Problems on using evalution

[heroding77]

When predicting the MNIST task with CNN, In the period of train, in conf.json, I use "encode_label=true", I checked the logs and found that the training process was fine, but the results in the evaluation stage were poor, I guessed that the one-hot processing of y was not performed in the evaluation stage, but I don't know how to deal with it. Can someone help me?

[heroding77]

I got the answer!

"evaluation_0": {
  "eval_tpye"="multi"
}