speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

update paraformer online python websocket code

Browse Source
This commit is contained in:
haoneng.lhn 2023-09-25 17:45:03 +08:00
parent 111d50159d
commit 97eb2bd656
3 changed files with 19 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
funasr/runtime/python/websocket/funasr_client_api.py
View File

@ -51,7 +51,8 @@ class Funasr_websocket_recognizer():
stride = int(60 * chunk_size[1]/ chunk_interval / 1000 * 16000 * 2) stride = int(60 * chunk_size[1]/ chunk_interval / 1000 * 16000 * 2)
chunk_num = (len(audio_bytes) - 1) // stride + 1 chunk_num = (len(audio_bytes) - 1) // stride + 1
message = json.dumps({"mode": mode, "chunk_size": chunk_size, "chunk_interval": chunk_interval, message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "encoder_chunk_look_back": 4,
"decoder_chunk_look_back": 1, "chunk_interval": args.chunk_interval,
"wav_name": wav_name, "is_speaking": True}) "wav_name": wav_name, "is_speaking": True})
self.websocket.send(message) self.websocket.send(message)
@ -131,4 +132,4 @@ if __name__ == '__main__':
print("text",text) print("text",text)

11
funasr/runtime/python/websocket/funasr_wss_client.py
View File

@ -29,6 +29,14 @@ parser.add_argument("--chunk_size",
type=str, type=str,
default="5, 10, 5", default="5, 10, 5",
help="chunk") help="chunk")
parser.add_argument("--encoder_chunk_look_back",
type=int,
default=4,
help="number of chunks to lookback for encoder self-attention")
parser.add_argument("--decoder_chunk_look_back",
type=int,
default=1,
help="number of encoder chunks to lookback for decoder cross-attention")
parser.add_argument("--chunk_interval", parser.add_argument("--chunk_interval",
type=int, type=int,
default=10, default=10,
@ -99,7 +107,8 @@ async def record_microphone():
input=True, input=True,
frames_per_buffer=CHUNK) frames_per_buffer=CHUNK)
message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "chunk_interval": args.chunk_interval, message = json.dumps({"mode": args.mode, "chunk_size": args.chunk_size, "encoder_chunk_look_back": args.encoder_chunk_look_back,
"decoder_chunk_look_back": args.decoder_chunk_look_back, "chunk_interval": args.chunk_interval,
"wav_name": "microphone", "is_speaking": True}) "wav_name": "microphone", "is_speaking": True})
#voices.put(message) #voices.put(message)
await websocket.send(message) await websocket.send(message)

8
funasr/runtime/python/websocket/funasr_wss_server.py
View File

@ -103,8 +103,8 @@ inference_pipeline_asr_online = pipeline(
model=args.asr_model_online, model=args.asr_model_online,
ngpu=args.ngpu, ngpu=args.ngpu,
ncpu=args.ncpu, ncpu=args.ncpu,
model_revision='v1.0.4', model_revision='v1.0.7',
update_model='v1.0.4', update_model='v1.0.7',
mode='paraformer_streaming') mode='paraformer_streaming')
print("model loaded! only support one client at the same time now!!!!") print("model loaded! only support one client at the same time now!!!!")
@ -159,6 +159,10 @@ async def ws_serve(websocket, path):
websocket.wav_name = messagejson.get("wav_name") websocket.wav_name = messagejson.get("wav_name")
if "chunk_size" in messagejson: if "chunk_size" in messagejson:
websocket.param_dict_asr_online["chunk_size"] = messagejson["chunk_size"] websocket.param_dict_asr_online["chunk_size"] = messagejson["chunk_size"]
if "encoder_chunk_look_back" in messagejson:
websocket.param_dict_asr_online["encoder_chunk_look_back"] = messagejson["encoder_chunk_look_back"]
if "decoder_chunk_look_back" in messagejson:
websocket.param_dict_asr_online["decoder_chunk_look_back"] = messagejson["decoder_chunk_look_back"]
if "mode" in messagejson: if "mode" in messagejson:
websocket.mode = messagejson["mode"] websocket.mode = messagejson["mode"]
if len(frames_asr_online) > 0 or len(frames_asr) > 0 or not isinstance(message, str): if len(frames_asr_online) > 0 or len(frames_asr) > 0 or not isinstance(message, str):