ruby-dnn 0.13.4 → 0.14.0

data/lib/dnn/core/cnn_layers.rb

@@ -83,7 +83,6 @@ module DNN
       end
     end
 
-
     class Conv2D < Connection
       include Conv2DUtils
 
@@ -114,7 +113,7 @@ module DNN
 
       def build(input_shape)
         unless input_shape.length == 3
-          raise DNN_ShapeError.new("Input shape is #{input_shape}. But input shape must be 3 dimensional.")
+          raise DNN_ShapeError, "Input shape is #{input_shape}. But input shape must be 3 dimensional."
         end
         super
         prev_h, prev_w, num_prev_filters = *input_shape
@@ -122,12 +121,12 @@ module DNN
         @bias.data = Xumo::SFloat.new(@num_filters) if @bias
         init_weight_and_bias
         @pad_size = if @padding == true
-          calc_conv2d_padding_size(prev_h, prev_w, *@filter_size, @strides)
-        elsif @padding.is_a?(Array)
-          @padding
-        else
-          [0, 0]
-        end
+                      calc_conv2d_padding_size(prev_h, prev_w, *@filter_size, @strides)
+                    elsif @padding.is_a?(Array)
+                      @padding
+                    else
+                      [0, 0]
+                    end
         @out_size = calc_conv2d_out_size(prev_h, prev_w, *@filter_size, *@pad_size, @strides)
       end
 
@@ -186,7 +185,6 @@ module DNN
       end
     end
 
-
     class Conv2DTranspose < Connection
       include Conv2DUtils
 
@@ -217,7 +215,7 @@ module DNN
 
       def build(input_shape)
         unless input_shape.length == 3
-          raise DNN_ShapeError.new("Input shape is #{input_shape}. But input shape must be 3 dimensional.")
+          raise DNN_ShapeError, "Input shape is #{input_shape}. But input shape must be 3 dimensional."
         end
         super
         prev_h, prev_w, num_prev_filters = *input_shape
@@ -225,12 +223,12 @@ module DNN
         @bias.data = Xumo::SFloat.new(@num_filters) if @bias
         init_weight_and_bias
         @pad_size = if @padding == true
-          calc_conv2d_transpose_padding_size(prev_h, prev_w, *@filter_size, @strides)
-        elsif @padding.is_a?(Array)
-          @padding
-        else
-          [0, 0]
-        end
+                      calc_conv2d_transpose_padding_size(prev_h, prev_w, *@filter_size, @strides)
+                    elsif @padding.is_a?(Array)
+                      @padding
+                    else
+                      [0, 0]
+                    end
         @out_size = calc_conv2d_transpose_out_size(prev_h, prev_w, *@filter_size, *@pad_size, @strides)
       end
 
@@ -291,7 +289,6 @@ module DNN
       end
     end
 
-
     # Super class of all pooling2D class.
     class Pool2D < Layer
       include Conv2DUtils
@@ -308,27 +305,27 @@ module DNN
         super()
         @pool_size = pool_size.is_a?(Integer) ? [pool_size, pool_size] : pool_size
         @strides = if strides
-          strides.is_a?(Integer) ? [strides, strides] : strides
-        else
-          @pool_size.clone
-        end
+                     strides.is_a?(Integer) ? [strides, strides] : strides
+                   else
+                     @pool_size.clone
+                   end
         @padding = padding.is_a?(Integer) ? [padding, padding] : padding
       end
 
       def build(input_shape)
         unless input_shape.length == 3
-          raise DNN_ShapeError.new("Input shape is #{input_shape}. But input shape must be 3 dimensional.")
+          raise DNN_ShapeError, "Input shape is #{input_shape}. But input shape must be 3 dimensional."
         end
         super
         prev_h, prev_w = input_shape[0..1]
         @num_channel = input_shape[2]
         @pad_size = if @padding == true
-          calc_conv2d_padding_size(prev_h, prev_w, *@pool_size, @strides)
-        elsif @padding.is_a?(Array)
-          @padding
-        else
-          [0, 0]
-        end
+                      calc_conv2d_padding_size(prev_h, prev_w, *@pool_size, @strides)
+                    elsif @padding.is_a?(Array)
+                      @padding
+                    else
+                      [0, 0]
+                    end
         @out_size = calc_conv2d_out_size(prev_h, prev_w, *@pool_size, *@pad_size, @strides)
       end
 
@@ -345,10 +342,8 @@ module DNN
       def load_hash(hash)
         initialize(hash[:pool_size], strides: hash[:strides], padding: hash[:padding])
       end
-
     end
 
-
     class MaxPool2D < Pool2D
       def forward(x)
         x = zero_padding(x, @pad_size) if @padding
@@ -368,7 +363,6 @@ module DNN
       end
     end
 
-
     class AvgPool2D < Pool2D
       def forward(x)
         x = zero_padding(x, @pad_size) if @padding
@@ -391,7 +385,6 @@ module DNN
       end
     end
 
-
     class UnPool2D < Layer
       include Conv2DUtils
 
@@ -405,7 +398,7 @@ module DNN
 
       def build(input_shape)
         unless input_shape.length == 3
-          raise DNN_ShapeError.new("Input shape is #{input_shape}. But input shape must be 3 dimensional.")
+          raise DNN_ShapeError, "Input shape is #{input_shape}. But input shape must be 3 dimensional."
         end
         super
         prev_h, prev_w = input_shape[0..1]

data/lib/dnn/core/embedding.rb

@@ -21,9 +21,9 @@ module DNN
         @weight_regularizer = weight_regularizer
       end
 
-      def call(input)
+      def call(input_tensor)
         build unless built?
-        [forward(input), Link.new(nil, self)]
+        Tensor.new(forward(input_tensor.data), Link.new(nil, self))
       end
 
       def build
@@ -56,6 +56,24 @@ module DNN
         @weight_regularizer ? [@weight_regularizer] : []
       end
 
+      def to_proc
+        method(:call).to_proc
+      end
+
+      def >>(layer)
+        if RUBY_VERSION < "2.6.0"
+          raise DNN_Error, "Function composition is not supported before ruby version 2.6.0."
+        end
+        to_proc >> layer
+      end
+
+      def <<(layer)
+        if RUBY_VERSION < "2.6.0"
+          raise DNN_Error, "Function composition is not supported before ruby version 2.6.0."
+        end
+        to_proc << layer
+      end
+
       def to_hash
         super(input_shape: @input_shape, input_length: @input_length,
               weight_initializer: @weight_initializer.to_hash, weight_regularizer: @weight_regularizer&.to_hash)

data/lib/dnn/core/error.rb

@@ -2,6 +2,4 @@ module DNN
   class DNN_Error < StandardError; end
 
   class DNN_ShapeError < DNN_Error; end
-
-  class DNN_UnknownEventError < DNN_Error; end
 end

data/lib/dnn/core/initializers.rb

@@ -6,7 +6,7 @@ module DNN
         return nil unless hash
         initializer_class = DNN.const_get(hash[:class])
         initializer = initializer_class.allocate
-        raise DNN_Error.new("#{initializer.class} is not an instance of #{self} class.") unless initializer.is_a?(self)
+        raise DNN_Error, "#{initializer.class} is not an instance of #{self} class." unless initializer.is_a?(self)
         initializer.load_hash(hash)
         initializer
       end
@@ -21,7 +21,7 @@ module DNN
       # @param [DNN::Layers::Layer] layer Layer that owns learning parameters.
       # @param [DNN::Param] param Learning parameter to be initialized.
       def init_param(layer, param)
-        raise NotImplementedError.new("Class '#{self.class.name}' has implement method 'init_param'")
+        raise NotImplementedError, "Class '#{self.class.name}' has implement method 'init_param'"
       end
 
       def to_hash(merge_hash = nil)
@@ -35,14 +35,12 @@ module DNN
       end
     end
 
-
     class Zeros < Initializer
       def init_param(layer, param)
         param.data = param.data.fill(0)
       end
     end
 
-
     class Const < Initializer
       attr_reader :const
 
@@ -65,7 +63,6 @@ module DNN
       end
     end
 
-
     class RandomNormal < Initializer
       attr_reader :mean
       attr_reader :std
@@ -92,7 +89,6 @@ module DNN
       end
     end
 
-
     class RandomUniform < Initializer
       attr_reader :min
       attr_reader :max
@@ -119,7 +115,6 @@ module DNN
       end
     end
 
-
     class Xavier < Initializer
       def initialize(seed: true)
         super
@@ -132,7 +127,6 @@ module DNN
       end
     end
 
