B3LYP geoopt crash


I encountered this problem recently while optimising some organic molecules with B3LYP.
Here is the error report.

Traceback (most recent call last):
  File "geoopt.py", line 53, in <module>
    E, wf = psi4.optimize('B3LYP', molecule=mol, return_wfn=True)
  File "/home/federif1/bin/psi4/lib/psi4/driver/driver.py", line 1052, in optimize
    G, wfn = gradient(lowername, return_wfn=True, molecule=moleculeclone, **kwargs)
  File "/home/federif1/bin/psi4/lib/psi4/driver/driver.py", line 691, in gradient
    wfn = procedures['gradient'][lowername](lowername, molecule=molecule, **kwargs)
  File "/home/federif1/bin/psi4/lib/psi4/driver/procrouting/proc.py", line 2064, in run_scf_gradient
    ref_wfn = run_scf(name, **kwargs)
  File "/home/federif1/bin/psi4/lib/psi4/driver/procrouting/proc.py", line 2002, in run_scf
    scf_wfn = scf_helper(name, post_scf=False, **kwargs)
  File "/home/federif1/bin/psi4/lib/psi4/driver/procrouting/proc.py", line 1289, in scf_helper
    old_wfn = core.Wavefunction.from_file(read_filename)
  File "/home/federif1/bin/psi4/lib/psi4/driver/p4util/python_helpers.py", line 156, in _core_wavefunction_from_file
    wfn_data = np.load(wfn_data).item()
  File "/share/apps2/anaconda/anaconda3/latest/lib/python3.6/site-packages/numpy/lib/npyio.py", line 447, in load
  File "/share/apps2/anaconda/anaconda3/latest/lib/python3.6/site-packages/numpy/lib/format.py", line 692, in read_array
    raise ValueError("Object arrays cannot be loaded when "
ValueError: Object arrays cannot be loaded when allow_pickle=False

Any clue?



I am not up to date on the discussion about it, but your problem seems related to this:

As someone who is up-to-date on the discussion: Yes, your problem is exactly what @hokru linked.

Internally, Psi uses the NumPy module. As of two days ago, NumPy changed the default argument to one of their functions, which broke some important functions in Psi. We have the functions working again on the developer version of Psi, and they will be backported as part of Psi 1.3.2.

I’d need to know how you installed Psi (installer, conda, or source?) to have a recommendation for you.

On the cluster I used the anaconda3 module to get python3 and Numpy. Then I compiled psi4 from source, using intel compiler and MKL.
Should I switch to dev branch and recompile?

If you want Psi to work again ASAP, either switch to dev branch and recompile, or downgrade your version of NumPy to before 1.6.3. You can also wait until we release 1.3.2, which should be soon, but I don’t have a precise date.