In Linux cluster system, ValueError: numpy.dtype has the wrong size, try recompiling. Expected 88, got 96

Question

While trying to do simulation in the Scientific Linux, I met the following error.

$ python plot_distribt.py
Loaded libmkl_rt.so for dgesvd
Loaded libmkl_rt.so for zgesvd
Traceback (most recent call last):
  File "plot_distribt.py", line 32, in <module>
    from mps.mps import iMPS
  File "/work/wangb/DMRG/mps/mps.py", line 13, in <module>
    from algorithms.linalg import np_conserved as npc
  File "/work/wangb/DMRG/algorithms/linalg/np_conserved.py", line 294, in <module>
    from algorithms.linalg import npc_helper
  File "__init__.pxd", line 164, in init npc_helper
ValueError: numpy.dtype has the wrong size, try recompiling. Expected 88, got 96

After creating a virtual environment, I was able to upgrade packages via pip2 install --upgrade numpy --user python, and now I have numpy-1.16.6, scipy-1.2.3 and pandas-0.24.2. However, the above ValueError still occurred.

I have also checked that after opening python within the newly-created environment, I get

Python 2.7.18 |Anaconda, Inc.| (default, Apr 23 2020, 22:42:48)
[GCC 7.3.0] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy
>>> numpy.version.version
'1.7.1'
>>> numpy.__file__
'/usr/lib64/python2.7/site-packages/numpy/__init__.pyc'
>>>

Thus, the key might be how to avoid using the numpy-1.7.1? Note that I cannot change anything in /usr/lib64/python2.7/site-packages as I am using the cluster systems from our university.

Thanks a lot in advance!