Datasets:

lmqg

/

qg_jaquad

like 5

Thanks for fixing the meta info! Could you explain a bit why the task_ids should be language-modeling? Also would you be able to share the available list of task_categories and task_ids? I don't think I fully understand the right way of writing the meta info, so it should be very useful if you could share it to me :)

task_categories:
- text-generation
task_ids:
- language-modeling

Hi @asahi417 ,

I was just making a reasonable guess while trying to fix all datastes with wrong tasks.

Please, feel free to see the list of valid tasks here and chose the ones you find suit better: https://github.com/huggingface/hub-docs/blob/main/js/src/lib/interfaces/Types.ts#L85

• first level keys correspond to valid task_categories values
• and second level keys (under "subtasks") correspond to valid task_ids

Hi again.

Independently of the specific values for task_categories and task_ids, please note their type should be an array (each item in the array preceded by an hyphen -) and not a string as it is right now.

I'm merging this PR so that their type is fixed as array, but feel free to change the elements in the array to other values if you think there are more suitable.

Once the type fixed, these errors will disappear from your dataset card:

YAML Metadata Error: "task_categories" must be an array 
YAML Metadata Error: "task_ids" must be an array