
       ASSERT_FALSE(bits.Check(i));
     }
   }
-
-  {
-    // regression test for correct index type.
-    std::vector<RBitField8::value_type> storage(33, 0);
-    storage[32] = static_cast<uint8_t>(1);
-    auto bits = RBitField8({storage.data(), storage.size()});
-    ASSERT_TRUE(bits.Check(256));
-  }
 }
 
 template <typename BitFieldT, typename VT = typename BitFieldT::value_type>

tests/cpp/common/test_hist_util.cu

@@ -392,52 +392,6 @@ TEST(HistUtil, AdapterSketchSlidingWindowWeightedMemory) {
   EXPECT_GE(dh::GlobalMemoryLogger().PeakMemory(), bytes_required);
 }
 
-void TestCategoricalSketchAdapter(size_t n, size_t num_categories,
-                                  int32_t num_bins, bool weighted) {
-  auto h_x = GenerateRandomCategoricalSingleColumn(n, num_categories);
-  thrust::device_vector<float> x(h_x);
-  auto adapter = AdapterFromData(x, n, 1);
-  MetaInfo info;
-  info.num_row_ = n;
-  info.num_col_ = 1;
-  info.feature_types.HostVector().push_back(FeatureType::kCategorical);
-
-  if (weighted) {
-    std::vector<float> weights(n, 0);
-    SimpleLCG lcg;
-    SimpleRealUniformDistribution<float> dist(0, 1);
-    for (auto& v : weights) {
-      v = dist(&lcg);
-    }
-    info.weights_.HostVector() = weights;
-  }
-
-  ASSERT_EQ(info.feature_types.Size(), 1);
-  SketchContainer container(info.feature_types, num_bins, 1, n, 0);
-  AdapterDeviceSketch(adapter.Value(), num_bins, info,
-                      std::numeric_limits<float>::quiet_NaN(), &container);
-  HistogramCuts cuts;
-  container.MakeCuts(&cuts);
-
-  thrust::sort(x.begin(), x.end());
-  auto n_uniques = thrust::unique(x.begin(), x.end()) - x.begin();
-  ASSERT_NE(n_uniques, x.size());
-  ASSERT_EQ(cuts.TotalBins(), n_uniques);
-  ASSERT_EQ(n_uniques, num_categories);
-
-  auto& values = cuts.cut_values_.HostVector();
-  ASSERT_TRUE(std::is_sorted(values.cbegin(), values.cend()));
-  auto is_unique = (std::unique(values.begin(), values.end()) - values.begin()) == n_uniques;
-  ASSERT_TRUE(is_unique);
-
-  x.resize(n_uniques);
-  h_x.resize(n_uniques);
-  thrust::copy(x.begin(), x.end(), h_x.begin());
-  for (decltype(n_uniques) i = 0; i < n_uniques; ++i) {
-    ASSERT_EQ(h_x[i], values[i]);
-  }
-}
-
 TEST(HistUtil, AdapterDeviceSketchCategorical) {
   int categorical_sizes[] = {2, 6, 8, 12};
   int num_bins = 256;
@@ -450,8 +404,6 @@ TEST(HistUtil, AdapterDeviceSketchCategorical) {
       auto adapter = AdapterFromData(x_device, n, 1);
       ValidateBatchedCuts(adapter, num_bins, adapter.NumColumns(),
                           adapter.NumRows(), dmat.get());
-      TestCategoricalSketchAdapter(n, num_categories, num_bins, true);
-      TestCategoricalSketchAdapter(n, num_categories, num_bins, false);
     }
   }
 }

