lightgbm 0.1.5 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: dad09a50916e229119d99de66aa676868224af44d8958916284718edb7fb3a3f
-  data.tar.gz: 2f7d299a7037455a4ebd1ccaf41bc102ec66808b9e806a59194fd14699204364
+  metadata.gz: 2b3405aa832aaa365ce38d59bf5a36b318f8d02c1a236ea144eb9b102bf7f07e
+  data.tar.gz: 5be378b94486a668fffdc276a73ff62f24a24b644932bd776fd877f4b69b157a
 SHA512:
-  metadata.gz: aa415946a82be07f1b9b8823b60fc75030f9e5547859078459583fa64d9212aad11db39d30e8e327b1fe23f48fc2ce9c6ddbd0445b52f73b44374a07c4724f5c
-  data.tar.gz: 99357888fde412672ecd804396cbd3a524b7dee6318d46d92a07064b858cd5f62f144f6b680a25ae0f1ae567df7d4670fcf38ba706fa31cbdfe58fd79eabc62d
+  metadata.gz: 914379fe9271306117fc222534915f3c19de1c6b165ecc053f49b9089da0b85594c8430958f03f28a23d98fcf2879c4948c8d3e5fecee920c2984700928380ec
+  data.tar.gz: 4f93ae6b4a2c3123d452ecf41d7ada8473af397f12ddafb4e34b4e552aae4b6ef23ecb75b54ce3f7f64ba0985402c6172070de4afaca2ea614875f6d91c2de1a

data/CHANGELOG.md

@@ -1,11 +1,35 @@
-## 0.1.5
+## 0.2.0 (2020-08-31)
+
+- Updated LightGBM to 3.0.0
+
+## 0.1.9 (2020-06-10)
+
+- Added support for Rover
+- Improved performance of Numo datasets
+
+## 0.1.8 (2020-05-09)
+
+- Improved error message when OpenMP not found on Mac
+- Fixed `Cannot add validation data` error
+
+## 0.1.7 (2019-12-05)
+
+- Updated LightGBM to 2.3.1
+- Switched to doubles for datasets and predictions
+
+## 0.1.6 (2019-09-29)
+
+- Updated LightGBM to 2.3.0
+- Fixed error with JRuby
+
+## 0.1.5 (2019-09-03)
 
 - Packaged LightGBM with gem
 - Added support for missing values
 - Added `feature_names` to datasets
 - Fixed Daru training and prediction
 
-## 0.1.4
+## 0.1.4 (2019-08-19)
 
 - Friendlier message when LightGBM not found
 - Added `Ranker`
@@ -13,22 +37,22 @@
 - Free memory when objects are destroyed
 - Removed unreleased `dump_text` method
 
-## 0.1.3
+## 0.1.3 (2019-08-16)
 
 - Added Scikit-Learn API
 - Added support for Daru and Numo::NArray
 
-## 0.1.2
+## 0.1.2 (2019-08-15)
 
 - Added `cv` method
 - Added early stopping
 - Fixed multiclass classification
 
-## 0.1.1
+## 0.1.1 (2019-08-14)
 
 - Added training API
 - Added many methods
 
-## 0.1.0
+## 0.1.0 (2019-08-13)
 
 - First release

data/LICENSE.txt

@@ -1,22 +1,22 @@
-Copyright (c) 2019 Andrew Kane
+The MIT License (MIT)
 
-MIT License
+Copyright (c) Microsoft Corporation
+Copyright (c) 2019-2020 Andrew Kane
 
-Permission is hereby granted, free of charge, to any person obtaining
-a copy of this software and associated documentation files (the
-"Software"), to deal in the Software without restriction, including
-without limitation the rights to use, copy, modify, merge, publish,
-distribute, sublicense, and/or sell copies of the Software, and to
-permit persons to whom the Software is furnished to do so, subject to
-the following conditions:
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
 
-The above copyright notice and this permission notice shall be
-included in all copies or substantial portions of the Software.
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
 
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
-EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
-NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
-LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
-OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
-WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

data/README.md

@@ -1,8 +1,6 @@
 # LightGBM
 
-[LightGBM](https://github.com/microsoft/LightGBM) - the high performance machine learning library - for Ruby
-
-:fire: Uses the C API for blazing performance
+[LightGBM](https://github.com/microsoft/LightGBM) - high performance gradient boosting - for Ruby
 
