error duing compilation of user-deepmd module in lammps

[hamelseb]

hello

I compiled the deepmd-kit (using tensorflow-1.5.0 and gnu-4.9.3) and made the lammps module. the serial version of lammps without the deepmd module compiles fine.
however when I tried to compile lammps with the deepmd module I'm seeing the following errors:

g++ -g -O3 -DLAMMPS_GZIP -DLAMMPS_MEMALIGN=64 -std=c++11 -DHIGH_PREC -I/g/g99/hamel2/build_tf/1.5.0/gpu/include -I/g/g99/hamel2/deepmd-kit/include/deepmd -std=c++11 -DHIGH_PREC -I/g/g99/hamel2/build_tf/1.5.0/gpu/include -I/g/g99/hamel2/deepmd-kit/include/deepmd -std=c++11 -DHIGH_PREC -I/g/g99/hamel2/build_tf/1.5.0/gpu/include -I/g/g99/hamel2/deepmd-kit/include/deepmd -std=c++11 -DHIGH_PREC -I/g/g99/hamel2/build_tf/1.5.0/gpu/include -I/g/g99/hamel2/deepmd-kit/include/deepmd -I../STUBS -c ../pair_nnp.cpp
../pair_nnp.cpp: In member function 'virtual void LAMMPS_NS::PairNNP::compute(int, int)':
../pair_nnp.cpp:33:20: error: invalid use of incomplete type 'class LAMMPS_NS::Atom'
double **x = atom->x;
^
In file included from ../pointers.h:26:0,
from ../memory.h:18,
from ../pair_nnp.cpp:3:
../lammps.h:29:9: error: forward declaration of 'class LAMMPS_NS::Atom'
class Atom *atom; // atom-based quantities
^
...

any suggestions?

[deepmodeling]

Which lammps version did you use?

[hamelseb]

I downloaded the latest stable release

[deepmodeling]

I can reproduce the reported error. Let me work on it.

A fast solution for you should be switching to lammps-31Mar17.

[hamelseb]

ok thx. i'm will try that now.

[hamelseb]

ok that compiled. interesting. good luck on finding the issue with the newer version.

[deepmodeling]

bug fixed.