Loading from existing datasets error

, ,
1

Hello Ines,
I have been trying to load in our images annotated data from the current existing dataset that is being processed by our labelers. I always get this error :

✘ Error while validating stream: no first example
This likely means that your stream is empty.

I actually db-out the current dataset being worked on, filter out the accept with a grep command, than pass that through a py script to change the text values in case they are the same.
Funny enough, loading from the dataset works with datasets that were created with a b64 jsonl file, that got encoded previously.
Is there a way to bypass the encoding and load the dataset that currently is being worked on?
Thank you

2

Hi! That message means that there are no new examples available for annotation, which could be caused by different things: for example, if there's no "image" value in the data, or if all examples in the stream have already been annotated in the current dataset (in which case Prodigy skips them so you're not being asked the same question twice). So if you're re-annotating examples, make sure you're using a new dataset.

1 Like