With the increasing importance of data privacy protection, federated learning is becoming more widely used, and there are more frameworks for federated learning. However, the centralization of federated learning has always restricted the development of federated learning and the federated learning framework. Although there are some decentralized federated learning algorithms, these algorithms have some shortcomings and there is no framework that can quickly use these algorithms. In this paper,we proposed and implemented a blockchain-based decentralized federated learning framework called GFL \cite{GFL} and integrated two new blockchain-based decentralized federated learning mechanisms to try to try to make the decentralized federated learning algorithm better landed.