diff options
| author | Cathal Corbett <cathal.corbett@arm.com> | 2022-05-13 09:55:59 +0100 |
|---|---|---|
| committer | Cathal Corbett <cathal.corbett@arm.com> | 2022-05-13 15:09:18 +0000 |
| commit | 4452baf3d295164877c5810a3867b1d2d79b04f3 (patch) | |
| tree | 861b248dd0a603140eb384e48cd9ef97603a296b /src/backends/neon/NeonBackend.cpp | |
| parent | c46b90ef031a8041fe25a7191de799eca9eff8e5 (diff) | |
| download | armnn-4452baf3d295164877c5810a3867b1d2d79b04f3.tar.gz | |
IVGCVSW-6260 ConstTensorsAsInput: Fully Connected Cl and Neon support.
* IVGCVSW-6940 ConstTensorsAsInput: DepthwiseConvolution2d - Complete Neon and Cl Bug Fix
* Bug fix to enable Cl and Neon Backend Compatibility ConstantTensorsAsInputs
* Updated Cl and Neon FullyConnected workloads to handle constant
weights and bias as inputs rather than reading from member variables.
* Prevent non const weights and biases passing CL and NEON validate
for Depthwise Convolution.
Signed-off-by: Cathal Corbett <cathal.corbett@arm.com>
Change-Id: I0f505ff5998a183152f843d0f6cc74327ba920e7
Diffstat (limited to 'src/backends/neon/NeonBackend.cpp')
| -rw-r--r-- | src/backends/neon/NeonBackend.cpp | 10 |
1 files changed, 6 insertions, 4 deletions
diff --git a/src/backends/neon/NeonBackend.cpp b/src/backends/neon/NeonBackend.cpp index 7a258c38e0..39ad4b9f32 100644 --- a/src/backends/neon/NeonBackend.cpp +++ b/src/backends/neon/NeonBackend.cpp @@ -250,17 +250,19 @@ OptimizationViews NeonBackend::OptimizeSubgraphView(const SubgraphView& subgraph else if (base.GetType() == LayerType::FullyConnected) { FullyConnectedLayer* baseLayer = PolymorphicDowncast<FullyConnectedLayer*>(&base); - Optional<TensorInfo> biases; + FullyConnectedDescriptor descriptor = baseLayer->GetParameters(); - if (baseLayer->GetParameters().m_BiasEnabled) + // As bias is optional only try to get TensorInfo from input if bias is enabled. + Optional<TensorInfo> biases; + if (descriptor.m_BiasEnabled) { - biases = baseLayer->m_Bias->GetTensorInfo(); + biases = baseLayer->GetInputSlot(2).GetConnectedOutputSlot()->GetTensorInfo(); } arm_compute::Status status = NeonFullyConnectedWorkloadValidate( baseLayer->GetInputSlot(0).GetConnectedOutputSlot()->GetTensorInfo(), activationLayer->GetInputSlot(0).GetConnectedOutputSlot()->GetTensorInfo(), - baseLayer->m_Weight->GetTensorInfo(), + baseLayer->GetInputSlot(1).GetConnectedOutputSlot()->GetTensorInfo(), biases, baseLayer->GetParameters(), &activationDesc); |