Support for long audio (#961)

夜雨飘零

2023-09-18 025116995df04c752b504052754547cec15c7e6a

Support for long audio (#961)

 funasr/runtime/python/http/README.md

@@ -21,9 +21,11 @@
--host [host ip] \
--port [server port] \
--asr_model [asr model_name] \
--vad_model [vad model_name] \
--punc_model [punc model_name] \
--ngpu [0 or 1] \
--ncpu [1 or 4] \
--hotword_path [path of hot word txt] \
--certfile [path of certfile for ssl] \
--keyfile [path of keyfile for ssl] \
--temp_dir [upload file temp dir] 

 funasr/runtime/python/http/server.py

@@ -28,11 +28,15 @@
parser.add_argument("--asr_model",
                    type=str,
                    default="damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",
                    help="model from modelscope")
                    help="offline asr model from modelscope")
parser.add_argument("--vad_model",
                    type=str,
                    default="damo/speech_fsmn_vad_zh-cn-16k-common-pytorch",
                    help="vad model from modelscope")
parser.add_argument("--punc_model",
                    type=str,
                    default="damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727",
                    help="model from modelscope")
                    default="damo/punc_ct-transformer_cn-en-common-vocab471067-large",
                    help="punc model from modelscope")
parser.add_argument("--ngpu",
                    type=int,
                    default=1,
@@ -41,6 +45,10 @@
                    type=int,
                    default=4,
                    help="cpu cores")
parser.add_argument("--hotword_path",
                    type=str,
                    default=None,
                    help="hot word txt path, only the hot word model works")
parser.add_argument("--certfile",
                    type=str,
                    default=None,
@@ -66,18 +74,21 @@
os.makedirs(args.temp_dir, exist_ok=True)

print("model loading")
param_dict = {}
if args.hotword_path is not None and os.path.exists(args.hotword_path):
    param_dict['hotword'] = args.hotword_path
# asr
inference_pipeline_asr = pipeline(task=Tasks.auto_speech_recognition,
                                  model=args.asr_model,
                                  vad_model=args.vad_model,
                                  ngpu=args.ngpu,
                                  ncpu=args.ncpu,
                                  model_revision=None)
                                  param_dict=param_dict)
print(f'loaded asr models.')

if args.punc_model != "":
    inference_pipeline_punc = pipeline(task=Tasks.punctuation,
                                       model=args.punc_model,
                                       model_revision="v1.0.2",
                                       ngpu=args.ngpu,
                                       ncpu=args.ncpu)
    print(f'loaded pun models.')