error "QNN execution provider is not supported in this build" when testing ONNX QNN EP

[shawnyxf]

Recently, i'm trying to use QNN EP in ONNX to make my model run faster on Android platform:

So i compiled the onnx + qnn android dynamic library according to https://onnxruntime.ai/docs/build/eps.html#qnn

ONNX VERSION: 1.16.3
QNN SDK: 2.20.0.240223

After finishing the compile, i add some related code in my project (i wanna test on cpu backend)：

std::string backend = "QnnCpu.so";
std::vector<const char*> options_keys = {"backend_path"};
std::vector<const char*> options_values = {backend.c_str()};

ORT_ABORT_ON_ERROR(g_ort->SessionOptionsAppendExecutionProvider(st->session_options, "QNN", options_keys.data(), options_values.data(), options_keys.size()));

then i run my project on android, i get the error "QNN execution provider is not supported in this build".

Can anyone help me solve my problem?

[Francis235]

Hello, I want to know what qualcomm soc platform you use?

[shawnyxf]

Hello, I want to know what qualcomm soc platform you use?

Hi, thanks for replying, i use qualcomm 8295

[shawnyxf]

Recently, i'm trying to use QNN EP in ONNX to make my model run faster on Android platform:

So i compiled the onnx + qnn android dynamic library according to https://onnxruntime.ai/docs/build/eps.html#qnn

ONNX VERSION: 1.16.3 QNN SDK: 2.20.0.240223

After finishing the compile, i add some related code in my project (i wanna test on cpu backend)：

std::string backend = "QnnCpu.so";
std::vector<const char*> options_keys = {"backend_path"};
std::vector<const char*> options_values = {backend.c_str()};

ORT_ABORT_ON_ERROR(g_ort->SessionOptionsAppendExecutionProvider(st->session_options, "QNN", options_keys.data(), options_values.data(), options_keys.size()));

then i run my project on android, i get the error "QNN execution provider is not supported in this build".

Can anyone help me solve my problem?

Just change the QNN sdk version to 2.18.0.240101, it works

[HectorSVC]

Ort 1.16.3 is an very old version. Please use the main branch which has more most recent features. The newer QNN 2.20 should definitely be worked.