IVGCVSW-3291 Add L2Normalization epsilon value to serialization

Signed-off-by: Ferran Balaguer <ferran.balaguer@arm.com>
Change-Id: Icfff3fb596a03c126a42b1d0c254a68e498df734

2 files changed, 143 insertions, 0 deletions

diff --git a/src/armnnDeserializer/Deserializer.cpp b/src/armnnDeserializer/Deserializer.cpp
index b23ed97b4c..90ca6d3f9a 100644
--- a/src/armnnDeserializer/Deserializer.cpp
+++ b/src/armnnDeserializer/Deserializer.cpp
@@ -1187,6 +1187,7 @@ void Deserializer::ParseL2Normalization(GraphPtr graph, unsigned int layerIndex)
     auto layerName = GetLayerName(graph, layerIndex);
     armnn::L2NormalizationDescriptor descriptor;
     descriptor.m_DataLayout = ToDataLayout(flatBufferDescriptor->dataLayout());
+    descriptor.m_Eps = flatBufferDescriptor->eps();
 
     IConnectableLayer* layer = m_Network->AddL2NormalizationLayer(descriptor, layerName.c_str());
     layer->GetOutputSlot(0).SetTensorInfo(outputInfo);
diff --git a/src/armnnDeserializer/test/DeserializeL2Normalization.cpp b/src/armnnDeserializer/test/DeserializeL2Normalization.cpp
new file mode 100644
index 0000000000..d8604a5cd5
--- /dev/null
+++ b/src/armnnDeserializer/test/DeserializeL2Normalization.cpp
@@ -0,0 +1,142 @@
+//
+// Copyright © 2017 Arm Ltd. All rights reserved.
+// SPDX-License-Identifier: MIT
+//
+
+#include <boost/test/unit_test.hpp>
+#include "ParserFlatbuffersSerializeFixture.hpp"
+#include "../Deserializer.hpp"
+
+#include <string>
+#include <iostream>
+
+BOOST_AUTO_TEST_SUITE(Deserializer)
+
+struct L2NormalizationFixture : public ParserFlatbuffersSerializeFixture
+{
+    explicit L2NormalizationFixture(const std::string &inputShape,
+                                    const std::string &outputShape,
+                                    const std::string &dataType,
+                                    const std::string &dataLayout,
+                                    const std::string epsilon)
+    {
+        m_JsonString = R"(
+    {
+        inputIds: [0],
+        outputIds: [2],
+        layers: [
+           {
+            layer_type: "InputLayer",
+            layer: {
+                base: {
+                    layerBindingId: 0,
+                    base: {
+                        index: 0,
+                        layerName: "InputLayer",
+                        layerType: "Input",
+                        inputSlots: [{
+                            index: 0,
+                            connection: {sourceLayerIndex:0, outputSlotIndex:0 },
+                            }],
+                        outputSlots: [{
+                            index: 0,
+                            tensorInfo: {
+                                dimensions: )" + inputShape + R"(,
+                                dataType: ")" + dataType + R"(",
+                                quantizationScale: 0.5,
+                                quantizationOffset: 0
+                                },
+                            }]
+                        },
+                    }
+                },
+            },
+        {
+        layer_type: "L2NormalizationLayer",
+        layer : {
+            base: {
+                index:1,
+                layerName: "L2NormalizationLayer",
+                layerType: "L2Normalization",
+                inputSlots: [{
+                        index: 0,
+                        connection: {sourceLayerIndex:0, outputSlotIndex:0 },
+                   }],
+                outputSlots: [{
+                    index: 0,
+                    tensorInfo: {
+                        dimensions: )" + outputShape + R"(,
+                        dataType: ")" + dataType + R"("
+                    },
+                    }],
+                },
+            descriptor: {
+                dataLayout: ")" + dataLayout + R"(",
+                eps: )" + epsilon + R"(
+                },
+            },
+        },
+        {
+        layer_type: "OutputLayer",
+        layer: {
+            base:{
+                layerBindingId: 0,
+                base: {
+                    index: 2,
+                    layerName: "OutputLayer",
+                    layerType: "Output",
+                    inputSlots: [{
+                        index: 0,
+                        connection: {sourceLayerIndex:1, outputSlotIndex:0 },
+                    }],
+                    outputSlots: [ {
+                        index: 0,
+                        tensorInfo: {
+                            dimensions: )" + outputShape + R"(,
+                            dataType: ")" + dataType + R"("
+                        },
+                    }],
+                }
+            }},
+        }]
+    }
+)";
+        Setup();
+    }
+};
+
+struct L2NormFixture : L2NormalizationFixture
+{
+    // Using a non standard epsilon value of 1e-8
+    L2NormFixture():L2NormalizationFixture("[ 1, 3, 1, 1 ]",
+                                           "[ 1, 3, 1, 1 ]",
+                                           "Float32",
+                                           "NCHW",
+                                           "0.00000001"){}
+};
+
+BOOST_FIXTURE_TEST_CASE(L2NormalizationFloat32, L2NormFixture)
+{
+    // 1 / sqrt(1^2 + 2^2 + 3^2)
+    const float approxInvL2Norm = 0.267261f;
+
+    RunTest<4, armnn::DataType::Float32>(0,
+                                         {{"InputLayer", { 1.0f, 2.0f, 3.0f }}},
+                                         {{"OutputLayer",{ 1.0f * approxInvL2Norm,
+                                                           2.0f * approxInvL2Norm,
+                                                           3.0f * approxInvL2Norm }}});
+}
+
+BOOST_FIXTURE_TEST_CASE(L2NormalizationEpsilonLimitFloat32, L2NormFixture)
+{
+    // 1 / sqrt(1e-8)
+    const float approxInvL2Norm = 10000;
+
+    RunTest<4, armnn::DataType::Float32>(0,
+                                         {{"InputLayer", { 0.00000001f, 0.00000002f, 0.00000003f }}},
+                                         {{"OutputLayer",{ 0.00000001f * approxInvL2Norm,
+                                                           0.00000002f * approxInvL2Norm,
+                                                           0.00000003f * approxInvL2Norm }}});
+}
+
+BOOST_AUTO_TEST_SUITE_END()