worstchan/Belle_1.4M-SLAM-Omni

Belle_1.4M数据集是一个用于支持SLAM-Omni论文复现的数据集。数据集经过过滤，移除了数据过长的样本。此外，使用CosyVoice合成了语音响应的token，并将其作为模型训练的目标。数据集中还包含了使用CosyVoice合成的用户指令语音，其音色从seed-tts-eval子集的1010个中文提示中随机选择。数据集来源于Belle_train_3.5M_CN。

The Belle_1.4M dataset is used to support the reproduction of the SLAM-Omni paper. The dataset has been filtered to remove samples with excessively long data. In addition, speech response tokens have been synthesized using CosyVoice and included as model training targets. The dataset also includes user instruction speech synthesized with CosyVoice, with timbres randomly selected from 1,010 Chinese prompts in the seed-tts-eval subset. The dataset is sourced from Belle_train_3.5M_CN.