Implementing CUDA Unified Memory in the PyTorch Framework | IEEE Conference Publication | IEEE Xplore