Trelis/pilotgpt-test-v2
收藏Hugging Face2026-01-30 更新2026-02-07 收录
下载链接：
https://hf-mirror.com/datasets/Trelis/pilotgpt-test-v2
下载链接
链接失效反馈官方服务：
资源简介：
pilotgpt-test-v2是一个由Trelis Studio准备的语音数据集，专为Whisper模型优化。包含音频段和文本转录，特别提供了带时间戳的转录信息。数据集经过VAD处理，去除了静音部分，确保训练数据与推理行为一致。统计显示有2个源文件，42个验证样本，总时长为4.6分钟。适用于Whisper时间戳训练，建议采用两桶方法进行训练。
Speech dataset prepared with Trelis Studio, optimized for Whisper models. Includes audio segments and text transcriptions, with special timestamped transcription information. Processed with VAD to remove silence, ensuring training data matches inference behavior. Statistics show 2 source files, 42 validation samples, and a total duration of 4.6 minutes. Suitable for Whisper timestamp training, with a recommended two-bucket approach for training.
提供机构：
Trelis
