meetween/mumospee_librispeech

该数据集是LibriSpeech语料库的修改版本，转换为parquet格式以提高高性能计算环境中的I/O效率。LibriSpeech广泛用于语音识别研究，包含超过1,000小时的英语朗读语音。此修改保留了原始数据属性，同时改进了大规模模型分布式训练的数据处理。

This dataset is a modified version of the LibriSpeech corpus, converted into parquet format to enhance I/O efficiency in high-performance computing environments. It contains over 1,000 hours of English read speech, suitable for speech recognition research. The modifications retain the original data attributes and optimize data handling for distributed training on large-scale models.