Weird (bad) behaviour while starting experiment!🔧

[daviddelriod]

I have been trying to start a custom FL experiment in two different ways:

• Via Visual Studio in local:
  The cell gets stuck after submitting the experiment to the director. The problem is related with some folders or requirements that cannot be parsed correctly. Moreover, as you can see in the last part of the video, large amount of files are generated (+1k) while trying to start the experiment.
  https://user-images.githubusercontent.com/64251001/223852024-7d016624-80f8-47a1-b11f-9031fd4da4d4.mp4 (ignore the football match and other tabs in the video😅)

• Via Google Colab (launched locally):
  When starting the experiment, I get an error while building the object FrameworkAdapterPlugin. The error I'm getting is related with the previous one: [WinError 183] Cannot create a file when that file already exists: 'temp_clash/workspace'. Important to mention that in the first attempt that I tried starting the experiment, I did not get the error but the experiment was stuck (it did not start training).

I can not figure where is the problem. Any help would be appreciated. Feel free to ask for additional info!

[daviddelriod]

The issue seemed to be related with training of the net. There were some errors in the loop`but it was very difficult to solve them as the debug message was almost unreadable. However, it still threw the error of the requirements:

But the plan kept running so I just omitted the error.