You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
Unexpected behavior observed when simulating the different straggler handling scenarios.
Scenario:
Any one collaborator manually terminated during the execution of the experiment.
Parameters for CutoffTimeBasedStragglerHandling policy:
Cutoff time: 10 seconds
Minimum reporting: 1
Total collaborators: 2
Expected behavior:
After the cutoff time (10 seconds) has elapsed aggregator should move to 2nd round as it has received results from minimum required collaborators (1 collaborator)
Observed behavior:
Aggregator keeps waiting for the results from terminated collaborator even though Cutofftime has expired and Minimum reporting collaborators have reported the results
To Reproduce
Steps to reproduce the behavior:
Go to: openfl/component/straggler_handling_functions/cutoff_time_based_straggler_handling.py
Change: straggler_cutoff_time parameter default value from np.inf to 10 in CutoffTimeBasedStragglerHandling class constructor.
Tutorial: Modify openfl-tutorials/interactive_api/PyTorch_LinearRegression tutorial to run with 2 collaborators and run the experiment
While the experiment is in progress terminate any one collaborator before it sends the results of the particular round.
See error: Aggregator will not start next round and will keep waiting for the terminated collaborators to send results.
Desktop (please complete the following information):
Describe the bug
Unexpected behavior observed when simulating the different straggler handling scenarios.
Scenario:
Cutoff time: 10 seconds
Minimum reporting: 1
Total collaborators: 2
Expected behavior:
Observed behavior:
To Reproduce
Steps to reproduce the behavior:
Desktop (please complete the following information):
straggler-handling-policy-logs.zip
The text was updated successfully, but these errors were encountered: