How to supress layer output to the console?

[Ilya Zhenin]

I'm training a net with last SIgmoid layer that used only on phase of TEST, and Sigmoid Cross Entropy Loss, since loss has its own sigmoid layer within.

And that last sigmoid layer prints its whole training batch output to the console, I do not understand why he does that. 

Here is that piece of net:

layer {
  name: "sigmoid"
  bottom: "ip2"
  top: "en1neuron"
  type: "Sigmoid"
}

layer {
  name: "accuracy"
  type: "Accuracy"
  bottom: "en1neuron"
  bottom: "label"
  top: "accuracy"
  include {
    phase: TEST
  }
}
layer {
  name: "loss"
  type: "SigmoidCrossEntropyLoss"
  bottom: "ip2"
  bottom: "label"
  top: "loss"
}

 

[Ilya Zhenin]

Oh, figured that, it was quiet obvious, just got confused since fought that same cofiguration in previous time do not produced any such output.

Just make sigmoid layer included only on TEST phase solved the problem.

вторник, 17 мая 2016 г., 11:37:18 UTC+3 пользователь Ilya Zhenin написал: