tensorflow.python.client.session.BaseSession.run

Function

Executes a computational graph.

Prototype

def run(self, fetches, feed_dict=None, options=None, run_metadata=None)

Parameters

Parameter	Type	Mandatory/Optional	Description
fetches	str tf.Operation tf.Variable tf.Tensor tf.sparse.SparseTensor list tuple dict	Mandatory	Runs the operation or obtains the tensor.
feed_dict	tf.Variable tf.Tensor tf.sparse.SparseTensor list tuple dict	Optional	Overwrites the tensor value in the graph.
options	tf.compat.v1.RunOptions	Optional	Controls the behavior of a specific step.
run_metadata	tf.compat.v1.RunMetadata	Optional	Collects non-tensor output at a specific step.

Return Value

Success: If fetches is a single element, a single element value is returned. If fetches is a list, a list is returned. If fetches is a dict, a dict with the same key is returned.
Failure: An exception is thrown.

Example

The following provides only an example of the usage process.

#1. Import required libraries.
import tensorflow as tf
from mx_rec.util.initialize import init
# 2. Build a computational graph.
# ...
# 3. Call the API for training.
with tf.compat.v1.Session() as sess:
     sess.run([train_ops])      # train_ops is the training operator built in the computational graph.

Parent topic: TensorFlow APIs