-
     class He < Initializer
       def initialize(seed: true)
         super

data/lib/dnn/core/iterator.rb

@@ -1,13 +1,18 @@
 module DNN
   # This class manages input datas and output datas together.
   class Iterator
+    attr_reader :num_datas
+    attr_reader :last_round_down
+
     # @param [Numo::SFloat] x_datas input datas.
     # @param [Numo::SFloat] y_datas output datas.
     # @param [Boolean] random Set true to return batches randomly. Setting false returns batches in order of index.
-    def initialize(x_datas, y_datas, random: true)
+    # @param [Boolean] last_round_down Set true to round down for last batch data when call foreach.
+    def initialize(x_datas, y_datas, random: true, last_round_down: false)
       @x_datas = x_datas
       @y_datas = y_datas
       @random = random
+      @last_round_down = last_round_down
       @num_datas = x_datas.is_a?(Array) ? x_datas[0].shape[0] : x_datas.shape[0]
       reset
     end
@@ -15,7 +20,7 @@ module DNN
     # Return the next batch.
     # @param [Integer] batch_size Required batch size.
     def next_batch(batch_size)
-      raise DNN_Error.new("This iterator has not next batch. Please call reset.") unless has_next?
+      raise DNN_Error, "This iterator has not next batch. Please call reset." unless has_next?
       if @indexes.length <= batch_size
         batch_indexes = @indexes
         @has_next = false
@@ -23,15 +28,15 @@ module DNN
         batch_indexes = @indexes.shift(batch_size)
       end
       x_batch = if @x_datas.is_a?(Array)
-        @x_datas.map { |datas| datas[batch_indexes, false] }
-      else
-        @x_datas[batch_indexes, false]
-      end
+                  @x_datas.map { |datas| datas[batch_indexes, false] }
+                else
+                  @x_datas[batch_indexes, false]
+                end
       y_batch = if @y_datas.is_a?(Array)
-        @y_datas.map { |datas| datas[batch_indexes, false] }
-      else
-        @y_datas[batch_indexes, false]
-      end
+                  @y_datas.map { |datas| datas[batch_indexes, false] }
+                else
+                  @y_datas[batch_indexes, false]
+                end
       [x_batch, y_batch]
     end
 
@@ -48,11 +53,10 @@ module DNN
     end
 
     def foreach(batch_size, &block)
-      step = 0
-      while has_next?
+      steps = @last_round_down ? @num_datas / batch_size : (@num_datas.to_f / batch_size).ceil
+      steps.times do |step|
         x_batch, y_batch = next_batch(batch_size)
         block.call(x_batch, y_batch, step)
-        step += 1
       end
       reset
     end

data/lib/dnn/core/layers.rb

@@ -7,15 +7,16 @@ module DNN
       attr_reader :input_shape
 
       def self.call(x, *args)
-        self.new(*args).(x)
+        new(*args).(x)
       end
 
       def self.from_hash(hash)
         return nil unless hash
         layer_class = DNN.const_get(hash[:class])
         layer = layer_class.allocate
-        raise DNN_Error.new("#{layer.class} is not an instance of #{self} class.") unless layer.is_a?(self)
+        raise DNN_Error, "#{layer.class} is not an instance of #{self} class." unless layer.is_a?(self)
         layer.load_hash(hash)
+        layer.name = hash[:name]&.to_sym
         layer
       end
 
@@ -25,13 +26,15 @@ module DNN
       end
 
       # Forward propagation and create a link.
-      # @param [Array] input Array of the form [x_input_data, prev_link].
-      def call(input)
-        x, prev_link = *input
+      # @param [Tensor] input_tensor Input tensor.
+      # @return [Tensor] Output tensor.
+      def call(input_tensor)
+        x = input_tensor.data
+        prev_link = input_tensor.link
         build(x.shape[1..-1]) unless built?
         y = forward(x)
         link = Link.new(prev_link, self)
-        [y, link]
+        Tensor.new(y, link)
       end
 
       # Build the layer.
@@ -49,13 +52,13 @@ module DNN
       # Forward propagation.
       # @param [Numo::SFloat] x Input data.
       def forward(x)
-        raise NotImplementedError.new("Class '#{self.class.name}' has implement method 'forward'")
+        raise NotImplementedError, "Class '#{self.class.name}' has implement method 'forward'"
       end
 
