modelscope / FunASR

A Fundamental End-to-End Speech Recognition Toolkit and Open Source SOTA Pretrained Models, Supporting Speech Recognition, Voice Activity Detection, Text Post-processing etc.
https://www.funasr.com
Other
5.81k stars 629 forks source link

报错 websocketpp.processor:4 (A message was too large) #1756

Open dfengpo opened 3 months ago

dfengpo commented 3 months ago

🐛 Bug

To Reproduce

Steps to reproduce the behavior (always include the command you ran):

  1. Run cmd '....' 我通过websocket 上传文件大小为106mb,时长为57分钟,我已经使用vad模型。我收到了报错信息: [error] consume error: websocketpp.processor:4 (A message was too large)
  2. See error image

Code sample

Expected behavior

Environment

docoker 部署包registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.4 1,调用方式为html5,在这个网站上把https://www.funasr.com/static/offline/index.html 中的asr服务器地址改成我本地部署的地址。 2,使用这个官网示例,也是同样的报错。https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/csharp/ws-client/FunASRWSClient_Offline/WebScoketClient.cs

测试文件为链接: https://pan.baidu.com/s/1QNNNyAR1aPDhfhZP_0pjTg?pwd=zj3q 提取码: zj3q

Additional context

与这里描述的差距好大,可直接对时长为数小时音频进行识别 Paraformer-large长音频模型集成VAD、ASR、标点与时间戳功能,可直接对时长为数小时音频进行识别,并输出带标点文字与时间戳: ASR模型:Parformer-large模型结构为非自回归语音识别模型,多个中文公开数据集上取得SOTA效果,可快速地基于ModelScope对模型进行微调定制和推理。

lyblsgo commented 3 months ago

直接在这上传文件,网盘我打不开

dfengpo commented 3 months ago

直接在这上传文件,网盘我打不开

不支持直接上传文件,用下面地址可以打开 http://[110.41.57.176:5005/assets/wav/6.wav](http://110.41.57.176:5005/assets/wav/6.wav) @lyblsgo

lyblsgo commented 3 months ago
image

我这边拿你的音频测试,服务端是可以的,建议使用c++ 或者 python的客户端测试一下,这是官方维护的客户端

dfengpo commented 3 months ago

https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/csharp/ws-client/FunASRWSClient_Offline/WebScoketClient.cs

这个也不行,这个不是官方的demo吗,还有h5 https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/csharp/ws-client/FunASRWSClient_Offline/WebScoketClient.cs

lyblsgo commented 3 months ago

https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/csharp/ws-client/FunASRWSClient_Offline/WebScoketClient.cs

这个也不行,这个不是官方的demo吗,还有h5 https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/csharp/ws-client/FunASRWSClient_Offline/WebScoketClient.cs

只有c++ 和 python是官方维护的,其他都是社区用户贡献的

ericg108 commented 1 week ago

@lyblsgo 您好,我用的是官方的python client,发现处理长视频时,会报错(如下),我的视频长度是2小时57分 (base) [root@hw-hrzyxt-hd2jr ~]# python3 funasr_wss_client.py --host "127.0.0.1" --port 10095 --mode offline --audio_in "longv.mp4" --output_dir "./results" --use_itn 0 Namespace(audio_fs=16000, audio_in='longv.mp4', chunk_interval=10, chunk_size=[5, 10, 5], host='127.0.0.1', hotword='', mode='offline', output_dir='./results', port=10095, send_without_sleep=True, ssl=1, thread_num=1, use_itn=0, words_max_print=10000) connect to wss://127.0.0.1:10095 Exception: sent 1009 (message too big); no close frame received

机器配置是8核16G,不知怎么解决,谢谢