Building TensorFlow Systems from Components

Thank you!

planspace.org

@planarrowspace

$ pip install --upgrade tensorflow
$ pip freeze | grep tensorflow
## tensorflow==1.1.0

$ pip install --upgrade tensorflow
$ pip freeze | grep tensorflow
## tensorflow==1.1.0

THE BIG IDEA

use what you need

THE PLAN

• one short work
• three longer works

Do what you want!

Intro by Logo

short work

draw a system!

• block diagram
• add detail
• pseudocode?

short work over

Let's do hard AI!

(build)

• getting data in
• distributed programs
• high-level ML APIs

What even is TensorFlow?

graph or not graph?

• getting data in

• getting data in
  • to the graph
  • TFRecords

(notebook)

long work

work with data!

• your own system?
• planspace.org: mystery.tfrecords
• move to graph

long work over

• distributed programs

• distributed programs
  • command-line arguments
  • MapReduce example

command-line arguments

$ python script.py --color red
a red flower

import sys

def main():
    assert sys.argv[1] == '--color'
    print('a {} flower'.format(sys.argv[2]))

if __name__ == '__main__':
    main()

import sys
import tensorflow as tf

flags = tf.app.flags
flags.DEFINE_string(flag_name='color',
                    default_value='green',
                    docstring='the color to make a flower')

def main():
    flags.FLAGS._parse_flags(args=sys.argv[1:])
    print('a {} flower'.format(flags.FLAGS.color))

if __name__ == '__main__':
    main()

import sys
import gflags

gflags.DEFINE_string(name='color',
                     default='green',
                     help='the color to make a flower')

def main():
    gflags.FLAGS(sys.argv)
    print('a {} flower'.format(gflags.FLAGS.color))

if __name__ == '__main__':
    main()

import sys
import argparse

parser = argparse.ArgumentParser()
parser.add_argument('--color',
                    default='green',
                    help='the color to make a flower')

def main():
    args = parser.parse_args(sys.argv[1:])
    print('a {} flower'.format(args.color))

if __name__ == '__main__':
    main()

import tensorflow as tf

def main(args):
    assert args[1] == '--color'
    print('a {} flower'.format(args[2]))

if __name__ == '__main__':
    tf.app.run()

import tensorflow as tf

flags = tf.app.flags
flags.DEFINE_string(flag_name='color',
                    default_value='green',
                    docstring='the color to make a flower')

def main(args):
    print('a {} flower'.format(flags.FLAGS.color))

if __name__ == '__main__':
    tf.app.run()

import google.apputils.app
import gflags

gflags.DEFINE_string(name='color',
                     default='green',
                     help='the color to make a flower')

def main(args):
    print('a {} flower'.format(gflags.FLAGS.color))

if __name__ == '__main__':
    google.apputils.app.run()

import argparse

parser = argparse.ArgumentParser()
parser.add_argument('--color',
                    default='green',
                    help='the color to make a flower')

def main():
    args = parser.parse_args()
    print('a {} flower'.format(args.color))

if __name__ == '__main__':
    main()

Whew!

MapReduce example

tf.device()

with tf.device('/cpu:0'):
    # Do something.

with tf.device('/gpu:0'):
    # Do something.

with tf.device('/job:ps/task:0'):
    # Do something.

(demo)

(code)

long work

make something happen!

• your own system?
• different distributed functionality?
• add command-line?

long work over

• high-level ML APIs

• high-level ML APIs
  • training an Estimator
  • pre-trained Keras

(notebook)

long work

make something happen!

• your own system?
• flip regression to classification?
• classify some images?

long work over

What else?

debugging, optimizing (XLA, low-precision, etc.), serving, building custom network architectures, embeddings, recurrent, generative, bazel, protobuf, gRPC, queues, threading...

Thanks!

planspace.org

@planarrowspace