 [![Build Status](https://travis-ci.org/ankane/lightgbm.svg?branch=master)](https://travis-ci.org/ankane/lightgbm)
 
@@ -14,22 +12,12 @@ Add this line to your application’s Gemfile:
 gem 'lightgbm'
 ```
 
-LightGBM is packaged with the gem, so no need to install it separately. On Mac, also run:
+On Mac, also install OpenMP:
 
 ```sh
 brew install libomp
 ```
 
-## Getting Started
-
-This library follows the [Python API](https://lightgbm.readthedocs.io/en/latest/Python-API.html). A few differences are:
-
-- The `get_` and `set_` prefixes are removed from methods
-- The default verbosity is `-1`
-- With the `cv` method, `stratified` is set to `false`
-
-Some methods and options are also missing at the moment. PRs welcome!
-
 ## Training API
 
 Prep your data
@@ -141,16 +129,22 @@ Data can be an array of arrays
 [[1, 2, 3], [4, 5, 6]]
 ```
 
-Or a Daru data frame
+Or a Numo array
 
 ```ruby
-Daru::DataFrame.from_csv("houses.csv")
+Numo::NArray.cast([[1, 2, 3], [4, 5, 6]])
 ```
 
-Or a Numo NArray
+Or a Rover data frame
 
 ```ruby
-Numo::DFloat.new(3, 2).seq
+Rover.read_csv("houses.csv")
+```
+
+Or a Daru data frame
+
+```ruby
+Daru::DataFrame.from_csv("houses.csv")
 ```
 
 ## Helpful Resources
@@ -160,12 +154,18 @@ Numo::DFloat.new(3, 2).seq
 
 ## Related Projects
 
-- [Xgb](https://github.com/ankane/xgb) - XGBoost for Ruby
-- [Eps](https://github.com/ankane/eps) - Machine Learning for Ruby
+- [XGBoost](https://github.com/ankane/xgboost) - XGBoost for Ruby
+- [Eps](https://github.com/ankane/eps) - Machine learning for Ruby
 
 ## Credits
 
-Thanks to the [xgboost](https://github.com/PairOnAir/xgboost-ruby) gem for serving as an initial reference, and Selva Prabhakaran for the [test datasets](https://github.com/selva86/datasets).
+This library follows the [Python API](https://lightgbm.readthedocs.io/en/latest/Python-API.html). A few differences are:
+
+- The `get_` and `set_` prefixes are removed from methods
+- The default verbosity is `-1`
+- With the `cv` method, `stratified` is set to `false`
+
+Thanks to the [xgboost](https://github.com/PairOnAir/xgboost-ruby) gem for showing how to use FFI.
 
 ## History
 
@@ -179,3 +179,13 @@ Everyone is encouraged to help improve this project. Here are a few ways you can
 - Fix bugs and [submit pull requests](https://github.com/ankane/lightgbm/pulls)
 - Write, clarify, or fix documentation
 - Suggest or add new features
+
+To get started with development:
+
+```sh
+git clone https://github.com/ankane/lightgbm.git
+cd lightgbm
+bundle install
+bundle exec rake vendor:all
+bundle exec rake test
+```

data/lib/lightgbm.rb

@@ -36,6 +36,8 @@ module LightGBM
           booster.train_data_name = name || "training"
           valid_contain_train = true
         else
+          # ensure the validation set references the training set
+          data.reference = train_set
           booster.add_valid(data, name || "valid_#{i}")
         end
       end
@@ -59,16 +61,14 @@ module LightGBM
           # print results
           messages = []
 
+          eval_valid = booster.eval_valid
           if valid_contain_train
-            # not sure why reversed in output
-            booster.eval_train.reverse.each do |res|
-              messages << "%s's %s: %g" % [res[0], res[1], res[2]]
-            end
+            eval_valid = eval_valid + booster.eval_train
           end
-
-          eval_valid = booster.eval_valid
           # not sure why reversed in output
-          eval_valid.reverse.each do |res|
+          eval_valid.reverse!
+
+          eval_valid.each do |res|
             messages << "%s's %s: %g" % [res[0], res[1], res[2]]
           end
 
@@ -133,6 +133,7 @@ module LightGBM
       if early_stopping_rounds
         best_score = {}
         best_iter = {}
+        best_iteration = nil
       end
 
       num_boost_round.times do |iteration|
@@ -172,6 +173,7 @@ module LightGBM
               best_score[k] = score
               best_iter[k] = iteration
             elsif iteration - best_iter[k] >= early_stopping_rounds
+              best_iteration = best_iter[k]
               stop_early = true
               break
             end
@@ -180,6 +182,14 @@ module LightGBM
         end
       end
 
