FunASR的本地使用及部署

代码链接附上

GitHub - 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.

我之前部署过两个TTS，ASR是我第一次接触，所以想写文记录一下，和之前的TTS不同的是FunASR没有webui，只有api的调用，所以本地部署的时候有一点手足无措。因此进行记录。话不多说，现在开始。

1. 克隆仓促至本地：

        在本地储存位置打开git bash：

git clone https://github.com/modelscope/FunASR.git

2. 创建新环境：

conda create -n funasr python=3.9
 
conda activate funasr
 
conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=11.8 -c pytorch -c nvidia
 

建议在安装torch的时候使用pip命令，conda命令一般都安不上。

pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu118

3. 安装funasr

pip install -U funasr

4. 使用预训练模型，这一步有两个方式，一是在每次运行脚本的时候让其下载模型，二是直接下载预训练模型在本地调用。我在本地运行的时候是使用的前者。因此需要安装modelscope与huggingface_hub。

pip install -U modelscope huggingface huggingface_hub

5. 快速开始

（1）命令行开始

funasr ++model=paraformer-zh ++punc_model="ct-punc"  ++input="example/asr_example_en.wav"

其中，input是输入的音频路径，默认是根据提供的模型目录，自动从ModelScope平台下载所需的模型和示例数据。但是仓库中的原命令是跑不通的，因此我将input转化为本地路径，就可以了。

（2）脚本

新建Python文件运行下列代码，同样，我将音频文件下载到本地调用。

from funasr import AutoModel
from funasr.utils.postprocess_utils import rich_transcription_postprocess

model_dir = "iic/SenseVoiceSmall"

model = AutoModel(
    model=model_dir,
    vad_model="fsmn-vad",
    vad_kwargs={"max_single_segment_time": 30000},
    device="cuda:0",
)

# en
res = model.generate(
    input=f"example/asr_example_en.wav",
    cache={},
    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
    use_itn=True,
    batch_size_s=60,
    merge_vad=True,  #
    merge_length_s=15,
)
text = rich_transcription_postprocess(res[0]["text"])
print(text)

input设置为自己的本地路径，就可成功运行。

6 流式输出

上面是非流式输出的实现，下面是流式输出的实现。

（a）脚本

from funasr import AutoModel

chunk_size = [0, 10, 5] #[0, 10, 5] 600ms, [0, 8, 4] 480ms
encoder_chunk_look_back = 4 #number of chunks to lookback for encoder self-attention
decoder_chunk_look_back = 1 #number of encoder chunks to lookback for decoder cross-attention

model = AutoModel(model='iic/SenseVoiceSmall')

import soundfile
import os

wav_file = "example/asr_example.wav"
speech, sample_rate = soundfile.read(wav_file)
chunk_stride = chunk_size[1] * 960 # 600ms

cache = {}
total_chunk_num = int(len((speech)-1)/chunk_stride+1)
for i in range(total_chunk_num):
    speech_chunk = speech[i*chunk_stride:(i+1)*chunk_stride]
    is_final = i == total_chunk_num - 1
    res = model.generate(input=speech_chunk, cache=cache, is_final=is_final, chunk_size=chunk_size, encoder_chunk_look_back=encoder_chunk_look_back, decoder_chunk_look_back=decoder_chunk_look_back)
    print(res)

其中，model = AutoModel(model='iic/SenseVoiceSmall') 和 wav_file = "example/asr_example.wav"分别为模型和输入的路径，如果需要换输入音频就直接改为自己音频的路径，若需要换个模型看效果，则需要在funasr的github中进入模型仓库-modelscope中查看路径。因为每次运行都是从modelscope上下载模型。

剩下的一些api和docker流程以及模型用法我会慢慢全部更新完成。

 

