Wrong loss function?

[mengcz13]

I notice that the code uses CrossEntropyLoss for local training:

LG-FedAvg/models/Update.py

Line 28 in 7af0568

self.loss_func = nn.CrossEntropyLoss()

And it accepts the log-probabilities as input:

LG-FedAvg/models/Update.py

Line 50 in 7af0568

loss = self.loss_func(log_probs, labels)

The output of CNN networks is also logsoftmax:

LG-FedAvg/models/Nets.py

Line 104 in 7af0568

return F.log_softmax(x, dim=1)

But according to the doc of PyTorch, CrossEntropyLoss already has logsoftmax inside:
https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html#torch.nn.CrossEntropyLoss

I think the loss should be calculated via NLLLoss instead if used with input after logsoftmax (https://pytorch.org/docs/stable/generated/torch.nn.NLLLoss.html#torch.nn.NLLLoss).

Or is there any reason why the code uses logsoftmax twice for calculating the loss?

[CarlBye]

Same issue