Strange issue : OOV when testing, while training is solid

[Chia-Hsuan-Lee]

A weird situation using dev branch on tensorflow1.2

I use command below train python -m basic.cli --mode train --noload --len_opt --cluster and this command to test python -m basic.cli --len_opt --cluster

In the same GPU device (which is fully available), I soundly finished training However, OOV issue just ruin the testing process

Does anyone have the same issue ? And how to solve it ? Very appreciate it !

[aneesh-joshi]

The repo doesn't support anything other than tf 0.11. So, you can expect unexpected stuff.

I was unable to port it myself from 0.11 to 1.2 Were you able to resolve the problem and could you provide a link to the ported version? @chiahsuan156