def test_get_default_graph(self):
graph1 = graph.get_default_graph()
graph2 = graph.get_default_graph()
self.assertEqual(graph1.__class__, Graph)
self.assertEqual(graph2.__class__, Graph)
self.assertEqual(graph1, graph2)
def __init__(self, variableOp, value, name="UpdateVariableOp"):
self.variableOp = variableOp
self.value = value
self.name = name
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
def __init__(self, value, is_trainable=True, name="VariableOp"):
super(VariableOp, self).__init__(name)
self._value = value
self._is_trainable = is_trainable
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
if self._is_trainable:
self._graph.add_to_trainable_variable(self.get_name(), self)
def __init__(self, dtype=None, shape=None, name="PlaceholderOp"):
super(PlaceholderOp, self).__init__(name)
# TODO
self._dtype = dtype
self._shape = shape
self._value = None
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, input, name="Cubic"):
if not isinstance(input, Op):
self.op = ConstantOp(input)
else:
self.op = input
self.name = name
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
def __init__(self, input, power, name="PowerOp"):
super(PowerOp, self).__init__(name)
if not isinstance(input, Op):
self._op = ConstantOp(input)
else:
self._op = input
self._power = power
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, *inputs):
self.name = "MultipleN"
self.ops = []
for input in inputs:
if not isinstance(input, Op):
input = ConstantOp(input)
self.ops.append(input)
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
def __init__(self, dtype=None, shape=None, name="Placeholder"):
super(PlaceholderOp, self).__init__(name)
# TODO: Use dtype and shape
self._dtype = dtype
self._shape = shape
# The value is None util Session.run() with feed_dict parameter
self._value = None
# TODO: Support other graph instance
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, input_nodes=[], name=None):
self.input_nodes = input_nodes
for input in input_nodes:
input.output_nodes.append(self)
self.output_nodes = []
self.output_value = None
self.name = name
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
self._isscalar = False
def __init__(self, input1, input2, name="MultipleOp"):
super(MultipleOp, self).__init__(self)
if not isinstance(input1, Op):
self._op1 = ConstantOp(input1)
else:
self._op1 = input1
if not isinstance(input2, Op):
self._op2 = ConstantOp(input2)
else:
self._op2 = input2
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, *inputs):
# TODO: Deprecated op
# TODO: Support user defined name in the parameter
self.name = "AddN"
self.ops = []
for input in inputs:
if not isinstance(input, Op):
input = ConstantOp(input)
self.ops.append(input)
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
def get_variable(name="Variable",
value=None,
shape=None,
dtype=None,
initializer=None,
regularizer=None,
reuse=None,
trainable=None):
_graph = graph.get_default_graph()
if name in _graph.get_name_op_map():
return _graph.get_name_op_map()[name]
else:
variable = VariableOp(value=value, name=name)
return variable
def __init__(self, name="LocalVariablesInitializer"):
super(LocalVariablesInitializerOp, self).__init__(name)
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, variableop_value_map, name="UpdateVariableNOp"):
self.variableop_value_map = variableop_value_map
self.name = name
self.graph = graph.get_default_graph()
self.graph.add_to_graph(self)
def __init__(self, value, name="Constant"):
super(ConstantOp, self).__init__(name)
self._value = value
self._graph = graph.get_default_graph()
self._graph.add_to_graph(self)
def __init__(self, learning_rate=0.01, name="GradientDescent"):
super(GradientDescentOptimizer, self).__init__(name)
self.learning_rate = learning_rate
self.graph = graph.get_default_graph()
