While building the NeuralNetwork object (Using the code snippet below)
NeuralNetwork network = new SNPE.NeuralNetworkBuilder(application)
.setDebugEnabled(false)
.setOutputLayers(outputLayerNames)
.setModel(assetInputStream, assetInputStream.available())
.setPerformanceProfile(NeuralNetwork.PerformanceProfile.HIGH_PERFORMANCE)
.setRuntimeOrder(selectedRuntime)
.setCpuFallbackEnabled(needsCpuFallback)
.build();
I am facing the below Exception :
java.lang.IllegalStateException: Unable to create network! Cause: error_code=300; error_message=Model parsing has failed.; error_component=Dl Container; line_no=319; thread_id=510488369816
Please let me know what modifications I need to do to successfully build the network.
I have been able to resolve this issue by ensuring the below points:
1. The issue could be because you might be creating the neural network object in the main thread.
One way of resolving this would be to create an Async Task and create the object of the NeuralNetwork in doInBackground() method.Below is an example snippet of building the network :
final SNPE.NeuralNetworkBuilder builder = new SNPE.NeuralNetworkBuilder(mApplication)
.setDebugEnabled(false)
.setRuntimeOrder(mTargetRuntime)
.setModel(mInputstream, mInputstream.available())
.setCpuFallbackEnabled(true)
.setUseUserSuppliedBuffers(mTensorFormat != MainActivity.SupportedTensorFormat.FLOAT);
network = builder.build();
Please also note that, in this NeuralNetwork object we would not pass the output layer. Instead we would create the input stream of the model (.dlc) in the main thread.
2. Putting the model in the /raw directory instead of the /asset directory.
3. Please ensure that you use the snpe-release.aar as snpe library.
You can get the file by following the steps mentioned in the setup instructions below:
https://developer.qualcomm.com/docs/snpe/android_tutorial.html