DeSTA-ntu/DeSTA-AQA5M-FROM-Llama3.1-8B-Instruct
收藏Hugging Face2025-08-07 更新2025-09-13 收录
下载链接：
https://hf-mirror.com/datasets/DeSTA-ntu/DeSTA-AQA5M-FROM-Llama3.1-8B-Instruct
下载链接
链接失效反馈官方服务：
资源简介：
DeSTA-AQA5M数据集包含50个语音、环境声音和音乐数据集，总时长超过7000小时。它由500万个音频-指令-响应三元组组成，通过使用LLM模型从音频元数据生成的文本描述来自动生成响应，用于微调大型音频语言模型。
DeSTA-AQA5M dataset consists of 50 speech, environmental sound, and music datasets, totaling more than 7,000 hours of audio. It comprises 5 million audio-instruction-response triplets constructed through the use of a Large Language Model (LLM) to self-generate responses from textual descriptions derived from audio metadata, for fine-tuning Large Audio Language Models (LALM).
提供机构：
DeSTA-ntu
