Very basic stuff

Hi,

I’m trying to do something very simple as a “hello world” example.
I have a very simple .tflite model that just multiplies 1 input by 2.
I want to try and infer that on the NPU just to test the integration.
I do everything in native with a simple Android application wrapper with 1 button to trigger my native function.
I compile and link my library with tensorflowlite, neuronusdk_adapter and neuron_buffer_allocator.
Now, we get a crash in somewhere from inside the libneuronusdk_adapter.mtk.so during its initialization.

Now to my question - we are doing this on production build on Dimensity9400,
could it be because we don’t have access to the neuron runtime and runtime API, in the basic SDK? Do I need to link with them or something?
Our company has NDA but I have no idea where do I get the NP Basic.
I’m using the NeuroPilotTFLiteShim.h to interface with NP.

Am I missing something, is my approach not correct for what I’m trying to do?
Any pointers would be welcome

Hi Mark,

Thank you for the inquiry.
Could you help to share the error you got when doing the model initialisation? It will be good if our teams can analyse the log and do the troubleshooting :slight_smile:

Also, developers not only need to sign the NDA but also apply for the NeuroPilot Account for access to NeuroPilot Documentation.

Here is the topic How-to-Get-An-Neuropolit-Account that might be helpful.