GitHub #543 Problem Parsing Mixed-Precision Model

 * Fixed bug when converting Constants with Per-Axis Quantization

Signed-off-by: Mike Kelly <mike.kelly@arm.com>
Change-Id: Ifbea23e60483746ec987da491dae96e74cb33af4

3 files changed, 59 insertions, 55 deletions

diff --git a/src/armnnTfLiteParser/TfLiteParser.cpp b/src/armnnTfLiteParser/TfLiteParser.cpp
index 0484c6f478..191cfd2b48 100644
--- a/src/armnnTfLiteParser/TfLiteParser.cpp
+++ b/src/armnnTfLiteParser/TfLiteParser.cpp
@@ -316,6 +316,14 @@ std::vector<unsigned int> GetUIntBuffer(armnn::TensorInfo info,
         ::memcpy(uint64Buffer.data(), bufferPtr->data.data(), bufferPtr->data.size());
         buffer.assign(std::begin(uint64Buffer), std::end(uint64Buffer));
     }
+    else
+    {
+        CheckLocation location = CHECK_LOCATION();
+        throw ParseException(
+                fmt::format("Unsupported data type for uint buffer {}, only Signed 32 or Signed 64 are supported. {}",
+                            GetDataTypeName(info.GetDataType()),
+                            location.AsString()));
+    }
     return buffer;
 }
 
@@ -911,42 +919,16 @@ INetworkPtr TfLiteParserImpl::CreateNetworkFromModel()
     return std::move(m_Network);
 }
 
-std::unique_ptr<float[]> AsFloatArray(TfLiteParserImpl::BufferRawPtr bufferPtr,
-                                      const TensorInfo& tensorInfo)
+bool TfLiteParserImpl::ShouldConstantTensorBeConverted(TfLiteParserImpl::TensorRawPtr tensorPtr,
+                                                       armnn::DataType inputDataType,
+                                                       armnn::DataType tensorDataType)
 {
-    if (tensorInfo.GetDataType() == DataType::QAsymmS8 || tensorInfo.GetDataType() == DataType::QSymmS8 ||
-        tensorInfo.GetDataType() == DataType::QAsymmU8)
-    {
-        std::unique_ptr<float[]> buffer(new float[tensorInfo.GetNumElements()]);
-
-        if (tensorInfo.HasPerAxisQuantization())
-        {
-            unsigned int axis = tensorInfo.GetQuantizationDim().value();
-            auto axisDimensionality = tensorInfo.GetShape()[axis];
-            auto axisFactor = armnnUtils::GetNumElementsAfter(tensorInfo.GetShape(), axis);
-
-            for (unsigned int i = 0; i < tensorInfo.GetNumDimensions(); ++i)
-            {
-                unsigned int axisIndex = (i / axisFactor) % axisDimensionality;
-                buffer[i] = Dequantize<int8_t>(bufferPtr->data[i], tensorInfo.GetQuantizationScales()[axisIndex],
-                                               tensorInfo.GetQuantizationOffset());
-            }
-        }
-        else
-        {
-            for (unsigned int i = 0; i < tensorInfo.GetNumElements(); ++i)
-            {
-                buffer[i] = Dequantize<int8_t>(bufferPtr->data[i], tensorInfo.GetQuantizationScale(),
-                                               tensorInfo.GetQuantizationOffset());
-            }
-        }
-        return buffer;
-    }
-    throw ParseException(
-            fmt::format("Unsupported input/weights combination:  Input {} not supported with Weights {}",
-                        GetDataTypeName(DataType::Float32),
-                        GetDataTypeName(tensorInfo.GetDataType()),
-                        CHECK_LOCATION().AsString()));
+    return (TfLiteParserImpl::IsConstTensor(tensorPtr) && inputDataType == DataType::Float32 &&
+            (tensorDataType == DataType::QAsymmU8 ||
+             tensorDataType == DataType::QAsymmS8 ||
+             tensorDataType == DataType::QSymmS8 ||
+             tensorDataType == DataType::Signed32 ||
+             tensorDataType == DataType::Signed64));
 }
 
 void TfLiteParserImpl::RegisterProducerOfTensor(size_t subgraphIndex,
@@ -1136,9 +1118,7 @@ void TfLiteParserImpl::ParseConv2D(size_t subgraphIndex, size_t operatorIndex)
     auto layerName = fmt::format("Conv2D:{}:{}", subgraphIndex, operatorIndex);
     armnn::IConnectableLayer* layer = m_Network->AddConvolution2dLayer(desc, layerName.c_str());
 
-    if (IsConstTensor(inputs[1]) && inputTensorInfo.GetDataType() == DataType::Float32 &&
-        (filterTensorInfo.GetDataType() == DataType::QAsymmU8 ||
-            filterTensorInfo.GetDataType() == DataType::QAsymmS8))
+    if (ShouldConstantTensorBeConverted(inputs[1], inputTensorInfo.GetDataType(), filterTensorInfo.GetDataType()))
     {
         m_ConstantsToDequantize.emplace_back(inputs[1]->buffer);
     }
@@ -1150,9 +1130,7 @@ void TfLiteParserImpl::ParseConv2D(size_t subgraphIndex, size_t operatorIndex)
         // Add the biases input to the registration list, a constant layer will be added by SetupConstantLayers.
         tensorIndexesToRegister.emplace_back(inputTensorIndexes[2]);
 
-        if (IsConstTensor(inputs[2]) && inputTensorInfo.GetDataType() == DataType::Float32 &&
-            (filterTensorInfo.GetDataType() == DataType::QAsymmU8 ||
-                filterTensorInfo.GetDataType() == DataType::QAsymmS8))
+        if (ShouldConstantTensorBeConverted(inputs[2], inputTensorInfo.GetDataType(), biasTensorInfo.GetDataType()))
         {
             m_ConstantsToDequantize.emplace_back(inputs[2]->buffer);
         }
@@ -3112,9 +3090,7 @@ void TfLiteParserImpl::ParseFullyConnected(size_t subgraphIndex, size_t operator
     // Add the weights input to the registration list, constant layers will be added by SetupConstantLayers if constant.
     tensorIndexesToRegister.emplace_back(inputTensorIndexes[1]);
 
-    if (desc.m_ConstantWeights && inputTensorInfo.GetDataType() == DataType::Float32 &&
-        (filterTensorInfo.GetDataType() == DataType::QAsymmU8 ||
-         filterTensorInfo.GetDataType() == DataType::QAsymmS8))
+    if (ShouldConstantTensorBeConverted(inputs[1], inputTensorInfo.GetDataType(), filterTensorInfo.GetDataType()))
     {
         m_ConstantsToDequantize.emplace_back(inputs[1]->buffer);
     }
@@ -3127,9 +3103,7 @@ void TfLiteParserImpl::ParseFullyConnected(size_t subgraphIndex, size_t operator
         // Add the biases input to the registration list, constant layer will be added by SetupConstantLayers.
         tensorIndexesToRegister.emplace_back(inputTensorIndexes[2]);
 
-        if (desc.m_ConstantWeights && inputTensorInfo.GetDataType() == DataType::Float32 &&
-            (biasTensorInfo.GetDataType() == DataType::QAsymmU8 ||
-             biasTensorInfo.GetDataType() == DataType::QAsymmS8))
+        if (ShouldConstantTensorBeConverted(inputs[2], inputTensorInfo.GetDataType(), biasTensorInfo.GetDataType()))
         {
             m_ConstantsToDequantize.emplace_back(inputs[2]->buffer);
         }
@@ -4925,11 +4899,22 @@ TfLiteParserImpl::CreateConstTensorNonPermuted(TensorRawPtr tensorPtr,
     // Make sure isConstant flag is set.
     tensorInfo.SetConstant();
 
-    if (inputDataType == DataType::Float32 &&  tensorInfo.GetDataType() != DataType::Float32)
+    if (inputDataType == DataType::Float32 && tensorInfo.GetDataType() != DataType::Float32)
     {
-        TensorInfo constTensorInfo(tensorInfo.GetShape(), DataType::Float32, 0.0f, 0, true);
-        std::unique_ptr<float[]> data = AsFloatArray(bufferPtr, tensorInfo);
-        return std::make_pair(ConstTensor(constTensorInfo, data.get()), std::move(data));
+        try
+        {
+            TensorInfo constTensorInfo(tensorInfo.GetShape(), DataType::Float32, 0.0f, 0, true);
+            std::unique_ptr<float[]> data = armnnUtils::ToFloatArray(bufferPtr->data, tensorInfo);
+            return std::make_pair(ConstTensor(constTensorInfo, data.get()), std::move(data));
+        }
+        catch (armnn::InvalidArgumentException)
+        {
+            throw ParseException(
+                    fmt::format("Unsupported input/weights combination:  Input {} not supported with Weights {}",
+                                GetDataTypeName(DataType::Float32),
+                                GetDataTypeName(tensorInfo.GetDataType()),
+                                CHECK_LOCATION().AsString()));
+        }
     }
     else
     {
@@ -4950,9 +4935,20 @@ TfLiteParserImpl::CreateConstTensorPtr(TensorRawPtr tensorPtr, armnn::TensorInfo
 
