26 questions
Score of 1
1 answer
98 views
SIGSEGV in aidl-service-armnn-gpu when initializing TFLite NNAPI Delegate on specific Vendor Devices (Vivo/Mali)
I am developing a custom C++ Inference Engine for Android using the TensorFlow Lite C API. The engine is compiled as a shared library (.so) and loaded via JNI.
The engine works perfectly on Pixel and ...
- reputation score 3
Score of 2
1 answer
587 views
"Tensorflow + NNAPI" runs very slow on Samsung Galaxy S21
Samsung Galaxy S21 is the only device runs slow when using tensor-flow lite and NNAPI together. Did anyone encounter this issue before?
Background
I used tensor-flow lite and NNAPI and built my app on ...
- reputation score 501
Score of 1
1 answer
2041 views
Tensorflow Lite Android: Both GPU delegate and NNAPI delegate are slower than CPU
I am currently evaluating and comparing the performance of some tensorflow models on different smartphones. I am testing the MNIST and CIFAR10 databases. The strange thing is, when i try to speed up ...
- reputation score 11
Score of 0
1 answer
191 views
Execute different operations of a NN model graph with different hardware accelerators while using Onnxruntime
I was trying to utilize NNAPI via OnnxRuntime for NN model inferencing on an Android device. Based on the youtube video here: https://www.youtube.com/watch?v=Ij5MoUnLQ0E it is possible to specify the ...
- reputation score 379
Score of -1
1 answer
224 views
please help me NNAPI analysis
I'm studying android platform, especially nnapi.
I found out Arm nn driver processes requests from application through NNAPI.
I was trying to figure out the flow of the function in the code base, and ...
- reputation score 1
Score of 0
1 answer
88 views
Access Android C++ NNAPIs - ANeuralNetworksCompilation_createForDevices gives error
Trying to Access GPU using Android C++ NNAPIs ANeuralNetworksCompilation_createForDevices gives error
ANeuralNetworksCompilation_createForDevices gives error
Linking fails.
- reputation score 1
Score of 0
1 answer
582 views
tflite model outputs different predictions on CPU and NNAPI
I trained and quantized a Tensorflow model on a Ubuntu 18.04 machine and I converted it to tflite format. Then I deployed it on a Linux Yocto board equipped with a NPU accelerator, tflite_runtime and ...
- reputation score 320
Score of 1
0 answers
95 views
Does Hiai foundation available for native shell program?
I am interested alot in huawei NPU(I am using kirin9000, 990, 810) and I just found that when using hiai foundation api to use NPU, hiai foundation is not working when i build as native shell program ...
- reputation score 11
Score of 0
1 answer
260 views
How do I use HAL version 1.2 on Tensorflow Lite for Android?
I have a quantized TensorflowLite model that I'm loading onto a Pixel 3 running Android 11. I built the model using Tensorflow Lite 2.5 and I'm using the nightly builds of Tensorflow for Android.
I'm ...
- reputation score 5986
Score of 0
1 answer
244 views
Multiple instances of tensorflow lite with NNAPI delegate
NNAPI Delegate in Tensorflow lite uses shared memory for input and output tensors of the graph. However the name of the shared memory pool is hardcoded ("input_pool" and "otput_pool&...
- reputation score 133
Score of -1
2 answers
574 views
where is the dynamic library file(.so) for sharedmem.h of Android NDK?
I am trying to build NNAPI c++ native program, and I need to use sharedmem library.
I found dynamic library file for NeuralNetwork.h however I cannot find dynamic library file for sharedmem.h. Thus I ...
- reputation score 11
Score of 0
0 answers
111 views
How are android nn_cts test cases created?
NN_CTS test cases are auto-generated using some script. I'm trying to understand if there is any source code available for those test cases. The documentation is not very clear too.
Is there any ...
- reputation score 17
Score of 1
1 answer
659 views
How can I let Kirin 990's NPU work on TensorFlow Lite?
I successfully converted TensorFlow model to TensorFlow Lite float16 model according to Post-training float16 quantization.
The below is a diagram of the converted model.
And I ran successfully it on ...
- reputation score 724
Score of 0
1 answer
73 views
Question about conv2D in Android NN: the output shape limit
When I was trying the android nn code, found these in Conv2D.cpp:
tflite::Dims<4> im2colDim; \
im2colDim.sizes[3] = (int)getSizeOfDimension(...
- reputation score 1
Score of 0
2 answers
679 views
What will happen if I try to use GPU delegate under android 8.1
Here below is the system architecture for NNAPI.
enter image description here
The NNAPI is available on Android 8.1 (API level27) or higher.
What will happen if I try to use GPU delegate under ...
- reputation score 1