BEE-spoke-data/FLAN-compressed
收藏Hugging Face2025-12-29 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/BEE-spoke-data/FLAN-compressed
下载链接
链接失效反馈官方服务：
资源简介：
FLAN数据集的压缩版本，通过去重技术将原始数据量从378M行减少到116M行。该数据集仅对`inputs`字段进行了minhash去重，并移除了所有涉及翻译任务的行。数据集包含多种任务，如文本生成、情感分析等，适用于文本到文本生成的任务类别。
A compressed version of the FLAN dataset, which reduces the original data volume from 378M rows to 116M rows through deduplication. This dataset has been deduplicated on the `inputs` field only using minhash and has removed all rows involving translation tasks. The dataset includes various tasks such as text generation, sentiment analysis, etc., and is suitable for text-to-text generation task categories.
提供机构：
BEE-spoke-data
