Here you can find example input files: http://lammps.sandia.gov/scripts.html
The input file in.graphene will generate a xyz file of a graphene sheet.
You could use the Tersoff potential for the interactions. But, I suggest you go through the literature and check what other people consider to be the best potential to reproduce the elastic response of graphene.
I can think two reasons to that problemas: 1-bad physics : check if there are overlaps in your atoms positions. You can use VMD package for visualization. It would be good if you make some minimization before doing your simulation and check the forces. Probably, I believe, some atom is experiencing a huge force and is traveling too fast to lammps. 2 - your neighbor's list must be updated more frequently. But as you are working with graphene (and not some liquid or gas), it would not be necessary, so I believe that the reason is the first one