def testTrainNetwork(self, distribution, optimizer_fn, is_tpu):
with distribution.scope():
single_loss_step, layer = single_loss_example(
optimizer_fn,
distribution,
use_bias=True,
iterations_per_step=2)
if context.executing_eagerly():
single_loss_step.initialize()
run_step = single_loss_step
else:
with self.cached_session() as sess:
sess.run(single_loss_step.initialize())
run_step = sess.make_callable(single_loss_step())
self.evaluate(variables.global_variables_initializer())
weights, biases = [], []
for _ in range(5):
run_step()
weights.append(self.evaluate(layer.kernel))
biases.append(self.evaluate(layer.bias))
error = abs(
numpy.add(numpy.squeeze(weights), numpy.squeeze(biases)) - 1)
is_not_increasing = all(y <= x for x, y in zip(error, error[1:]))
self.assertTrue(is_not_increasing)
def testNotPassingASessionInGraph(self):
distribution = one_device_strategy.OneDeviceStrategy(
"/device:CPU:0")
step_function, _ = single_loss_example(
lambda: gradient_descent.GradientDescentOptimizer(0.2), distribution)
with context.graph_mode(), ops.Graph().as_default():
with self.assertRaisesRegexp(ValueError, "Should provide"):
_ = monitor_lib.Monitor(step_function, session=None)
def testPassingASessionInEager(self):
distribution = one_device_strategy.OneDeviceStrategy(
"/device:CPU:0")
step_function, _ = single_loss_example(
lambda: gradient_descent.GradientDescentOptimizer(0.2), distribution)
with session.Session() as sess, context.eager_mode():
with self.assertRaisesRegexp(ValueError, "Should not provide"):
_ = monitor_lib.Monitor(step_function, sess)
def testNotPassingASessionInGraph(self):
distribution = one_device_strategy.OneDeviceStrategy(
"/device:CPU:0")
step_function, _ = single_loss_example(
lambda: gradient_descent.GradientDescentOptimizer(0.2), distribution)
with context.graph_mode(), ops.Graph().as_default():
with self.assertRaisesRegexp(ValueError, "Should provide"):
_ = monitor_lib.Monitor(step_function, session=None)
def testPassingASessionInEager(self):
distribution = one_device_strategy.OneDeviceStrategy(
"/device:CPU:0")
step_function, _ = single_loss_example(
lambda: gradient_descent.GradientDescentOptimizer(0.2), distribution)
with session.Session() as sess, context.eager_mode():
with self.assertRaisesRegexp(ValueError, "Should not provide"):
_ = monitor_lib.Monitor(step_function, sess)
def testTrainNetwork(self, distribution, optimizer_fn):
with distribution.scope():
single_loss_step, layer = single_loss_example(optimizer_fn, distribution)
if context.executing_eagerly():
monitor = monitor_lib.Monitor(single_loss_step, None)
else:
with self.cached_session() as sess:
monitor = monitor_lib.Monitor(single_loss_step, sess)
monitor.run_steps(1)
self.assertEqual(1, len(layer.trainable_variables))
mirrored_weight_variable = layer.trainable_variables[0]
start_error = self.evaluate(mirrored_weight_variable)
start_error = abs(numpy.array(start_error) - 1)
monitor.run_steps(9)
end_error = self.evaluate(mirrored_weight_variable)
end_error = abs(numpy.array(end_error) - 1)
self.assertGreaterEqual(start_error, end_error)
def testTrainNetwork(self, distribution, optimizer_fn):
with distribution.scope():
single_loss_step, layer = single_loss_example(optimizer_fn, distribution)
if context.executing_eagerly():
monitor = monitor_lib.Monitor(single_loss_step, None)
else:
with self.cached_session() as sess:
monitor = monitor_lib.Monitor(single_loss_step, sess)
monitor.run_steps(1)
self.assertEqual(1, len(layer.trainable_variables))
mirrored_weight_variable = layer.trainable_variables[0]
start_error = self.evaluate(mirrored_weight_variable)
start_error = abs(numpy.array(start_error) - 1)
monitor.run_steps(9)
end_error = self.evaluate(mirrored_weight_variable)
end_error = abs(numpy.array(end_error) - 1)
self.assertGreaterEqual(start_error, end_error)
def testTrainNetwork(self, distribution, optimizer_fn, is_tpu):
with distribution.scope():
single_loss_step, layer = single_loss_example(
optimizer_fn, distribution, use_bias=True, iterations_per_step=2)
if context.executing_eagerly():
single_loss_step.initialize()
run_step = single_loss_step
else:
with self.cached_session() as sess:
sess.run(single_loss_step.initialize())
run_step = sess.make_callable(single_loss_step())
self.evaluate(variables.global_variables_initializer())
weights, biases = [], []
for _ in range(5):
run_step()
weights.append(self.evaluate(layer.kernel))
biases.append(self.evaluate(layer.bias))
error = abs(numpy.add(numpy.squeeze(weights), numpy.squeeze(biases)) - 1)
is_not_increasing = all(y <= x for x, y in zip(error, error[1:]))
self.assertTrue(is_not_increasing)