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
Until we move the questions to another medium, feel free to use this to submit your question:
Question
I run the FLsim cifar10 example, change the model from simple CNN to Resnet, and run the same dataset. The evaluate accuracy is very low. The report as:
Train finished Global Round: 2
(round = 2, epoch = 1, global round = 2), Loss/Training: 1.919056011840796
(round = 2, epoch = 1, global round = 2), Accuracy/Training: 29.19
(round = 2, epoch = 1, global round = 2), Loss/Aggregation: 2.3162186018220936
(round = 2, epoch = 1, global round = 2), Accuracy/Aggregation: 12.362
(round = 2, epoch = 1, global round = 2): Evaluates global model on all data of eval users
(round = 2, epoch = 1, global round = 2), Loss/Eval: 2.315063210050012
(round = 2, epoch = 1, global round = 2), Accuracy/Eval: 12.2
Current eval accuracy: {'Accuracy': 12.2}%, Best so far: {'Accuracy': 10.01}%
❓ Questions and Help
Until we move the questions to another medium, feel free to use this to submit your question:
Question
I run the FLsim cifar10 example, change the model from simple CNN to Resnet, and run the same dataset. The evaluate accuracy is very low. The report as:
Train finished Global Round: 2
(round = 2, epoch = 1, global round = 2), Loss/Training: 1.919056011840796
(round = 2, epoch = 1, global round = 2), Accuracy/Training: 29.19
(round = 2, epoch = 1, global round = 2), Loss/Aggregation: 2.3162186018220936
(round = 2, epoch = 1, global round = 2), Accuracy/Aggregation: 12.362
(round = 2, epoch = 1, global round = 2): Evaluates global model on all data of eval users
(round = 2, epoch = 1, global round = 2), Loss/Eval: 2.315063210050012
(round = 2, epoch = 1, global round = 2), Accuracy/Eval: 12.2
Current eval accuracy: {'Accuracy': 12.2}%, Best so far: {'Accuracy': 10.01}%
The text was updated successfully, but these errors were encountered: