Xiaoquan Kong

@bhack This issue is CRF related, so please label it with `crf`.

@xuxingya Also, you may find this PR #2552 helps. It uses Custom Training Loop programming style instead of Custom Training Step. Both methods can work.

Hi @WindQAQ, thank you for your detailed report. Since this issue was written about 10 months ago, is there any update about those Pros and Cons of both choices? Could...

@bhack Thank you! I will try to do this by myself.

Regarding to `returns the probabilities for every tag for each token in sequence`. Do you means you want the emission probability of each words? if so, I think you can...

@Ghostvv Thank you for the information you provided. I will do some investigation on the PR. Then, hopefully, do some work based on that.

Hi @YangHan-Morningstar, This problem is caused by incorrect use of CRF. I will give a small demonstration of how to use CRF correctly. I need some time to build this...

Hi @YangHan-Morningstar I just submit a **WIP** tutorial for CRF layer at #2552 . It will show you how to use CRF layer. If you still have any questions about...

Hi @YangHan-Morningstar if my tutorial in the previous comment works for you, please close this PR for me. If it does not work for you or if you have any...

@DLPerf Thank you for your reporting. I will take a closer look on it and I will keep you posted.