Datasets:

CohereLabs
/

aya_dataset

Tasks:

Modalities:

Formats:

Languages:

Standard Arabic

Moroccan Arabic

Size:

ArXiv:

Libraries:

License:

Dataset card Data Studio Files Files and versions Community

MarziehFadaee commited on Feb 12, 2024

Commit

1d53341

·

verified ·

1 Parent(s): f07417d

Update README.md

Files changed (1) hide show

README.md +1 -1

README.md CHANGED Viewed

@@ -158,7 +158,7 @@ This dataset can be used to train, finetune, evaluate multilingual LLMs.
   | Name | Explanation |
   |------|--------------|
   | [aya_dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset) | Human-annotated multilingual instruction finetuning dataset, comprising over 204K instances across 65 languages. |
-  | [aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection) | Created by applying instruction-style templates from fluent speakers to 44 datasets, including translations of 19 instruction-style datasets into 101 languages, providing 513k instances for various tasks.|
   | [aya_evaluation_suite](https://huggingface.co/datasets/CohereForAI/aya_evaluation_suite) | A diverse evaluation set for multilingual open-ended generation, featuring 250 culturally grounded prompts in 7 languages, 200 translated prompts in 24 languages, and human-edited versions selected for cross-cultural relevance from English Dolly in 6 languages.|
 # Dataset

   | Name | Explanation |
   |------|--------------|
   | [aya_dataset](https://huggingface.co/datasets/CohereForAI/aya_dataset) | Human-annotated multilingual instruction finetuning dataset, comprising over 204K instances across 65 languages. |
+  | [aya_collection](https://huggingface.co/datasets/CohereForAI/aya_collection) | Created by applying instruction-style templates from fluent speakers to 44 datasets, including translations of 19 instruction-style datasets into 101 languages, providing 513M instances for various tasks.|
   | [aya_evaluation_suite](https://huggingface.co/datasets/CohereForAI/aya_evaluation_suite) | A diverse evaluation set for multilingual open-ended generation, featuring 250 culturally grounded prompts in 7 languages, 200 translated prompts in 24 languages, and human-edited versions selected for cross-cultural relevance from English Dolly in 6 languages.|
 # Dataset