Training always ends

[yiren556]

Excuse me?
Why does the process still end when the training reaches epoch9 after the configuration file is modified

[yiren556]

This is my modified configuration file

[byeonghu-na]

Hello, sorry for late reply.

I found that the baseline code also occurs same problem. (FangShancheng/ABINet#11, FangShancheng/ABINet#52, FangShancheng/ABINet#79)

From these replies and my trial, you should change the sum of optimizer.scheduler.periods in the configuration file to epochs, e.g. [400, 100] for 500 epochs.

[byeonghu-na]

For convenience, I add assertion of scheduler periods and epochs. (12fd00a)
Thank you for your contribution!