I got an error with c10::Error: Couldn't find an operator for aten::dropout(Tensor input, float p, bool train) when I tried to load a torchscript model by libtorch c++ api. The linking libtorch library is compiled with script/build_pytorch_android.sh script.
target platform: android arm64-v8a
pytorch tag: v1.6.0
And if I linked the program with x86 prebuilt libtorch library, and run it in x86 platform, all worked well.