Federated Fine-tuning on Federated Fine-tuning Simulation Environment
1.86Time per Round (min)FFA-LoRA
Evaluation results
| Method | Links | |
|---|---|---|
| FFA-LoRABackbone=Llama3.2-3B2025.10 | 1.86 | |
| FedAvgBackbone=Llama3.2-3B2025.10 | 1.92 | |
| Fed-DAREBackbone=Llama3.2-3B2025.10 | 1.92 | |
| FedProxBackbone=Llama3.2-3B2025.10 | 2.1 | |
| FedSRD-eBackbone=Llama3.2-3B2025.10 | 2.1 | |
| FFA-LoRABackbone=Qwen2-7B2025.10 | 3 | |
| FedAvgBackbone=Qwen2-7B2025.10 | 3.06 | |
| Fed-DAREBackbone=Qwen2-7B2025.10 | 3.06 | |
| FedSRD-eBackbone=Qwen2-7B2025.10 | 3.15 | |
| FedProxBackbone=Qwen2-7B2025.10 | 3.18 | |
| FlexLoRABackbone=Llama3.2-3B2025.10 | 7.2 | |
| FedSRDBackbone=Llama3.2-3B2025.10 | 7.92 | |
| FlexLoRABackbone=Qwen2-7B2025.10 | 12.6 | |
| FedSRDBackbone=Qwen2-7B2025.10 | 13.5 | |
| FedICUBackbone=Llama3.2-3B2025.10 | 30.36 | |
| FedICUBackbone=Qwen2-7B2025.10 | 51.6 |