I found stack trace for the segfault:
```
#0  0x00002aaaab037706 in __memcpy_ssse3_back () from /usr/lib64/libc.so.6                                                                                         [298/337]
#1  0x00002aab2c6b2f2c in MPIDI_CH3U_Buffer_copy () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#2  0x00002aab2c6ded44 in MPIDI_Isend_self () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#3  0x00002aab2c6d8b18 in MPID_Isend () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#4  0x00002aab2c692d0b in MPIC_Isend () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#5  0x00002aab2c494fca in MPIR_Alltoall_Scatter_dest_MV2 () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#6  0x00002aab2c496f4e in MPIR_Alltoall_MV2 () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#7  0x00002aab2c447ca9 in MPIR_Alltoall_impl () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#8  0x00002aab2c448373 in PMPI_Alltoall () from /opt/mvapich2/gnu/6.3/2.2/lib/libmpi.so.12
#9  0x00002aaab553742f in __pyx_pf_6mpi4py_3MPI_4Comm_96Alltoall (__pyx_v_self=0x2aaab564a8a0, __pyx_v_recvbuf=<optimized out>, __pyx_v_sendbuf=<optimized out>) at src/mpi4
py.MPI.c:107259
#10 __pyx_pw_6mpi4py_3MPI_4Comm_97Alltoall (__pyx_v_self=0x2aaab564a8a0, __pyx_args=<optimized out>, __pyx_kwds=<optimized out>) at src/mpi4py.MPI.c:41665
#11 0x00005555556673d4 in _PyCFunction_FastCallDict () at /tmp/build/80754af9/python_1540319457073/work/Objects/methodobject.c:231
#12 0x00005555556f0c1c in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4837
#13 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#14 0x00005555556ead8b in _PyFunction_FastCall (globals=<optimized out>, nargs=8, args=<optimized out>, co=<optimized out>) at /tmp/build/80754af9/python_1540319457073/work
/Python/ceval.c:4919
#15 fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4954
#16 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858
#17 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#18 0x00005555556ead8b in _PyFunction_FastCall (globals=<optimized out>, nargs=2, args=<optimized out>, co=<optimized out>) at /tmp/build/80754af9/python_154031945[316/337]
/Python/ceval.c:4919
#19 fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4954
#20 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858                                                    [319/337]
#21 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#22 0x00005555556ea124 in _PyEval_EvalCodeWithName () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4166
#23 0x00005555556eb49c in _PyFunction_FastCallDict () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:5070
#24 0x000055555566779f in _PyObject_FastCallDict () at /tmp/build/80754af9/python_1540319457073/work/Objects/abstract.c:2310
#25 0x000055555566c303 in _PyObject_Call_Prepend () at /tmp/build/80754af9/python_1540319457073/work/Objects/abstract.c:2373
#26 0x00005555556671de in PyObject_Call () at /tmp/build/80754af9/python_1540319457073/work/Objects/abstract.c:2261
#27 0x0000555555714d9a in do_call_core (kwdict=0x0, callargs=0x2aaab79ab318, func=0x2aaab56e2cc8) at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:5106
#28 _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3404
#29 0x00005555556ead8b in _PyFunction_FastCall (globals=<optimized out>, nargs=1, args=<optimized out>, co=<optimized out>) at /tmp/build/80754af9/python_1540319457073/work
/Python/ceval.c:4919
#30 fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4954
#31 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858
#32 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#33 0x00005555556ead8b in _PyFunction_FastCall (globals=<optimized out>, nargs=1, args=<optimized out>, co=<optimized out>) at /tmp/build/80754af9/python_1540319457073/work
/Python/ceval.c:4919
#34 fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4954
#35 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858
#36 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#37 0x00005555556ea42e in _PyEval_EvalCodeWithName () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4166
#38 0x00005555556eafc1 in fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4978
#39 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858
#40 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335                                         [336/337]
#41 0x00005555556ead8b in _PyFunction_FastCall (globals=<optimized out>, nargs=0, args=<optimized out>, co=<optimized out>) at /tmp/build/80754af9/python_1540319457073/work
/Python/ceval.c:4919
#42 fast_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4954
#43 0x00005555556f0cf5 in call_function () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4858
#44 0x000055555571371a in _PyEval_EvalFrameDefault () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:3335
#45 0x00005555556ebad9 in _PyEval_EvalCodeWithName (qualname=0x0, name=<optimized out>, closure=0x0, kwdefs=0x0, defcount=0, defs=0x0, kwstep=2, kwcount=<optimized out>, kw
args=0x0, kwnames=0x0, argcount=0, args=0x0, locals=0x2aaaaab963f0, globals=0x2aaaaab963f0, _co=0x2aaab211ced0) at /tmp/build/80754af9/python_1540319457073/work/Python/ceva
l.c:4166
#46 PyEval_EvalCodeEx () at /tmp/build/80754af9/python_1540319457073/work/Python/ceval.c:4187
#47 0x00005555556ec87c in PyEval_EvalCode (co=co@entry=0x2aaab211ced0, globals=globals@entry=0x2aaaaab963f0, locals=locals@entry=0x2aaaaab963f0) at /tmp/build/80754af9/pyth
on_1540319457073/work/Python/ceval.c:731
#48 0x000055555576d074 in run_mod () at /tmp/build/80754af9/python_1540319457073/work/Python/pythonrun.c:1025
#49 0x000055555576d471 in PyRun_FileExFlags () at /tmp/build/80754af9/python_1540319457073/work/Python/pythonrun.c:978
#50 0x000055555576d673 in PyRun_SimpleFileExFlags () at /tmp/build/80754af9/python_1540319457073/work/Python/pythonrun.c:419
#51 0x000055555576d77d in PyRun_AnyFileExFlags () at /tmp/build/80754af9/python_1540319457073/work/Python/pythonrun.c:81
#52 0x00005555557711b0 in run_file (p_cf=0x7fffffffb80c, filename=0x5555558ac3e0 L"train_imagenet.py", fp=0x555555962930) at /tmp/build/80754af9/python_1540319457073/work/M
odules/main.c:340
#53 Py_Main () at /tmp/build/80754af9/python_1540319457073/work/Modules/main.c:811
#54 0x0000555555638b4e in main () at /tmp/build/80754af9/python_1540319457073/work/Programs/python.c:69
#55 0x00002aaaaaf0cc05 in __libc_start_main () from /usr/lib64/libc.so.6
#56 0x000055555571a1a8 in _start () at ../sysdeps/x86_64/elf/start.S:103