📝 New preprint released! FLAME: Towards Federated Fine-Tuning Large Language Models Through Adaptive SMoE