The operation mode 1.eager
# In the operation mode can be directly eager
x = [[3.]]
m = tf.matmul(x, x)
print(m.numpy())
a = tf.constant([[1,9],[3,6]])
print(a)
b = tf.add(a, 2)
print(b)
print(a*b)
import numpy as np
s = np.multiply(a,b)
print(s)
[[9.]]
tf.Tensor(
[[1 9]
[3 6]], shape=(2, 2), dtype=int32)
tf.Tensor(
[[ 3 11]
[ 5 8]], shape=(2, 2), dtype=int32)
tf.Tensor(
[[ 3 99]
[15 48]], shape=(2, 2), dtype=int32)
[[ 3 99]
[15 48]]
2. Dynamic flow control
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):
Surely = tf.constant (whether)
if int (whether% 3) == 0 and int (or 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 (16)
1
2
Fizz
4
Buzz
Fizz
7
8
Fizz
Buzz
11
Fizz
13
14
FizzBuzz
16
3. Build Model
# If the layer must be enforced, self.dynamic = True is set in the constructor:
class MySimpleLayer(tf.keras.layers.Layer):
def __init__(self, output_units):
super(MySimpleLayer, self).__init__()
self.output_units = output_units
self.dynamic = True
def build(self, input_shape):
self.kernel = self.add_variable(
"kernel", [input_shape[-1], self.output_units])
def call(self, input):
return tf.matmul(input, self.kernel)
# Construct a model
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, inputs):
"""Run the model."""
result = self.dense1(inputs)
result = self.dense2(result)
result = self.dense2(result) # reuse variables from dense2 layer
return result
model = MNISTModel()
4. Use the training mode eager
# Compute the gradient
w = tf.Variable([[1.0]])
with tf.GradientTape() as tape:
loss = w*w
grad = tape.gradient(loss, w)
print (city)
tf.Tensor([[2.]], shape=(1, 1), dtype=float32)
# A training model
(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)
mnist_model = tf.keras.Sequential([
tf.keras.layers.Conv2D(16,[3,3], activation='relu',
input_shape=(None, None, 1)),
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())
optimizer = tf.keras.optimizers.Adam ()
loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
loss_history = []
for (batch, (images, labels)) in enumerate(dataset.take(400)):
if batch % 10 == 0:
print('.', end='')
with tf.GradientTape() as tape:
logits = mnist_model(images, training=True)
loss_value = loss_object(labels, logits)
loss_history.append(loss_value.numpy().mean())
grads = tape.gradient(loss_value, mnist_model.trainable_variables)
optimizer.apply_gradients(zip(grads, mnist_model.trainable_variables))
import matplotlib.pyplot as plt
plt.plot(loss_history)
plt.xlabel('Batch #')
plt.ylabel('Loss [entropy]')
Logits: [[-0.05677134 0.00057287 0.00242344 0.04665339 -0.07706797 0.00323912
0.01508885 -0.07409166 -0.01701452 0.05076526]]
........................................
Text(0, 0.5, 'Loss [entropy]')
The optimization variables derivative
MyModel class (tf.keras.Model):
def __init__(self):
super (MyModel, 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 = MyModel ()
optimizer = tf.keras.optimizers.SGD(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]))
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: 68.801
Loss at step 000: 66.129
Loss at step 020: 30.134
Loss at step 040: 14.026
Loss at step 060: 6.816
Loss at step 080: 3.589
Loss at step 100: 2.144
Loss at step 120: 1.497
Loss at step 140: 1.207
Loss at step 160: 1.078
Loss at step 180: 1.020
Loss at step 200: 0.994
Loss at step 220: 0.982
Loss at step 240: 0.977
Loss at step 260: 0.974
Loss at step 280: 0.973
Final loss: 0.973
W = 3.0180840492248535, B = 2.0045783519744873
Objects under 6.eager mode
# Variable that is subject
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
Save Object #
x = tf.Variable(6.0)
checkpoint = tf.train.Checkpoint(x=x)
x.assign(1.0)
checkpoint.save('./ckpt/')
x.assign(8.0)
checkpoint.restore(tf.train.latest_checkpoint('./ckpt/'))
print(x)
# Model remains
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.keras.optimizers.Adam(learning_rate=0.001)
checkpoint_dir = './ck_model_dir'
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
checkpoint_prefix = os.path.join(checkpoint_dir, "ckpt")
root = tf.train.Checkpoint(optimizer=optimizer,
model=model)
root.save(checkpoint_prefix)
root.restore(tf.train.latest_checkpoint(checkpoint_dir))
# Object-oriented indicators
m = tf.keras.metrics.Mean('loss')
m(0)
m(5)
print(m.result()) # => 2.5
m([8, 9])
print(m.result()) # => 5.5
tf.Tensor(2.5, shape=(), dtype=float32)
tf.Tensor(5.5, shape=(), dtype=float32)
7. Advanced automatic differentiation theme
Dynamic Model #
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
# Custom Gradient
@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
# Custom gradient may provide a numerically stable gradient
def log1pexp(x):
return tf.math.log(1 + tf.exp(x))
def grad_log1pexp(x):
with tf.GradientTape() as tape:
tape.watch(x)
value = log1pexp(x)
return tape.gradient(value, x)
print(grad_log1pexp(tf.constant(0.)).numpy())
# However, x = 100 fails because of numerical instability.
print(grad_log1pexp(tf.constant(100.)).numpy())
0.5
nan Wuxi crowd Hospital Which http://www.wxbhnkyy120.com/
Here, log1pexp function can be used to customize a gradient analysis to simplify the derivation. Following a preceding value reuse tf.exp (x) is calculated during the transfer - it more efficient by eliminating redundant calculations:
@tf.custom_gradient
def log1pexp(x):
e = tf.exp(x)
def grad(dy):
two return * (1 - 1 / (1 + e))
return tf.math.log(1 + e), grad
def grad_log1pexp(x):
with tf.GradientTape() as tape:
tape.watch(x)
value = log1pexp(x)
return tape.gradient(value, x)
print(grad_log1pexp(tf.constant(0.)).numpy())
print(grad_log1pexp(tf.constant(100.)).numpy())
0.5
1.0
## 8. improve performance gpu
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 tf.test.is_gpu_available():
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: 1.5692901611328125 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