     if (inputTensorInfo.GetDataType() == DataType::Float32 && tensorInfo.GetDataType() != DataType::Float32)
     {
-        TensorInfo constTensorInfo(tensorInfo.GetShape(), DataType::Float32, 0.0f, 0, true);
-        std::unique_ptr<float[]> data = AsFloatArray(bufferPtr, tensorInfo);
-        return std::make_pair(new ConstTensor(constTensorInfo, data.get()), std::move(data));
+        try
+        {
+            TensorInfo constTensorInfo(tensorInfo.GetShape(), DataType::Float32, 0.0f, 0, true);
+            std::unique_ptr<float[]> data = armnnUtils::ToFloatArray(bufferPtr->data, tensorInfo);
+            return std::make_pair(new ConstTensor(constTensorInfo, data.get()), std::move(data));
+        }
+        catch (armnn::InvalidArgumentException)
+        {
+            throw ParseException(
+                    fmt::format("Unsupported input/weights combination:  Input {} not supported with Weights {}",
+                                GetDataTypeName(DataType::Float32),
+                                GetDataTypeName(tensorInfo.GetDataType()),
+                                CHECK_LOCATION().AsString()));
+        }
     }
     else
     {
diff --git a/src/armnnTfLiteParser/TfLiteParser.hpp b/src/armnnTfLiteParser/TfLiteParser.hpp
index f8ddc55649..7eb6c48501 100644
--- a/src/armnnTfLiteParser/TfLiteParser.hpp
+++ b/src/armnnTfLiteParser/TfLiteParser.hpp
@@ -242,7 +242,13 @@ private:
     };
 
     bool ShouldConstantTensorBeCreated(unsigned int tensorIndex);
+
     bool IsConstTensor(TensorRawPtr tensorPtr);
+
+    bool ShouldConstantTensorBeConverted(TfLiteParserImpl::TensorRawPtr tensorPtr,
+                                         armnn::DataType inputDataType,
+                                         armnn::DataType filterDataType);
+
     armnn::ConstTensor CreateConstTensorNonPermuted(TensorRawPtr tensorPtr,
                                                     armnn::TensorInfo& tensorInfo);
 
@@ -250,6 +256,7 @@ private:
     CreateConstTensorPermuted(TensorRawPtr tensorPtr,
                               armnn::TensorInfo& tensorInfo,
                               armnn::Optional<armnn::PermutationVector&> permutationVector);
+
     std::pair<armnn::ConstTensor, std::unique_ptr<float[]>>
     CreateConstTensorNonPermuted(TensorRawPtr tensorPtr,
                                  armnn::TensorInfo& tensorInfo,
@@ -261,6 +268,7 @@ private:
                                   TfLiteParserImpl::TensorRawPtr tensorPtr,
                                   armnn::TensorInfo& tensorInfo,
                                   armnn::Optional<armnn::PermutationVector&> permutationVector);
+
     std::pair<armnn::ConstTensor*, std::unique_ptr<float[]>>
     CreateConstTensorPtr(TensorRawPtr tensorPtr,
                          armnn::TensorInfo& inputTensorInfo);
diff --git a/src/armnnTfLiteParser/test/Conv2D.cpp b/src/armnnTfLiteParser/test/Conv2D.cpp
index 45c4a43519..334c102344 100644
--- a/src/armnnTfLiteParser/test/Conv2D.cpp
+++ b/src/armnnTfLiteParser/test/Conv2D.cpp
@@ -673,7 +673,7 @@ struct Conv2FloatWithInt8WeightsAndBiasesFixture : Conv2DWithBiasesFixture
                                       "[ 1, 2, 2, 1 ]",    // filterShape
                                       "[ 2,1, 0,6 ]",      // filterData
                                       "[ 1 ]",             // biasShape
-                                      "[ 10, 0, 0, 0 ]",   // biasData
+                                      "[ 10 ]",            // biasData
                                       "1",                 // stride w and h
                                       "NONE",              // activation
                                       "1.0",               // filterScale