Display exported concrete and polymorphic functions in SavedModel CLI

[captain-pool]

Adding features to saved_model_cli to show SavedModel2.0 Functions.
SavedModelCLI doesn't show details of Polymorphic Functions or multiple Concrete Functions. This Pull request addresses this problem,
The Sample output is provided here: #30752 (comment)
This PR is made for Google Summer of Code 2019

CC: @vbardiovskyg @srjoglekar246

[vbardiovskyg]

Thanks for contributing. I think this is a great first step to start showing defined functions on SavedModelV2s that don't expose any v1 style signatures.

Could you please copy/paste a typical output of the CLI feature you are adding into the review comment?

[captain-pool]

As Requested, Output of SavedModel2.0

Model

import tensorflow.compat.v2 as tf
tf.enable_v2_behaviour()
class A(tf.train.Checkpoint):
  @tf.function
  def func1(self, a, b, c):
    if c:
      return a + b 
    else:
      return a * b
  @tf.function(input_signature=[tf.TensorSpec(shape=(2, 2), dtype=tf.float32)])
  def func2(self, x):
    return x + 2
  @tf.function
  def __call__(self, y, c=7):
    return y + 2 * c

a = A()
a.func1(tf.constant(5), tf.constant(9), {"key1": [{"k1":tf.constant(5), "k2":1}, 2], "key2": 2})
a.func1(tf.constant(5), tf.constant(9), False)
a(tf.constant(5))
tf.saved_model.save(a, "/tmp/model")

Run

$ saved_model_cli show --dir /tmp/model --all

Output

MetaGraphDef with tag-set: 'serve' contains the following SignatureDefs:

signature_def['__saved_model_init_op']:
  The given SavedModel SignatureDef contains the following input(s):
  The given SavedModel SignatureDef contains the following output(s):
    outputs['__saved_model_init_op'] tensor_info:
        dtype: DT_INVALID
        shape: unknown_rank
        name: NoOp
  Method name is: 

signature_def['serving_default']:
  The given SavedModel SignatureDef contains the following input(s):
    inputs['x'] tensor_info:
        dtype: DT_FLOAT
        shape: (2, 2)
        name: serving_default_x:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['output_0'] tensor_info:
        dtype: DT_FLOAT
        shape: (2, 2)
        name: PartitionedCall:0
  Method name is: tensorflow/serving/predict

Defined Functions:
  Function Name: 'func2'
    Option #1
      Callable with:
        Argument #1
          x: TensorSpec(shape=(2, 2), dtype=tf.float32, name='x')

  Function Name: 'func1'
    Option #1
      Callable with:
        Argument #1
          a: TensorSpec(shape=(), dtype=tf.int32, name='a')
        Argument #2
          b: TensorSpec(shape=(), dtype=tf.int32, name='b')
        Argument #3
          DType: bool
          Value: True
    Option #2
      Callable with:
        Argument #1
          a: TensorSpec(shape=(), dtype=tf.int32, name='a')
        Argument #2
          b: TensorSpec(shape=(), dtype=tf.int32, name='b')
        Argument #3
          DType: dict
          Value: {'key1': [{'k1': TensorSpec(shape=(), dtype=tf.int32, name='c/key1/0/k1'), 'k2': 1}, 2], 'key2': 2} 

  Function Name: '__call__'
    Option #1
      Callable with:
        Argument #1
          y: TensorSpec(shape=(), dtype=tf.int32, name='y')
        Argument #2
          DType: int
          Value: 7

[vbardiovskyg]

Sorry for the delay.

The PR looks good except for the single comment and also the missing test. Please add at least one to saved_model_cli_test. Most of that file is loading a model from saved_model/testdata/half_plus_two/00000123, but I think you can just create a model and save it in temp directory, then load it, all from one test function.

[captain-pool]

Added the test, @vbardiovskyg please take a look :)

[captain-pool]

@vbardiovskyg

[vbardiovskyg]

Thanks for the change!

[captain-pool]

@vbardiovskyg

[vbardiovskyg]

Looks good to me.

[captain-pool]

@vbardiovskyg , This should remove the linter issues.

[vbardiovskyg]

Not sure why you didn't break the line using "" or brackets, but approving since it is not a big deal.