tests/cpp/gbm/test_gbtree.cc

@@ -452,47 +452,4 @@ TEST(GBTree, FeatureScore) {
   test_eq("gain");
   test_eq("cover");
 }
-
-TEST(GBTree, PredictRange) {
-  size_t n_samples = 1000, n_features = 10, n_classes = 4;
-  auto m = RandomDataGenerator{n_samples, n_features, 0.5}.GenerateDMatrix(true, false, n_classes);
-
-  std::unique_ptr<Learner> learner{Learner::Create({m})};
-  learner->SetParam("num_class", std::to_string(n_classes));
-
-  learner->Configure();
-  for (size_t i = 0; i < 2; ++i) {
-    learner->UpdateOneIter(i, m);
-  }
-  HostDeviceVector<float> out_predt;
-  ASSERT_THROW(learner->Predict(m, false, &out_predt, 0, 3), dmlc::Error);
-
-  auto m_1 =
-      RandomDataGenerator{n_samples, n_features, 0.5}.GenerateDMatrix(true, false, n_classes);
-  HostDeviceVector<float> out_predt_full;
-  learner->Predict(m_1, false, &out_predt_full, 0, 0);
-  ASSERT_TRUE(std::equal(out_predt.HostVector().begin(), out_predt.HostVector().end(),
-                         out_predt_full.HostVector().begin()));
-
-  {
-    // inplace predict
-    HostDeviceVector<float> raw_storage;
-    auto raw = RandomDataGenerator{n_samples, n_features, 0.5}.GenerateArrayInterface(&raw_storage);
-    std::shared_ptr<data::ArrayAdapter> x{new data::ArrayAdapter{StringView{raw}}};
-
-    HostDeviceVector<float>* out_predt;
-    learner->InplacePredict(x, nullptr, PredictionType::kValue,
-                            std::numeric_limits<float>::quiet_NaN(), &out_predt, 0, 2);
-    auto h_out_predt = out_predt->HostVector();
-    learner->InplacePredict(x, nullptr, PredictionType::kValue,
-                            std::numeric_limits<float>::quiet_NaN(), &out_predt, 0, 0);
-    auto h_out_predt_full = out_predt->HostVector();
-
-    ASSERT_TRUE(std::equal(h_out_predt.begin(), h_out_predt.end(), h_out_predt_full.begin()));
-
-    ASSERT_THROW(learner->InplacePredict(x, nullptr, PredictionType::kValue,
-                                         std::numeric_limits<float>::quiet_NaN(), &out_predt, 0, 3),
-                 dmlc::Error);
-  }
-}
 }  // namespace xgboost

tests/python-gpu/test_from_cudf.py

@@ -186,37 +186,6 @@ Arrow specification.'''
         assert len(Xy.feature_types) == X.shape[1]
         assert all(t == "c" for t in Xy.feature_types)
 
-        # test missing value
-        X = cudf.DataFrame({"f0": ["a", "b", np.NaN]})
-        X["f0"] = X["f0"].astype("category")
-        df, cat_codes, _, _ = xgb.data._transform_cudf_df(
-            X, None, None, enable_categorical=True
-        )
-        for col in cat_codes:
-            assert col.has_nulls
-
-        y = [0, 1, 2]
-        with pytest.raises(ValueError):
-            xgb.DMatrix(X, y)
-        Xy = xgb.DMatrix(X, y, enable_categorical=True)
-        assert Xy.num_row() == 3
-        assert Xy.num_col() == 1
-
-        with pytest.raises(ValueError):
-            xgb.DeviceQuantileDMatrix(X, y)
-
-        Xy = xgb.DeviceQuantileDMatrix(X, y, enable_categorical=True)
-        assert Xy.num_row() == 3
-        assert Xy.num_col() == 1
-
-        X = X["f0"]
-        with pytest.raises(ValueError):
-            xgb.DMatrix(X, y)
-
-        Xy = xgb.DMatrix(X, y, enable_categorical=True)
-        assert Xy.num_row() == 3
-        assert Xy.num_col() == 1
-
 
