Bdtjtk

I have a savedModel exported using estimator.export_saved_model and indicate how to parse the input using a function like below:

def serving_example_input_receiver_fn():

    serialized_tf_example = tf.placeholder(dtype=tf.string)

    receiver_tensors = {'examples': serialized_tf_example}

    features = tf.parse_example(serialized_tf_example, FEATURE_SCHEMA)

    return tf.estimator.export.ServingInputReceiver(features, receiver_tensors)

I serve this model using tensorflow_model_server with docker. I can get right predict result through gRpc request, but when I use RESTful API, I can't assemble the right request. I got error below every time:

{ "error": "Could not parse example input, value:
\'b\'\n\xb5\x02\n\x14\n\x0bisorderdish\x12\x05\x1a\x03\n\x01\x00\n\x12\n\tisnewuser\x12\x05\x1a\x03\n\x01\x00\n\x13\n\ncitylevel1\x12\x05\x1a\x03\n\x01\x01\n\x11\n\x08isiphone\x12\x05\x1a\x03\n\x01\x01\n\x17\n\x0eistakeawayshop\x12\x05\x1a\x03\n\x01\x00\n\x0f\n\x06cityid\x12\x05\x1a\x03\n\x01\x01\n\x0f\n\x06iswifi\x12\x05\x1a\x03\n\x01\x00\n\x14\n\x0bgoodreviews\x12\x05\x1a\x03\n\x01\x06\n\x13\n\nbadreviews\x12\x05\x1a\x03\n\x01\x03\n\x12\n\tistopshop\x12\x05\x1a\x03\n\x01\x01\n\x11\n\x06shopid\x12\x07\x1a\x05\n\x03\xb3\xfe3\n\x13\n\nisdealshop\x12\x05\x1a\x03\n\x01\x00\n\x14\n\nshopregion\x12\x06\x1a\x04\n\x02\xca\x06\n\x11\n\x08shopcat0\x12\x05\x1a\x03\n\x01\n\n\x16\n\risbookingshop\x12\x05\x1a\x03\n\x01\x00\'\'nt
[[{{node ParseExample/ParseExample}} = ParseExample[Ndense=0,
Nsparse=15, Tdense=, _output_shapes=[[?,2], [?,2], [?,2], [?,2],
[?,2], ..., [2], [2], [2], [2], [2]], dense_shapes=,
sparse_types=[DT_INT64, DT_INT64, DT_INT64, DT_INT64, DT_INT64, ...,
DT_INT64, DT_INT64, DT_INT64, DT_INT64, DT_INT64],
_device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_input_example_tensor_0_0,
ParseExample/ParseExample/names,
ParseExample/ParseExample/sparse_keys_0,
ParseExample/ParseExample/sparse_keys_1,
ParseExample/ParseExample/sparse_keys_2,
ParseExample/ParseExample/sparse_keys_3,
ParseExample/ParseExample/sparse_keys_4,
ParseExample/ParseExample/sparse_keys_5,
ParseExample/ParseExample/sparse_keys_6,
ParseExample/ParseExample/sparse_keys_7,
ParseExample/ParseExample/sparse_keys_8,
ParseExample/ParseExample/sparse_keys_9,
ParseExample/ParseExample/sparse_keys_10,
ParseExample/ParseExample/sparse_keys_11,
ParseExample/ParseExample/sparse_keys_12,
ParseExample/ParseExample/sparse_keys_13,
ParseExample/ParseExample/sparse_keys_14)]]" }'

The Rest API predict request format is like below:



{

  "signature_name": "predict_method_signature",

  "instances": [

    {

      "examples": str(my_tf_example1.SerializeToString())  //python code to produce the serialized string

    },

    {

      "examples": str(my_tf_example2.SerializeToString())

    }

  ]

}

I cant find useful document or examples in this case. Please help.