'MemoryError' for pb.invert

[Coastal0]

Problem description

When attempting to invert a dataset, a 'MemoryError' exception is thrown.

Your environment

Windows
Python 3.6
pyGIMLI latest wheel
Wheel-based Winpython install

Steps to reproduce

import pybert as pb
ert = pb.ERTManager()
data = ert.loadData('Stable_015mday_fwdModel_wa_5pcNoise.txt')
model = ert.invert(data = data, lam=30)

Stable_015mday_fwdModel_wa_5pcNoise.txt

Expected behavior

A perfect replication of the model (or something similar).

Actual behavior

I run out of memory.

model = ert.invert(data = data, lam=30)
Traceback (most recent call last):

  File "<ipython-input-16-381264132134>", line 1, in <module>
    model = ert.invert(data = data, lam=30)

  File "F:\WinPython-64bit-3.6.3.0Qt5\python-3.6.3.amd64\lib\site-packages\pybert\manager\ertmanager.py", line 392, in invert
    model = self.inv.run()

MemoryError

Additional Notes

I have 64 Gb of RAM on this workstation, and the implementation of BERT outside python appears to work fine.

(I wasn't sure whether to post here or on BERT, sorry!)

[halbmy]

Well this seems to be a clear BERT issue as the code is entirely from there.

[halbmy]

As said before there is a default mechanism for using (2d) topography but it is not foolproof or tested. In rather recommend using the command line to prepare mesh and primary potentials.

[Coastal0]

Good point. I can test with meshes I've already generated from earlier in the script using pyGIMLi.

Is it worth posting in the BERT tracker instead?

[florian-wagner]

64 GB of RAM should be more than enough, especially as you are working in 2D. I cannot reproduce the problem. Your code snippet works with the attached data set and produces the following result (rrms = 100%, chi^2 = 0.77).

It's definitely not a memory problem (I have 16 GB).

[Coastal0]

Thanks guys. This looks like it's specific to my install of Spyder. I'll close the issue (can we delete it?)