The file QGRNN-for 5 node cyclic graph simulation contains the results of the paper: https://arxiv.org/abs/1909.12264. Also I took help from pennylane tutorial to write the code.
The only difference I followed is instead of considering the true ground state I took random initial state and trained the circuit accordingly. Fortunately that trick doen't effect the actual result and the method can efficiently learn the actual hamiltonian for any size of graph. Here I took particularly 5 node cyclic graph to replicate the results of the actual paper given above....cheers