我正在尝试为我的模型创建一个推理应用程序(语言分类(,并且我收到错误 Channel
对象没有属性unary_unary
。我找不到有关此问题的任何信息,因此这篇文章。我是Python和Tensorflow区域的新手,我仍在学习。错误日志看起来像这样(几行(
2019-07-30T12:34:12.24+0200 [APP/PROC/WEB/0] ERR File "app.py", line 189, in do_inference
2019-07-30T12:34:12.24+0200 [APP/PROC/WEB/0] ERR stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
2019-07-30T12:34:12.24+0200 [APP/PROC/WEB/0] ERR File "/home/vcap/deps/0/python/lib/python3.6/site-packages/tensorflow_serving/apis/prediction_service_pb2_grpc.py", line 40, in __init__
2019-07-30T12:34:12.24+0200 [APP/PROC/WEB/0] ERR self.Classify = channel.unary_unary(
2019-07-30T12:34:12.24+0200 [APP/PROC/WEB/0] ERR AttributeError: 'Channel' object has no attribute 'unary_unary'
我正在使用烧瓶来创建消耗模型的Web服务。
@app.route('/LangDet', methods=['POST'])
def do_inference():
# get deployed model details
token = get_access_token()
model_name = request.path[1:]
query_string = {"modelName": model_name}
headers = {
'Authorization': token
}
res = requests.get(deployment_url, headers=headers, params=query_string)
model_info = json.loads(res.text)
if int(model_info["count"]) < 1:
return Response('404 Not Found: Model ' + model_name + ' is unavailable.', status=404)
else:
latest_version = [0, 0]
for index, model in enumerate(model_info["modelServers"]):
if int(model["specs"]["models"][0]["modelVersion"]) > latest_version[0]:
latest_version = [int(model["specs"]["models"][0]["modelVersion"]), index]
model_host = model_info["modelServers"][latest_version[1]]["endpoints"][0]
credentials = implementations.ssl_channel_credentials(root_certificates=bytes(model_host["caCrt"], 'ascii'))
channel = implementations.secure_channel(str(model_host["host"]), int(model_host["port"]), credentials)
stub = prediction_service_pb2_grpc.PredictionServiceStub(channel)
声明存根后,应用程序引发异常。这里似乎是一个问题,我该怎么办才能解决?
看来原始海报弄清楚了,对于任何遇到此问题的人来说,这是解决方案。
而不是:
channel = grpc.beta.implementations.insecure_channel(host, int(port))
使用:
channel = grpc.insecure_channel(server)
请注意,尽管Beta实现分别接受主机和端口,但第二种方法将它们作为单个参数接受。例如,
insecure_channel(0.0.0.0, 9999)
变成
insecure_channel(0.0.0.0:9999)