Tensorflow服务客户端问题

Question

我已经使用Keras训练并保存了TF模型，并且能够将Keras模型(Hdf5)转换为Tensorflow服务。

设置详细信息，模型是在哪里训练和保存的: Python版本: 2.7.5 Keras版本: 2.1.4使用TensorFlow后端。Tensorflow CPU : 1.5.0

当我尝试使用sample_client.py运行客户端时，我有用于Tensorflow服务的Docker镜像，它失败了，错误如下。有没有人能建议我如何继续下去？

/root/.local/lib/python2.7/site-packages/h5py/__init__.py:36: FutureWarning:         
Conversion of the second argument of issubdtype from `float` to 
`np.floating` is deprecated. In future, it will be treated as `np.float64 == 
 np.dtype(float).type`.
 from ._conv import register_converters as _register_converters
 Using TensorFlow backend.
(60000, 28, 28)
6
Traceback (most recent call last):
File "sample_client.py", line 62, in <module>
main()
File "sample_client.py", line 54, in main
result = stub.Predict(request, request_timeout)
File "/usr/lib64/python2.7/site-packages/grpc/beta/_client_adaptations.py", 
line 309, in __call__
self._request_serializer, self._response_deserializer)
File "/usr/lib64/python2.7/site-packages/grpc/beta/_client_adaptations.py", 
line 195, in _blocking_unary_unary
raise _abortion_error(rpc_error_call)
grpc.framework.interfaces.face.face.AbortionError: 
AbortionError(code=StatusCode.INVALID_ARGUMENT, details="NodeDef mentions 
attr 'dilations' not in Op<name=Conv2D; signature=input:T, filter:T -> 
output:T; attr=T:type,allowed=[DT_HALF, DT_FLOAT]; attr=strides:list(int); 
attr=use_cudnn_on_gpu:bool,default=true; attr=padding:string,allowed=
["SAME", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", 
"NCHW"]>; NodeDef: conv2d_1/convolution = Conv2D[T=DT_FLOAT, _output_shapes=
 [[?,26,26,32]], data_format="NHWC", dilations=[1, 1, 1, 1], 
 padding="VALID", strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, 
_device="/job:localhost/replica:0/task:0/cpu:0"](_arg_conv2d_1_input_0_0, 
conv2d_1/kernel/read). (Check whether your GraphDef-interpreting binary is 
up to date with your GraphDef-generating binary.).
     [[Node: conv2d_1/convolution = Conv2D[T=DT_FLOAT, _output_shapes=
[[?,26,26,32]], data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID", 
strides=[1, 1, 1, 1], use_cudnn_on_gpu=true, 
_device="/job:localhost/replica:0/task:0/cpu:0"](_arg_conv2d_1_input_0_0, 
conv2d_1/kernel/read)]]")


sample_client.py:
import numpy
from keras.datasets import mnist
from grpc.beta import implementations
import tensorflow as tf
from predict_client import predict_pb2
from predict_client import prediction_service_pb2
tf.app.flags.DEFINE_string("host", "0.0.0.0", "gRPC server host")
tf.app.flags.DEFINE_integer("port", 9000, "gRPC server port")
tf.app.flags.DEFINE_string("model_name", "mnist", "TensorFlow model name")
tf.app.flags.DEFINE_integer("model_version", -1, "TensorFlow model version")
tf.app.flags.DEFINE_float("request_timeout", 10.0, "Timeout of gRPC     request")
FLAGS = tf.app.flags.FLAGS

(x_train, y_train), (x_test, y_test) = mnist.load_data()

print(numpy.shape(x_train))
idx = 2324
img = x_train[idx,:,:]
label = y_train[idx]
img = numpy.resize(img, (1, 28, 28, 1))
print(label)


def main():
  host = FLAGS.host
  port = FLAGS.port
  model_name = FLAGS.model_name
  model_version = FLAGS.model_version
  request_timeout = FLAGS.request_timeout

  # Generate inference data
  keys = numpy.asarray([1, 2, 3])
  keys_tensor_proto = tf.contrib.util.make_tensor_proto(keys, dtype=tf.int32)
  features_tensor_proto = tf.contrib.util.make_tensor_proto(img,
                                                        dtype=tf.float32)

  # Create gRPC client and request
  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
  if model_version > 0:
    request.model_spec.version.value = model_version
  request.inputs['inputs'].CopyFrom(features_tensor_proto)
  request.model_spec.signature_name = 'predict'
  #request.inputs['features'].CopyFrom(features_tensor_proto)

  # Send request
  result = stub.Predict(request, request_timeout)
  response = numpy.array(result.outputs['outputs'].float_val)
  prediction = numpy.argmax(response)

  print(prediction)


if __name__ == '__main__':
  main()

提前感谢

关于Latha

Answer 1

该错误是由于tensorflow_model_server中的TensorFlow版本只是早于1.5，因此不知道如何实例化该运算符。

可以尝试在足够旧的TensorFlow版本上重新构建模型，也可以尝试使用较新的TensorFlow版本创建tensorflow_model_server。

只需遵循此处给出的步骤，即可使用Tensorflow服务的最新源代码构建docker。https://www.tensorflow.org/serving/docker