use dataset_loader to load data

include/LightGBM/application.h

@@ -8,6 +8,7 @@
 
 namespace LightGBM {
 
+class DatasetLoader;
 class Dataset;
 class Boosting;
 class ObjectiveFunction;
@@ -59,6 +60,8 @@ private:
 
   /*! \brief All configs */
   OverallConfig config_;
+  /*! \brief Dataset loader */
+  DatasetLoader* dataset_loader_;
   /*! \brief Training data */
   Dataset* train_data_;
   /*! \brief Validation data */

include/LightGBM/config.h

@@ -93,7 +93,6 @@ public:
   std::string output_model = "LightGBM_model.txt";
   std::string output_result = "LightGBM_predict_result.txt";
   std::string input_model = "";
-  std::string input_init_score = "";
   int verbosity = 1;
   int num_model_predict = -1;
   bool is_pre_partition = false;
@@ -318,7 +317,6 @@ struct ParameterAlias {
       { "model_out", "output_model" },
       { "model_input", "input_model" },
       { "model_in", "input_model" },
-      { "init_score", "input_init_score"},
       { "predict_result", "output_result" },
       { "prediction_result", "output_result" },
       { "valid", "valid_data" },

include/LightGBM/dataset.h

-#ifndef LIGHTGBM_DATA_H_
-#define LIGHTGBM_DATA_H_
+#ifndef LIGHTGBM_DATASET_H_
+#define LIGHTGBM_DATASET_H_
 
 #include <LightGBM/utils/random.h>
 #include <LightGBM/utils/text_reader.h>
@@ -18,6 +18,7 @@ namespace LightGBM {
 /*! \brief forward declaration */
 class Feature;
 class BinMapper;
+class DatasetLoader;
 
 /*!
 * \brief This class is used to store some meta(non-feature) data for training data,
@@ -44,13 +45,7 @@ public:
   * \param init_score_filename Filename of initial score
   * \param num_class Number of classes
   */
-  void Init(const char* data_filename, const char* init_score_filename, const int num_class);
-  /*!
-  * \brief Initialize, only load initial score
-  * \param init_score_filename Filename of initial score
-  * \param num_class Number of classes
-  */
-  void Init(const char* init_score_filename, const int num_class);
+  void Init(const char* data_filename, const int num_class);
   /*!
   * \brief Initial with binary memory
   * \param memory Pointer to memory
@@ -177,11 +172,10 @@ public:
   * \return Pointer of initial scores
   */
   inline const float* init_score() const { return init_score_; }
-  
-  /*! \brief Load initial scores from file */
-  void LoadInitialScore();
 
 private:
+  /*! \brief Load initial scores from file */
+  void LoadInitialScore();
   /*! \brief Load wights from file */
   void LoadWeights();
   /*! \brief Load query boundaries from file */
@@ -190,8 +184,6 @@ private:
   void LoadQueryWeights();
   /*! \brief Filename of current data */
   const char* data_filename_;
-  /*! \brief Filename of initial scores */
-  const char* init_score_filename_;
   /*! \brief Number of data */
   data_size_t num_data_;
   /*! \brief Number of classes */
@@ -251,79 +243,16 @@ using PredictFunction =
 */
 class Dataset {
 public:
-  /*!
-  * \brief Constructor
-  * \param data_filename Filename of dataset
-  * \param init_score_filename Filename of initial score
-  * \param io_config configs for IO
-  * \param predict_fun Used for initial model, will give a prediction score based on this function, then set as initial score
-  */
-  Dataset(const char* data_filename, const char* init_score_filename,
-    const IOConfig& io_config, const PredictFunction& predict_fun);
+  friend DatasetLoader;
 
-  /*!
-  * \brief Constructor
-  * \param data_filename Filename of dataset
-  * \param io_config configs for IO
-  * \param predict_fun Used for initial model, will give a prediction score based on this function, then set as initial score
-  */
-  Dataset(const char* data_filename,
-    const IOConfig& io_config, const PredictFunction& predict_fun)
-    : Dataset(data_filename, "", io_config, predict_fun) {
-  }
-
-  /*!
-  * \brief Constructor, without filename, used to load data from memory 
-  * \param io_config configs for IO
-  * \param predict_fun Used for initial model, will give a prediction score based on this function, then set as initial score
-  */
-  Dataset(const IOConfig& io_config, const PredictFunction& predict_fun);
+  Dataset();
 
   /*! \brief Destructor */
   ~Dataset();
 
