- tensorflow-GPU 1.10.0
- 张量流服务器 1.10.0
我已经部署了一个为多个模型提供服务的张量流服务器。 客户端代码与此类似client.py，我调用预测函数。
channel = implementations.insecure_channel(host, port)
stub = prediction_service_pb2.beta_create_PredictionService_stub(channel)
request = predict_pb2.PredictRequest()
def predict(data, shape, model_name, signature_name="predict"):
request.model_spec.name = model_name
request.model_spec.signature_name = signature_name
request.inputs['image'].CopyFrom(tf.contrib.util.make_tensor_proto(data, shape=shape))
result = stub.Predict(request, 10.0)
return result.outputs['prediction'].float_val[0]
我有大约 100 个具有相同配置的客户端。 下面是调用predict函数的示例代码：
from client import predict
while True:
print(predict(data, shape, model_name))
# time.sleep some while
起初，当我运行客户端代码时，我可以正确接收响应。 但是几个小时后，客户端因错误而崩溃
_Rendezvous of RPC that terminated with (StatusCode.UNAVAILABLE, Socket closed)
我尝试修改我的客户端代码以
def predict(data, shape, model_name, signature_name="predict"):
channel = implementations.insecure_channel(host, port)
stub = prediction_service_pb2.beta_create_PredictionService_stub(channel)
request = predict_pb2.PredictRequest()
request.model_spec.name = model_name
request.model_spec.signature_name = signature_name
request.inputs['image'].CopyFrom(tf.contrib.util.make_tensor_proto(data, shape=shape))
result = stub.Predict(request, 10.0)
return result.outputs['prediction'].float_val[0]
这意味着每次调用predict函数时，我都会尝试与 tfs 服务器建立连接。但是这段代码也像以前一样失败了。
那么我应该怎么做才能处理这种情况呢？
最后我在return之前添加了一个channel.close()，它工作正常。