Not sure what's happening here. I tried using --loader jsonl but that argument was not accepted so I think the Prodigy documents need to be updated.
Anyway I ignored that and just set the stream in my custom recipe.
Setting the stream in my custom recipe I do:
JSONL(jsonl_file_path)
This is the error I get.
This is my JSONL file
Hi! Can you share your custom recipe code? I don't think the problem here is related to the JSON, which seems to load fine. It's more likely that there's a problem with the interface used and the data it expects, vs. the data that's loaded in.
If you're using a custom recipe, make sure your recipe accepts a --loader argument. The recipe arguments are based on what the recipe defines. So the recipes that document a --loader argument define that in the @prodigy.recipe decorator and then use the provided value to decide which loader to use (e.g. JSONL vs. something else). You can read more about this here: Custom Recipes · Prodigy · An annotation tool for AI, Machine Learning & NLP
Ok I actually just hardcoded the JSONL loader into the recipe and did not add the option for a --loader argument. I initialize the stream as JSONL(Filename).
I'll get back to you @ines with the recipe code that caused this error, sorry I'm away from my computer tonight. We didn't need to implement the URL loading in the end and so my more recent ticket I opened is much more important to me.
