datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
live-streaming-content
直播内容训练数据集
数据集简介
这个数据集包含直播记录的对话训练数据，适用于：
微调语言模型以生成直播风格的内容
构建直播内容问答系统
RAG (检索增强生成) 系统训练
数据格式
1. SFT 格式 (training_sft.jsonl)
对话式训练数据，适用于监督微调：
{
"messages": [
{"role": "system", "content": "你是一个专业的直播内容助手..."},
{"role": "user", "content": "请介绍一下..."},
{"role": "assistant", "content": "..."}
]
}
2. RAG 格式 (training_rag.jsonl)
问答对格式，包含上下文：
{
"question": "...",
"context": "...",
"answer": "...",
"metadata": {...}
}… See the full description on the dataset page: https://huggingface.co/datasets/licco123/live-streaming-content.Project-Kyra-LiveStreams-RAWDate of scrape: 17/04/2024
4666 streams from 17 different vtubers
Transcribed by YouTube auto captions
This data has somewhat been cleaned:
Removed all under 12 chars
Removed most duplicates
Merged multiple lines together
This will probably need to be processed further to be used
Channels Scraped
Vtuber Name
Scraped Channel ID
Tenma
UC3K7pmiHsNSx1y0tdx2bbCw
Kirsche
UC4uVQv3D4-DtUQOvG1Fo6Jg
Chibidoki
UC6eWvFoYe5nOia-hyPxFPZw
Akuma Nihmune
UC79Ekr3umIkWmgZ-an01LfQ… See the full description on the dataset page: https://huggingface.co/datasets/kyratron/Project-Kyra-LiveStreams-RAW.