Reputation: 77
Hello everyone,
I have installed Openmdao, pyOpt and pyoptsparse on my computer. As my program works with the Scipy optimizer, I tried it with a random optimizer of pyoptsparse (that was 'ALPSO'). It worked and I was happy. But it turns out it seems to be the only one working.
Every time I try to use another one (like 'SLSQP', which is the default optimizer !), i get this message "pyOptSparse Error : There was an error importing the compiled SLSQP module", inside a frame made of '-' and '+'.
Does anybody know what to do ? I am using Ubuntu if it changes something.
Upvotes: 0
Views: 878
Reputation: 77
Thanks to swryan for the link, i found the answer.
One of the possible answer was to put libgfortran.so.3 in anaconda2/lib, but i already had it.
They were also saying the problem was solved when installing anaconda 4.0+ but i also had the latest version.
What worked for me was to run : conda update libgfortran --force
Doing it without the --force retrogrades the scipy, which seems to disable scipy.optimize.least_squares. If you did that, you can then run conda update scipy --force
Upvotes: 1
Reputation: 754
Did you get any errors when you ran 'python setup.py install' for pyoptsparse? I looked at the setup.py file for pyoptsparse and there are some instructions in there that you could try:
print("\nTo install, run: python setup.py install --user\n\n"
"To build, run: python setup.py build_ext --inplace\n\n"
"For help on C-compiler options run: python setup.py build --help-compiler\n\n"
"For help on Fortran-compiler options run: python setup.py build --help-fcompiler\n\n"
"To specify a Fortran compiler to use run: python setup.py install --user --fcompiler=<fcompiler name>\n\n"
"For further help run: python setup.py build --help"
)
Upvotes: 0