Loss decreases but the acc is always 0?

[Piaoyyt]

In the trainig progress, the trainig loss is decreasing but the accuracy on the training dataset is always zero. I don't know why this happens?

[Jia-Xueyu]

In the trainig progress, the trainig loss is decreasing but the accuracy on the training dataset is always zero. I don't know why this happens?

Hello, I have the same question. Do you solve it now?

[Rui-coder]

Are you sure that the Loss Model is in the training state? Auto-grad-discrese should also apply to the loss model !!!

[yulongPan]

请问解决了吗