Segmentation fault (core dumped)

[njzjz]

When I use dp_train, an error occurred:

2018-08-07 11:39:39.538273: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
Segmentation fault (core dumped)

I don't know what went wrong.

[amcadmus]

Could you provide more details on your training? what are your training parameters and data?

[njzjz]

Here are all files I used:
C1111.tar.gz

I want to train complex systems containing different types of atoms. I used conda install deepmd -c deepmd.kit -y to install DeePMD. Then I converted raw data to numpy data and ran dp_train C1111.json command at the train directory.

[amcadmus]

For O, you used the axis rule [0, 0, 0, 0, 0, 1], which asks for at least two C(type 0) neighbours. Some of your molecules have only one carbon, so the dp_train failed to make the axes for them.

[amcadmus]

Btw, there is not virial in your data, please set the virial prefactors to 0.

[njzjz]

I have changed axis_rule to [0,-2,0,0,-2,1,0,-2,0,0,-2,1,0,-2,0,0,-2,1] but it still occurred even in the latest version. I analyzed core dump file with gdb:

Using host libthread_db library "/lib64/libthread_db.so.1".
Missing separate debuginfo for /opt/conda3/lib/python3.6/site-packages/numpy/../../../libiomp5.so
Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/50/126f8244a53ed88a1531546fcaa8dedc4bc85c.debug
Core was generated by `python3 /opt/deepmd/bin/dp_train ../train/C1111.json'.
Program terminated with signal 11, Segmentation fault.
#0  0x00007f9c856c241c in compute_descriptor(std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> >&, std::vector<double, std::allocator<double> > const&, int const&, std::vector<int, std::allocator<int> > const&, SimulationRegion<double> const&, bool const&, int const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, std::vector<int, std::allocator<int> > const&, int, int, int, int) ()
   from /opt/deepmd/lib/deepmd/libop_abi.so
Missing separate debuginfos, use: debuginfo-install glibc-2.17-260.el7.x86_64 libuuid-2.23.2-59.el7.x86_64

It seems the error occurred in compute_descriptor but I have no idea about it.