Small cleanup for the model IO. (#11602)

- Remove binary-IO related code.
- Split up the approx tree SHAP implementation.

Author: trivialfis

R-package/src/Makevars.in

@@ -85,7 +85,7 @@ OBJECTS= \
     $(PKGROOT)/src/data/iterative_dmatrix.o \
     $(PKGROOT)/src/predictor/predictor.o \
     $(PKGROOT)/src/predictor/cpu_predictor.o \
-    $(PKGROOT)/src/predictor/cpu_treeshap.o \
+    $(PKGROOT)/src/predictor/treeshap.o \
     $(PKGROOT)/src/tree/constraints.o \
     $(PKGROOT)/src/tree/param.o \
     $(PKGROOT)/src/tree/fit_stump.o \

R-package/src/Makevars.win.in

@@ -84,7 +84,7 @@ OBJECTS= \
     $(PKGROOT)/src/data/iterative_dmatrix.o \
     $(PKGROOT)/src/predictor/predictor.o \
     $(PKGROOT)/src/predictor/cpu_predictor.o \
-    $(PKGROOT)/src/predictor/cpu_treeshap.o \
+    $(PKGROOT)/src/predictor/treeshap.o \
     $(PKGROOT)/src/tree/constraints.o \
     $(PKGROOT)/src/tree/param.o \
     $(PKGROOT)/src/tree/fit_stump.o \

include/xgboost/linalg.h

@@ -15,8 +15,8 @@
 
 #include <algorithm>
 #include <cassert>
-#include <cinttypes>  // for int32_t
-#include <cstddef>    // for size_t
+#include <cstddef>  // for size_t
+#include <cstdint>  // for int32_t
 #include <limits>
 #include <string>
 #include <tuple>  // for make_tuple

include/xgboost/tree_model.h

@@ -1,14 +1,12 @@
 /**
  * Copyright 2014-2025, XGBoost Contributors
- * \file tree_model.h
+ *
  * \brief model structure for tree
  * \author Tianqi Chen
  */
 #ifndef XGBOOST_TREE_MODEL_H_
 #define XGBOOST_TREE_MODEL_H_
 
-#include <dmlc/io.h>
-#include <dmlc/parameter.h>
 #include <xgboost/base.h>
 #include <xgboost/data.h>
 #include <xgboost/feature_map.h>
@@ -28,67 +26,24 @@
 namespace xgboost {
 class Json;
 
-// FIXME(trivialfis): Once binary IO is gone, make this parameter internal as it should
-// not be configured by users.
-/*! \brief meta parameters of the tree */
-struct TreeParam : public dmlc::Parameter<TreeParam> {
-  /*! \brief (Deprecated) number of start root */
-  int deprecated_num_roots{1};
-  /*! \brief total number of nodes */
-  int num_nodes{1};
-  /*!\brief number of deleted nodes */
-  int num_deleted{0};
-  /*! \brief maximum depth, this is a statistics of the tree */
-  int deprecated_max_depth{0};
-  /*! \brief number of features used for tree construction */
+/** @brief meta parameters of the tree */
+struct TreeParam {
+  /** @brief The number of nodes */
+  bst_node_t num_nodes{1};
+  /** @brief The number of deleted nodes */
+  bst_node_t num_deleted{0};
+  /** @brief The number of features used for tree construction */
   bst_feature_t num_feature{0};
-  /*!
-   * \brief leaf vector size, used for vector tree
-   * used to store more than one dimensional information in tree
-   */
+  /** @brief leaf vector size. Used by the vector leaf. */
   bst_target_t size_leaf_vector{1};
-  /*! \brief reserved part, make sure alignment works for 64bit */
-  int reserved[31];
-  /*! \brief constructor */
-  TreeParam() {
-    // assert compact alignment
-    static_assert(sizeof(TreeParam) == (31 + 6) * sizeof(int), "TreeParam: 64 bit align");
-    std::memset(reserved, 0, sizeof(reserved));
-  }
-
-  // Swap byte order for all fields. Useful for transporting models between machines with different
-  // endianness (big endian vs little endian)
-  [[nodiscard]] TreeParam ByteSwap() const {
-    TreeParam x = *this;
-    dmlc::ByteSwap(&x.deprecated_num_roots, sizeof(x.deprecated_num_roots), 1);
-    dmlc::ByteSwap(&x.num_nodes, sizeof(x.num_nodes), 1);
-    dmlc::ByteSwap(&x.num_deleted, sizeof(x.num_deleted), 1);
-    dmlc::ByteSwap(&x.deprecated_max_depth, sizeof(x.deprecated_max_depth), 1);
-    dmlc::ByteSwap(&x.num_feature, sizeof(x.num_feature), 1);
-    dmlc::ByteSwap(&x.size_leaf_vector, sizeof(x.size_leaf_vector), 1);
-    dmlc::ByteSwap(x.reserved, sizeof(x.reserved[0]), sizeof(x.reserved) / sizeof(x.reserved[0]));
-    return x;
-  }
-
-  // declare the parameters
-  DMLC_DECLARE_PARAMETER(TreeParam) {
-    // only declare the parameters that can be set by the user.
-    // other arguments are set by the algorithm.
-    DMLC_DECLARE_FIELD(num_nodes).set_lower_bound(1).set_default(1);
-    DMLC_DECLARE_FIELD(num_feature)
-        .set_default(0)
-        .describe("Number of features used in tree construction.");
-    DMLC_DECLARE_FIELD(num_deleted).set_default(0);
-    DMLC_DECLARE_FIELD(size_leaf_vector)
-        .set_lower_bound(0)
-        .set_default(1)
-        .describe("Size of leaf vector, reserved for vector tree");
-  }
 
