I am a graduate student at Beijing Jiaotong University, China. I am currently working on reproducing the results presented in your paper, "FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction", using the corresponding code.
However, I encountered an issue regarding the performance when I ran the code. Specifically, I noticed a significant gap between the local accuracy and the global accuracy. Despite following the instructions carefully, the global accuracy does not reach the results reported in the paper.
The first running:
The second running:
Could you kindly provide any insights or suggestions on what might be causing this discrepancy? I would greatly appreciate any guidance you can offer.
Thank you for your time and your excellent work on this project. I look forward to hearing from you.
I am a graduate student at Beijing Jiaotong University, China. I am currently working on reproducing the results presented in your paper, "FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction", using the corresponding code.
However, I encountered an issue regarding the performance when I ran the code. Specifically, I noticed a significant gap between the local accuracy and the global accuracy. Despite following the instructions carefully, the global accuracy does not reach the results reported in the paper.
The first running:
The second running:
Could you kindly provide any insights or suggestions on what might be causing this discrepancy? I would greatly appreciate any guidance you can offer.
Thank you for your time and your excellent work on this project. I look forward to hearing from you.