Reinforcement Learning with Sparse Rewards using Guidance from Offline Demonstration