+      if early_stopping_rounds
+        # use best iteration from first metric if not stopped early
+        best_iteration ||= best_iter[best_iter.keys.first]
+        eval_hist.each_key do |k|
+          eval_hist[k] = eval_hist[k].first(best_iteration + 1)
+        end
+      end
+
       eval_hist
     end
 

data/lib/lightgbm/booster.rb

@@ -30,7 +30,7 @@ module LightGBM
 
     def current_iteration
       out = ::FFI::MemoryPointer.new(:int)
-      check_result FFI::LGBM_BoosterGetCurrentIteration(handle_pointer, out)
+      check_result FFI.LGBM_BoosterGetCurrentIteration(handle_pointer, out)
       out.read_int
     end
 
@@ -38,12 +38,13 @@ module LightGBM
       num_iteration ||= best_iteration
       buffer_len = 1 << 20
       out_len = ::FFI::MemoryPointer.new(:int64)
-      out_str = ::FFI::MemoryPointer.new(:string, buffer_len)
-      check_result FFI.LGBM_BoosterDumpModel(handle_pointer, start_iteration, num_iteration, buffer_len, out_len, out_str)
-      actual_len = out_len.read_int64
+      out_str = ::FFI::MemoryPointer.new(:char, buffer_len)
+      feature_importance_type = 0 # TODO add option
+      check_result FFI.LGBM_BoosterDumpModel(handle_pointer, start_iteration, num_iteration, feature_importance_type, buffer_len, out_len, out_str)
+      actual_len = read_int64(out_len)
       if actual_len > buffer_len
-        out_str = ::FFI::MemoryPointer.new(:string, actual_len)
-        check_result FFI.LGBM_BoosterDumpModel(handle_pointer, start_iteration, num_iteration, actual_len, out_len, out_str)
+        out_str = ::FFI::MemoryPointer.new(:char, actual_len)
+        check_result FFI.LGBM_BoosterDumpModel(handle_pointer, start_iteration, num_iteration, feature_importance_type, actual_len, out_len, out_str)
       end
       out_str.read_string
     end
@@ -85,12 +86,13 @@ module LightGBM
       num_iteration ||= best_iteration
       buffer_len = 1 << 20
       out_len = ::FFI::MemoryPointer.new(:int64)
-      out_str = ::FFI::MemoryPointer.new(:string, buffer_len)
-      check_result FFI.LGBM_BoosterSaveModelToString(handle_pointer, start_iteration, num_iteration, buffer_len, out_len, out_str)
-      actual_len = out_len.read_int64
+      out_str = ::FFI::MemoryPointer.new(:char, buffer_len)
+      feature_importance_type = 0 # TODO add option
+      check_result FFI.LGBM_BoosterSaveModelToString(handle_pointer, start_iteration, num_iteration, feature_importance_type, buffer_len, out_len, out_str)
+      actual_len = read_int64(out_len)
       if actual_len > buffer_len
-        out_str = ::FFI::MemoryPointer.new(:string, actual_len)
-        check_result FFI.LGBM_BoosterSaveModelToString(handle_pointer, start_iteration, num_iteration, actual_len, out_len, out_str)
+        out_str = ::FFI::MemoryPointer.new(:char, actual_len)
+        check_result FFI.LGBM_BoosterSaveModelToString(handle_pointer, start_iteration, num_iteration, feature_importance_type, actual_len, out_len, out_str)
       end
       out_str.read_string
     end
@@ -104,18 +106,18 @@ module LightGBM
 
     def num_model_per_iteration
       out = ::FFI::MemoryPointer.new(:int)
-      check_result FFI::LGBM_BoosterNumModelPerIteration(handle_pointer, out)
+      check_result FFI.LGBM_BoosterNumModelPerIteration(handle_pointer, out)
       out.read_int
     end
 
     def num_trees
       out = ::FFI::MemoryPointer.new(:int)
-      check_result FFI::LGBM_BoosterNumberOfTotalModel(handle_pointer, out)
+      check_result FFI.LGBM_BoosterNumberOfTotalModel(handle_pointer, out)
       out.read_int
     end
 
