Elmo Linking Rule Latency

[andrew-j-luong]

Hi,

We are using the native AllenNLP Elmo Linking in the Tagging/Linking Rules, and it seems to stall at this step. Have you seen any historical issues with this step?

I think some clarity on how long the step takes to run? OR how we may speed up the process would be greatly appreciated.

Thank you!

[andrew-j-luong]

Hi,

Follow-up on the question - I have found a way to spell out the time.
By utilizing the tqdm in the rules.py you can add a progress bar to each of the tr.apply() or lr.apply() file

Code:

**from tqdm.auto import tqdm
tqdm.pandas()**

class TaggingRule:
    def apply(self, instances):
        for instance in **tqdm**(instances):

This way you can show how long an apply for tagging/linking rule can take per instance (sentence token).

Thanks

[safranchik]

Thank you for your suggestion!

Adding the tqdm_auto wrapper looks like the best way to display the progress to users.