Closed hmaster6 closed 1 hour ago
I set that error. pFedHN is incompatible with parallel training due to the hypernetwork. pFedHN needs the hypernetwork to generate model parameters. So pFedHN needs to update its hypernetwork right before the next client's training.
BTW, thanks for pointing out the typo. It will be fixed in no time.
Maybe you can change the update way of hypernetwork to aggregate gradients from more than 1 clients like FedAvg does. That can make pFedHN training parallelized, but the results is not guaranteed.
Thank you for your response, and I will try this.^_^
I encountered an error while using the pFedHN algorithm: parallel computing is not supported. I want to know if I have a way to use parallel computing by changing the configuration.Thank you! By the way, the word 'parallel' below seems to be misspelled.