websocket python offline/online 2pass demo

游雁

2023-05-08 59f920f17c781be3a9b31d2b82dbda28c9b0c362

websocket python offline/online 2pass demo

 funasr/bin/punctuation_infer_vadrealtime.py

@@ -61,7 +61,7 @@
            text_name="text",
            non_linguistic_symbols=train_args.non_linguistic_symbols,
        )
        print("start decoding!!!")
        

    @torch.no_grad()
    def __call__(self, text: Union[list, str], cache: list, split_size=20):

 funasr/runtime/python/websocket/README.md

@@ -33,11 +33,9 @@

#### ASR offline/online 2pass server

[//]: # (```shell)

[//]: # (python ws_server_online.py --host "0.0.0.0" --port 10095 --asr_model "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch")

[//]: # (```)
```shell
python ws_server_2pass.py --port 10095 --asr_model "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"  --asr_model_online "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online"
```

## For the client

@@ -49,6 +47,7 @@
```

### Start client

#### ASR offline client
##### Recording from mircrophone
```shell
@@ -60,6 +59,7 @@
# --chunk_interval, "10": 600/10=60ms, "5"=600/5=120ms, "20": 600/12=30ms
python ws_client.py --host "0.0.0.0" --port 10095 --chunk_interval 10 --words_max_print 100 --audio_in "./data/wav.scp" --send_without_sleep --output_dir "./results"
```

#### ASR streaming client
##### Recording from mircrophone
```shell
@@ -73,7 +73,16 @@
```

#### ASR offline/online 2pass client

##### Recording from mircrophone
```shell
# --chunk_size, "5,10,5"=600ms, "8,8,4"=480ms
python ws_client.py --host "0.0.0.0" --port 10095 --chunk_size "8,8,4" --words_max_print 10000
```
##### Loadding from wav.scp(kaldi style)
```shell
# --chunk_size, "5,10,5"=600ms, "8,8,4"=480ms
python ws_client.py --host "0.0.0.0" --port 10095 --chunk_size "8,8,4" --audio_in "./data/wav.scp" --words_max_print 10000 --output_dir "./results"
```
## Acknowledge
1. This project is maintained by [FunASR community](https://github.com/alibaba-damo-academy/FunASR).
2. We acknowledge [zhaoming](https://github.com/zhaomingwork/FunASR/tree/fix_bug_for_python_websocket) for contributing the websocket service.

 funasr/runtime/python/websocket/ws_client.py

@@ -10,6 +10,10 @@
from multiprocessing import Process
from funasr.fileio.datadir_writer import DatadirWriter

import logging

logging.basicConfig(level=logging.ERROR)

parser = argparse.ArgumentParser()
parser.add_argument("--host",
                    type=str,
@@ -158,25 +162,40 @@
async def message(id):
    global websocket
    text_print = ""
    text_print_2pass_online = ""
    text_print_2pass_offline = ""
    while True:
        try:
            meg = await websocket.recv()
            meg = json.loads(meg)
            # print(meg, end = '')
            # print("\r")
            # print(meg)
            wav_name = meg.get("wav_name", "demo")
            print(wav_name)
            # print(wav_name)
            text = meg["text"]
            if ibest_writer is not None:
                ibest_writer["text"][wav_name] = text
            
            if meg["mode"] == "online":
                text_print += " {}".format(text)
            else:
                text_print = text_print[-args.words_max_print:]
                os.system('clear')
                print("\rpid"+str(id)+": "+text_print)
            elif meg["mode"] == "online":
                text_print += "{}".format(text)
            text_print = text_print[-args.words_max_print:]
            os.system('clear')
            print("\rpid"+str(id)+": "+text_print)
                text_print = text_print[-args.words_max_print:]
                os.system('clear')
                print("\rpid"+str(id)+": "+text_print)
            else:
                if meg["mode"] == "2pass-online":
                    text_print_2pass_online += " {}".format(text)
                    text_print = text_print_2pass_offline + text_print_2pass_online
                else:
                    text_print_2pass_online = " "
                    text_print = text_print_2pass_offline + "{}".format(text)
