Piros/CamStories_full
收藏Hugging Face2025-08-10 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/Piros/CamStories_full
下载链接
链接失效反馈官方服务：
资源简介：
CamStories-Full是一个完整的、经过清洗和合并的儿童小说故事语料库，来源于TinyStories GPT-4变体和SimpleStories数据集。它采用了与CamStories-10k相同的清洗、名称标准化和语法过滤流程，保留了31,244个token的完整未大小写敏感词汇表，是CamStories-10k的超集，具有更多的词汇多样性。同时，为每个词汇项提供了音频向量，支持两种音频格式：原始Kokoro 24 kHz合成和8 kHz、半秒（4096样本）向量。
CamStories-Full is the full cleaned and combined corpus of children’s fiction stories from the TinyStories GPT-4 variant and SimpleStories datasets. It uses the same extensive cleaning, name normalization, and grammar filtering pipeline as CamStories-10k, but retains the full uncased word vocabulary of 31,244 tokens, making it a superset of CamStories-10k with more lexical variety. Audio vectors for each vocabulary item are provided in two formats: Original Kokoro 24 kHz synthesis and 8 kHz, half-second (4096-sample) vectors.
提供机构：
Piros
