Eager Execution，一个灵活的机器学习平台

!pip install -q --upgrade tensorflow==1.11

You are using pip version 18.0, however version 18.1 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.

from __future__ import absolute_import, division, print_function

import tensorflow as tf

tf.enable_eager_execution()

tf.executing_eagerly()

True

x = [[2.]]
m = tf.matmul(x, x)
print("hello, {}".format(m))

hello, [[4.]]

a = tf.constant([[1, 2],
                 [3, 4]])
print(a)

tf.Tensor(
[[1 2]
[3 4]], shape=(2, 2), dtype=int32)

# Broadcasting support
b = tf.add(a, 1)
print(b)

tf.Tensor(
[[2 3]
[4 5]], shape=(2, 2), dtype=int32)

# Operator overloading is supported
print(a * b)

tf.Tensor(
[[ 2  6]
[12 20]], shape=(2, 2), dtype=int32)

# Use NumPy values
import numpy as np

c = np.multiply(a, b)
print(c)

[[ 2  6]
[12 20]]

# Obtain numpy value from a tensor:
print(a.numpy())
# => [[1 2]
#     [3 4]]

[[1 2]
[3 4]]

tfe = tf.contrib.eager

def fizzbuzz(max_num):
  counter = tf.constant(0)
  max_num = tf.convert_to_tensor(max_num)
  for num in range(1, max_num.numpy()+1):
    num = tf.constant(num)
    if int(num % 3) == 0 and int(num % 5) == 0:
      print('FizzBuzz')
    elif int(num % 3) == 0:
      print('Fizz')
    elif int(num % 5) == 0:
      print('Buzz')
    else:
      print(num.numpy())
    counter += 1

fizzbuzz(15)

1
2
Fizz
4
Buzz
Fizz
7
8
Fizz
Buzz
11
Fizz
13
14
FizzBuzz

class MySimpleLayer(tf.keras.layers.Layer):
  def __init__(self, output_units):
    super(MySimpleLayer, self).__init__()
    self.output_units = output_units

  def build(self, input_shape):
    # The build method gets called the first time your layer is used.
    # Creating variables on build() allows you to make their shape depend
    # on the input shape and hence removes the need for the user to specify
    # full shapes. It is possible to create variables during __init__() if
    # you already know their full shapes.
    self.kernel = self.add_variable(
      "kernel", [input_shape[-1], self.output_units])

  def call(self, input):
    # Override call() instead of __call__ so we can perform some bookkeeping.
    return tf.matmul(input, self.kernel)

model = tf.keras.Sequential([
  tf.keras.layers.Dense(10, input_shape=(784,)),  # must declare input shape
  tf.keras.layers.Dense(10)
])

class MNISTModel(tf.keras.Model):
  def __init__(self):
    super(MNISTModel, self).__init__()
    self.dense1 = tf.keras.layers.Dense(units=10)
    self.dense2 = tf.keras.layers.Dense(units=10)

  def call(self, input):
    """Run the model."""
    result = self.dense1(input)
    result = self.dense2(result)
    result = self.dense2(result)  # reuse variables from dense2 layer
    return result

model = MNISTModel()

w = tf.Variable([[1.0]])
with tf.GradientTape() as tape:
  loss = w * w

grad = tape.gradient(loss, w)
print(grad)  # => tf.Tensor([[ 2.]], shape=(1, 1), dtype=float32)

tf.Tensor([[2.]], shape=(1, 1), dtype=float32)

# Fetch and format the mnist data
(mnist_images, mnist_labels), _ = tf.keras.datasets.mnist.load_data()

dataset = tf.data.Dataset.from_tensor_slices(
  (tf.cast(mnist_images[...,tf.newaxis]/255, tf.float32),
   tf.cast(mnist_labels,tf.int64)))
dataset = dataset.shuffle(1000).batch(32)

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step

# Build the model
mnist_model = tf.keras.Sequential([
  tf.keras.layers.Conv2D(16,[3,3], activation='relu'),
  tf.keras.layers.Conv2D(16,[3,3], activation='relu'),
  tf.keras.layers.GlobalAveragePooling2D(),
  tf.keras.layers.Dense(10)
])

for images,labels in dataset.take(1):
  print("Logits: ", mnist_model(images[0:1]).numpy())

Logits:  [[-1.9521490e-02  2.2975644e-02  2.8935237e-02  2.0388789e-02
 -1.8511273e-02 -6.4317137e-05  6.0662534e-03 -1.7174225e-02
  5.4899108e-02 -2.8871424e-02]]

optimizer = tf.train.AdamOptimizer()

loss_history = []

for (batch, (images, labels)) in enumerate(dataset.take(400)):
  if batch % 80 == 0:
    print()
  print('.', end='')
  with tf.GradientTape() as tape:
    logits = mnist_model(images, training=True)
    loss_value = tf.losses.sparse_softmax_cross_entropy(labels, logits)

  loss_history.append(loss_value.numpy())
  grads = tape.gradient(loss_value, mnist_model.variables)
  optimizer.apply_gradients(zip(grads, mnist_model.variables),
                            global_step=tf.train.get_or_create_global_step())

................................................................................
................................................................................

import matplotlib.pyplot as plt

plt.plot(loss_history)
plt.xlabel('Batch #')
plt.ylabel('Loss [entropy]')

Text(0, 0.5, 'Loss [entropy]')

class Model(tf.keras.Model):
  def __init__(self):
    super(Model, self).__init__()
    self.W = tf.Variable(5., name='weight')
    self.B = tf.Variable(10., name='bias')
  def call(self, inputs):
    return inputs * self.W + self.B

# A toy dataset of points around 3 * x + 2
NUM_EXAMPLES = 2000
training_inputs = tf.random_normal([NUM_EXAMPLES])
noise = tf.random_normal([NUM_EXAMPLES])
training_outputs = training_inputs * 3 + 2 + noise

# The loss function to be optimized
def loss(model, inputs, targets):
  error = model(inputs) - targets
  return tf.reduce_mean(tf.square(error))

def grad(model, inputs, targets):
  with tf.GradientTape() as tape:
    loss_value = loss(model, inputs, targets)
  return tape.gradient(loss_value, [model.W, model.B])

# Define:
# 1. A model.
# 2. Derivatives of a loss function with respect to model parameters.
# 3. A strategy for updating the variables based on the derivatives.
model = Model()
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01)

print("Initial loss: {:.3f}".format(loss(model, training_inputs, training_outputs)))

# Training loop
for i in range(300):
  grads = grad(model, training_inputs, training_outputs)
  optimizer.apply_gradients(zip(grads, [model.W, model.B]),
                            global_step=tf.train.get_or_create_global_step())
  if i % 20 == 0:
    print("Loss at step {:03d}: {:.3f}".format(i, loss(model, training_inputs, training_outputs)))

print("Final loss: {:.3f}".format(loss(model, training_inputs, training_outputs)))
print("W = {}, B = {}".format(model.W.numpy(), model.B.numpy()))

Initial loss: 69.940
Loss at step 000: 67.176
Loss at step 020: 30.184
Loss at step 040: 13.858
Loss at step 060: 6.650
Loss at step 080: 3.467
Loss at step 100: 2.060
Loss at step 120: 1.438
Loss at step 140: 1.163
Loss at step 160: 1.041
Loss at step 180: 0.987
Loss at step 200: 0.963
Loss at step 220: 0.953
Loss at step 240: 0.948
Loss at step 260: 0.946
Loss at step 280: 0.945
Final loss: 0.945
W = 3.00280499458313, B = 2.0040433406829834

if tf.test.is_gpu_available():
  with tf.device("gpu:0"):
    v = tf.Variable(tf.random_normal([1000, 1000]))
    v = None  # v no longer takes up GPU memory

x = tf.Variable(10.)
checkpoint = tf.train.Checkpoint(x=x)

x.assign(2.)   # Assign a new value to the variables and save.
checkpoint_path = './ckpt/'
checkpoint.save('./ckpt/')

'./ckpt/-1'

x.assign(11.)  # Change the variable after saving.

# Restore values from the checkpoint
checkpoint.restore(tf.train.latest_checkpoint(checkpoint_path))

print(x)  # => 2.0

import os

model = tf.keras.Sequential([
  tf.keras.layers.Conv2D(16,[3,3], activation='relu'),
  tf.keras.layers.GlobalAveragePooling2D(),
  tf.keras.layers.Dense(10)
])
optimizer = tf.train.AdamOptimizer(learning_rate=0.001)
checkpoint_dir = '/path/to/model_dir'
os.makedirs(checkpoint_dir, exist_ok=True)
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
root = tf.train.Checkpoint(optimizer=optimizer,
                           model=model,
                           optimizer_step=tf.train.get_or_create_global_step())

root.save(checkpoint_prefix)
root.restore(tf.train.latest_checkpoint(checkpoint_dir))

m = tfe.metrics.Mean("loss")
m(0)
m(5)
m.result()  # => 2.5
m([8, 9])
m.result()  # => 5.5

global_step = tf.train.get_or_create_global_step()

logdir = "./tb/"
writer = tf.contrib.summary.create_file_writer(logdir)
writer.set_as_default()

for _ in range(10):
  global_step.assign_add(1)
  # Must include a record_summaries method
  with tf.contrib.summary.record_summaries_every_n_global_steps(100):
    # your model code goes here
    tf.contrib.summary.scalar('global_step', global_step)

ls tb/

events.out.tfevents.1540951101.9fdaf0d0cf8e.v2

def line_search_step(fn, init_x, rate=1.0):
  with tf.GradientTape() as tape:
    # Variables are automatically recorded, but manually watch a tensor
    tape.watch(init_x)
    value = fn(init_x)
  grad = tape.gradient(value, init_x)
  grad_norm = tf.reduce_sum(grad * grad)
  init_value = value
  while value > init_value - rate * grad_norm:
    x = init_x - rate * grad
    value = fn(x)
    rate /= 2.0
  return x, value

def square(x):
  return tf.multiply(x, x)

grad = tfe.gradients_function(square)

square(3.).numpy()

9.0

grad(3.)[0].numpy()

6.0

# The second-order derivative of square:
gradgrad = tfe.gradients_function(lambda x: grad(x)[0])
gradgrad(3.)[0].numpy()

2.0

# The third-order derivative is None:
gradgradgrad = tfe.gradients_function(lambda x: gradgrad(x)[0])
gradgradgrad(3.)

[None]

# With flow control:
def abs(x):
  return x if x > 0. else -x

grad = tfe.gradients_function(abs)

grad(3.)[0].numpy()

1.0

grad(-3.)[0].numpy()

-1.0

@tf.custom_gradient
def clip_gradient_by_norm(x, norm):
  y = tf.identity(x)
  def grad_fn(dresult):
    return [tf.clip_by_norm(dresult, norm), None]
  return y, grad_fn

def log1pexp(x):
  return tf.log(1 + tf.exp(x))
grad_log1pexp = tfe.gradients_function(log1pexp)

# The gradient computation works fine at x = 0.
grad_log1pexp(0.)[0].numpy()

0.5

# However, x = 100 fails because of numerical instability.
grad_log1pexp(100.)[0].numpy()

nan

@tf.custom_gradient
def log1pexp(x):
  e = tf.exp(x)
  def grad(dy):
    return dy * (1 - 1 / (1 + e))
  return tf.log(1 + e), grad

grad_log1pexp = tfe.gradients_function(log1pexp)

# As before, the gradient computation works fine at x = 0.
grad_log1pexp(0.)[0].numpy()

0.5

# And the gradient computation also works at x = 100.
grad_log1pexp(100.)[0].numpy()

1.0

import time

def measure(x, steps):
  # TensorFlow initializes a GPU the first time it's used, exclude from timing.
  tf.matmul(x, x)
  start = time.time()
  for i in range(steps):
    x = tf.matmul(x, x)
  # tf.matmul can return before completing the matrix multiplication
  # (e.g., can return after enqueing the operation on a CUDA stream).
  # The x.numpy() call below will ensure that all enqueued operations
  # have completed (and will also copy the result to host memory,
  # so we're including a little more than just the matmul operation
  # time).
  _ = x.numpy()
  end = time.time()
  return end - start

shape = (1000, 1000)
steps = 200
print("Time to multiply a {} matrix by itself {} times:".format(shape, steps))

# Run on CPU:
with tf.device("/cpu:0"):
  print("CPU: {} secs".format(measure(tf.random_normal(shape), steps)))

# Run on GPU, if available:
if tfe.num_gpus() > 0:
  with tf.device("/gpu:0"):
    print("GPU: {} secs".format(measure(tf.random_normal(shape), steps)))
else:
  print("GPU: not found")

Time to multiply a (1000, 1000) matrix by itself 200 times:
CPU: 0.4751627445220947 secs
GPU: not found

if tf.test.is_gpu_available():
  x = tf.random_normal([10, 10])

  x_gpu0 = x.gpu()
  x_cpu = x.cpu()

  _ = tf.matmul(x_cpu, x_cpu)    # Runs on CPU
  _ = tf.matmul(x_gpu0, x_gpu0)  # Runs on GPU:0

  if tfe.num_gpus() > 1:
    x_gpu1 = x.gpu(1)
    _ = tf.matmul(x_gpu1, x_gpu1)  # Runs on GPU:1

def my_py_func(x):
  x = tf.matmul(x, x)  # You can use tf ops
  print(x)  # but it's eager!
  return x

with tf.Session() as sess:
  x = tf.placeholder(dtype=tf.float32)
  # Call eager function in graph!
  pf = tfe.py_func(my_py_func, [x], tf.float32)

  sess.run(pf, feed_dict={x: [[2.0]]})  # [[4.0]]

tf.Tensor([[4.]], shape=(1, 1), dtype=float32)