7 服务端客户端的运行

这里我测试的是runtime\python\websocket中的funasr_wss_server.py和funasr_wss_client.py的运行。我在windows中运行服务端代码的时候没有运行成功，错误信息：RuntimeError: no running event loop
表明 asyncio.get_running_loop() 失败了，通常是因为当前环境下 asyncio 没有正确启动。Windows 下 asyncio 事件循环机制可能导致此问题。因此需要修改后面的代码，完整代码如下：

import asyncio
import json
import websockets
import time
import logging
import tracemalloc
import numpy as np
import argparse
import ssl


parser = argparse.ArgumentParser()
parser.add_argument(
    "--host", type=str, default="0.0.0.0", required=False, help="host ip, localhost, 0.0.0.0"
)
parser.add_argument("--port", type=int, default=10095, required=False, help="grpc server port")
parser.add_argument(
    "--asr_model",
    type=str,
    default="iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
    help="model from modelscope",
)
parser.add_argument("--asr_model_revision", type=str, default="v2.0.4", help="")
parser.add_argument(
    "--asr_model_online",
    type=str,
    default="iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online",
    help="model from modelscope",
)
parser.add_argument("--asr_model_online_revision", type=str, default="v2.0.4", help="")
parser.add_argument(
    "--vad_model",
    type=str,
    default="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",
    help="model from modelscope",
)
parser.add_argument("--vad_model_revision", type=str, default="v2.0.4", help="")
parser.add_argument(
    "--punc_model",
    type=str,
    default="iic/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727",
    help="model from modelscope",
)
parser.add_argument("--punc_model_revision", type=str, default="v2.0.4", help="")
parser.add_argument("--ngpu", type=int, default=1, help="0 for cpu, 1 for gpu")
parser.add_argument("--device", type=str, default="cuda", help="cuda, cpu")
parser.add_argument("--ncpu", type=int, default=4, help="cpu cores")
parser.add_argument(
    "--certfile",
    type=str,
    default="../../ssl_key/server.crt",
    required=False,
    help="certfile for ssl",
)

parser.add_argument(
    "--keyfile",
    type=str,
    default="../../ssl_key/server.key",
    required=False,
    help="keyfile for ssl",
)
args = parser.parse_args()


websocket_users = set()

print("model loading")
from funasr import AutoModel

# asr
model_asr = AutoModel(
    model=args.asr_model,
    model_revision=args.asr_model_revision,
    ngpu=args.ngpu,
    ncpu=args.ncpu,
    device=args.device,
    disable_pbar=True,
    disable_log=True,
)
# asr
model_asr_streaming = AutoModel(
    model=args.asr_model_online,
    model_revision=args.asr_model_online_revision,
    ngpu=args.ngpu,
    ncpu=args.ncpu,
    device=args.device,
    disable_pbar=True,
    disable_log=True,
)
# vad
model_vad = AutoModel(
    model=args.vad_model,
    model_revision=args.vad_model_revision,
    ngpu=args.ngpu,
    ncpu=args.ncpu,
    device=args.device,
    disable_pbar=True,
    disable_log=True,
    # chunk_size=60,
)

if args.punc_model != "":
    model_punc = AutoModel(
        model=args.punc_model,
        model_revision=args.punc_model_revision,
        ngpu=args.ngpu,
        ncpu=args.ncpu,
        device=args.device,
        disable_pbar=True,
        disable_log=True,
    )
else:
    model_punc = None


print("model loaded! only support one client at the same time now!!!!")


async def ws_reset(websocket):
    print("ws reset now, total num is ", len(websocket_users))

    websocket.status_dict_asr_online["cache"] = {}
    websocket.status_dict_asr_online["is_final"] = True
    websocket.status_dict_vad["cache"] = {}
    websocket.status_dict_vad["is_final"] = True
    websocket.status_dict_punc["cache"] = {}

    await websocket.close()


async def clear_websocket():
    for websocket in websocket_users:
        await ws_reset(websocket)
    websocket_users.clear()


async def ws_serve(websocket, path="dafault_path"):
    frames = []
    frames_asr = []
    frames_asr_online = []
    global websocket_users
    # await clear_websocket()
    websocket_users.add(websocket)
    websocket.status_dict_asr = {}
    websocket.status_dict_asr_online = {"cache": {}, "is_final": False}
    websocket.status_dict_vad = {"cache": {}, "is_final": False}
    websocket.status_dict_punc = {"cache": {}}
    websocket.chunk_interval = 10
    websocket.vad_pre_idx = 0
    speech_start = False
    speech_end_i = -1
    websocket.wav_name = "microphone"
    websocket.mode = "2pass"
    print("new user connected", flush=True)

    try:
        async for message in websocket:
            if isinstance(message, str):
                messagejson = json.loads(message)

                if "is_speaking" in messagejson:
                    websocket.is_speaking = messagejson["is_speaking"]
                    websocket.status_dict_asr_online["is_final"] = not websocket.is_speaking
                if "chunk_interval" in messagejson:
                    websocket.chunk_interval = messagejson["chunk_interval"]
                if "wav_name" in messagejson:
                    websocket.wav_name = messagejson.get("wav_name")
                if "chunk_size" in messagejson:
                    chunk_size = messagejson["chunk_size"]
                    if isinstance(chunk_size, str):
                        chunk_size = chunk_size.split(",")
                    websocket.status_dict_asr_online["chunk_size"] = [int(x) for x in chunk_size]
                if "encoder_chunk_look_back" in messagejson:
                    websocket.status_dict_asr_online["encoder_chunk_look_back"] = messagejson[
                        "encoder_chunk_look_back"
                    ]
                if "decoder_chunk_look_back" in messagejson:
                    websocket.status_dict_asr_online["decoder_chunk_look_back"] = messagejson[
                        "decoder_chunk_look_back"
                    ]
                if "hotword" in messagejson:
                    websocket.status_dict_asr["hotword"] = messagejson["hotwords"]
                if "mode" in messagejson:
                    websocket.mode = messagejson["mode"]

            websocket.status_dict_vad["chunk_size"] = int(
                websocket.status_dict_asr_online["chunk_size"][1] * 60 / websocket.chunk_interval
            )
            if len(frames_asr_online) > 0 or len(frames_asr) >= 0 or not isinstance(message, str):
                if not isinstance(message, str):
                    frames.append(message)
                    duration_ms = len(message) // 32
                    websocket.vad_pre_idx += duration_ms

                    # asr online
                    frames_asr_online.append(message)
                    websocket.status_dict_asr_online["is_final"] = speech_end_i != -1
                    if (
                        len(frames_asr_online) % websocket.chunk_interval == 0
                        or websocket.status_dict_asr_online["is_final"]
                    ):
                        if websocket.mode == "2pass" or websocket.mode == "online":
                            audio_in = b"".join(frames_asr_online)
                            try:
                                await async_asr_online(websocket, audio_in)
                            except:
                                print(f"error in asr streaming, {websocket.status_dict_asr_online}")
                        frames_asr_online = []
                    if speech_start:
                        frames_asr.append(message)
                    # vad online
                    try:
                        speech_start_i, speech_end_i = await async_vad(websocket, message)
                    except:
                        print("error in vad")
                    if speech_start_i != -1:
                        speech_start = True
                        beg_bias = (websocket.vad_pre_idx - speech_start_i) // duration_ms
                        frames_pre = frames[-beg_bias:]
                        frames_asr = []
                        frames_asr.extend(frames_pre)
                # asr punc offline
                if speech_end_i != -1 or not websocket.is_speaking:
                    # print("vad end point")
                    if websocket.mode == "2pass" or websocket.mode == "offline":
                        audio_in = b"".join(frames_asr)
                        try:
                            await async_asr(websocket, audio_in)
                        except:
                            print("error in asr offline")
                    frames_asr = []
                    speech_start = False
                    frames_asr_online = []
                    websocket.status_dict_asr_online["cache"] = {}
                    if not websocket.is_speaking:
                        websocket.vad_pre_idx = 0
                        frames = []
                        websocket.status_dict_vad["cache"] = {}
                    else:
                        frames = frames[-20:]

    except websockets.ConnectionClosed:
        print("ConnectionClosed...", websocket_users, flush=True)
        await ws_reset(websocket)
        websocket_users.remove(websocket)
    except websockets.InvalidState:
        print("InvalidState...")
    except Exception as e:
        print("Exception:", e)


async def async_vad(websocket, audio_in):

    segments_result = model_vad.generate(input=audio_in, **websocket.status_dict_vad)[0]["value"]
    # print(segments_result)

    speech_start = -1
    speech_end = -1

    if len(segments_result) == 0 or len(segments_result) > 1:
        return speech_start, speech_end
    if segments_result[0][0] != -1:
        speech_start = segments_result[0][0]
    if segments_result[0][1] != -1:
        speech_end = segments_result[0][1]
    return speech_start, speech_end


async def async_asr(websocket, audio_in):
    if len(audio_in) > 0:
        # print(len(audio_in))
        rec_result = model_asr.generate(input=audio_in, **websocket.status_dict_asr)[0]
        # print("offline_asr, ", rec_result)
        if model_punc is not None and len(rec_result["text"]) > 0:
            # print("offline, before punc", rec_result, "cache", websocket.status_dict_punc)
            rec_result = model_punc.generate(
                input=rec_result["text"], **websocket.status_dict_punc
            )[0]
            # print("offline, after punc", rec_result)
        if len(rec_result["text"]) > 0:
            # print("offline", rec_result)
            mode = "2pass-offline" if "2pass" in websocket.mode else websocket.mode
            message = json.dumps(
                {
                    "mode": mode,
                    "text": rec_result["text"],
                    "wav_name": websocket.wav_name,
                    "is_final": websocket.is_speaking,
                }
            )
            await websocket.send(message)

    else:
        mode = "2pass-offline" if "2pass" in websocket.mode else websocket.mode
        message = json.dumps(
            {
                "mode": mode,
                "text": "",
                "wav_name": websocket.wav_name,
                "is_final": websocket.is_speaking,
            }
        )
        await websocket.send(message)    

async def async_asr_online(websocket, audio_in):
    if len(audio_in) > 0:
        # print(websocket.status_dict_asr_online.get("is_final", False))
        rec_result = model_asr_streaming.generate(
            input=audio_in, **websocket.status_dict_asr_online
        )[0]
        # print("online, ", rec_result)
        if websocket.mode == "2pass" and websocket.status_dict_asr_online.get("is_final", False):
            return
            #     websocket.status_dict_asr_online["cache"] = dict()
        if len(rec_result["text"]):
            mode = "2pass-online" if "2pass" in websocket.mode else websocket.mode
            message = json.dumps(
                {
                    "mode": mode,
                    "text": rec_result["text"],
                    "wav_name": websocket.wav_name,
                    "is_final": websocket.is_speaking,
                }
            )
            await websocket.send(message)


async def main():
    if len(args.certfile) > 0:
        # 加载SSL证书
        ssl_context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
        ssl_cert = args.certfile
        ssl_key = args.keyfile
        ssl_context.load_cert_chain(ssl_cert, keyfile=ssl_key)
        server = await websockets.serve(
            ws_serve,
            args.host,
            args.port,
            subprotocols=["binary"],
            ping_interval=None,
            ssl=ssl_context
        )
    else:
        # 无SSL
        server = await websockets.serve(
            ws_serve,
            args.host,
            args.port,
            subprotocols=["binary"],
            ping_interval=None,
            max_queue=10
        )
    
    print(f"Server started on {args.host}:{args.port}")
    await server.wait_closed()

if __name__ == "__main__":
    asyncio.run(main())  # 统一入口

在修改最后的运行代码之后，服务端运行成功，在运行客户端的时候又报了错：ConnectionClosedError: received 1011 (internal error); then sent 1011 (internal error) end

此时服务端报错：TypeError: ws_serve() missing 1 required positional argument: 'path' ERROR:websockets.server:connection handler failed Traceback (most recent call last): File "D:\anaconda3\envs\funasr\lib\site-packages\websockets\asyncio\server.py", line 374, in conn_handler await self.handler(connection) TypeError: ws_serve() missing 1 required positional argument: 'path'

说明 websockets.serve() 期望 ws_serve 这个函数接受两个参数（websocket, path），但ws_serve 可能只定义了一个参数（websocket）。所以在服务端代码中修改了

async def ws_serve(websocket, path="default_path"):

就可运行成功，成功之后就可连接电脑麦克风进行实时语音转文字。