     # TODO support different prediction types
-    def predict(input, num_iteration: nil, **params)
+    def predict(input, start_iteration: nil, num_iteration: nil, **params)
       input =
         if daru?(input)
           input.map_rows(&:to_a)
@@ -126,18 +128,19 @@ module LightGBM
       singular = !input.first.is_a?(Array)
       input = [input] if singular
 
+      start_iteration ||= 0
       num_iteration ||= best_iteration
       num_class ||= num_class()
 
       flat_input = input.flatten
       handle_missing(flat_input)
-      data = ::FFI::MemoryPointer.new(:float, input.count * input.first.count)
-      data.put_array_of_float(0, flat_input)
+      data = ::FFI::MemoryPointer.new(:double, input.count * input.first.count)
+      data.write_array_of_double(flat_input)
 
       out_len = ::FFI::MemoryPointer.new(:int64)
       out_result = ::FFI::MemoryPointer.new(:double, num_class * input.count)
-      check_result FFI.LGBM_BoosterPredictForMat(handle_pointer, data, 0, input.count, input.first.count, 1, 0, num_iteration, params_str(params), out_len, out_result)
-      out = out_result.read_array_of_double(out_len.read_int64)
+      check_result FFI.LGBM_BoosterPredictForMat(handle_pointer, data, 1, input.count, input.first.count, 1, 0, start_iteration, num_iteration, params_str(params), out_len, out_result)
+      out = out_result.read_array_of_double(read_int64(out_len))
       out = out.each_slice(num_class).to_a if num_class > 1
 
       singular ? out.first : out
@@ -145,7 +148,8 @@ module LightGBM
 
     def save_model(filename, num_iteration: nil, start_iteration: 0)
       num_iteration ||= best_iteration
-      check_result FFI.LGBM_BoosterSaveModel(handle_pointer, start_iteration, num_iteration, filename)
+      feature_importance_type = 0 # TODO add
+      check_result FFI.LGBM_BoosterSaveModel(handle_pointer, start_iteration, num_iteration, feature_importance_type, filename)
       self # consistent with Python API
     end
 
@@ -168,17 +172,19 @@ module LightGBM
 
     def eval_counts
       out = ::FFI::MemoryPointer.new(:int)
-      check_result FFI::LGBM_BoosterGetEvalCounts(handle_pointer, out)
+      check_result FFI.LGBM_BoosterGetEvalCounts(handle_pointer, out)
       out.read_int
     end
 
     def eval_names
       eval_counts ||= eval_counts()
       out_len = ::FFI::MemoryPointer.new(:int)
+      out_buffer_len = ::FFI::MemoryPointer.new(:size_t)
       out_strs = ::FFI::MemoryPointer.new(:pointer, eval_counts)
-      str_ptrs = eval_counts.times.map { ::FFI::MemoryPointer.new(:string, 255) }
-      out_strs.put_array_of_pointer(0, str_ptrs)
-      check_result FFI.LGBM_BoosterGetEvalNames(handle_pointer, out_len, out_strs)
+      buffer_len = 255
+      str_ptrs = eval_counts.times.map { ::FFI::MemoryPointer.new(:char, buffer_len) }
+      out_strs.write_array_of_pointer(str_ptrs)
+      check_result FFI.LGBM_BoosterGetEvalNames(handle_pointer, eval_counts, out_len, buffer_len, out_buffer_len, out_strs)
       str_ptrs.map(&:read_string)
     end
 
@@ -198,10 +204,15 @@ module LightGBM
 
     def num_class
       out = ::FFI::MemoryPointer.new(:int)
-      check_result FFI::LGBM_BoosterGetNumClasses(handle_pointer, out)
+      check_result FFI.LGBM_BoosterGetNumClasses(handle_pointer, out)
       out.read_int
     end
 
+    # read_int64 not available on JRuby
+    def read_int64(ptr)
+      ptr.read_array_of_int64(1).first
+    end
+
     include Utils
   end
 end

data/lib/lightgbm/dataset.rb

@@ -4,51 +4,16 @@ module LightGBM
 
     def initialize(data, label: nil, weight: nil, group: nil, params: nil, reference: nil, used_indices: nil, categorical_feature: "auto", feature_names: nil)
       @data = data
+      @label = label
+      @weight = weight
+      @group = group
+      @params = params
+      @reference = reference
+      @used_indices = used_indices
+      @categorical_feature = categorical_feature
+      @feature_names = feature_names
 
-      # TODO stringify params
-      params ||= {}
-      if categorical_feature != "auto" && categorical_feature.any?
-        params["categorical_feature"] ||= categorical_feature.join(",")
-      end
-      set_verbosity(params)
-
-      @handle = ::FFI::MemoryPointer.new(:pointer)
-      parameters = params_str(params)
-      reference = reference.handle_pointer if reference
-      if used_indices
-        used_row_indices = ::FFI::MemoryPointer.new(:int32, used_indices.count)
-        used_row_indices.put_array_of_int32(0, used_indices)
-        check_result FFI.LGBM_DatasetGetSubset(reference, used_row_indices, used_indices.count, parameters, @handle)
-      elsif data.is_a?(String)
-        check_result FFI.LGBM_DatasetCreateFromFile(data, parameters, reference, @handle)
-      else
-        if matrix?(data)
-          nrow = data.row_count
-          ncol = data.column_count
-          flat_data = data.to_a.flatten
-        elsif daru?(data)
-          nrow, ncol = data.shape
-          flat_data = data.map_rows(&:to_a).flatten
-        elsif narray?(data)
-          nrow, ncol = data.shape
-          flat_data = data.flatten.to_a
-        else
-          nrow = data.count
-          ncol = data.first.count
-          flat_data = data.flatten
-        end
-
-        handle_missing(flat_data)
-        c_data = ::FFI::MemoryPointer.new(:float, nrow * ncol)
-        c_data.put_array_of_float(0, flat_data)
-        check_result FFI.LGBM_DatasetCreateFromMat(c_data, 0, nrow, ncol, 1, parameters, reference, @handle)
-      end
-      ObjectSpace.define_finalizer(self, self.class.finalize(handle_pointer)) unless used_indices
-
-      self.label = label if label
-      self.weight = weight if weight
-      self.group = group if group
-      self.feature_names = feature_names if feature_names
+      construct
     end
 
     def label
@@ -59,34 +24,50 @@ module LightGBM
       field("weight")
     end
 
-    def label=(label)
-      set_field("label", label)
-    end
-
     def feature_names
       # must preallocate space
       num_feature_names = ::FFI::MemoryPointer.new(:int)
-      out_strs = ::FFI::MemoryPointer.new(:pointer, 1000)
-      str_ptrs = 1000.times.map { ::FFI::MemoryPointer.new(:string, 255) }
-      out_strs.put_array_of_pointer(0, str_ptrs)
-      check_result FFI.LGBM_DatasetGetFeatureNames(handle_pointer, out_strs, num_feature_names)
+      out_buffer_len = ::FFI::MemoryPointer.new(:size_t)
+      len = 1000
+      out_strs = ::FFI::MemoryPointer.new(:pointer, len)
+      buffer_len = 255
+      str_ptrs = len.times.map { ::FFI::MemoryPointer.new(:char, buffer_len) }
+      out_strs.write_array_of_pointer(str_ptrs)
+      check_result FFI.LGBM_DatasetGetFeatureNames(handle_pointer, len, num_feature_names, buffer_len, out_buffer_len, out_strs)
       str_ptrs[0, num_feature_names.read_int].map(&:read_string)
     end
 
+    def label=(label)
+      @label = label
+      set_field("label", label)
+    end
+
     def weight=(weight)
+      @weight = weight
       set_field("weight", weight)
     end
 
     def group=(group)
+      @group = group
       set_field("group", group, type: :int32)
     end
 
     def feature_names=(feature_names)
+      @feature_names = feature_names
       c_feature_names = ::FFI::MemoryPointer.new(:pointer, feature_names.size)
       c_feature_names.write_array_of_pointer(feature_names.map { |v| ::FFI::MemoryPointer.from_string(v) })
       check_result FFI.LGBM_DatasetSetFeatureNames(handle_pointer, c_feature_names, feature_names.size)
     end
 
+    # TODO only update reference if not in chain
+    def reference=(reference)
+      if reference != @reference
+        @reference = reference
+        free_handle
+        construct
+      end
+    end
+
     def num_data
       out = ::FFI::MemoryPointer.new(:int)
       check_result FFI.LGBM_DatasetGetNumData(handle_pointer, out)
@@ -103,11 +84,6 @@ module LightGBM
       check_result FFI.LGBM_DatasetSaveBinary(handle_pointer, filename)
     end
 
