new SFT datasets

[theblackcat102]

Added translation and additional datasets to the pool:

translation:

• WMT 2019
• TED Talk translation
• DiveMT : expert written translation dataset

safety:

• Prosocial dialogue in both dialogue generation and explanation format

others:

• @Rallio67 instruction tuning dataset instruct_tuning
• Other summarization dataset : debate_sum, tldr_news

WMT and TED translation can be a big ( few GB per pair ), so becareful not to include too many language pairs.

[github-actions]

❌ pre-commit failed.
Please run pre-commit run --all-files locally and commit the changes.
Find more information in the repository's CONTRIBUTING.md

[github-actions]

❌ pre-commit failed.
Please run pre-commit run --all-files locally and commit the changes.
Find more information in the repository's CONTRIBUTING.md

[sanagno]

Looks great, I also agree that we should upload datasets into the hub. Perhaps we can have a dataset_preprocess folder where we create the hub datasets and then we simply download the final versions.