Type tf.keras.experimental
Namespace tensorflow
Methods
- export_saved_model
- export_saved_model
- load_from_saved_model
- load_from_saved_model_dyn
- terminate_keras_multiprocessing_pools
- terminate_keras_multiprocessing_pools_dyn
Properties
Public static methods
void export_saved_model(Model model, string saved_model_path, IDictionary<string, object> custom_objects, bool as_text, TensorSpec input_signature, bool serving_only)
Exports a
tf.keras.Model as a Tensorflow SavedModel. (deprecated) Warning: THIS FUNCTION IS DEPRECATED. It will be removed in a future version.
Instructions for updating:
Please use `model.save(..., save_format="tf")` or `tf.keras.models.save_model(..., save_format="tf")`. Note that at this time, subclassed models can only be saved using
`serving_only=True`. The exported `SavedModel` is a standalone serialization of Tensorflow objects,
and is supported by TF language APIs and the Tensorflow Serving system.
To load the model, use the function
tf.keras.experimental.load_from_saved_model. The `SavedModel` contains: 1. a checkpoint containing the model weights.
2. a `SavedModel` proto containing the Tensorflow backend graph. Separate
graphs are saved for prediction (serving), train, and evaluation. If
the model has not been compiled, then only the graph computing predictions
will be exported.
3. the model's json config. If the model is subclassed, this will only be
included if the model's `get_config()` method is overwritten. Example:
Parameters
-
Modelmodel - A
tf.keras.Modelto be saved. If the model is subclassed, the flag `serving_only` must be set to True. -
stringsaved_model_path - a string specifying the path to the SavedModel directory.
-
IDictionary<string, object>custom_objects - Optional dictionary mapping string names to custom classes or functions (e.g. custom loss functions).
-
boolas_text - bool, `False` by default. Whether to write the `SavedModel` proto in text format. Currently unavailable in serving-only mode.
-
TensorSpecinput_signature - A possibly nested sequence of
tf.TensorSpecobjects, used to specify the expected model inputs. Seetf.functionfor more details. -
boolserving_only - bool, `False` by default. When this is true, only the prediction graph is saved.
Show Example
import tensorflow as tf # Create a tf.keras model. model = tf.keras.Sequential() model.add(tf.keras.layers.Dense(1, input_shape=[10])) model.summary() # Save the tf.keras model in the SavedModel format. path = '/tmp/simple_keras_model' tf.keras.experimental.export_saved_model(model, path) # Load the saved keras model back. new_model = tf.keras.experimental.load_from_saved_model(path) new_model.summary()
void export_saved_model(Sequential model, string saved_model_path, IDictionary<string, object> custom_objects, bool as_text, TensorSpec input_signature, bool serving_only)
Exports a
tf.keras.Model as a Tensorflow SavedModel. (deprecated) Warning: THIS FUNCTION IS DEPRECATED. It will be removed in a future version.
Instructions for updating:
Please use `model.save(..., save_format="tf")` or `tf.keras.models.save_model(..., save_format="tf")`. Note that at this time, subclassed models can only be saved using
`serving_only=True`. The exported `SavedModel` is a standalone serialization of Tensorflow objects,
and is supported by TF language APIs and the Tensorflow Serving system.
To load the model, use the function
tf.keras.experimental.load_from_saved_model. The `SavedModel` contains: 1. a checkpoint containing the model weights.
2. a `SavedModel` proto containing the Tensorflow backend graph. Separate
graphs are saved for prediction (serving), train, and evaluation. If
the model has not been compiled, then only the graph computing predictions
will be exported.
3. the model's json config. If the model is subclassed, this will only be
included if the model's `get_config()` method is overwritten. Example:
Parameters
-
Sequentialmodel - A
tf.keras.Modelto be saved. If the model is subclassed, the flag `serving_only` must be set to True. -
stringsaved_model_path - a string specifying the path to the SavedModel directory.
-
IDictionary<string, object>custom_objects - Optional dictionary mapping string names to custom classes or functions (e.g. custom loss functions).
-
boolas_text - bool, `False` by default. Whether to write the `SavedModel` proto in text format. Currently unavailable in serving-only mode.
-
TensorSpecinput_signature - A possibly nested sequence of
tf.TensorSpecobjects, used to specify the expected model inputs. Seetf.functionfor more details. -
boolserving_only - bool, `False` by default. When this is true, only the prediction graph is saved.
