Yizhi Zhou

Have you solved the problem?I have the same problem.

> i had solved the problem, pull and make it. Dear author,Could you please share your solution? thank you.

> @timohaha add HSWISH and HSIGMODE to darknet.h i know。but when i changed darknet.h file. Other problems arose. `./src/yolo_layer.c: In function ‘forward_yolo_layer’: ./src/yolo_layer.c:479:25: warning: variable ‘best_match_t’ set but not used...

这是我复现过程中的完整参数设置，除了展示的fedprox外，其他方法存在同样问题。 {"option": {"sample": "md", "aggregate": "uniform", "num_rounds": 100, "proportion": 0.6, "learning_rate_decay": 0.998, "lr_scheduler": -1, "early_stop": -1, "num_epochs": 2, "num_steps": -1, "learning_rate": 0.1, "batch_size": 64.0, "optimizer": "SGD", "clip_grad": 0.0, "momentum": 0.0,...

是的，确实像您所说我所构建数据分布是non-iid且imbalance的。很大概率会导致某次反向传播出现了nan数值。是否可以在聚合时加上判断，来消除这类错误。