custom build from source about my model, build is ok. but using libpytorch_jni.so on Android device will failed.
First, I build libpytorch_jni.so without custom build. Build is ok, and model can inference on android
conmmand is : BUILD_LITE_INTERPRETER=0 scripts/build_pytorch_android.sh arm64-v8a.
Second. I build libpytorch_jni.so with custom build. Build is ok, but model failed in inferencing on android
command is : SELECTED_OP_LIST=mymodel.yaml BUILD_LITE_INTERPRETER=0 scripts/build_pytorch_android.sh arm64-v8a
Error message :
ld.lld: error: undefined symbol: torch::jit::Module::train(bool)
ld.lld: error: undefined symbol: c10::GradMode::set_enabled(bool)
ld.lld: error: undefined symbol: at::Tensor::to(c10::TensorOptions, bool, bool, c10::optionalc10::MemoryFormat) const
ld.lld: error: undefined symbol: c10::UndefinedTensorImpl::_singleton
ld.lld: error: undefined symbol: c10::IValue::reportToTensorTypeError() const