Missing module error in Federated Learning with NVFlare tutorial

[wyli]

Discussed in #376

^{Originally posted by sorensenmarius September 28, 2021}
I have followed each step of the federated learning with NVFlare without Docker tutorial, and have run into a problem when I want to start the server and the clients. When I do the "Run federated learning" step in the README I am getting this error from sub_start.sh:
/tutorials/federated_learning/nvflare/nvflare_example/nvflare_monai/bin/python3: No module named nvflare.private.fed.app.client.client_train
and
/tutorials/federated_learning/nvflare/nvflare_example/nvflare_monai/bin/python3: No module named nvflare.private.fed.app.server.server_train

I have tried redoing the whole installation and manually installing nvflare with both pip and pip3, however these tell me that all requirements are satisfied.

Has anyone run into this problem? I would love some input on how to move forward with this.

[LSnyd]

Hi @wyli,
I got it working by updating the versions of the monai and nvflare packages in the requirement.txt file:

monai==0.7.0
nvflare==1.0.2

Hope that solves it for you as well!

[wyli]

thanks, this is addressed by #291.

[TaocsZhang]

I do have the same issue, even if I use monai==0.7.0 and nvflare==1.0.2.

So the quickstart tutorial in nvflare is working or not?

[TaocsZhang]

This is the thing displaying when i run ./poc/server/startup/start.sh or ./poc/site-1/startup/start.sh site-1 localhost.

Does anyone run into the same issue?

Thanks.

The following files are not secure content.
fed_server.json
server.crt
server.key
rootCA.pem
System is in trouble and unable to start the task!!!!!

[Anand-afk]

@TaocsZhang I am facing the same issue, how you got any resolution to it ?
I am facing something like below :

The following files are not secure content.
fed_server.json
start fl because process of 2026 does not exist
System is in trouble and unable to start the task!!!!!

Thanks.

[pipi9baby]

@Anand-afk I solved this problem by using the same version of nvflare in every environment.
In my case, my provision tool works in nvflare==2.0.16
my federated leaning server works in nvflare==2.1.2

source code
I found that signature_filename is signature.json in 2.1.2
In 2.0.16, the signature_filename is signature.pkl