Kemal Kurniawan

Hi, I'm not sure I understand the question. Basically, `num_tags` should be set to the size of possible tags for a token. So, if your task is NER with 3...

Thanks for this. I have allowed the tests to run, and it seems they're all succeeding! Another thing left is then to update the minimum requirement in `docs/index.rst` to say...

Thanks for the PR! Looks good but I have to confirm it works for earlier PyTorch versions. This is usually done by Travis CI but since the move to travis.com...

Hi, I've setup Github CI to replace Travis, so do you mind rebasing this PR onto the master branch?

> project requirements don't say anything about pytorch version, this is only set in the test config Yes, this is intentional as there are several ways to install pytorch depending...

Thanks @myedibleenso. The builds for PyTorch v1.1 have failed. I can't see the log any more so I'm not sure why. I can't seem to re-run the builds either ☹️

Hi, thanks for using the library. One potential issue is that you convert the scores to long for the CRF. The scores are supposed to be floats. Can you try...

Hi, just letting you know that ACL Anthology is now hosted on https://aclanthology.org/. The old Anthology URL still works, but the new one is preferred. It'd be great if the...

@TidorP is right. Please set those indices to a value between 0-8 before passing it through the CRF layer. You can restore them afterwards.

@siddharthtumre Just remove the [CLS] token before feeding into the CRF layer. So something like ```python scores = scores[:, 1:] tags = tags[:, 1:] ``` should work (assuming the first...