tf.config.experimental.list_logical_devices

TensorFlow 1 version View source on GitHub

Return a list of logical devices created by runtime.

tf.config.experimental.list_logical_devices(
    device_type=None
)

Logical devices may correspond to physical devices or remote devices in the cluster. Operations and tensors may be placed on these devices by using the name of the LogicalDevice.

For example:

logical_devices = tf.config.experimental.list_logical_devices('GPU')
# Allocate on GPU:0
with tf.device(logical_devices[0].name):
  one = tf.constant(1)
# Allocate on GPU:1
with tf.device(logical_devices[1].name):
  two = tf.constant(2)

device_type (optional) Device type to filter by such as "CPU" or "GPU"

List of LogicalDevice objects