 @pytest.mark.skipif(**tm.no_cudf())
 @pytest.mark.skipif(**tm.no_cupy())

tests/python-gpu/test_gpu_linear.py

@@ -1,6 +1,7 @@
 import sys
-from hypothesis import strategies, given, settings, assume, note
+from hypothesis import strategies, given, settings, assume
 import pytest
+import numpy
 import xgboost as xgb
 sys.path.append("tests/python")
 import testing as tm
@@ -16,14 +17,10 @@ parameter_strategy = strategies.fixed_dictionaries({
     'top_k': strategies.integers(1, 10),
 })
 
-
 def train_result(param, dmat, num_rounds):
     result = {}
-    booster = xgb.train(
-        param, dmat, num_rounds, [(dmat, 'train')], verbose_eval=False,
-        evals_result=result
-    )
-    assert booster.num_boosted_rounds() == num_rounds
+    xgb.train(param, dmat, num_rounds, [(dmat, 'train')], verbose_eval=False,
+              evals_result=result)
     return result
 
 
@@ -36,7 +33,6 @@ class TestGPULinear:
         param['updater'] = 'gpu_coord_descent'
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         assert tm.non_increasing(result)
 
     # Loss is not guaranteed to always decrease because of regularisation parameters
@@ -53,7 +49,6 @@ class TestGPULinear:
         param['lambda'] = lambd
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         assert tm.non_increasing([result[0], result[-1]])
 
     @pytest.mark.skipif(**tm.no_cupy())

tests/python/test_linear.py

@@ -32,7 +32,6 @@ class TestLinear:
         param.update(coord_param)
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         assert tm.non_increasing(result, 5e-4)
 
     # Loss is not guaranteed to always decrease because of regularisation parameters
@@ -49,7 +48,6 @@ class TestLinear:
         param.update(coord_param)
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         assert tm.non_increasing([result[0], result[-1]])
 
     @given(parameter_strategy, strategies.integers(10, 50),
@@ -59,7 +57,6 @@ class TestLinear:
         param['updater'] = 'shotgun'
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         # shotgun is non-deterministic, so we relax the test by only using first and last
         # iteration.
         if len(result) > 2:
@@ -78,5 +75,4 @@ class TestLinear:
         param['lambda'] = lambd
         param = dataset.set_params(param)
         result = train_result(param, dataset.get_dmat(), num_rounds)['train'][dataset.metric]
-        note(result)
         assert tm.non_increasing([result[0], result[-1]])

tests/python/test_with_dask.py

@@ -705,7 +705,8 @@ async def run_from_dask_array_asyncio(scheduler_address: str) -> xgb.dask.TrainR
 async def run_dask_regressor_asyncio(scheduler_address: str) -> None:
     async with Client(scheduler_address, asynchronous=True) as client:
         X, y, _ = generate_array()
-        regressor = await xgb.dask.DaskXGBRegressor(verbosity=1, n_estimators=2)
+        regressor = await xgb.dask.DaskXGBRegressor(verbosity=1,
+                                                    n_estimators=2)
         regressor.set_params(tree_method='hist')
         regressor.client = client
         await regressor.fit(X, y, eval_set=[(X, y)])

tests/python/test_with_pandas.py

@@ -138,22 +138,9 @@ class TestPandas:
             X, enable_categorical=True
         )
 
+        assert np.issubdtype(transformed[:, 0].dtype, np.integer)
         assert transformed[:, 0].min() == 0
 
-        # test missing value
-        X = pd.DataFrame({"f0": ["a", "b", np.NaN]})
-        X["f0"] = X["f0"].astype("category")
-        arr, _, _ = xgb.data._transform_pandas_df(X, enable_categorical=True)
-        assert not np.any(arr == -1.0)
-
-        X = X["f0"]
-        with pytest.raises(ValueError):
-            xgb.DMatrix(X, y)
-
-        Xy = xgb.DMatrix(X, y, enable_categorical=True)
-        assert Xy.num_row() == 3
-        assert Xy.num_col() == 1
-
     def test_pandas_sparse(self):
         import pandas as pd
         rows = 100