Federated Learning Loss not logged correctly?

[LSnyd]

Hi there,

I am currently trying to get a Federated Learning training running according to the Spleen Example.

I got the training process running and get a somehow meaningful output. However, when I am checking my loss on tensorboard (from fl_workspace/server/run_1/mmar_client1/custom/models)

I am receiving the following output:

It seems like the loss is not correctly logged over the 200 global epochs ( 5 local epochs). Could anyone help me out with that?

[LSnyd]

Hi @yiheng-wang-nv,
as you were tagged in my other FL problem, I am assuming that you are the FL expert :)
Could you maybe also help me out with this problem?

[LSnyd]

I just found out that I am getting this result, since multiple tensorboard logfiles are included and overlapped in this plot. It looks fine, when I am only keeping the log results of one run.

[yiheng-wang-nv]

Hi @LSnyd , sorry for late reply, for this question do you still have other questions?