这时节点定义找不到NodeDef attr 'dilations' not in，说明执行版本的NodeDef不在节点定义上，两个不一致，分别是执行inference的代码和生成静态图节点不一致（当然，因为执行环境和生成环境不一样，也就是版本问题），

解决方案是：如果是云端部署，那就在云端生成模型graph，然后也在云端将ckpt和graph合在一起生成静态图。如果是嵌入式，就是将本地版本调整一致（可能也有其他原因可以解决）。

具体错误是：

2018-08-30 22:26:51.668185: E tensorflow/core/common_runtime/executor.cc:643] Executor failed to create kernel. Invalid argument: NodeD
ons attr 'dilations' not in Op<name=Conv2D; signature=input:T, filter:T -> output:T; attr=T:type,allowed=[DT_HALF, DT_FLOAT]; attr=stri
(int); attr=use_cudnn_on_gpu:bool,default=true; attr=padding:string,allowed=["SAME", "VALID"]; attr=data_format:string,default="NHWC",a
"NHWC", "NCHW"]>; NodeDef: resnet_v1_101/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID",
=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/we
(Check whether your GraphDef-interpreting binary is up to date with your GraphDef-generating binary.).
     [[Node: resnet_v1_101/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID", strides=[
 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/weights)]]
Traceback (most recent call last):
  File "time_task_of_belt_detection_new.py", line 186, in <module>
    tf.app.run()
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "time_task_of_belt_detection_new.py", line 175, in main
    [predictions], feed_dict={image_tensor: image_np_expanded})
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 889, in run
    run_metadata_ptr)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1120, in _run
    feed_dict_tensor, options, run_metadata)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1317, in _do_run
    options, run_metadata)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1336, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: NodeDef mentions attr 'dilations' not in Op<name=Conv2D; signature=input:
r: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
allowed=["SAME", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", "NCHW"]>; NodeDef: resnet_v1_101/conv1/Conv2D = Conv
FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/job:localhos
a:0/task:0/device:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/weights). (Check whether your GraphDef-interpreting binary is up to da
your GraphDef-generating binary.).
     [[Node: resnet_v1_101/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID", strides=[
 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/weights)]]

Caused by op u'resnet_v1_101/conv1/Conv2D', defined at:
  File "time_task_of_belt_detection_new.py", line 186, in <module>
    tf.app.run()
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 48, in run
    _sys.exit(main(_sys.argv[:1] + flags_passthrough))
  File "time_task_of_belt_detection_new.py", line 98, in main
    tf.import_graph_def(od_graph_def, name='')
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/importer.py", line 313, in import_graph_def
    op_def=op_def)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 2956, in create_op
    op_def=op_def)
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 1470, in __init__
    self._traceback = self._graph._extract_stack()  # pylint: disable=protected-access

InvalidArgumentError (see above for traceback): NodeDef mentions attr 'dilations' not in Op<name=Conv2D; signature=input:T, filter:T ->
T; attr=T:type,allowed=[DT_HALF, DT_FLOAT]; attr=strides:list(int); attr=use_cudnn_on_gpu:bool,default=true; attr=padding:string,allowe
", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", "NCHW"]>; NodeDef: resnet_v1_101/conv1/Conv2D = Conv2D[T=DT_FLOAT,
rmat="NHWC", dilations=[1, 1, 1, 1], padding="VALID", strides=[1, 2, 2, 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/ta
ice:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/weights). (Check whether your GraphDef-interpreting binary is up to date with your G
generating binary.).
     [[Node: resnet_v1_101/conv1/Conv2D = Conv2D[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="VALID", strides=[
 1], use_cudnn_on_gpu=true, _device="/job:localhost/replica:0/task:0/device:GPU:0"](resnet_v1_101/Pad, resnet_v1_101/conv1/weights)]]