How to handle multi tensors input?

[doubler]

The mnist example shows how to use the model with one input tensor x and one output tensor y.
While I have two input, tensor x and a tf.placeholder named dropout_prob.
If I don't specify the dropout_prob value, it will produce error output log like

W external/tf/tensorflow/core/common_runtime/executor.cc:1157] 0x7fcf1c0023c0 Compute status: Invalid argument: You must feed a value for placeholder tensor 'dropout_prob' with dtype float
     [[Node: dropout_prob = Placeholder[dtype=DT_FLOAT, shape=[], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

Would you please show some example with multi input tensors or multi output tensors?

[doubler]

I have solved my problem by pre-define dropout_prob as a tf.constant before export the model.
While I'm still looking forward the multi input and output example. :)

[fangweili]

Unfortunately, neither our standard ClassificationSignature nor RegressionSignature supports multiple input binding for now. You can however, use GenericSignature to specify arbitrary number of inputs and outputs.

You can use exporter.generic_signature() to create it. Then provide it as default_signature in exporter.export().

[doubler]

Got it. Thanks for explanation.

[insectatorious]

@doubler would it be possible for you to share a code example of how you set the dropout_prob using tf.constant? I'm confused about how to fit this into the export step.

[doubler]

@insectatorious no need to set the dropout_prob, just feed it with 1.0 as input when serving online

[fabienbaradel]

@doubler I don't understand your explanation. What do you mean by "feed it with 1.0 as input"?

Below is the inception signature for a standard image classification:

      classification_signature = (
          tf.saved_model.signature_def_utils.build_signature_def(
              inputs={
                  tf.saved_model.signature_constants.CLASSIFY_INPUTS:
                      classify_inputs_tensor_info
              },
              outputs={
                  tf.saved_model.signature_constants.CLASSIFY_OUTPUT_CLASSES:
                      classes_output_tensor_info,
                  tf.saved_model.signature_constants.CLASSIFY_OUTPUT_SCORES:
                      scores_output_tensor_info
              },
              method_name=tf.saved_model.signature_constants.
              CLASSIFY_METHOD_NAME))

Could you show how is yours with two inputs? Do you have to redefine the tf.saved_mode.signature_constants?

Thanks for help.