NNAPI Execution Provider
Accelerate ONNX models on Android devices with ONNX Runtime and the NNAPI execution provider. Android Neural Networks API (NNAPI) is a unified interface to CPU, GPU, and NN accelerators on Android.
Contents
Requirements
The NNAPI Execution Provider (EP) requires Android devices with Android 8.1 or higher. It is recommended to use Android devices with Android 9 or higher to achieve optimal performance.
Install
Pre-built packages of ONNX Runtime with NNAPI EP for Android are published on Maven.
See here for installation instructions.
Build
Please see the Build Android EP for instructions on building a package that includes the NNAPI EP.
Usage
The ONNX Runtime API details are here.
The NNAPI EP can be used via the C, C++ or Java APIs
The NNAPI EP must be explicitly registered when creating the inference session. For example:
Ort::Env env = Ort::Env{ORT_LOGGING_LEVEL_ERROR, "Default"}; Ort::SessionOptions so; uint32_t nnapi_flags = 0; Ort::ThrowOnError(OrtSessionOptionsAppendExecutionProvider_Nnapi(so, nnapi_flags)); Ort::Session session(env, model_path, so); Configuration Options
There are several run time options available for the NNAPI EP.
To use the NNAPI EP run time options, create an unsigned integer representing the options, and set each individual option by using the bitwise OR operator.
uint32_t nnapi_flags = 0; nnapi_flags |= NNAPI_FLAG_USE_FP16; Available Options
NNAPI_FLAG_USE_FP16
Use fp16 relaxation in NNAPI EP.
This may improve performance but can also reduce accuracy due to the lower precision.
NNAPI_FLAG_USE_NCHW
Use the NCHW layout in NNAPI EP.
This is only available for Android API level 29 and higher. Please note that for now, NNAPI might have worse performance using NCHW compared to using NHWC.
NNAPI_FLAG_CPU_DISABLED
Prevent NNAPI from using CPU devices.
NNAPI is more efficient using GPU or NPU for execution, however NNAPI might fall back to its CPU implementation for operations that are not supported by GPU/NPU. The CPU implementation of NNAPI (which is called nnapi-reference) is often less efficient than the optimized versions of the operation of ORT. Due to this, it may be advantageous to disable the NNAPI CPU fallback and handle execution using ORT kernels.
For some models, if NNAPI would use CPU to execute an operation, and this flag is set, the execution of the model may fall back to ORT kernels.
This option is only available for Android API level 29 and higher, and will be ignored for Android API level 28 and lower.
For NNAPI device assignments, see https://developer.android.com/ndk/guides/neuralnetworks#device-assignment
For NNAPI CPU fallback, see https://developer.android.com/ndk/guides/neuralnetworks#cpu-fallback
NNAPI_FLAG_CPU_ONLY
Using CPU only in NNAPI EP, this may decrease the perf but will provide reference output value without precision loss, which is useful for validation.
This option is only available for Android API level 29 and higher, and will be ignored for Android API level 28 and lower.
Supported ops
Following ops are supported by the NNAPI Execution Provider,
| Operator | Note |
|---|---|
| ai.onnx:Abs | |
| ai.onnx:Add | |
| ai.onnx:AveragePool | Only 2D Pool is supported. |
| ai.onnx:BatchNormalization | |
| ai.onnx:Cast | |
| ai.onnx:Clip | |
| ai.onnx:Concat | |
| ai.onnx:Conv | Only 2D Conv is supported. Weights and bias should be constant. |
| ai.onnx:DepthToSpace | Only DCR mode DepthToSpace is supported. |
| ai.onnx:DequantizeLinear | All quantization scales and zero points should be constant. |
| ai.onnx:Div | |
| ai.onnx:Elu | |
| ai.onnx:Exp | |
| ai.onnx:Flatten | |
| ai.onnx:Floor | |
| ai.onnx:Gather | Input indices should be constant if not int32 type. |
| ai.onnx:Gemm | If input B is not constant, transB should be 1. |
| ai.onnx:GlobalAveragePool | Only 2D Pool is supported. |
| ai.onnx:GlobalMaxPool | Only 2D Pool is supported. |
| ai.onnx:Identity | |
| ai.onnx:LeakyRelu | |
| ai.onnx:Log | |
| ai.onnx:LRN | |
| ai.onnx:MatMul | |
| ai.onnx:MaxPool | Only 2D Pool is supported. |
| ai.onnx:Max | |
| ai.onnx:Min | |
| ai.onnx:Mul | |
| ai.onnx:Neg | |
| ai.onnx:Pad | Only constant mode Pad is supported. Input pads and constant_value should be constant. Input pads values should be non-negative. |
| ai.onnx:Pow | |
| ai.onnx:PRelu | |
| ai.onnx:QLinearConv | Only 2D Conv is supported. Weights and bias should be constant. All quantization scales and zero points should be constant. |
| ai.onnx:QLinearMatMul | All quantization scales and zero points should be constant. |
| ai.onnx:QuantizeLinear | All quantization scales and zero points should be constant. |
| ai.onnx:ReduceMean | |
| ai.onnx:Relu | |
| ai.onnx:Reshape | |
| ai.onnx:Resize | Only 2D Resize is supported. |
| ai.onnx:Sigmoid | |
| ai.onnx:Sin | |
| ai.onnx:Slice | |
| ai.onnx:Softmax | |
| ai.onnx:Split | Number of splits must evenly divide split axis size. Input split should be constant if provided. |
| ai.onnx:Sqrt | |
| ai.onnx:Squeeze | Input axes should be constant. |
| ai.onnx:Sub | |
| ai.onnx:Tanh | |
| ai.onnx:Transpose | |
| ai.onnx:Unsqueeze | Input axes should be constant. |
| com.microsoft:QLinearAdd | All quantization scales and zero points should be constant. |
| com.microsoft:QLinearAveragePool | Only 2D Pool is supported. All quantization scales and zero points should be constant. |
| com.microsoft:QLinearSigmoid | All quantization scales and zero points should be constant. |
