first commit

conf/config.json

@@ -0,0 +1,6 @@
+{
+	"zmq_url" : "tcp://127.0.0.1:9999",
+	"model_dir" : "/d/ymq/models/funasr/SenseVoiceSmall",
+	"model_dir" : "turbo",
+	"device" : "cuda:0"
+}

requirements.txt

@@ -0,0 +1,2 @@
+git+https://git.kaiyuancloud.cn/yumoqing/apppublic
+openai-whisper

run.sh

@@ -0,0 +1,4 @@
+#!/bin/sh
+
+r=$HOME/ve/rtc/bin/python
+$r $*

zmq_client.py

@@ -0,0 +1,41 @@
+import json
+import os
+
+from appPublic.dictObject import DictObject
+from appPublic.zmq_reqrep import ZmqRequester
+from appPublic.jsonConfig import getConfig
+
+zmq_url = "tcp://127.0.0.1:9999"
+from time import time
+
+class ASRClient:
+    def __init__(self, zmq_url):
+        self.zmq_url = zmq_url
+        self.requester = ZmqRequester(self.zmq_url)
+
+    def generate(self, audio_file):
+        d = {
+            "audio_file":audio_file
+        }
+        msg = json.dumps(d)
+        resp = self.requester.send(msg)
+        if resp != None:
+            ret = json.loads(resp)
+            print(f'response={ret}')
+        else:
+            print(f'response is None')
+
+    def run(self):
+        print(f'running {self.zmq_url}')
+        while True:
+            print('input audio_file:')
+            af = input()
+            if len(af) > 0:
+                self.generate(af)
+        print('ended ...')
+
+if __name__ == '__main__':
+    workdir = os.getcwd()
+    config = getConfig(workdir)
+    asr = ASRClient(config.zmq_url or zmq_url)
+    asr.run()

zmq_whisper.py

@@ -0,0 +1,45 @@
+import json
+import os
+
+from appPublic.dictObject import DictObject
+from appPublic.zmq_reqrep import ZmqReplier
+from appPublic.jsonConfig import getConfig
+
+from time import time
+import whisper
+
+class ZmqASR:
+    def __init__(self, zmq_url, model_dir, device):
+        self.model = whisper.load_model(model_dir, device=device)
+        self.model_dir = model_dir
+        self.device = device
+        self.zmq_url = zmq_url
+        self.replier = ZmqReplier(self.zmq_url, self.generate)
+
+    def run(self):
+        print(f'running {self.zmq_url}')
+        self.replier._run()
+        print('ended ...')
+
+    def generate(self, d):
+        msg= d.decode('utf-8')
+        data = DictObject(**json.loads(msg))
+        t1 = time()
+        res = self.model.transcribe(data.audio_file)
+        t2 = time()
+        text = res['text']
+        d = {k:v for k, v in res.items() if k!='text'}
+        d['content'] = text
+        d['time_cost'] = t2 - t1
+        print(f'{d}')
+        return json.dumps(d)
+
+if __name__ == '__main__':
+    workdir = os.getcwd()
+    config = getConfig(workdir)
+    print(f'{config=}')
+    asr = ZmqASR(config.zmq_url or zmq_url,
+                    config.model_dir or model_dir,
+                    config.device or device)
+    print('here')
+    asr.run()