Show Example
import tensorflow as tf # Create a tf.keras model. model = tf.keras.Sequential() model.add(tf.keras.layers.Dense(1, input_shape=[10])) model.summary() # Save the tf.keras model in the SavedModel format. path = '/tmp/simple_keras_model' tf.keras.experimental.export_saved_model(model, path) # Load the saved keras model back. new_model = tf.keras.experimental.load_from_saved_model(path) new_model.summary()
object load_from_saved_model(object saved_model_path, object custom_objects)
Loads a keras Model from a SavedModel created by `export_saved_model()`. (deprecated) Warning: THIS FUNCTION IS DEPRECATED. It will be removed in a future version.
Instructions for updating:
The experimental save and load functions have been deprecated. Please switch to
tf.keras.models.load_model. This function reinstantiates model state by:
1) loading model topology from json (this will eventually come
from metagraph).
2) loading model weights from checkpoint. Example:
Parameters
-
objectsaved_model_path - a string specifying the path to an existing SavedModel.
-
objectcustom_objects - Optional dictionary mapping names (strings) to custom classes or functions to be considered during deserialization.
Returns
-
object - a keras.Model instance.
Show Example
import tensorflow as tf # Create a tf.keras model. model = tf.keras.Sequential() model.add(tf.keras.layers.Dense(1, input_shape=[10])) model.summary() # Save the tf.keras model in the SavedModel format. path = '/tmp/simple_keras_model' tf.keras.experimental.export_saved_model(model, path) # Load the saved keras model back. new_model = tf.keras.experimental.load_from_saved_model(path) new_model.summary()
object load_from_saved_model_dyn(object saved_model_path, object custom_objects)
Loads a keras Model from a SavedModel created by `export_saved_model()`. (deprecated) Warning: THIS FUNCTION IS DEPRECATED. It will be removed in a future version.
Instructions for updating:
The experimental save and load functions have been deprecated. Please switch to
tf.keras.models.load_model. This function reinstantiates model state by:
1) loading model topology from json (this will eventually come
from metagraph).
2) loading model weights from checkpoint. Example:
Parameters
-
objectsaved_model_path - a string specifying the path to an existing SavedModel.
-
objectcustom_objects - Optional dictionary mapping names (strings) to custom classes or functions to be considered during deserialization.
Returns
-
object - a keras.Model instance.
Show Example
import tensorflow as tf # Create a tf.keras model. model = tf.keras.Sequential() model.add(tf.keras.layers.Dense(1, input_shape=[10])) model.summary() # Save the tf.keras model in the SavedModel format. path = '/tmp/simple_keras_model' tf.keras.experimental.export_saved_model(model, path) # Load the saved keras model back. new_model = tf.keras.experimental.load_from_saved_model(path) new_model.summary()
IList<string> terminate_keras_multiprocessing_pools(double grace_period, bool use_sigkill)
Destroy Keras' multiprocessing pools to prevent deadlocks. In general multiprocessing.Pool can interact quite badly with other, seemingly
unrelated, parts of a codebase due to Pool's reliance on fork. This method
cleans up all pools which are known to belong to Keras (and thus can be safely
terminated).
Parameters
-
doublegrace_period - Time (in seconds) to wait for process cleanup to propagate.
-
booluse_sigkill - Boolean of whether or not to perform a cleanup pass using SIGKILL.
Returns
-
IList<string> - A list of human readable strings describing all issues encountered. It is up to the caller to decide whether to treat this as an error condition.
object terminate_keras_multiprocessing_pools_dyn(ImplicitContainer<T> grace_period, ImplicitContainer<T> use_sigkill)
Destroy Keras' multiprocessing pools to prevent deadlocks. In general multiprocessing.Pool can interact quite badly with other, seemingly
unrelated, parts of a codebase due to Pool's reliance on fork. This method
cleans up all pools which are known to belong to Keras (and thus can be safely
terminated).
Parameters
-
ImplicitContainer<T>grace_period - Time (in seconds) to wait for process cleanup to propagate.
-
ImplicitContainer<T>use_sigkill - Boolean of whether or not to perform a cleanup pass using SIGKILL.
Returns
-
object - A list of human readable strings describing all issues encountered. It is up to the caller to decide whether to treat this as an error condition.