Acceleration for Compressed Gradient Descent in Distributed Optimization