   bool operator==(const TreeParam& b) const {
     return num_nodes == b.num_nodes && num_deleted == b.num_deleted &&
            num_feature == b.num_feature && size_leaf_vector == b.size_leaf_vector;
   }
+
+  void FromJson(Json const& in);
+  void ToJson(Json* p_out) const;
 };
 
 /*! \brief node statistics used in regression tree */
@@ -109,16 +64,6 @@ struct RTreeNodeStat {
     return loss_chg == b.loss_chg && sum_hess == b.sum_hess &&
            base_weight == b.base_weight && leaf_child_cnt == b.leaf_child_cnt;
   }
-  // Swap byte order for all fields. Useful for transporting models between machines with different
-  // endianness (big endian vs little endian)
-  [[nodiscard]] RTreeNodeStat ByteSwap() const {
-    RTreeNodeStat x = *this;
-    dmlc::ByteSwap(&x.loss_chg, sizeof(x.loss_chg), 1);
-    dmlc::ByteSwap(&x.sum_hess, sizeof(x.sum_hess), 1);
-    dmlc::ByteSwap(&x.base_weight, sizeof(x.base_weight), 1);
-    dmlc::ByteSwap(&x.leaf_child_cnt, sizeof(x.leaf_child_cnt), 1);
-    return x;
-  }
 };
 
 /**
@@ -166,12 +111,11 @@ class RegTree : public Model {
    public:
     XGBOOST_DEVICE Node()  {
       // assert compact alignment
-      static_assert(sizeof(Node) == 4 * sizeof(int) + sizeof(Info),
-                    "Node: 64 bit align");
+      static_assert(sizeof(Node) == 4 * sizeof(int) + sizeof(Info), "Node: 64 bit align");
     }
-    Node(int32_t cleft, int32_t cright, int32_t parent,
-         uint32_t split_ind, float split_cond, bool default_left) :
-        parent_{parent}, cleft_{cleft}, cright_{cright} {
+    Node(int32_t cleft, int32_t cright, int32_t parent, uint32_t split_ind, float split_cond,
+         bool default_left)
+        : parent_{parent}, cleft_{cleft}, cright_{cright} {
       this->SetParent(parent_);
       this->SetSplit(split_ind, split_cond, default_left);
     }
@@ -261,16 +205,6 @@ class RegTree : public Model {
              info_.leaf_value == b.info_.leaf_value;
     }
 
-    [[nodiscard]] Node ByteSwap() const {
-      Node x = *this;
-      dmlc::ByteSwap(&x.parent_, sizeof(x.parent_), 1);
-      dmlc::ByteSwap(&x.cleft_, sizeof(x.cleft_), 1);
-      dmlc::ByteSwap(&x.cright_, sizeof(x.cright_), 1);
-      dmlc::ByteSwap(&x.sindex_, sizeof(x.sindex_), 1);
-      dmlc::ByteSwap(&x.info_, sizeof(x.info_), 1);
-      return x;
-    }
-
    private:
     /*!
      * \brief in leaf node, we have weights, in non-leaf nodes,
@@ -320,7 +254,6 @@ class RegTree : public Model {
   }
 
   RegTree() {
-    param_.Init(Args{});
     nodes_.resize(param_.num_nodes);
     stats_.resize(param_.num_nodes);
     split_types_.resize(param_.num_nodes, FeatureType::kNumerical);
@@ -589,14 +522,6 @@ class RegTree : public Model {
     bool has_missing_;
   };
 