-  /*! \brief Init Dataset with specific binmapper */
-  void InitByBinMapper(std::vector<const BinMapper*> bin_mappers, data_size_t num_data);
-
-  /*! \brief push raw data into dataset */
-  void PushData(const std::vector<std::vector<std::pair<int, float>>>& datas, data_size_t start_idx, bool is_finished);
 
   void SetField(const char* field_name, const void* field_data, data_size_t num_element, int type);
 
-  /*!
-  * \brief Load training data on parallel training
-  * \param rank Rank of local machine
-  * \param num_machines Total number of all machines
-  * \param is_pre_partition True if data file is pre-partitioned
-  * \param use_two_round_loading True if need to use two round loading
-  */
-  void LoadTrainData(int rank, int num_machines, bool is_pre_partition,
-                                           bool use_two_round_loading);
-
-  /*!
-  * \brief Load training data on single machine training
-  * \param use_two_round_loading True if need to use two round loading
-  */
-  inline void LoadTrainData(bool use_two_round_loading) {
-    LoadTrainData(0, 1, false, use_two_round_loading);
-  }
-
-  /*!
-  * \brief Load data and use bin mapper from other data set, general this function is used to extract feature for validation data
-  * \param train_set Other loaded data set
-  * \param use_two_round_loading True if need to use two round loading
-  */
-  void LoadValidationData(const Dataset* train_set, bool use_two_round_loading);
-
-  /*!
-  * \brief Load data set from binary file
-  * \param bin_filename filename of bin data 
-  * \param rank Rank of local machine
-  * \param num_machines Total number of all machines
-  * \param is_pre_partition True if data file is pre-partitioned
-  */
-  void LoadDataFromBinFile(const char* bin_filename, int rank, int num_machines, bool is_pre_partition);
-
   /*!
   * \brief Save current dataset into binary file, will save to "filename.bin"
   */
@@ -331,6 +260,8 @@ public:
 
   std::vector<const BinMapper*> GetBinMappers() const;
 
+  void CopyFeatureMetadataTo(Dataset *dataset, bool is_enable_sparse) const;
+
   /*!
   * \brief Get a feature pointer for specific index
   * \param i Index for feature
@@ -365,57 +296,7 @@ public:
   Dataset(const Dataset&) = delete;
 
 private:
-  /*!
-  * \brief Load data content on memory. if num_machines > 1 and !is_pre_partition, will partition data
-  * \param rank Rank of local machine
-  * \param num_machines Total number of all machines
-  * \param is_pre_partition True if data file is pre-partitioned
-  */
-  void LoadDataToMemory(int rank, int num_machines, bool is_pre_partition);
-
-  /*!
-  * \brief Sample data from memory, need load data to memory first
-  * \param out_data Store the sampled data
-  */
-  void SampleDataFromMemory(std::vector<std::string>* out_data);
-
-  /*!
-  * \brief Sample data from file
-  * \param rank Rank of local machine
-  * \param num_machines Total number of all machines
-  * \param is_pre_partition True if data file is pre-partitioned
-  * \param out_data Store the sampled data
-  */
-  void SampleDataFromFile(int rank, int num_machines,
-    bool is_pre_partition, std::vector<std::string>* out_data);
-
-  /*!
-  * \brief Get feature bin mapper from sampled data.
-  * if num_machines > 1, differnt machines will construct bin mapper for different features, then have a global sync up
-  * \param rank Rank of local machine
-  * \param num_machines Total number of all machines
-  */
-  void ConstructBinMappers(int rank, int num_machines,
-         const std::vector<std::string>& sample_data);
-
-  /*! \brief Extract local features from memory */
-  void ExtractFeaturesFromMemory();
-
-  /*! \brief Extract local features from file */
-  void ExtractFeaturesFromFile();
-
-  /*! \brief Check can load from binary file */
-  void CheckCanLoadFromBin();
-
-  /*! \brief Check this data set is null or not */
-  void CheckDataset();
-
-  /*! \brief Filename of data */
   const char* data_filename_;
-  /*! \brief A reader class that can read text data */
-  TextReader<data_size_t>* text_reader_;
-  /*! \brief A parser class that can parse data */
-  Parser* parser_;
   /*! \brief Store used features */
   std::vector<Feature*> features_;
   /*! \brief Mapper from real feature index to used index*/
@@ -430,32 +311,12 @@ private:
   int num_class_;
   /*! \brief Store some label level data*/
   Metadata metadata_;
-  /*! \brief Random generator*/
-  Random random_;
-  /*! \brief The maximal number of bin that feature values will bucket in */
-  int max_bin_;
-  /*! \brief True if enable sparse */
-  bool is_enable_sparse_;
   /*! \brief True if dataset is loaded from binary file */
   bool is_loading_from_binfile_;
-  /*! \brief Number of global data, used for distributed learning */
-  size_t global_num_data_ = 0;
-  /*! \brief used to local used data indices */
-  std::vector<data_size_t> used_data_indices_;
-  /*! \brief prediction function for initial model */
-  const PredictFunction& predict_fun_;
   /*! \brief index of label column */
   int label_idx_ = 0;
-  /*! \brief index of weight column */
-  int weight_idx_ = -1;
-  /*! \brief index of group column */
-  int group_idx_ = -1;
-  /*! \brief Mapper from real feature index to used index*/
-  std::unordered_set<int> ignore_features_;
   /*! \brief store feature names */
   std::vector<std::string> feature_names_;
-  /*! \brief store feature names */
-  int bin_construct_sample_cnt_;
 };
 
