返回全部动态
Dasheng：大规模通用音频编码器
原标题：mispeech/dasheng-base
AI 摘要
Hugging Face 上发布了 Dasheng（大声）通用音频编码器模型 mispeech/dasheng-base，采用 Apache-2.0 许可。该模型基于 272,356 小时多样化音频数据自监督训练，参数量 12 亿，覆盖语音、音乐和环境声。在 HEAR 基准及 CREMA-D、LibriCount、Speech Commands、VoxLingua 等任务上优于此前工作，并提供了推理与微调示例。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
--- license: apache-2.0 tags: - audio - audio-classification --- # Dasheng: a large scale general-purpose audio encoder Dasheng (**D**eep **A**udio-**S**ignal **H**olistic **E**mbeddi**ng**s), or “大声” ("great sound"), is a general-purpose audio encoder trained on a large-scale self-supervised learning task. Dasheng is designed to capture rich audio information across various domains, including speech, music, and environmental sounds. The model is trained on 272,356 hours of diverse audio data
发布时间：2026-10-25 16:51
抓取时间：2026-10-25 16:52
来源机构：Hugging Face