-    # not released yet
-    # def dump_text(filename)
-    #   check_result FFI.LGBM_DatasetDumpText(handle_pointer, filename)
-    # end
-
     def subset(used_indices, params: nil)
       # categorical_feature passed via params
       params ||= self.params
@@ -129,6 +105,70 @@ module LightGBM
 
     private
 
+    def construct
+      data = @data
+      used_indices = @used_indices
+
+      # TODO stringify params
+      params = @params || {}
+      if @categorical_feature != "auto" && @categorical_feature.any?
+        params["categorical_feature"] ||= @categorical_feature.join(",")
+      end
+      set_verbosity(params)
+
+      @handle = ::FFI::MemoryPointer.new(:pointer)
+      parameters = params_str(params)
+      reference = @reference.handle_pointer if @reference
+      if used_indices
+        used_row_indices = ::FFI::MemoryPointer.new(:int32, used_indices.count)
+        used_row_indices.write_array_of_int32(used_indices)
+        check_result FFI.LGBM_DatasetGetSubset(reference, used_row_indices, used_indices.count, parameters, @handle)
+      elsif data.is_a?(String)
+        check_result FFI.LGBM_DatasetCreateFromFile(data, parameters, reference, @handle)
+      else
+        if matrix?(data)
+          nrow = data.row_count
+          ncol = data.column_count
+          flat_data = data.to_a.flatten
+        elsif daru?(data)
+          nrow, ncol = data.shape
+          flat_data = data.map_rows(&:to_a).flatten
+        elsif numo?(data) || rover?(data)
+          data = data.to_numo if rover?(data)
+          nrow, ncol = data.shape
+        else
+          nrow = data.count
+          ncol = data.first.count
+          flat_data = data.flatten
+        end
+
+        c_data = ::FFI::MemoryPointer.new(:double, nrow * ncol)
+        if numo?(data)
+          c_data.write_bytes(data.cast_to(Numo::DFloat).to_string)
+        else
+          handle_missing(flat_data)
+          c_data.write_array_of_double(flat_data)
+        end
+
+        check_result FFI.LGBM_DatasetCreateFromMat(c_data, 1, nrow, ncol, 1, parameters, reference, @handle)
+      end
+      ObjectSpace.define_finalizer(self, self.class.finalize(handle_pointer)) unless used_indices
+
+      self.label = @label if @label
+      self.weight = @weight if @weight
+      self.group = @group if @group
+      self.feature_names = @feature_names if @feature_names
+    end
+
+    def free_handle
+      FFI.LGBM_DatasetFree(handle_pointer)
+      ObjectSpace.undefine_finalizer(self)
+    end
+
+    def dump_text(filename)
+      check_result FFI.LGBM_DatasetDumpText(handle_pointer, filename)
+    end
+
     def field(field_name)
       num_data = self.num_data
       out_len = ::FFI::MemoryPointer.new(:int)
@@ -142,11 +182,11 @@ module LightGBM
       data = data.to_a unless data.is_a?(Array)
       if type == :int32
         c_data = ::FFI::MemoryPointer.new(:int32, data.count)
-        c_data.put_array_of_int32(0, data)
+        c_data.write_array_of_int32(data)
         check_result FFI.LGBM_DatasetSetField(handle_pointer, field_name, c_data, data.count, 2)
       else
         c_data = ::FFI::MemoryPointer.new(:float, data.count)
-        c_data.put_array_of_float(0, data)
+        c_data.write_array_of_float(data)
         check_result FFI.LGBM_DatasetSetField(handle_pointer, field_name, c_data, data.count, 0)
       end
     end

data/lib/lightgbm/ffi.rb

@@ -5,11 +5,11 @@ module LightGBM
     begin
       ffi_lib LightGBM.ffi_lib
     rescue LoadError => e
-      raise e if ENV["LIGHTGBM_DEBUG"]
-      if e.message.include?("libomp")
-        raise LoadError, "Could not find OpenMP"
+      if e.message.include?("Library not loaded: /usr/local/opt/libomp/lib/libomp.dylib") && e.message.include?("Reason: image not found")
+        raise LoadError, "OpenMP not found. Run `brew install libomp`"
+      else
+        raise e
       end
-      raise LoadError, "Could not find LightGBM"
     end
 
     # https://github.com/microsoft/LightGBM/blob/master/include/LightGBM/c_api.h
@@ -23,10 +23,10 @@ module LightGBM
     attach_function :LGBM_DatasetCreateFromMat, %i[pointer int int32 int32 int string pointer pointer], :int
     attach_function :LGBM_DatasetGetSubset, %i[pointer pointer int32 string pointer], :int
     attach_function :LGBM_DatasetSetFeatureNames, %i[pointer pointer int], :int
-    attach_function :LGBM_DatasetGetFeatureNames, %i[pointer pointer pointer], :int
+    attach_function :LGBM_DatasetGetFeatureNames, %i[pointer int pointer size_t pointer pointer], :int
     attach_function :LGBM_DatasetFree, %i[pointer], :int
     attach_function :LGBM_DatasetSaveBinary, %i[pointer string], :int
-    # attach_function :LGBM_DatasetDumpText, %i[pointer string], :int
+    attach_function :LGBM_DatasetDumpText, %i[pointer string], :int
     attach_function :LGBM_DatasetSetField, %i[pointer string pointer int int], :int
     attach_function :LGBM_DatasetGetField, %i[pointer string pointer pointer pointer], :int
     attach_function :LGBM_DatasetGetNumData, %i[pointer pointer], :int
@@ -44,13 +44,13 @@ module LightGBM
     attach_function :LGBM_BoosterNumModelPerIteration, %i[pointer pointer], :int
     attach_function :LGBM_BoosterNumberOfTotalModel, %i[pointer pointer], :int
     attach_function :LGBM_BoosterGetEvalCounts, %i[pointer pointer], :int
-    attach_function :LGBM_BoosterGetEvalNames, %i[pointer pointer pointer], :int
+    attach_function :LGBM_BoosterGetEvalNames, %i[pointer int pointer size_t pointer pointer], :int
     attach_function :LGBM_BoosterGetNumFeature, %i[pointer pointer], :int
     attach_function :LGBM_BoosterGetEval, %i[pointer int pointer pointer], :int
-    attach_function :LGBM_BoosterPredictForMat, %i[pointer pointer int int32 int32 int int int string pointer pointer], :int
-    attach_function :LGBM_BoosterSaveModel, %i[pointer int int string], :int
-    attach_function :LGBM_BoosterSaveModelToString, %i[pointer int int int64 pointer pointer], :int
-    attach_function :LGBM_BoosterDumpModel, %i[pointer int int int64 pointer pointer], :int
+    attach_function :LGBM_BoosterPredictForMat, %i[pointer pointer int int32 int32 int int int int string pointer pointer], :int
+    attach_function :LGBM_BoosterSaveModel, %i[pointer int int int string], :int
+    attach_function :LGBM_BoosterSaveModelToString, %i[pointer int int int int64 pointer pointer], :int
+    attach_function :LGBM_BoosterDumpModel, %i[pointer int int int int64 pointer pointer], :int
     attach_function :LGBM_BoosterFeatureImportance, %i[pointer int int pointer], :int
   end
 end

data/lib/lightgbm/utils.rb

@@ -37,8 +37,12 @@ module LightGBM
       defined?(Daru::DataFrame) && data.is_a?(Daru::DataFrame)
     end
 
-    def narray?(data)
+    def numo?(data)
       defined?(Numo::NArray) && data.is_a?(Numo::NArray)
     end
+
+    def rover?(data)
+      defined?(Rover::DataFrame) && data.is_a?(Rover::DataFrame)
+    end
   end
 end

data/lib/lightgbm/version.rb

@@ -1,3 +1,3 @@
 module LightGBM
-  VERSION = "0.1.5"
+  VERSION = "0.2.0"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: lightgbm
 version: !ruby/object:Gem::Version
-  version: 0.1.5
+  version: 0.2.0
 platform: ruby
 authors:
 - Andrew Kane
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2019-09-03 00:00:00.000000000 Z
+date: 2020-09-01 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi
@@ -80,20 +80,6 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
-- !ruby/object:Gem::Dependency
-  name: numo-narray
-  requirement: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
-  type: :development
-  prerelease: false
-  version_requirements: !ruby/object:Gem::Requirement
-    requirements:
-    - - ">="
-      - !ruby/object:Gem::Version
-        version: '0'
 description: 
 email: andrew@chartkick.com
 executables: []
@@ -136,8 +122,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.0.3
+rubygems_version: 3.1.2
 signing_key: 
 specification_version: 4
-summary: LightGBM - the high performance machine learning library - for Ruby
+summary: High performance gradient boosting for Ruby
 test_files: []