As computational graph and serialized description of computation

General Discussion
, ,
#1

I am studying the way the TFsaved data.
For saved_model format, one textbook mentions it saves “computational graph” wherein , in Training checkpoints  |  TensorFlow Core, it says it saves “serialized description of computation”. Are they referring to same? I am assuming so. Thanks.,

#3

Take a look at:

1 Like
#5

I see following files are created as saved_model:
p297_saved_model.h5
├── p297_saved_model.h5/assets
├── p297_saved_model.h5/saved_model.pb
└── p297_saved_model.h5/variables
├── p297_saved_model.h5/variables/variables.data-00000-of-00001
└── p297_saved_model.h5/variables/variables.index

Is there a tool to read back without loading? I see those are binary files.

#6

Read back? What is your goal?

#7

Well, i was thinking to see what is there without loading the model. But it may not be necessary.
Regardless, I found an answer to my question while reading a back. It turns out tf comes with handy cli tool (output below):


saved_model_cli show --dir p297 --all

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['flatten_input'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 28, 28)
        name: serving_default_flatten_input:0
  The given SavedModel SignatureDef contains the following output(s):
    outputs['dense_2'] tensor_info:
        dtype: DT_FLOAT
        shape: (-1, 30)
        name: StatefulPartitionedCall:0
  Method name is: tensorflow/serving/predict

Defined Functions:
  Function Name: '__call__'
    Option #1
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #2
      Callable with:
        Argument #1
          flatten_input: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='flatten_input')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #3
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None
    Option #4
      Callable with:
        Argument #1
          flatten_input: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='flatten_input')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None

  Function Name: '_default_save_signature'
    Option #1
      Callable with:
        Argument #1
          flatten_input: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='flatten_input')

  Function Name: 'call_and_return_all_conditional_losses'
    Option #1
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #2
      Callable with:
        Argument #1
          inputs: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='inputs')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None
    Option #3
      Callable with:
        Argument #1
          flatten_input: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='flatten_input')
        Argument #2
          DType: bool
          Value: False
        Argument #3
          DType: NoneType
          Value: None
    Option #4
      Callable with:
        Argument #1
          flatten_input: TensorSpec(shape=(None, 28, 28), dtype=tf.float32, name='flatten_input')
        Argument #2
          DType: bool
          Value: True
        Argument #3
          DType: NoneType
          Value: None
2 Likes