[ONNX][Autotune] Replace CUDA memory management from CUDART to PyTorch#998

gcunhase merged 6 commits intoNVIDIA:mainNVIDIA/Model-Optimizer:mainfrom

gcunhase:dev/gcunhasergio/autotune_remove_cudartgcunhase/TensorRT-Model-Optimizer:dev/gcunhasergio/autotune_remove_cudartCopy head branch name to clipboard