       # Backward propagation.
       # @param [Numo::SFloat] dy Differential value of output data.
       def backward(dy)
-        raise NotImplementedError.new("Class '#{self.class.name}' has implement method 'backward'")
+        raise NotImplementedError, "Class '#{self.class.name}' has implement method 'backward'"
       end
 
       # Please reimplement this method as needed.
@@ -77,7 +80,6 @@ module DNN
       end
     end
 
-
     # This class is a superclass of all classes with learning parameters.
     class HasParamLayer < Layer
       # @return [Boolean] Setting false prevents learning of parameters.
@@ -90,15 +92,13 @@ module DNN
 
       # @return [Array] The parameters of the layer.
       def get_params
-        raise NotImplementedError.new("Class '#{self.class.name}' has implement method 'get_params'")
+        raise NotImplementedError, "Class '#{self.class.name}' has implement method 'get_params'"
       end
     end
 
-
     class InputLayer < Layer
-      def self.call(input)
-        shape = input.is_a?(Array) ? input[0].shape : input.shape
-        self.new(shape[1..-1]).(input)
+      def self.call(input_tensor)
+        new(input_tensor.data.shape[1..-1]).(input_tensor)
       end
 
       # @param [Array] input_dim_or_shape Setting the shape or dimension of the input data.
@@ -107,16 +107,9 @@ module DNN
         @input_shape = input_dim_or_shape.is_a?(Array) ? input_dim_or_shape : [input_dim_or_shape]
       end
 
-      def call(input)
+      def call(input_tensor)
         build unless built?
-        if input.is_a?(Array)
-          x, prev_link = *input
-        else
-          x = input
-          prev_link = nil
-        end
-        link = prev_link ? Link.new(prev_link, self) : Link.new(nil, self)
-        [forward(x), link]
+        Tensor.new(forward(input_tensor.data), Link.new(input_tensor&.link, self))
       end
 
       def build
@@ -125,7 +118,7 @@ module DNN
 
       def forward(x)
         unless x.shape[1..-1] == @input_shape
-          raise DNN_ShapeError.new("The shape of x does not match the input shape. input shape is #{@input_shape}, but x shape is #{x.shape[1..-1]}.")
+          raise DNN_ShapeError, "The shape of x does not match the input shape. input shape is #{@input_shape}, but x shape is #{x.shape[1..-1]}."
         end
         x
       end
@@ -134,6 +127,24 @@ module DNN
         dy
       end
 
+      def to_proc
+        method(:call).to_proc
+      end
+
+      def >>(layer)
+        if RUBY_VERSION < "2.6.0"
+          raise DNN_Error, "Function composition is not supported before ruby version 2.6.0."
+        end
+        to_proc >> layer
+      end
+
+      def <<(layer)
+        if RUBY_VERSION < "2.6.0"
+          raise DNN_Error, "Function composition is not supported before ruby version 2.6.0."
+        end
+        to_proc << layer
+      end
+
       def to_hash
         super(input_shape: @input_shape)
       end
@@ -143,7 +154,6 @@ module DNN
       end
     end
 
-
     # It is a superclass of all connection layers.
     class Connection < HasParamLayer
       attr_reader :weight
@@ -206,7 +216,6 @@ module DNN
       end
     end
 
-
     class Dense < Connection
       attr_reader :num_nodes
 
@@ -224,7 +233,7 @@ module DNN
 
       def build(input_shape)
         unless input_shape.length == 1
-          raise DNN_ShapeError.new("Input shape is #{input_shape}. But input shape must be 1 dimensional.")
+          raise DNN_ShapeError, "Input shape is #{input_shape}. But input shape must be 1 dimensional."
         end
         super
         num_prev_nodes = input_shape[0]
@@ -266,7 +275,6 @@ module DNN
       end
     end
 
-
     class Flatten < Layer
       def forward(x)
         x.reshape(x.shape[0], *output_shape)
@@ -281,8 +289,9 @@ module DNN
       end
     end
 
-
     class Reshape < Layer
+      attr_reader :output_shape
+
       def initialize(output_shape)
         super()
         @output_shape = output_shape
@@ -296,10 +305,6 @@ module DNN
         dy.reshape(dy.shape[0], *@input_shape)
       end
 
-      def output_shape
-        @output_shape
-      end
-
       def to_hash
         super(output_shape: @output_shape)
       end
@@ -309,7 +314,6 @@ module DNN
       end
     end
 
-
     class Dropout < Layer
       attr_accessor :dropout_ratio
       attr_reader :use_scale