-  /*!
-   * \brief calculate the approximate feature contributions for the given root
-   * \param feat dense feature vector, if the feature is missing the field is set to NaN
-   * \param out_contribs output vector to hold the contributions
-   */
-  void CalculateContributionsApprox(const RegTree::FVec& feat,
-                                    std::vector<float>* mean_values,
-                                    bst_float* out_contribs) const;
   /*!
    * \brief dump the model in the requested format as a text string
    * \param fmap feature map that may help give interpretations of feature

src/gbm/gblinear_model.cc

@@ -6,9 +6,7 @@
 #include "xgboost/json.h"
 #include "gblinear_model.h"
 
-namespace xgboost {
-namespace gbm {
-
+namespace xgboost::gbm {
 void GBLinearModel::SaveModel(Json* p_out) const {
   auto& out = *p_out;
 
@@ -42,7 +40,4 @@ void GBLinearModel::LoadModel(Json const& in) {
     this->num_boosted_rounds = 0;
   }
 }
-
-DMLC_REGISTER_PARAMETER(DeprecatedGBLinearModelParam);
-}  // namespace gbm
-}  // namespace xgboost
+}  // namespace xgboost::gbm

src/gbm/gblinear_model.h

@@ -14,40 +14,14 @@
 #include "xgboost/feature_map.h"
 #include "xgboost/model.h"
 #include "xgboost/json.h"
-#include "xgboost/parameter.h"
 
 namespace xgboost {
 class Json;
 namespace gbm {
-// Deprecated in 1.0.0. model parameter.  Only staying here for compatible binary model IO.
-struct DeprecatedGBLinearModelParam : public dmlc::Parameter<DeprecatedGBLinearModelParam> {
-  // number of feature dimension
-  uint32_t deprecated_num_feature;
-  // deprecated. use learner_model_param_->num_output_group.
-  int32_t deprecated_num_output_group;
-  // reserved field
-  int32_t reserved[32];
-  // constructor
-  DeprecatedGBLinearModelParam() {
-    static_assert(sizeof(*this) == sizeof(int32_t) * 34,
-                  "Model parameter size can not be changed.");
-    std::memset(this, 0, sizeof(DeprecatedGBLinearModelParam));
-  }
-
-  DMLC_DECLARE_PARAMETER(DeprecatedGBLinearModelParam) {
-    DMLC_DECLARE_FIELD(deprecated_num_feature);
-    DMLC_DECLARE_FIELD(deprecated_num_output_group);
-  }
-};
-
 // model for linear booster
 class GBLinearModel : public Model {
- private:
-  // Deprecated in 1.0.0
-  DeprecatedGBLinearModelParam param_;
-
  public:
-  int32_t num_boosted_rounds{0};
+  std::int32_t num_boosted_rounds{0};
   LearnerModelParam const* learner_model_param;
 
  public:

src/gbm/gbtree_model.h

@@ -42,21 +42,13 @@ struct GBTreeModelParam : public dmlc::Parameter<GBTreeModelParam> {
   /**
    * \brief number of trees
    */
-  std::int32_t num_trees;
+  std::int32_t num_trees{0};
   /**
    * \brief Number of trees for a forest.
    */
-  std::int32_t num_parallel_tree;
-  /*! \brief reserved parameters */
-  int32_t reserved[38];
-
-  /*! \brief constructor */
-  GBTreeModelParam() {
-    std::memset(this, 0, sizeof(GBTreeModelParam));  // FIXME(trivialfis): Why?
-    static_assert(sizeof(GBTreeModelParam) == (4 + 2 + 2 + 32) * sizeof(int32_t),
-                  "64/32 bit compatibility issue");
-    num_parallel_tree = 1;
-  }
+  std::int32_t num_parallel_tree{1};
+
+  GBTreeModelParam() = default;
 
   // declare parameters, only declare those that need to be set.
   DMLC_DECLARE_PARAMETER(GBTreeModelParam) {
@@ -69,16 +61,6 @@ struct GBTreeModelParam : public dmlc::Parameter<GBTreeModelParam> {
             "Number of parallel trees constructed during each iteration."
             " This option is used to support boosted random forest.");
   }
-
-  // Swap byte order for all fields. Useful for transporting models between machines with different
-  // endianness (big endian vs little endian)
-  GBTreeModelParam ByteSwap() const {
-    GBTreeModelParam x = *this;
-    dmlc::ByteSwap(&x.num_trees, sizeof(x.num_trees), 1);
-    dmlc::ByteSwap(&x.num_parallel_tree, sizeof(x.num_parallel_tree), 1);
-    dmlc::ByteSwap(x.reserved, sizeof(x.reserved[0]), sizeof(x.reserved) / sizeof(x.reserved[0]));
-    return x;
-  }
 };
 
