No tasks available with ner.correct and prodigy 1.11.0

,
1

Hi,

I updated to prodigy 1.11.0 and I'm getting a "No tasks available using" ner.correct, although I have a lot of text still to annotate.(My corpus has about 5000 json lines). A similar issue started before the update while using the beta: prodigy sent me back to the beginning of the dataset and offered texts that I had already annotated.

The problem began, I think, after I trained a NER model and switched to ner.correct with --update. The database is SQLite.

1 Like
2

Hi! So just to confirm, the remaining examples in your dataset are all examples with an _input_hash that's not yet in the current dataset you're saving your annotations to? One simple way to check this would be to load your input examples, call prodigy.set_hashes() on each of them and then check the _input_hash values against the list of input hashes in the dataset.

3

@ines @jcbmyrstn for what it's worth, I observe the same.
I have a jsonl file that I have partially annotated with 1.10.8, annotations are kept in a dataset in a conda env.
I wanted to extend the dataset switching to 1.11.0, and for this I:

  • exported the dataset with db-out
  • imported the dataset with db-in in another conda env where I have 1.11.0 installed
  • trying ner.correct using 1.11.0

I also get "no tasks available". So far I completed 5000 records out of ~9000 examples so @ines input hashes in the dataset (which should be around 5000) can't exhaust all the 9000 examples. Also, ner.correct still streams outstanding examples happily when I do that in the 1.10.8 conda env.

Unfortunately it would take me time to manufacture a reproducible example (can try), and I can't share my original data/annotations :frowning_face:

4

I tried to reproduce this today but wasn't able to. There was a small bug in 1.11.0 with the *.correct recipe that was fixed in 1.11.1 that was released today. If you're able to try with this release that would great! Even if that doesn't fix your issue, we can at least rule that known bug out as the cause.

Thanks!

5

@kab thanks! I tried 1.11.1, and the error is not there anymore.

I accidentally lost track if my db-out and db-in steps need to be performed with 1.10, 1.11 or that doesn't make any difference, but it seems that I can continue annotating my dataset using 1.11.1 now, I don't know if that solves for @jcbmyrstn and others as well.

1 Like
6

Updating to 1.11.1 did solve the issue. Thanks!

2 Likes
7

Awesome!

1 Like