Torchscript inference is working on PC but is not working on Android

1

Hello All,

I have a torchscript that is working on PC ( PyTorch / libtorch ) for both trace mode and script mode, but when I run the same torchscript on my android application, it crashes without throwing explicit information.

Traceback of TorchScript, serialized code (most recent call last):
  File "code/__torch__/model_system.py", line 33, in forward
  def forward(self: __torch__.model_system.TransformerSystem,
    inputs: Tensor) -> Tensor:
    spec = (self.preprop).forward(inputs, )
            ~~~~~~~~~~~~~~~~~~~~~ <--- HERE
    outputs = (self.model).forward(spec, )
    outputs0 = torch.argmax(outputs)
  File "code/__torch__/torch/nn/modules/container/___torch_mangle_0.py", line 12, in forward
    _0 = getattr(self, "0")
    _1 = getattr(self, "1")
    input0 = (_0).forward(input, )
              ~~~~~~~~~~~ <--- HERE
    return (_1).forward(input0, )

Are there checkers or guidelines on how to land the torchscript to the Android from PC?