experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Remove run_v1_only annotation from proximal_adagrad_test, saved_model_experimental_test, legacy base layer test

Browse Source 
PiperOrigin-RevId: 320522977
Change-Id: I671006714ffb961be0ddbd33fe8f58f28571cfb0
This commit is contained in:
Pavithra Vijay 2020-07-09 18:41:28 -07:00 committed by TensorFlower Gardener
parent 2015320832
commit d13f4ec326
3 changed files with 77 additions and 84 deletions
tensorflow/python
keras
legacy_tf_layers
base_test.py
saving
saved_model_experimental_test.py
training
proximal_adagrad_test.py

122
tensorflow/python/keras/legacy_tf_layers/base_test.py
View File

@ -28,7 +28,6 @@ from tensorflow.python.eager import def_function
from tensorflow.python.framework import constant_op
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import test_util
from tensorflow.python.keras import combinations
from tensorflow.python.keras.engine import base_layer as keras_base_layer
from tensorflow.python.keras.engine import input_spec
@ -162,21 +161,22 @@ class BaseLayerTest(test.TestCase, parameterized.TestCase):
synchronization=variable_scope.VariableSynchronization.ON_READ,
trainable=True)
@test_util.run_v1_only('Legacy TF Base layer is supported only in V1.')
def testReusePartitionedVariablesAndRegularizers(self):
regularizer = lambda x: math_ops.reduce_sum(x) * 1e-3
partitioner = partitioned_variables.fixed_size_partitioner(3)
for reuse in [False, True]:
with variable_scope.variable_scope(variable_scope.get_variable_scope(),
partitioner=partitioner,
reuse=reuse):
layer = base_layers.Layer(name='my_layer')
_ = layer.add_variable(
'reg_part_var', [4, 4],
initializer=init_ops.zeros_initializer(),
regularizer=regularizer)
self.assertEqual(
len(ops.get_collection(ops.GraphKeys.REGULARIZATION_LOSSES)), 3)
with ops.Graph().as_default():
regularizer = lambda x: math_ops.reduce_sum(x) * 1e-3
partitioner = partitioned_variables.fixed_size_partitioner(3)
for reuse in [False, True]:
with variable_scope.variable_scope(
variable_scope.get_variable_scope(),
partitioner=partitioner,
reuse=reuse):
layer = base_layers.Layer(name='my_layer')
_ = layer.add_variable(
'reg_part_var', [4, 4],
initializer=init_ops.zeros_initializer(),
regularizer=regularizer)
self.assertEqual(
len(ops.get_collection(ops.GraphKeys.REGULARIZATION_LOSSES)), 3)
@combinations.generate(combinations.combine(mode=['graph', 'eager']))
def testCall(self):
@ -464,13 +464,13 @@ class BaseLayerTest(test.TestCase, parameterized.TestCase):
self.assertTrue(isinstance(result, dict))
self.assertEqual(set(['label', 'logits']), set(result.keys()))
@test_util.run_v1_only('Legacy TF Base layer is supported only in V1.')
def testActivityRegularizer(self):
regularizer = math_ops.reduce_sum
layer = base_layers.Layer(activity_regularizer=regularizer)
x = array_ops.placeholder('int32')
layer.apply(x)
self.assertEqual(len(layer.get_losses_for(x)), 1)
with ops.Graph().as_default():
regularizer = math_ops.reduce_sum
layer = base_layers.Layer(activity_regularizer=regularizer)
x = array_ops.placeholder('int32')
layer.apply(x)
self.assertEqual(len(layer.get_losses_for(x)), 1)
def testNameScopeIsConsistentWithVariableScope(self):
# Github issue 13429.
@ -553,7 +553,6 @@ class BaseLayerTest(test.TestCase, parameterized.TestCase):
self.assertEqual(len(layer.trainable_variables), 1)
self.assertEqual(layer.variables[0].graph, outer_graph)
@test_util.run_v1_only('Legacy TF Base layer is supported only in V1.')
def testGetUpdateFor(self):
class MyLayer(base_layers.Layer):
@ -575,30 +574,30 @@ class BaseLayerTest(test.TestCase, parameterized.TestCase):
inputs=True)
return inputs + 1
layer = MyLayer()
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
with ops.Graph().as_default():
layer = MyLayer()
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
self.assertEqual(len(layer.updates), 2)
self.assertEqual(len(layer.get_updates_for(None)), 1)
self.assertEqual(len(layer.get_updates_for([inputs])), 1)
self.assertEqual(len(layer.get_updates_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_updates_for([outputs])), 0)
self.assertEqual(len(layer.updates), 2)
self.assertEqual(len(layer.get_updates_for(None)), 1)
self.assertEqual(len(layer.get_updates_for([inputs])), 1)
self.assertEqual(len(layer.get_updates_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_updates_for([outputs])), 0)
# Call same layer on new input, creating one more conditional update
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
# Call same layer on new input, creating one more conditional update
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
self.assertEqual(len(layer.updates), 3)
self.assertEqual(len(layer.get_updates_for(None)), 1)
# Check that we are successfully filtering out irrelevant updates
self.assertEqual(len(layer.get_updates_for([inputs])), 1)
self.assertEqual(len(layer.get_updates_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_updates_for([outputs])), 0)
self.assertEqual(len(layer.updates), 3)
self.assertEqual(len(layer.get_updates_for(None)), 1)
# Check that we are successfully filtering out irrelevant updates
self.assertEqual(len(layer.get_updates_for([inputs])), 1)
self.assertEqual(len(layer.get_updates_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_updates_for([outputs])), 0)
@test_util.run_v1_only('Legacy TF Base layer is supported only in V1.')
def testGetLossesFor(self):
class MyLayer(base_layers.Layer):
@ -619,28 +618,29 @@ class BaseLayerTest(test.TestCase, parameterized.TestCase):
self.add_loss(inputs, inputs=True)
return inputs + 1
layer = MyLayer()
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
with ops.Graph().as_default():
layer = MyLayer()
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
self.assertEqual(len(layer.losses), 2)
self.assertEqual(len(layer.get_losses_for(None)), 1)
self.assertEqual(len(layer.get_losses_for([inputs])), 1)
self.assertEqual(len(layer.get_losses_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_losses_for([outputs])), 0)
self.assertEqual(len(layer.losses), 2)
self.assertEqual(len(layer.get_losses_for(None)), 1)
self.assertEqual(len(layer.get_losses_for([inputs])), 1)
self.assertEqual(len(layer.get_losses_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_losses_for([outputs])), 0)
# Call same layer on new input, creating one more conditional loss
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
# Call same layer on new input, creating one more conditional loss
inputs = array_ops.placeholder(dtypes.float32, (), 'inputs')
intermediate_inputs = inputs + 1
outputs = layer.apply(intermediate_inputs)
self.assertEqual(len(layer.losses), 3)
self.assertEqual(len(layer.get_losses_for(None)), 1)
# Check that we are successfully filtering out irrelevant losses
self.assertEqual(len(layer.get_losses_for([inputs])), 1)
self.assertEqual(len(layer.get_losses_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_losses_for([outputs])), 0)
self.assertEqual(len(layer.losses), 3)
self.assertEqual(len(layer.get_losses_for(None)), 1)
# Check that we are successfully filtering out irrelevant losses
self.assertEqual(len(layer.get_losses_for([inputs])), 1)
self.assertEqual(len(layer.get_losses_for([intermediate_inputs])), 1)
self.assertEqual(len(layer.get_losses_for([outputs])), 0)
class IdentityLayer(base_layers.Layer):

5
tensorflow/python/keras/saving/saved_model_experimental_test.py
View File

@ -31,7 +31,6 @@ from tensorflow.python.eager import context
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import tensor_spec
from tensorflow.python.framework import test_util
from tensorflow.python.keras.engine import training as model_lib
from tensorflow.python.keras.optimizer_v2 import adadelta
from tensorflow.python.keras.optimizer_v2 import rmsprop
@ -45,8 +44,6 @@ from tensorflow.python.saved_model import model_utils
from tensorflow.python.training import training as training_module
@test_util.run_v1_only(
'keras.experimental.load_from_saved_model is supported only in V1.')
class TestModelSavingandLoading(parameterized.TestCase, test.TestCase):
def _save_model_dir(self, dirname='saved_model'):
@ -280,8 +277,6 @@ def load_model(sess, path, mode):
return inputs, outputs, meta_graph_def
@test_util.run_v1_only(
'keras.experimental.export_saved_model is supported only in V1.')
class TestModelSavedModelExport(test.TestCase, parameterized.TestCase):
def _save_model_dir(self, dirname='saved_model'):

34
tensorflow/python/training/proximal_adagrad_test.py
View File

@ -23,7 +23,6 @@ import numpy as np
from tensorflow.python.framework import constant_op
from tensorflow.python.framework import dtypes
from tensorflow.python.framework import ops
from tensorflow.python.framework import test_util
from tensorflow.python.ops import embedding_ops
from tensorflow.python.ops import math_ops
from tensorflow.python.ops import resource_variable_ops
@ -36,7 +35,8 @@ from tensorflow.python.training import proximal_adagrad
class ProximalAdagradOptimizerTest(test.TestCase):
def doTestProximalAdagradwithoutRegularization(self, use_resource=False):
with self.cached_session() as sess:
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
var0 = variables.Variable([0.0, 0.0])
var1 = variables.Variable([0.0, 0.0])
grads0 = constant_op.constant([0.1, 0.2])
@ -65,17 +65,15 @@ class ProximalAdagradOptimizerTest(test.TestCase):
self.assertStartsWith(opt_vars[1].name, var1._shared_name)
self.assertEqual(2, len(opt_vars))
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testProximalAdagradwithoutRegularization(self):
self.doTestProximalAdagradwithoutRegularization(use_resource=False)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testResourceProximalAdagradwithoutRegularization(self):
self.doTestProximalAdagradwithoutRegularization(use_resource=True)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testProximalAdagradwithoutRegularization2(self):
with self.cached_session() as sess:
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
var0 = variables.Variable([1.0, 2.0])
var1 = variables.Variable([4.0, 3.0])
grads0 = constant_op.constant([0.1, 0.2])
@ -100,10 +98,10 @@ class ProximalAdagradOptimizerTest(test.TestCase):
self.assertAllClose(np.array([-1.60261, -2.296985]), v0_val)
self.assertAllClose(np.array([3.715679, 2.433051]), v1_val)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testMinimizeSparseResourceVariable(self):
for dtype in [dtypes.float32, dtypes.float64]:
with self.cached_session():
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
var0 = resource_variable_ops.ResourceVariable([[1.0, 2.0]], dtype=dtype)
x = constant_op.constant([[4.0], [5.0]], dtype=dtype)
pred = math_ops.matmul(embedding_ops.embedding_lookup([var0], [0]), x)
@ -119,9 +117,9 @@ class ProximalAdagradOptimizerTest(test.TestCase):
self.evaluate(var0),
atol=0.01)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testProximalAdagradWithL1(self):
with self.cached_session() as sess:
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
var0 = variables.Variable([1.0, 2.0])
var1 = variables.Variable([4.0, 3.0])
grads0 = constant_op.constant([0.1, 0.2])
@ -146,9 +144,9 @@ class ProximalAdagradOptimizerTest(test.TestCase):
self.assertAllClose(np.array([-6.663634, -9.190331]), v0_val)
self.assertAllClose(np.array([2.959304, 1.029232]), v1_val)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testProximalAdagradWithL1_L2(self):
with self.cached_session() as sess:
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
var0 = variables.Variable([1.0, 2.0])
var1 = variables.Variable([4.0, 3.0])
grads0 = constant_op.constant([0.1, 0.2])
@ -213,9 +211,9 @@ class ProximalAdagradOptimizerTest(test.TestCase):
v0_val, v1_val = self.evaluate([var0, var1])
return v0_val, v1_val
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testEquivAdagradwithoutRegularization(self):
with self.cached_session():
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
val0, val1 = self.applyOptimizer(
proximal_adagrad.ProximalAdagradOptimizer(
3.0,
@ -223,7 +221,7 @@ class ProximalAdagradOptimizerTest(test.TestCase):
l1_regularization_strength=0.0,
l2_regularization_strength=0.0))
with self.cached_session():
with ops.Graph().as_default(), self.cached_session():
val2, val3 = self.applyOptimizer(
adagrad.AdagradOptimizer(
3.0, initial_accumulator_value=0.1))
@ -231,9 +229,9 @@ class ProximalAdagradOptimizerTest(test.TestCase):
self.assertAllClose(val0, val2)
self.assertAllClose(val1, val3)
@test_util.run_v1_only("ProximalAdagradOptimizer is supported only in V1.")
def testEquivSparseAdagradwithoutRegularization(self):
with self.cached_session():
# ProximalAdagradOptimizer is supported only in V1.
with ops.Graph().as_default(), self.cached_session():
val0, val1 = self.applyOptimizer(
proximal_adagrad.ProximalAdagradOptimizer(
3.0,
@ -242,7 +240,7 @@ class ProximalAdagradOptimizerTest(test.TestCase):
l2_regularization_strength=0.0),
is_sparse=True)
with self.cached_session():
with ops.Graph().as_default(), self.cached_session():
val2, val3 = self.applyOptimizer(
adagrad.AdagradOptimizer(
3.0, initial_accumulator_value=0.1),