tf.compat.v1.train.Saver.restore

Function

Loads a model.

Prototype

def restore(self, sess, save_path)

Parameters

Parameter	Type	Mandatory/Optional	Description
sess	Session	Mandatory	Session of the TensorFlow model to be imported.
save_path	str	Mandatory	Path for storing the model checkpoint file. The local file system and HDFS file system are supported. The length range is [1, 1024]. When loading a model for multi-device training, you can set save_path to the same loading path (which stores the multi-device training result). Each device automatically loads the parameters of the local device. NOTE: The maximum size of a single file to be loaded is 500 GB. Concurrent reads may cause system OOM.

Return Value

Success: None
Failure: An exception is thrown.

Example

The following just briefly describes the usage. For details, see the little demo in the Rec SDK code repository.

#1. Import required libraries.
import tensorflow as tf
from mx_rec.util.initialize import init, get_rank_id
# 2. Build a computational graph.
# ...
# 3. Create a saver.
saver = tf.compat.v1.train.Saver() 
# 4. Obtain the rank ID.
rank_id = get_rank_id()
# 5. Set the number of training steps when the model to be loaded is saved.
latest_step = 200
with tf.compat.v1.Session() as sess:
    saver.restore(sess, f"./saved-model/model-{rank_id}-{latest_step}")

tf.compat.v1.train.Saver.restore

Function

Prototype

Parameters

Return Value

Example

See Also