 struct GBTreeModel : public Model {

src/learner.cc

@@ -41,7 +41,6 @@
 #include "common/random.h"                // for GlobalRandom
 #include "common/timer.h"                 // for Monitor
 #include "common/version.h"               // for Version
-#include "dmlc/endian.h"                  // for ByteSwap, DMLC_IO_NO_ENDIAN_SWAP
 #include "xgboost/base.h"                 // for Args, bst_float, GradientPair, bst_feature_t, ...
 #include "xgboost/context.h"              // for Context
 #include "xgboost/data.h"                 // for DMatrix, MetaInfo
@@ -84,22 +83,22 @@ T& UsePtr(T& ptr) {  // NOLINT
  */
 struct LearnerModelParamLegacy : public dmlc::Parameter<LearnerModelParamLegacy> {
   /* \brief global bias */
-  bst_float base_score;
+  bst_float base_score{ObjFunction::DefaultBaseScore()};
   /* \brief number of features  */
-  bst_feature_t num_feature;
+  bst_feature_t num_feature{0};
   /* \brief number of classes, if it is multi-class classification  */
-  std::int32_t num_class;
+  std::int32_t num_class{0};
   /*! \brief Model contain additional properties */
-  int32_t contain_extra_attrs;
+  int32_t contain_extra_attrs{0};
   /*! \brief Model contain eval metrics */
-  int32_t contain_eval_metrics;
+  int32_t contain_eval_metrics{0};
   /*! \brief the version of XGBoost. */
-  std::uint32_t major_version;
-  std::uint32_t minor_version;
+  std::int32_t major_version{std::get<0>(Version::Self())};
+  std::int32_t minor_version{std::get<1>(Version::Self())};
   /**
    * \brief Number of target variables.
    */
-  bst_target_t num_target;
+  bst_target_t num_target{1};
   /**
    * \brief Whether we should calculate the base score from training data.
    *
@@ -110,19 +109,8 @@ struct LearnerModelParamLegacy : public dmlc::Parameter<LearnerModelParamLegacy>
    *   of bool for the ease of serialization.
    */
   std::int32_t boost_from_average{true};
-  /*! \brief reserved field */
-  int reserved[25];
-  /*! \brief constructor */
-  LearnerModelParamLegacy() {
-    std::memset(this, 0, sizeof(LearnerModelParamLegacy));
-    base_score = ObjFunction::DefaultBaseScore();
-    num_target = 1;
-    major_version = std::get<0>(Version::Self());
-    minor_version = std::get<1>(Version::Self());
-    boost_from_average = true;
-    static_assert(sizeof(LearnerModelParamLegacy) == 136,
-                  "Do not change the size of this struct, as it will break binary IO.");
-  }
+
+  LearnerModelParamLegacy() = default;
 
   // Skip other legacy fields.
   [[nodiscard]] Json ToJson() const {
@@ -175,21 +163,6 @@ struct LearnerModelParamLegacy : public dmlc::Parameter<LearnerModelParamLegacy>
     from_chars(str.c_str(), str.c_str() + str.size(), base_score);
   }
 
-  [[nodiscard]] LearnerModelParamLegacy ByteSwap() const {
-    LearnerModelParamLegacy x = *this;
-    dmlc::ByteSwap(&x.base_score, sizeof(x.base_score), 1);
-    dmlc::ByteSwap(&x.num_feature, sizeof(x.num_feature), 1);
-    dmlc::ByteSwap(&x.num_class, sizeof(x.num_class), 1);
-    dmlc::ByteSwap(&x.contain_extra_attrs, sizeof(x.contain_extra_attrs), 1);
-    dmlc::ByteSwap(&x.contain_eval_metrics, sizeof(x.contain_eval_metrics), 1);
-    dmlc::ByteSwap(&x.major_version, sizeof(x.major_version), 1);
-    dmlc::ByteSwap(&x.minor_version, sizeof(x.minor_version), 1);
-    dmlc::ByteSwap(&x.num_target, sizeof(x.num_target), 1);
-    dmlc::ByteSwap(&x.boost_from_average, sizeof(x.boost_from_average), 1);
-    dmlc::ByteSwap(x.reserved, sizeof(x.reserved[0]), sizeof(x.reserved) / sizeof(x.reserved[0]));
-    return x;
-  }
-
   template <typename Container>
   Args UpdateAllowUnknown(Container const& kwargs) {
     // Detect whether user has made their own base score.