Alibaba's Tongyi Qianwen Opensources Qwen2-Audio Series Models

TapTechNews August 13th news, Alibaba Tongyi Qianwen open sources the two models of the Qwen2-Audio series, Qwen2-Audio-7B and Qwen2-Audio-7B-Instruct.

As a large-scale audio language model, Qwen2-Audio can accept various audio signal inputs and execute audio analysis or directly respond to text according to voice instructions. There are two different audio interaction modes:

Voice chat: Users can freely interact with Qwen2-Audio in voice, without the need for text input.

Audio analysis: Users can provide audio and text instructions during the interaction to analyze the audio.

The official conducted tests on a series of benchmark datasets, and Qwen2-Audio surpassed the previous best model.

TapTechNews attaches the following related links:

Trial link: https://huggingface.co/spaces/Qwen/Qwen2-Audio-Instruct-Demo

Paper address: https://arxiv.org/abs/2407.10759

Evaluation criteria: https://github.com/OFA-Sys/AIR-Bench

Open source code: https://github.com/QwenLM/Qwen2-Audio