A refined version of EuroParl dataset in The Pile by Data-Juicer. Removing some "bad" samples from the original dataset to make it higher-quality.
This dataset is usually used to pretrain a Large Language Model.
Notice: Here is a small subset for previewing. The whole dataset is available here (About 2.2GB).
Dataset Information
Number of samples: 61,601 (Keep ~88.23% from the original dataset)
Refining Recipe… See the full description on the dataset page: https://huggingface.co/datasets/datajuicer/the-pile-europarl-refined-by-data-juicer.
