Question about the step in wandb log about fedml HOT 4 CLOSED

You can edit the x-axis to "round" by clicking the button I highlighted in the picture above

from fedml.

Thanks for your answer @chaoyanghe . And I have another question about the cifar10 result. I didn't find the result of cifar10 in the standalone FedAvg readme document (https://github.com/FedML-AI/FedML/tree/master/fedml_experiments/standalone/fedavg). I notice in the benchmark that there is a cifar FedAvg log (https://wandb.ai/automl/fedml/runs/2cxam561/) but I tried the hyperparameter setting of this log and only got an acc of 0.35. I ran the code on 1 gpu. Is there anything I did wrong or the result of FedAvg on cifar10 is just as low as 0.35? BTW I notice there is cifar10 result in the standalone FedNova doc.

from fedml.

maybe the hyper-parameters you find is not the optimal one or some contributors changed it. You can tune hyper-parameters by yourself, like changing learning rate, batch size, etc...

from fedml.

it should be > 90%

from fedml.