
- 谁
- Hugging Face
- 时间
- 引文
Today we're open sourcing Rumble-2-Flash under the Apache 2.0 license.
发生了什么
Hugging Face 在 Apache 2.0 许可证下开源了 Rumble-2-Flash。这款 0.5B 参数的语音识别模型支持 40 多种语言,且无需自定义代码即可在 Transformers 和 vLLM 中运行。在 302 小时的呼叫中心基准测试中,其词错误率 (WER) 达到 7.86%,与 NVIDIA Nemotron 3.5 Transcribe (30B) 的 7.78% 表现接近。
推荐理由
Rumble-2-Flash 为 0.5B 参数模型带来了极具竞争力的多语言和电话转录准确性,使得这些成果能够通过标准的 Transformers 或 vLLM 运行时直接使用。
这件事到目前为止
Hugging Face has open-sourced its 0.5-billion-parameter speech recognition model, Rumble-2-Flash, under the Apache 2.0 license. The model supports more than 40 languages and functions directly within Transformers and vLLM environments without the need for custom code. On a 302-hour call-center benchmark, Rumble-2-Flash achieved a 7.86% word error rate. Hugging Face noted that this performance is comparable to the 30-billion-parameter NVIDIA Nemotron 3.5 Transcribe, which recorded a 7.78% word error rate on the same benchmark.
