[Layer] add "div layer"

- added "div layer" for division.

**Self evaluation:**
1. Build test:   [X]Passed [ ]Failed [ ]Skipped
2. Run test:     [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Seungbaek Hong <[email protected]>

api/ccapi/include/layer.h

@@ -40,6 +40,7 @@ enum LayerType {
   LAYER_ADD = ML_TRAIN_LAYER_TYPE_ADD,       /**< Add Layer type */
   LAYER_SUB = ML_TRAIN_LAYER_TYPE_SUB,       /**< Subtract Layer type */
   LAYER_MUL = ML_TRAIN_LAYER_TYPE_MUL,       /**< Multiply Layer type */
+  LAYER_DIV = ML_TRAIN_LAYER_TYPE_DIV,       /**< Divide Layer type */
   LAYER_FC = ML_TRAIN_LAYER_TYPE_FC,         /**< Fully Connected Layer type */
   LAYER_SWIGLU = ML_TRAIN_LAYER_TYPE_SWIGLU, /**< Swiglu Layer type */
   LAYER_BN = ML_TRAIN_LAYER_TYPE_BN, /**< Batch Normalization Layer type */
@@ -326,6 +327,14 @@ MulLayer(const std::vector<std::string> &properties = {}) {
   return createLayer(LayerType::LAYER_MUL, properties);
 }
 
+/**
+ * @brief Helper function to create div layer
+ */
+inline std::unique_ptr<Layer>
+DivLayer(const std::vector<std::string> &properties = {}) {
+  return createLayer(LayerType::LAYER_DIV, properties);
+}
+
 /**
  * @brief Helper function to create fully connected layer
  */

api/nntrainer-api-common.h

@@ -68,6 +68,7 @@ typedef enum {
   ML_TRAIN_LAYER_TYPE_ADD = 32,      /**< Add Layer type (Since 9.0)*/
   ML_TRAIN_LAYER_TYPE_SUB = 33,      /**< Sub Layer type (Since 9.0)*/
   ML_TRAIN_LAYER_TYPE_MUL = 34,      /**< Mul Layer type (Since 9.0)*/
+  ML_TRAIN_LAYER_TYPE_DIV = 35,      /**< Div Layer type (Since 9.0)*/
   ML_TRAIN_LAYER_TYPE_PREPROCESS_FLIP =
     300, /**< Preprocess flip Layer (Since 6.5) */
   ML_TRAIN_LAYER_TYPE_PREPROCESS_TRANSLATE =

nntrainer/app_context.cpp

@@ -42,6 +42,7 @@
 #include <conv2d_layer.h>
 #include <cross_entropy_sigmoid_loss_layer.h>
 #include <cross_entropy_softmax_loss_layer.h>
+#include <div_layer.h>
 #include <dropout.h>
 #include <embedding.h>
 #include <fc_layer.h>
@@ -257,6 +258,8 @@ static void add_default_object(AppContext &ac) {
                      LayerType::LAYER_SUB);
   ac.registerFactory(nntrainer::createLayer<MulLayer>, MulLayer::type,
                      LayerType::LAYER_MUL);
+  ac.registerFactory(nntrainer::createLayer<DivLayer>, DivLayer::type,
+                     LayerType::LAYER_DIV);
   ac.registerFactory(nntrainer::createLayer<FullyConnectedLayer>,
                      FullyConnectedLayer::type, LayerType::LAYER_FC);
   ac.registerFactory(nntrainer::createLayer<BatchNormalizationLayer>,

nntrainer/layers/div_layer.cpp

@@ -0,0 +1,97 @@
+// SPDX-License-Identifier: Apache-2.0
+/**
+ * Copyright (C) 2024 SeungBaek Hong <[email protected]>
+ *
+ * @file   div_layer.cpp
+ * @date   30 August 2024
+ * @see    https://github.com/nnstreamer/nntrainer
+ * @author SeungBaek Hong <[email protected]>
+ * @bug    No known bugs except for NYI items
+ * @brief  This is div layer class (operation layer)
+ *
+ */
+
+#include <div_layer.h>
+#include <nntrainer_error.h>
+#include <nntrainer_log.h>
+#include <node_exporter.h>
+#include <util_func.h>
+
+#include <layer_context.h>
+
+namespace nntrainer {
+
+static constexpr size_t SINGLE_INOUT_IDX = 0;
+
+void DivLayer::finalize(InitLayerContext &context) {
+  context.setOutputDimensions({context.getInputDimensions()[0]});
+}
+
+void DivLayer::forwarding(RunLayerContext &context, bool training) {
+  Tensor &hidden_ = context.getOutput(SINGLE_INOUT_IDX);
+
+  const Tensor &input0 = context.getInput(0);
+  const Tensor &input1 = context.getInput(1);
+
+  input0.divide(input1, hidden_);
+}
+
+void DivLayer::incremental_forwarding(RunLayerContext &context,
+                                      unsigned int from, unsigned int to,
+                                      bool training) {
+  Tensor &hidden_ = context.getOutput(SINGLE_INOUT_IDX);
+  TensorDim hidden_dim = hidden_.getDim();
+  TensorDim hidden_step_dim = hidden_dim;
+
+  if (from) {
+    NNTR_THROW_IF(to - from != 1, std::invalid_argument)
+      << "incremental step size is not 1";
+    from = 0;
+    to = 1;
+  }
+
+  hidden_step_dim.batch(1);
+  hidden_step_dim.height(to - from);
+
+  for (unsigned int b = 0; b < hidden_.batch(); ++b) {
+    Tensor hidden_step = hidden_.getSharedDataTensor(
+      hidden_step_dim, b * hidden_dim.getFeatureLen(), true);
+
+    const Tensor &input0 = context.getInput(0);
+    const Tensor &input1 = context.getInput(1);
+
+    TensorDim input_dim = input0.getDim();
+    TensorDim input_step_dim = input_dim;
+    input_step_dim.batch(1);
+    input_step_dim.height(to - from);
+
+    Tensor input0_step = input0.getSharedDataTensor(
+      input_step_dim, b * input_dim.getFeatureLen(), true);
+
+    Tensor input1_step = input1.getSharedDataTensor(
+      input_step_dim, b * input_dim.getFeatureLen(), true);
+
+    input0_step.divide(input1_step, hidden_step);
+  }
+}
+
+void DivLayer::calcDerivative(RunLayerContext &context) {
+  context.getOutgoingDerivative(0).copy(
+    context.getIncomingDerivative(SINGLE_INOUT_IDX)
+      .divide(context.getInput(1)));
+
+  context.getOutgoingDerivative(1).copy(
+    context.getIncomingDerivative(SINGLE_INOUT_IDX)
+      .multiply(context.getInput(0).multiply(-1))
+      .divide(context.getInput(1).pow(2)));
+}
+
+void DivLayer::setProperty(const std::vector<std::string> &values) {
+  auto remain_props = loadProperties(values, div_props);
+  if (!remain_props.empty()) {
+    std::string msg = "[DivLayer] Unknown Layer Properties count " +
+                      std::to_string(values.size());
+    throw exception::not_supported(msg);
+  }
+}
+} /* namespace nntrainer */

nntrainer/layers/div_layer.h

@@ -0,0 +1,103 @@
+// SPDX-License-Identifier: Apache-2.0
+/**
+ * Copyright (C) 2024 SeungBaek Hong <[email protected]>
+ *
+ * @file   div_layer.h
+ * @date   30 August 2024
+ * @see    https://github.com/nnstreamer/nntrainer
+ * @author SeungBaek Hong <[email protected]>
+ * @bug    No known bugs except for NYI items
+ * @brief  This is div layer class (operation layer)
+ *
+ */
+
+#ifndef __DIV_LAYER_H__
+#define __DIV_LAYER_H__
+#ifdef __cplusplus
+
+#include <common_properties.h>
+#include <layer_devel.h>
+
+namespace nntrainer {
+
+/**
+ * @class Div Layer
+ * @brief Div Layer
+ */
+class DivLayer : public Layer {
+public:
+  /**
+   * @brief Constructor of Div Layer
+   */
+  DivLayer() : Layer(), div_props(props::Print()) {}
+
+  /**
+   * @brief Destructor of Div Layer
+   */
+  ~DivLayer(){};
+
+  /**
+   *  @brief  Move constructor of Div Layer.
+   *  @param[in] DivLayer &&
+   */
+  DivLayer(DivLayer &&rhs) noexcept = default;
+
+  /**
+   * @brief  Move assignment operator.
+   * @parma[in] rhs DivLayer to be moved.
+   */
+  DivLayer &operator=(DivLayer &&rhs) = default;
+
+  /**
+   * @copydoc Layer::finalize(InitLayerContext &context)
+   */
+  void finalize(InitLayerContext &context) override;
+
+  /**
+   * @copydoc Layer::forwarding(RunLayerContext &context, bool training)
+   */
+  void forwarding(RunLayerContext &context, bool training) override;
+
+  /**
+   * @copydoc Layer::incremental_forwarding(RunLayerContext &context, unsigned
+   * int from, unsigned int to, bool training)
+   */
+  void incremental_forwarding(RunLayerContext &context, unsigned int from,
+                              unsigned int to, bool training) override;
+
+  /**
+   * @copydoc Layer::calcDerivative(RunLayerContext &context)
+   */
+  void calcDerivative(RunLayerContext &context) override;
+
+  /**
+   * @copydoc bool supportBackwarding() const
+   */
+  bool supportBackwarding() const override { return true; };
+
+  /**
+   * @copydoc Layer::exportTo(Exporter &exporter, ml::train::ExportMethods
+   * method)
+   */
+  void exportTo(Exporter &exporter,
+                const ml::train::ExportMethods &method) const override {}
+
+  /**
+   * @copydoc Layer::setProperty(const std::vector<std::string> &values)
+   */
+  void setProperty(const std::vector<std::string> &values) override;
+
+  /**
+   * @copydoc Layer::getType()
+   */
+  const std::string getType() const override { return DivLayer::type; };
+
+  std::tuple<props::Print> div_props;
+
+  inline static const std::string type = "div";
+};
+
+} // namespace nntrainer
+
+#endif /* __cplusplus */
+#endif /* __DIV_LAYER_H__ */

nntrainer/layers/meson.build

@@ -8,6 +8,7 @@ layer_sources = [
   'add_layer.cpp',
   'sub_layer.cpp',
   'mul_layer.cpp',
+  'div_layer.cpp',
   'addition_layer.cpp',
   'attention_layer.cpp',
   'mol_attention_layer.cpp',

test/ccapi/unittest_ccapi.cpp

@@ -73,6 +73,9 @@ TEST(ccapi_layer, construct_02_p) {
   EXPECT_NO_THROW(layer = ml::train::layer::MulLayer());
   EXPECT_EQ(layer->getType(), "mul");
 
+  EXPECT_NO_THROW(layer = ml::train::layer::DivLayer());
+  EXPECT_EQ(layer->getType(), "div");
+
   EXPECT_NO_THROW(layer = ml::train::layer::FullyConnected());
   EXPECT_EQ(layer->getType(), "fully_connected");
 

test/input_gen/genModelTests_v2.py

@@ -468,6 +468,19 @@ def forward(self, inputs, labels):
         return out, loss
 
 
+class DivOperation(torch.nn.Module):
+    def __init__(self):
+        super().__init__()
+        self.fc = torch.nn.Linear(2, 2)
+        self.loss = torch.nn.MSELoss()
+
+    def forward(self, inputs, labels):
+        out = self.fc(inputs[0])
+        out = inputs[0] / out
+        loss = self.loss(out, labels[0])
+        return out, loss
+
+
 if __name__ == "__main__":
     record_v2(
         ReduceMeanLast(),
@@ -772,5 +785,15 @@ def forward(self, inputs, labels):
         name="mul_operation",
     )
 
