Secure Federated Learning Across Heterogeneous Cloud and High-Performance Computing Resources -- A Case Study on Federated Fine-tuning of LLaMA 2

Abstract--Federated learning enables multiple data owners to collaboratively train robust machine learning models without transferring large or sensitive local datasets by only sharing the parameters of the locally trained models. In this paper, we elaborate on the design of our Advanced Privacy-Preserving Federated Learning (APPFL) framework, which streamlines end-to-end secure and reliable federated learning experiments across cloud computing facilities and high-performance computing resources by leveraging Globus Compute, a distributed function as a service platform, and Amazon Web Services. We further demonstrate the use case of APPFL in finetuning a LLaMA 2 7B model using several cloud resources and supercomputers. The server aggregates these model parameters and redistributes the updated parameters to the clients for further local training iterations. As FL does not require collecting and storing distributed client datasets together as a centralized dataset, it is becoming an increasingly promising approach to train a more robust machine learning model and alleviate the domain shift problem without compromising the privacy of local training datasets.3