GPU device plugins

TensorFlow's pluggable device architecture adds new device support as separate plug-in packages that are installed alongside the official TensorFlow package.

The mechanism requires no device-specific changes in the TensorFlow code. It relies on C APIs to communicate with the TensorFlow binary in a stable manner. Plug-in developers maintain separate code repositories and distribution packages for their plugins and are responsible for testing their devices.

Use device plugins

To use a particular device, like one would a native device in TensorFlow, users only have to install the device plug-in package for that device. The following code snippet shows how the plugin for a new demonstration device, Awesome Processing Unit (APU), is installed and used. For simplicity, this sample APU plug-in only has one custom kernel for ReLU:

# Install the APU example plug-in package
$ pip install tensorflow-apu-0.0.1-cp36-cp36m-linux_x86_64.whl
Successfully installed tensorflow-apu-0.0.1

With the plug-in installed, test that the device is visible and run an operation on the new APU device:

import tensorflow as tf   # TensorFlow registers PluggableDevices here.
tf.config.list_physical_devices()  # APU device is visible to TensorFlow.
[PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU'), PhysicalDevice(name='/physical_device:APU:0', device_type='APU')]

a = tf.random.normal(shape=[5], dtype=tf.float32)  # Runs on CPU.
b =  tf.nn.relu(a)         # Runs on APU.

with tf.device("/APU:0"):  # Users can also use 'with tf.device' syntax.
  c = tf.nn.relu(a)        # Runs on APU.

with tf.device("/CPU:0"):
  c = tf.nn.relu(a)        # Runs on CPU.

@tf.function  # Defining a tf.function
def run():
  d = tf.random.uniform(shape=[100], dtype=tf.float32)  # Runs on CPU.
  e = tf.nn.relu(d)        # Runs on APU.

run()  # PluggableDevices also work with tf.function and graph mode.

Available devices

Metal PluggableDevice for macOS GPUs:

DirectML PluggableDevice for Windows and WSL (preview):

Intel® Extension for TensorFlow PluggableDevice for Linux and WSL: