Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings

tf.compat.v1.saved_model.signature_def_utils.MethodNameUpdater

View source on GitHub

Updates the method name(s) of the SavedModel stored in the given path.

tf.compat.v1.saved_model.signature_def_utils.MethodNameUpdater(
    export_dir
)

The MethodNameUpdater class provides the functionality to update the method name field in the signature_defs of the given SavedModel. For example, it can be used to replace the predict method_name to regress.

Typical usages of the MethodNameUpdater

...
updater = tf.compat.v1.saved_model.MethodNameUpdater(export_dir)
# Update all signature_defs with key "foo" in all meta graph defs.
updater.replace_method_name(signature_key="foo", method_name="regress")
# Update a single signature_def with key "bar" in the meta graph def with
# tags ["serve"]
updater.replace_method_name(signature_key="bar", method_name="classify",
                            tags="serve")
updater.save(new_export_dir)

Args:

  • export_dir: Directory containing the SavedModel files.

Raises:

  • IOError: If the saved model file does not exist, or cannot be successfully parsed.

Methods

replace_method_name

View source

replace_method_name(
    signature_key, method_name, tags=None
)

Replaces the method_name in the specified signature_def.

This will match and replace multiple sig defs iff tags is None (i.e when multiple MetaGraphs have a signature_def with the same key). If tags is not None, this will only replace a single signature_def in the MetaGraph with matching tags.

Args:

  • signature_key: Key of the signature_def to be updated.
  • method_name: new method_name to replace the existing one.
  • tags: A tag or sequence of tags identifying the MetaGraph to update. If None, all meta graphs will be updated.

Raises:

  • ValueError: if signature_key or method_name are not defined or if no metagraphs were found with the associated tags or if no meta graph has a signature_def that matches signature_key.

save

View source

save(
    new_export_dir=None
)

Saves the updated SavedModel.

Args:

  • new_export_dir: Path where the updated SavedModel will be saved. If None, the input SavedModel will be overriden with the updates.

Raises: