exclude_by warning


I've been working with the ner.correct recipe for the last couple of weeks.
I started getting a new warning yesterday, when running the recipe on a new dataset:

:warning: Config setting 'exclude_by' defined in recipe is overwritten by a
different value set in the global or local prodigy.json. This may lead to
unexpected results and potentially changes to the core behavior of the recipe.
If that's surprising, you should probably remove the setting 'exclude_by' from
your prodigy.json.

I checked the prodigy.json file - the 'exclude_by' setting is set to 'task' which from what I understand should be the default value (?). Removing the setting all together didn't get rid of the warning. I haven't intentionally done anything to change this setting.

This is the command I used to run the recipe:

prodigy ner.correct my_new_empty_dataset my_previously_trained_model_based_on_en_core_web_lg my_new_input_jsonl_file --label LABEL_A,LABEL_B,LABEL_C,LABEL_D,LABEL_E

Do you have any idea why I'm getting this warning?

Hi! The warning here happens because the recipe you're using specifies an exclude_by setting and the prodigy.json can be used to override any defaults specified in the recipe. Because the exlusion setting modifies the recipe behaviour, often in a significant way, we show the warning because overwriting it is typically not what you want (and if it is, you can ignore the warning). So if you don't want to overwrite the setting, the solution is to just remove it from your prodigy.json :slightly_smiling_face: