The deep learning model was deployed on the vaa3D platform, but encountered an environment dependency problem. We use the Pytorch (python) training model and generate a .pth file, and then use libtorch (C++) to deploy on the Vaa3D platform. However, the vaa3d platform relies on the g++ 4.8 compiler, while libtorch recommends the g++ 4.9 compiler. To test this question, I did the following experiment.
1. A model , test.pth,  was trained.(pytorch 1.1)
2. Test on computer A (g++ 4.8, gcc 4.8, libtorch 1.1), failed.
3. Test on computer B (g++ 7.5.0, gcc 7.5.0, libtorch 1.1) using the same file, success!

How can I solve this problem, or is there another way to deploy the PyTorch model?

Thanks!