 }  // namespace LightGBM

src/application/application.cpp

@@ -5,6 +5,7 @@
 
 #include <LightGBM/network.h>
 #include <LightGBM/dataset.h>
+#include <LightGBM/dataset_loader.h>
 #include <LightGBM/boosting.h>
 #include <LightGBM/objective_function.h>
 #include <LightGBM/metric.h>
@@ -26,7 +27,7 @@
 namespace LightGBM {
 
 Application::Application(int argc, char** argv)
-  :train_data_(nullptr), boosting_(nullptr), objective_fun_(nullptr) {
+  :dataset_loader_(nullptr), train_data_(nullptr), boosting_(nullptr), objective_fun_(nullptr) {
   LoadParameters(argc, argv);
   // set number of threads for openmp
   if (config_.num_threads > 0) {
@@ -35,6 +36,7 @@ Application::Application(int argc, char** argv)
 }
 
 Application::~Application() {
+  if (dataset_loader_ != nullptr) { delete dataset_loader_; }
   if (train_data_ != nullptr) { delete train_data_; }
   for (auto& data : valid_datas_) {
     if (data != nullptr) { delete data; }
@@ -141,19 +143,17 @@ void Application::LoadData() {
     config_.io_config.data_random_seed =
        GlobalSyncUpByMin<int>(config_.io_config.data_random_seed);
   }
-  train_data_ = new Dataset(config_.io_config.data_filename.c_str(),
-                         config_.io_config.input_init_score.c_str(),
-                                                  config_.io_config,
-                                                       predict_fun);
+
+  dataset_loader_ = new DatasetLoader(config_.io_config, predict_fun);
+  dataset_loader_->SetHeadder(config_.io_config.data_filename.c_str());
   // load Training data
   if (config_.is_parallel_find_bin) {
     // load data for parallel training
-    train_data_->LoadTrainData(Network::rank(), Network::num_machines(),
-                                     config_.io_config.is_pre_partition,
-                               config_.io_config.use_two_round_loading);
+    train_data_ = dataset_loader_->LoadFromFile(config_.io_config.data_filename.c_str(),
+      Network::rank(), Network::num_machines());
   } else {
     // load data for single machine
-    train_data_->LoadTrainData(config_.io_config.use_two_round_loading);
+    train_data_ = dataset_loader_->LoadFromFile(config_.io_config.data_filename.c_str(), 0, 1);
   }
   // need save binary file
   if (config_.io_config.is_save_binary_file) {
@@ -173,13 +173,8 @@ void Application::LoadData() {
   // Add validation data, if it exists
   for (size_t i = 0; i < config_.io_config.valid_data_filenames.size(); ++i) {
     // add
-    valid_datas_.push_back(
-      new Dataset(config_.io_config.valid_data_filenames[i].c_str(),
-                                                  config_.io_config,
-                                                      predict_fun));
-    // load validation data like train data
-    valid_datas_.back()->LoadValidationData(train_data_,
-                config_.io_config.use_two_round_loading);
+    valid_datas_.push_back(dataset_loader_->LoadFromFileLikeOthers(config_.io_config.valid_data_filenames[i].c_str(),
+      train_data_));
     // need save binary file
     if (config_.io_config.is_save_binary_file) {
       valid_datas_.back()->SaveBinaryFile(nullptr);

src/c_api.cpp

-
 #include <LightGBM/c_api.h>
 #include <LightGBM/dataset.h>
 #include <LightGBM/boosting.h>
@@ -10,6 +9,7 @@
 #include <vector>
 #include <string>
 #include <cstring>
+#include <memory>
 
 namespace LightGBM {
 
@@ -100,3 +100,6 @@ private:
 };
 
 }
+
+using namespace LightGBM;
+

src/io/config.cpp

@@ -202,7 +202,6 @@ void IOConfig::Set(const std::unordered_map<std::string, std::string>& params) {
   GetString(params, "output_model", &output_model);
   GetString(params, "input_model", &input_model);
   GetString(params, "output_result", &output_result);
-  GetString(params, "input_init_score", &input_init_score);
   std::string tmp_str = "";
   if (GetString(params, "valid_data", &tmp_str)) {
     valid_data_filenames = Common::Split(tmp_str.c_str(), ',');

src/io/metadata.cpp

@@ -14,9 +14,8 @@ Metadata::Metadata()
 
 }
 
-void Metadata::Init(const char * data_filename, const char* init_score_filename, const int num_class) {
+void Metadata::Init(const char * data_filename, const int num_class) {
   data_filename_ = data_filename;
-  init_score_filename_ = init_score_filename;
   num_class_ = num_class;
   // for lambdarank, it needs query data for partition data in parallel learning
   LoadQueryBoundaries();
@@ -25,11 +24,6 @@ void Metadata::Init(const char * data_filename, const char* init_score_filename,
   LoadInitialScore();
 }
 
-void Metadata::Init(const char* init_score_filename, const int num_class) {
-  init_score_filename_ = init_score_filename;
-  num_class_ = num_class;
-  LoadInitialScore();
-}
 
 
 Metadata::~Metadata() {
@@ -294,10 +288,14 @@ void Metadata::LoadWeights() {
 
 void Metadata::LoadInitialScore() {
   num_init_score_ = 0;
-  if (init_score_filename_[0] == '\0') { return; }
-  TextReader<size_t> reader(init_score_filename_, false);
+  std::string init_score_filename(data_filename_);
+  // default weight file name
+  init_score_filename.append(".init");
+  TextReader<size_t> reader(init_score_filename.c_str(), false);
   reader.ReadAllLines();
-
+  if (reader.Lines().size() <= 0) {
+    return;
+  }
   Log::Info("Loading initial scores...");
   num_init_score_ = static_cast<data_size_t>(reader.Lines().size());
 

windows/LightGBM.vcxproj

@@ -161,6 +161,7 @@
     <ClInclude Include="..\include\LightGBM\config.h" />
     <ClInclude Include="..\include\LightGBM\c_api.h" />
     <ClInclude Include="..\include\LightGBM\dataset.h" />
+    <ClInclude Include="..\include\LightGBM\dataset_loader.h" />
     <ClInclude Include="..\include\LightGBM\feature.h" />
     <ClInclude Include="..\include\LightGBM\meta.h" />
     <ClInclude Include="..\include\LightGBM\metric.h" />
@@ -208,6 +209,7 @@
     <ClCompile Include="..\src\io\bin.cpp" />
     <ClCompile Include="..\src\io\config.cpp" />
     <ClCompile Include="..\src\io\dataset.cpp" />
+    <ClCompile Include="..\src\io\dataset_loader.cpp" />
     <ClCompile Include="..\src\io\metadata.cpp" />
     <ClCompile Include="..\src\io\parser.cpp" />
     <ClCompile Include="..\src\io\tree.cpp" />

windows/LightGBM.vcxproj.filters

@@ -168,6 +168,9 @@
     <ClInclude Include="..\include\LightGBM\c_api.h">
       <Filter>include\LightGBM</Filter>
     </ClInclude>
+    <ClInclude Include="..\include\LightGBM\dataset_loader.h">
+      <Filter>include\LightGBM</Filter>
+    </ClInclude>
   </ItemGroup>
   <ItemGroup>
     <ClCompile Include="..\src\application\application.cpp">
@@ -236,5 +239,8 @@
     <ClCompile Include="..\src\c_api.cpp">
       <Filter>src</Filter>
     </ClCompile>
+    <ClCompile Include="..\src\io\dataset_loader.cpp">
+      <Filter>src\io</Filter>
+    </ClCompile>
   </ItemGroup>
 </Project>
\ No newline at end of file