def testGradOfGrad(self):
def square(x):
return math_ops.multiply(x, x)
grad = tfe.gradients_function(square)
gradgrad = tfe.gradients_function(lambda x: grad(x)[0])
self.assertEquals([2], [x.numpy() for x in gradgrad(3)])
def testGradOfGrad(self):
def square(x):
return math_ops.multiply(x, x)
grad = tfe.gradients_function(square)
gradgrad = tfe.gradients_function(lambda x: grad(x)[0])
self.assertEquals([2], [x.numpy() for x in gradgrad(3.)])
def testGradients(self):
def square(x):
return math_ops.multiply(x, x)
grad = tfe.gradients_function(square)
self.assertEquals([6], [x.numpy() for x in grad(3.)])
def testGradients(self):
def square(x):
return math_ops.multiply(x, x)
grad = tfe.gradients_function(square)
self.assertEquals([6], [x.numpy() for x in grad(3)])
def reference_func():
gradval = tensors_to_numpy(tfe.gradients_function(func, params=wrt)(*args))
if preserve_result:
val = tensors_to_numpy(func(*args))
if isinstance(gradval, (tuple)):
return gradval + (val,)
return gradval, val
else:
return gradval
def testCustomGrad(self):
@tfe.custom_gradient
def f(x):
y = math_ops.multiply(x, x)
def grad_fn(_):
return [x + y]
return y, grad_fn
grad = tfe.gradients_function(f)
self.assertEquals([12], [x.numpy() for x in grad(3)])
def testCustomGrad(self):
@tfe.custom_gradient
def f(x):
y = math_ops.multiply(x, x)
def grad_fn(_):
return [x + y]
return y, grad_fn
grad = tfe.gradients_function(f)
self.assertEquals([12], [x.numpy() for x in grad(3.)])
def testCustomGrad(self):
@tfe.custom_gradient
def f(x):
y = math_ops.multiply(x, x)
def grad_fn(_):
return [x + y]
return y, grad_fn
# TODO(ashankar): This [0] should ideally not be needed.
grad = tfe.gradients_function(f, [0])
self.assertEquals([12], [x.numpy() for x in grad(3)])
def testCustomGrad(self):
@tfe.custom_gradient
def f(x):
y = math_ops.multiply(x, x)
def grad_fn(_):
return [x + y]
return y, grad_fn
# TODO(ashankar): This [0] should ideally not be needed.
grad = tfe.gradients_function(f, [0])
self.assertEquals([12], [x.numpy() for x in grad(3)])
# Module 10: New Features in Tensorflow
# Eagle Execution
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf
from tensorflow.contrib.eager.python import tfe
tfe.enable_eager_execution()
x = [[2.]]
m = tf.matmul(x, x)
print(m)
def square(x):
return tf.multiply(x, x)
grad = tfe.gradients_function(square)
print(square(3.))
print(grad(3.))
gradgrad = tfe.gradients_function(lambda x: grad(x)[0])
print(gradgrad(3.))
MIT License
Copyright (c) 2018. Victor I. Afolabi. All rights reserved.
"""
import tensorflow as tf
# import tensorflow.contrib.eager as tfe
from tensorflow.contrib.eager.python import tfe
from sklearn import datasets, preprocessing, model_selection
# Enable eager mode.
# tf.enable_eager_execution()
data = datasets.load_iris()
TARGET_NAMES = {i: l for i, l in enumerate(data['target_names'])}
features = preprocessing.MinMaxScaler(feature_range=(-1, 1)).fit_transform(
data['data'])
labels = preprocessing.OneHotEncoder(sparse=False).fit_transform(
data['target'].reshape(-1, 1))
X_train, X_test, y_train, y_test = model_selection.train_test_split(
features, labels, test_size=0.1)
def square_func(W):
return tf.square(W)
f_grad = tfe.gradients_function(square_func, params=['W'])
print(f_grad(tf.constant(0.3)))
def reference_func():
dxx = tfe.gradients_function(tfe.gradients_function(func))(*args)
return tensors_to_numpy(tuple(t.numpy() for t in dxx))
def reference_func():
return tensors_to_numpy(
tfe.gradients_function(func, params=wrt)(*args))
