chengjunyan1/smollm-10

cosmopedia-v2数据集包含prompt、text、token_length、audience、format和seed_data等特征，主要用于训练，包含4537737个样本。fineweb-edu-dedup数据集包含text、id和metadata等特征，metadata中包括dump、url、date、file_path、language、language_score、token_count、score和int_score等子特征，主要用于训练，包含22701367个样本。python-edu数据集包含blob_id、repo_name、path、length_bytes、score和int_score等特征，主要用于训练，包含864074个样本。

The cosmopedia-v2 dataset includes features such as prompt, text, token_length, audience, format, and seed_data, primarily used for training, containing 4,537,737 samples. The fineweb-edu-dedup dataset includes features such as text, id, and metadata, with metadata containing sub-features like dump, url, date, file_path, language, language_score, token_count, score, and int_score, primarily used for training, containing 22,701,367 samples. The python-edu dataset includes features such as blob_id, repo_name, path, length_bytes, score, and int_score, primarily used for training, containing 864,074 samples.