neural_compressor.model.tensorflow_model

Class for Tensorflow model.

Module Contents

Classes

TensorflowBaseModel	Build Tensorflow Base Model.
TensorflowSavedModelModel	Build Tensorflow saved model.
TensorflowQATModel	Build Tensorflow QAT model.
TensorflowCheckpointModel	Build Tensorflow checkpoint model.
TensorflowModel	A wrapper to construct a Tensorflow Model.

Functions

get_model_type(model)	Get Tensorflow mode type.
validate_graph_node(graph_def, node_names)	Validate nodes exist in the graph_def.
validate_and_inference_input_output(graph_def, ...)	Validate and inference the input and output tensor names of graph_def.
graph_session(model, input_tensor_names, ...)	Helper to build session with tf.compat.v1.Graph.
graph_def_session(model, input_tensor_names, ...)	Build session with tf.compat.v1.GraphDef.
frozen_pb_session(model, input_tensor_names, ...)	Build session with frozen pb.
load_saved_model(model, saved_model_tags, ...)	Load graph_def from saved model with the default serving signature key.
keras_session(model, input_tensor_names, ...)	Build session with keras model.
slim_session(model, input_tensor_names, ...)	Build session with slim model.
checkpoint_session(model, input_tensor_names, ...)	Build session with ckpt model.
estimator_session(model, input_tensor_names, ...)	Build session with estimator model.
saved_model_session(model, input_tensor_names, ...)	Build session with saved model.

neural_compressor.model.tensorflow_model.get_model_type(model)

Get Tensorflow mode type.

Parameters:

model (string or model object) – model path or model object.

Returns:

model type

Return type:

string

neural_compressor.model.tensorflow_model.validate_graph_node(graph_def, node_names)

Validate nodes exist in the graph_def.

Parameters:

• graph_def (tf.compat.v1.GraphDef) – tf.compat.v1.GraphDef object.

• node_names (list of string) – node names to be validated.

neural_compressor.model.tensorflow_model.validate_and_inference_input_output(graph_def, input_tensor_names, output_tensor_names)

Validate and inference the input and output tensor names of graph_def.

Parameters:

• graph_def (tf.compat.v1.GraphDef) – tf.compat.v1.GraphDef object.

• input_tensor_names (list of string) – input_tensor_names of graph_def.

• output_tensor_names (list of string) – output_tensor_names of graph_def.

Returns:

validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

input_tensor_names (list of string)

neural_compressor.model.tensorflow_model.graph_session(model, input_tensor_names, output_tensor_names, **kwargs)

Helper to build session with tf.compat.v1.Graph.

Parameters:

• model (tf.compat.v1.Graph) – tf.compat.v1.Graph object.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.graph_def_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with tf.compat.v1.GraphDef.

Parameters:

• model (tf.compat.v1.GraphDef) – tf.compat.v1.GraphDef object.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object input_tensor_names (list of string): validated input_tensor_names output_tensor_names (list of string): validated output_tensor_names

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.frozen_pb_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with frozen pb.

Parameters:

• model (string) – model path.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.load_saved_model(model, saved_model_tags, input_tensor_names, output_tensor_names)

Load graph_def from saved model with the default serving signature key.

Parameters:

• saved_model_dir – Directory of the SavedModel.

• saved_model_tags – Set of tags identifying the MetaGraphDef within the SavedModel to analyze.

Returns:

The loaded GraphDef. input_tensors: List of input tensors. output_tensors: List of output tensors.

Return type:

graph_def

neural_compressor.model.tensorflow_model.keras_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with keras model.

Parameters:

• model (string or tf.keras.Model) – model path or tf.keras.Model object.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.slim_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with slim model.

Parameters:

• model (string) – model path.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.checkpoint_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with ckpt model.

Parameters:

• model (string) – model path.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – validated output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.estimator_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with estimator model.

Parameters:

• model (tf.estimator.Estimator) – tf.estimator.Estimator object.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

• kwargs (dict) – other required parameters, like input_fn.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

neural_compressor.model.tensorflow_model.saved_model_session(model, input_tensor_names, output_tensor_names, **kwargs)

Build session with saved model.

Parameters:

• model (string) – model path.

• input_tensor_names (list of string) – input_tensor_names of model.

• output_tensor_names (list of string) – output_tensor_names of model.

Returns:

tf.compat.v1.Session object. input_tensor_names (list of string): validated input_tensor_names. output_tensor_names (list of string): validated output_tensor_names.

Return type:

sess (tf.compat.v1.Session)

class neural_compressor.model.tensorflow_model.TensorflowBaseModel(model, **kwargs)

Bases: neural_compressor.model.base_model.BaseModel

Build Tensorflow Base Model.

property name

Renturn name.

property weights

Return weights.

property q_config

Return q_config.

property workspace_path

Return workspace path.

property model_type

Return model type.

property model

Return model itself.

property graph_def

Return graph defination.

property graph_info

Return graph info.

property sess

Return Session object.

property graph

Return model graph.

property iter_op

Return model iter op list.

property input_tensor_names

Return input tensor names.

property output_tensor_names

Return output tensor names.

property input_node_names

Return input node names.

property output_node_names

Return output node names.

property input_tensor

Return input tensor.

property output_tensor

Return output tensor.

framework()

Return framework.

save(root=None)

Save Tensorflow model.

class neural_compressor.model.tensorflow_model.TensorflowSavedModelModel(model, **kwargs)

Bases: TensorflowBaseModel

Build Tensorflow saved model.

property model

Return model itself.

get_all_weight_names()

Get weight names of model.

Returns:

weight names list.

Return type:

list

update_weights(tensor_name, new_tensor)

Update model weights.

get_weight(tensor_name)

Return model wight with a given tensor name.

Parameters:

tensor_name (str) – name of a tensor.

report_sparsity()

Get sparsity of the model.

Returns:

DataFrame of sparsity of each weight. total_sparsity (float): total sparsity of model.

Return type:

df (DataFrame)

build_saved_model(root=None)

Build Tensorflow saved model.

Parameters:

root (str, optional) – path to saved model. Defaults to None.

Returns:

path to saved model. builder (tf.compat.v1.saved_model.builder.SavedModelBuilder): builds

the SavedModel protocol buffer and saves variables and assets.

Return type:

root (str)

save(root=None)

Save Tensorflow model.

class neural_compressor.model.tensorflow_model.TensorflowQATModel(model='', **kwargs)

Bases: TensorflowSavedModelModel

Build Tensorflow QAT model.

property model

Return model itself.

save(root=None)

Save Tensorflow QAT model.

class neural_compressor.model.tensorflow_model.TensorflowCheckpointModel(model, **kwargs)

Bases: TensorflowBaseModel

Build Tensorflow checkpoint model.

property graph_def

Return graph defination.

class neural_compressor.model.tensorflow_model.TensorflowModel

Bases: object

A wrapper to construct a Tensorflow Model.