                    text_print_2pass_offline += "{}".format(text)
                text_print = text_print[-args.words_max_print:]
                os.system('clear')
                print("\rpid" + str(id) + ": " + text_print)

        except Exception as e:
            print("Exception:", e)
            traceback.print_exc()
@@ -207,7 +226,7 @@
        await asyncio.gather(task, task2, task3)

def one_thread(id):
   asyncio.get_event_loop().run_until_complete(ws_client(id)) # 启动协程
   asyncio.get_event_loop().run_until_complete(ws_client(id))
   asyncio.get_event_loop().run_forever()



 funasr/runtime/python/websocket/ws_server_2pass.py

New file
@@ -0,0 +1,182 @@
import asyncio
import json
import websockets
import time
import logging
import tracemalloc
import numpy as np

from parse_args import args
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
from modelscope.utils.logger import get_logger
from funasr.runtime.python.onnxruntime.funasr_onnx.utils.frontend import load_bytes

tracemalloc.start()

logger = get_logger(log_level=logging.CRITICAL)
logger.setLevel(logging.CRITICAL)


websocket_users = set()

print("model loading")
# asr
inference_pipeline_asr = pipeline(
    task=Tasks.auto_speech_recognition,
    model=args.asr_model,
    ngpu=args.ngpu,
    ncpu=args.ncpu,
    model_revision=None)


# vad
inference_pipeline_vad = pipeline(
    task=Tasks.voice_activity_detection,
    model=args.vad_model,
    model_revision=None,
    output_dir=None,
    batch_size=1,
    mode='online',
    ngpu=args.ngpu,
    ncpu=args.ncpu,
)

if args.punc_model != "":
    inference_pipeline_punc = pipeline(
        task=Tasks.punctuation,
        model=args.punc_model,
        model_revision=None,
        ngpu=args.ngpu,
        ncpu=args.ncpu,
    )
else:
    inference_pipeline_punc = None

inference_pipeline_asr_online = pipeline(
    task=Tasks.auto_speech_recognition,
    model=args.asr_model_online,
    ngpu=args.ngpu,
    ncpu=args.ncpu,
    model_revision='v1.0.4')

print("model loaded")

async def ws_serve(websocket, path):
    frames = []
    frames_asr = []
    frames_asr_online = []
    global websocket_users
    websocket_users.add(websocket)
    websocket.param_dict_asr = {}
    websocket.param_dict_asr_online = {"cache": dict()}
    websocket.param_dict_vad = {'in_cache': dict(), "is_final": False}
    websocket.param_dict_punc = {'cache': list()}
    websocket.vad_pre_idx = 0
    speech_start = False

    try:
        async for message in websocket:
            message = json.loads(message)
            is_finished = message["is_finished"]
            if not is_finished:
                audio = bytes(message['audio'], 'ISO-8859-1')
                frames.append(audio)
                duration_ms = len(audio)//32
                websocket.vad_pre_idx += duration_ms

                is_speaking = message["is_speaking"]
                websocket.param_dict_vad["is_final"] = not is_speaking
                websocket.param_dict_asr_online["is_final"] = not is_speaking
                websocket.param_dict_asr_online["chunk_size"] = message["chunk_size"]
                websocket.wav_name = message.get("wav_name", "demo")
                # asr online
                frames_asr_online.append(audio)
                if len(frames_asr_online) % message["chunk_interval"] == 0:
                    audio_in = b"".join(frames_asr_online)
                    await async_asr_online(websocket, audio_in)
                    frames_asr_online = []
                if speech_start:
                    frames_asr.append(audio)
                # vad online
                speech_start_i, speech_end_i = await async_vad(websocket, audio)
                if speech_start_i:
                    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 or not is_speaking:
                    audio_in = b"".join(frames_asr)
                    await async_asr(websocket, audio_in)
                    frames_asr = []
                    speech_start = False
                    frames_asr_online = []
                    websocket.param_dict_asr_online = {"cache": dict()}
