Obtain Field Values from a Model

1 min

When defining Analytics processors to interact with a Machine Learning model (TensorFlow or TensorFlow Images), you need to fill out a number of fields using values from that model.

This procedure describes how you can obtain the relevant values from your model.

To obtain the relevant field values:

With TensorFlow installed on your local machine, enter this command in shell: saved_model_cli show --dir {PATH_TO_SAVE_MODEL}/{NAME_OF_SAVED_MODEL} --all This gives an output similar to the screenshot below.
TensorFLow output
Find the The given savedModel SignatureDef contains the following input(s) section.
Copy aside:
- The input tensor name (input operation) from the name parameter (serving_default_conv1d_input in the screenshot above).
- The window size and number of inputs expected from the shape parameter. (-1,48,19) in the screenshot above means that the window size is 48, and the number of inputs is 19.
Find the The given savedModel SignatureDef contains the following output(s) section.
Copy aside the output tensor name (output operation) from the name parameter (StatefulPartitionedCall in the screenshot above).