Newbie question about the mode=‘sow’ with a conda installation of v1.3.2, python3.6. I’m interested in parallelizing finite difference calculations with sow/reap for geometry optimizations (as suggested here). When using the pywrap-opt-sowreap sample, no evidence of sow/reap behavior is found: the calculation runs to completion, regardless of the number of threads specified (e.g. $ conda activate base; psi4 input.dat -n 5) and no BASIC files are generated. Happy to provide the output file if needed.
The sample prints the warning “This file will not work outside a cmake test environment,” and I’m pretty sure this warning is suggesting I do something that I’m currently not: any guidance would be greatly appreciated! I noticed that psi4.driver.optimize() and psi4.driver.hessian() both dispatch all finite displacements serially, however: is the desired behavior simply not implemented yet?
Additional, possibly helpful information:
The pywrap-db1 and -db3 samples run successfully, whereas the -db2 sample produces an error similar to that described in #833 after the three BASIC files are generated:
Traceback (most recent call last):
File “/home/wpm216/psi4conda/bin/psi4”, line 287, in
exec(content)
File “”, line 29, in
File “/home/wpm216/psi4conda/lib//python3.6/site-packages/psi4/driver/wrapper_database.py”, line 563, in database
pickle.dump(kwargs, freagent)TypeError: write() argument must be str, not bytes
-db2 runs successfully after removing {mode=‘sow’} from the db() command’s kwargs.
Happy to clarify anything. Thanks in advance!
-Wes