IVGCVSW-5826 Change weights layout for depthwise to [1,H,W,I*M]
* This change is necessary because tflite uses a [1,H,W,I*M] format
and uses the I*M dimension for per axis quantization. Our previous
layout [M,I,H,W] can't handle the correlating quantization scales.
* Updates Onnx-, TfLiteParser and TfliteDelegate
* Updates the CpuRef, CpuAcc and GpuAcc backends
* Adjusts unit tests
* Adds test to ensure models with old layout can still be read and
executed
* Adds conversion function to previous layout [1,H,W,I*M] --> [M,I,H,W]
which can be used by backend developers
!android-nn-driver:5553
Signed-off-by: Jan Eilers <jan.eilers@arm.com>
Change-Id: Ifef23368b8c3702cf315a5838d214f7dc13c0152
diff --git a/src/armnnOnnxParser/OnnxParser.hpp b/src/armnnOnnxParser/OnnxParser.hpp
index 7716e50..f618ff4 100644
--- a/src/armnnOnnxParser/OnnxParser.hpp
+++ b/src/armnnOnnxParser/OnnxParser.hpp
@@ -128,7 +128,9 @@
void ResetParser();
void Cleanup();
- std::pair<armnn::ConstTensor, std::unique_ptr<float[]>> CreateConstTensor(const std::string name);
+ std::pair<armnn::ConstTensor, std::unique_ptr<float[]>>
+ CreateConstTensor(const std::string name,
+ armnn::Optional<armnn::PermutationVector&> permutationVector = armnn::EmptyOptional());
template <typename TypeList, typename Location>
void ValidateInputs(const onnx::NodeProto& node,