Error in performing example/04-parallel_mol_hf.py

[jiangth1997]

I am trying to install and use mpi4pyscf to accelerate an ROHF calculation on a cluster containing (1784o,1265e). Now I've just installed mpi4py and mpi4pyscf, and trying a test calculation of UHF on water trimer, which is revised based on example/04-parallel_mol_hf.py.

#!/usr/bin/env python

import numpy
from pyscf import gto, scf
from mpi4pyscf import scf as mpi_scf

mol = gto.M(atom='''O -1.340565 0.901693 0.069532
H -1.199374 -0.057160 0.031951
H -1.953650 1.097896 -0.638937
O 1.436646 0.695966 -0.004172
H 0.571354 1.130319 -0.001360
H 2.096731 1.386248 -0.026654
O -0.108265 -1.612993 -0.063996
H 0.696797 -1.077090 -0.043847
H -0.017920 -2.236192 0.657113''',
basis='cc-pvtz')

mf = mpi_scf.UHF(mol)
mf.direct_scf_tol = 1e-9
mf.verbose = 6
mf.kernel()

The Python script is performed with the following bash command.

mpirun -np 2 python 04-parallel_mol_hf.py

I got the following errors.

(base) [1600011363@a7u09n08 mpi4pyscf_molecule]$ mpirun -np 2 python mol_mpi4pyscf.py > mol_mpirun.out
Traceback (most recent call last):
File "/gpfs/share/home/1600011363/.local/lib/python3.7/site-packages/mpi4pyscf/init.py", line 31, in
mpi.pool.wait()
File "/gpfs/share/home/1600011363/.local/lib/python3.7/site-packages/mpi4pyscf/tools/mpi_pool.py", line 112, in wait
ans = self.function(*task) # task = worker_args
File "/gpfs/share/home/1600011363/.local/lib/python3.7/site-packages/mpi4pyscf/tools/mpi.py", line 577, in _distribute_call
return fn(dev, *args, **kwargs)
File "/gpfs/share/home/1600011363/.local/lib/python3.7/site-packages/mpi4pyscf/scf/hf.py", line 43, in get_jk
vj, vk = _eval_jk(mf, dm, hermi, _jk_jobs_s8)
File "/gpfs/share/home/1600011363/.local/lib/python3.7/site-packages/mpi4pyscf/scf/hf.py", line 115, in _eval_jk
dm = numpy.asarray(dm).reshape(-1,nao,nao)
ValueError: cannot reshape array of size 1 into shape (174,174)
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 1
(base) [1600011363@a7u09n08 mpi4pyscf_molecule]$ *** Error in `/bin/srun': double free or corruption (fasttop): 0x00000000016f52b0 ***
======= Backtrace: =========
/lib64/libc.so.6(+0x81679)[0x7f37d5679679]
/usr/lib64/slurm/libslurmfull.so(slurm_xfree+0x25)[0x7f37d6274668]
/usr/lib64/slurm/libslurmfull.so(+0x5c482)[0x7f37d6159482]
/usr/lib64/slurm/libslurmfull.so(list_destroy+0xbc)[0x7f37d61ae649]
/usr/lib64/slurm/libslurmfull.so(client_io_handler_destroy+0xe3)[0x7f37d615b8a4]
/usr/lib64/slurm/libslurmfull.so(slurm_step_launch_wait_finish+0x4ba)[0x7f37d615fbb1]
/usr/lib64/slurm/launch_slurm.so(launch_p_step_wait+0x22)[0x7f37d456a681]
/bin/srun(launch_g_step_wait+0x2f)[0x40be1d]
/bin/srun[0x407c4d]
/bin/srun(srun+0x1527)[0x409278]
/bin/srun(main+0x9)[0x409730]
/lib64/libc.so.6(__libc_start_main+0xf5)[0x7f37d561a505]
/bin/srun[0x407079]
======= Memory map: ========
......

It seems mpi4pyscf has been installed, but an error occurs in mpi4pyscf.scf.hf. How to fix this problem?

The version of dependencies are listed as follows.
conda = 4.10.3 python = 3.7.1 mpi4py = 3.1.3 mpi4pyscf = 0.3.0
mpi4py was installed with OpenMPI (version = 3.0.0) compiled with Intel compilers (2018.0).