+    div_operation = DivOperation()
+    record_v2(
+        div_operation,
+        iteration=2,
+        input_dims=[(1, 2)],
+        input_dtype=[float],
+        label_dims=[(1, 2)],
+        name="div_operation",
+    )
+
     # Function to check the created golden test file
-    inspect_file("sub_operation.nnmodelgolden")
+    inspect_file("div_operation.nnmodelgolden")

test/unittest/layers/meson.build

@@ -50,6 +50,7 @@ test_target = [
   'unittest_layers_add.cpp',
   'unittest_layers_sub.cpp',
   'unittest_layers_mul.cpp',
+  'unittest_layers_div.cpp',
   'unittest_layers_multiout.cpp',
   'unittest_layers_rnn.cpp',
   'unittest_layers_rnncell.cpp',

test/unittest/layers/unittest_layers_div.cpp

@@ -0,0 +1,28 @@
+// SPDX-License-Identifier: Apache-2.0
+/**
+ * Copyright (C) 2024 SeungBaek Hong <[email protected]>
+ *
+ * @file unittest_layers_div.cpp
+ * @date 30 August 2024
+ * @brief Div Layer Test
+ * @see	https://github.com/nnstreamer/nntrainer
+ * @author SeungBaek Hong <[email protected]>
+ * @bug No known bugs except for NYI items
+ */
+#include <tuple>
+
+#include <gtest/gtest.h>
+
+#include <div_layer.h>
+#include <layers_common_tests.h>
+
+auto semantic_div = LayerSemanticsParamType(
+  nntrainer::createLayer<nntrainer::DivLayer>, nntrainer::DivLayer::type, {},
+  LayerCreateSetPropertyOptions::AVAILABLE_FROM_APP_CONTEXT, false, 1);
+
+auto semantic_div_multi = LayerSemanticsParamType(
+  nntrainer::createLayer<nntrainer::DivLayer>, nntrainer::DivLayer::type, {},
+  LayerCreateSetPropertyOptions::AVAILABLE_FROM_APP_CONTEXT, false, 2);
+
+GTEST_PARAMETER_TEST(Div, LayerSemantics,
+                     ::testing::Values(semantic_div, semantic_div_multi));

test/unittest/models/unittest_models.cpp

@@ -929,6 +929,25 @@ static std::unique_ptr<NeuralNetwork> makeMulOperation() {
   return nn;
 }
 
+static std::unique_ptr<NeuralNetwork> makeDivOperation() {
+  std::unique_ptr<NeuralNetwork> nn(new NeuralNetwork());
+
+  auto outer_graph =
+    makeGraph({{"input", {"name=in", "input_shape=1:1:2"}},
+               {"fully_connected", {"name=fc", "unit=2", "input_layers=in"}},
+               {"div", {"name=div_layer", "input_layers=in,fc"}},
+               {"mse", {"name=loss", "input_layers=div_layer"}}});
+
+  for (auto &node : outer_graph) {
+    nn->addLayer(node);
+  }
+
+  nn->setProperty({"batch_size=1"});
+  nn->setOptimizer(ml::train::createOptimizer("sgd", {"learning_rate=0.1"}));
+
+  return nn;
+}
+
 GTEST_PARAMETER_TEST(
   model, nntrainerModelTest,
   ::testing::ValuesIn({
@@ -995,6 +1014,7 @@ GTEST_PARAMETER_TEST(
     mkModelTc_V2(makeAddOperation, "add_operation", ModelTestOption::ALL_V2),
     mkModelTc_V2(makeSubOperation, "sub_operation", ModelTestOption::ALL_V2),
     mkModelTc_V2(makeMulOperation, "mul_operation", ModelTestOption::ALL_V2),
+    mkModelTc_V2(makeDivOperation, "div_operation", ModelTestOption::ALL_V2),
   }),
   [](const testing::TestParamInfo<nntrainerModelTest::ParamType> &info)
     -> const auto & { return std::get<1>(info.param); });