tf.saved_model.experimental.Fingerprint

The SavedModel fingerprint.

tf.saved_model.experimental.Fingerprint(
    saved_model_checksum: int = None,
    graph_def_program_hash: int = None,
    signature_def_hash: int = None,
    saved_object_graph_hash: int = None,
    checkpoint_hash: int = None,
    version: int = None
)

Each attribute of this class is named after a field name in the FingerprintDef proto and contains the value of the respective field in the protobuf.

saved_model_checksum Value of thesaved_model_checksum.
graph_def_program_hash Value of the graph_def_program_hash.
signature_def_hash Value of the signature_def_hash.
saved_object_graph_hash Value of the saved_object_graph_hash.
checkpoint_hash Value of the checkpoint_hash.
version Value of the producer field of the VersionDef.

saved_model_checksum A uint64 containing the saved_model_checksum.
graph_def_program_hash A uint64 containing graph_def_program_hash.
signature_def_hash A uint64 containing the signature_def_hash.
saved_object_graph_hash A uint64 containing the saved_object_graph_hash.
checkpoint_hash A uint64 containing thecheckpoint_hash.
version An int32 containing the producer field of the VersionDef.

Methods

from_proto

View source

@classmethod
from_proto(
    proto: fingerprint_pb2.FingerprintDef
) -> 'Fingerprint'

Constructs Fingerprint object from protocol buffer message.

singleprint

View source

singleprint() -> fingerprinting_pywrap.Singleprint

Canonical fingerprinting ID for a SavedModel.

Uniquely identifies a SavedModel based on the regularized fingerprint attributes. (saved_model_checksum is sensitive to immaterial changes and thus non-deterministic.)

Returns
The string concatenation of graph_def_program_hash, signature_def_hash, saved_object_graph_hash, and checkpoint_hash fingerprint attributes (separated by '/').

Raises
ValueError If the fingerprint fields cannot be used to construct the singleprint.

__eq__

View source

__eq__(
    other: Any
) -> bool

Return self==value.