Post-refactoring (update get_dataset_info & dataset scripts)

[pfliu-nlp]

No description provided.

[pfliu-nlp]

Regarding the dataset_info.jsonl format, the major difference comes from the change of task_categories (previously, task_category).
We could also keep the original format of dataset_info.jsonl unchanged by taking the value of task_categories[0].

I'm running get_dataset_info.py from scratch to collect the remaining errors or issues.

[neubig]

LGTM. I think we should definitely change to task_categories, no need to remove information from dataset_info.jsonl that might be useful later.

[pfliu-nlp]

This is the latest version of dataset_info.jsonl & some points I want to share

• So far, almost all ERRORs result from the use of google drive link, which can work sometimes but will fail as well sometimes. We can move them to S3 gradually (Since most of them are from summarization tasks, so maybe @yixinL7 and @xcfcode could help out with this part.
• languages for several datasets should be added.
• get_dataset_info.py now can print sub_dataset_name when errors happen instead of just printing ___NONE__