memory mechanics in pytorch. Override variable names or make new?

pytorch-memtest.py

import torch
@profile
def keep():
    x = torch.randn(10000,100)
    y = x+1
    z = y**2
    w = z**2
    return w
# keep()

@profile
def override():
    x = torch.randn(10000,100)
    x = x+1
    x = x**2
    x = x**2
    return x
override()

results.log

[~/Prylar/diverse]$ python -m memory_profiler pytorch-memtest.py
Filename: pytorch-memtest.py

Line #    Mem usage    Increment   Line Contents
================================================
     2   66.578 MiB   66.578 MiB   @profile
     3                             def keep():
     4   70.504 MiB    3.926 MiB       x = torch.randn(10000,100)
     5   74.363 MiB    3.859 MiB       y = x+1
     6   78.219 MiB    3.855 MiB       z = y**2
     7   82.035 MiB    3.816 MiB       w = z**2
     8   82.035 MiB    0.000 MiB       return w


[~/Prylar/diverse]$ python -m memory_profiler pytorch-memtest.py
Filename: pytorch-memtest.py

Line #    Mem usage    Increment   Line Contents
================================================
    11   66.594 MiB   66.594 MiB   @profile
    12                             def override():
    13   70.520 MiB    3.926 MiB       x = torch.randn(10000,100)
    14   74.383 MiB    3.863 MiB       x = x+1
    15   74.422 MiB    0.039 MiB       x = x**2
    16   74.422 MiB    0.000 MiB       x = x**2
    17   74.422 MiB    0.000 MiB       return x