                    if not is_speaking:
                        websocket.vad_pre_idx = 0
                        frames = []
                        websocket.param_dict_vad = {'in_cache': dict()}
                    else:
                        frames = frames[-20:]

     
    except websockets.ConnectionClosed:
        print("ConnectionClosed...", websocket_users)
        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 = inference_pipeline_vad(audio_in=audio_in, param_dict=websocket.param_dict_vad)

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


async def async_asr(websocket, audio_in):
            if len(audio_in) > 0:
                # print(len(audio_in))
                audio_in = load_bytes(audio_in)
                
                rec_result = inference_pipeline_asr(audio_in=audio_in,
                                                    param_dict=websocket.param_dict_asr)
                # print(rec_result)
                if inference_pipeline_punc is not None and 'text' in rec_result and len(rec_result["text"])>0:
                    rec_result = inference_pipeline_punc(text_in=rec_result['text'],
                                                         param_dict=websocket.param_dict_punc)
                    # print("offline", rec_result)
                message = json.dumps({"mode": "2pass-offline", "text": rec_result["text"], "wav_name": websocket.wav_name})
                await websocket.send(message)


async def async_asr_online(websocket, audio_in):
    if len(audio_in) > 0:
        audio_in = load_bytes(audio_in)
        rec_result = inference_pipeline_asr_online(audio_in=audio_in,
                                                   param_dict=websocket.param_dict_asr_online)
        if websocket.param_dict_asr_online["is_final"]:
            websocket.param_dict_asr_online["cache"] = dict()
        if "text" in rec_result:
            if rec_result["text"] != "sil" and rec_result["text"] != "waiting_for_more_voice":
                # print("online", rec_result)
                message = json.dumps({"mode": "2pass-online", "text": rec_result["text"], "wav_name": websocket.wav_name})
                await websocket.send(message)


start_server = websockets.serve(ws_serve, args.host, args.port, subprotocols=["binary"], ping_interval=None)
asyncio.get_event_loop().run_until_complete(start_server)
asyncio.get_event_loop().run_forever()

 funasr/runtime/python/websocket/ws_server_online.py

@@ -37,12 +37,10 @@


async def ws_serve(websocket, path):
    frames_online = []
    frames_asr_online = []
    global websocket_users
    websocket.send_msg = Queue()
    websocket_users.add(websocket)
    websocket.param_dict_asr_online = {"cache": dict()}
    websocket.speek_online = Queue()

    try:
        async for message in websocket:
@@ -56,11 +54,11 @@
                websocket.wav_name = message.get("wav_name", "demo")
                websocket.param_dict_asr_online["chunk_size"] = message["chunk_size"]
                
                frames_online.append(audio)
                if len(frames_online) % message["chunk_interval"] == 0 or not is_speaking:
                    audio_in = b"".join(frames_online)
                frames_asr_online.append(audio)
                if len(frames_asr_online) % message["chunk_interval"] == 0 or not is_speaking:
                    audio_in = b"".join(frames_asr_online)
                    await async_asr_online(websocket,audio_in)
                    frames_online = []
                    frames_asr_online = []


     
@@ -81,8 +79,6 @@
                    websocket.param_dict_asr_online["cache"] = dict()
                if "text" in rec_result:
                    if rec_result["text"] != "sil" and rec_result["text"] != "waiting_for_more_voice":
                        # if len(rec_result["text"])>0:
                        #     rec_result["text"][0]=rec_result["text"][0] #.replace(" ","")
                        message = json.dumps({"mode": "online", "text": rec_result["text"], "wav_name": websocket.wav_name})
                        await websocket.send(message)