test,py get different result with train,py

[ArtemisZGL]

when i load the checkpoint to use test.py to test the valid set, it get very higher wer and cer, which is much higher than valid cer showing in training. Why would this happen ？

[ArtemisZGL]

all of the result str from greedy search is the same

[shibo0]

Hello, I also encountered the same problem, the prediction results obtained from greedy search are only slightly different, wer and cer are very high. I don't know if you have solved this problem.

[ArtemisZGL]

@shibo0 sorry, I didn't solve this problem and now I use another repo.

[shibo0]

Can I take the liberty to ask your email address? I have some questions I want to discuss with you. Thank you.

[Chriszhangmw]

https://github.com/Chriszhangmw/end2end-asr-pytorch/blob/master/predict.py