def is_tensor(x): # pylint: disable=invalid-name
"""Checks whether `x` is a TF-native type that can be passed to many TF ops.
Use is_tensor to differentiate types that can ingested by TensorFlow ops
without any conversion (e.g., `tf.Tensor`, `tf.SparseTensor`, and
`tf.RaggedTensor`) from types that need to be converted into tensors before
they are ingested (e.g., numpy `ndarray` and Python scalars).
For example, in the following code block:
```python
if not tf.is_tensor(t):
t = tf.convert_to_tensor(t)
return t.dtype
```
we check to make sure that `t` is a tensor (and convert it if not) before
accessing its `shape` and `dtype`.
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor or "tensor-like", `False` if not.
"""
return (isinstance(x, internal.NativeObject) or ops.is_dense_tensor_like(x)
or getattr(x, "is_tensor_like", False))
def is_tensor(x): # pylint: disable=invalid-name
"""Check whether `x` is of tensor type.
Check whether an object is a tensor. Equivalent to
`isinstance(x, [tf.Tensor, tf.SparseTensor, tf.Variable])`.
Args:
x: An python object to check.
Returns:
`True` if `x` is a tensor, `False` if not.
"""
return isinstance(x, ops._TensorLike) or ops.is_dense_tensor_like(x) # pylint: disable=protected-access
def is_tensor(x): # pylint: disable=invalid-name
"""Checks whether `x` is a tensor or "tensor-like".
If `is_tensor(x)` returns `True`, it is safe to assume that `x` is a tensor or
can be converted to a tensor using `ops.convert_to_tensor(x)`.
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor or "tensor-like", `False` if not.
"""
return (isinstance(x, tensor_like._TensorLike) or # pylint: disable=protected-access
ops.is_dense_tensor_like(x) or getattr(x, "is_tensor_like", False))
def is_tensor(x): # pylint: disable=invalid-name
"""Checks whether `x` is a tensor or "tensor-like".
If `is_tensor(x)` returns `True`, it is safe to assume that `x` is a tensor or
can be converted to a tensor using `ops.convert_to_tensor(x)`.
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor or "tensor-like", `False` if not.
"""
return (isinstance(x, ops._TensorLike) or ops.is_dense_tensor_like(x) or # pylint: disable=protected-access
getattr(x, "is_tensor_like", False))
def is_tensor(x): # pylint: disable=invalid-name
"""Check whether `x` is of tensor type.
Check whether an object is a tensor. This check is equivalent to calling
`isinstance(x, (tf.Tensor, tf.SparseTensor, tf.Variable))` and also checks
if all the component variables of a MirroredVariable or a TowerLocalVariable
are tensors.
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor, `False` if not.
"""
return (isinstance(x, ops._TensorLike) or ops.is_dense_tensor_like(x) or # pylint: disable=protected-access
(hasattr(x, "is_tensor_like") and x.is_tensor_like))
def is_tensor(x): # pylint: disable=invalid-name
"""Check whether `x` is of tensor type.
Check whether an object is a tensor. This check is equivalent to calling
`isinstance(x, (tf.Tensor, tf.SparseTensor, tf.Variable))` and also checks
if all the component variables of a MirroredVariable or a TowerLocalVariable
are tensors.
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor, `False` if not.
"""
return (isinstance(x, ops._TensorLike) or ops.is_dense_tensor_like(x) or # pylint: disable=protected-access
(hasattr(x, "is_tensor_like") and x.is_tensor_like))
def is_tensor(x): # pylint: disable=invalid-name
"""Checks whether `x` is a tensor or "tensor-like".
If `is_tensor(x)` returns `True`, it is safe to assume that `x` is a tensor or
can be converted to a tensor using `ops.convert_to_tensor(x)`.
Usage example:
>>> tf.is_tensor(tf.constant([[1,2,3],[4,5,6],[7,8,9]]))
True
>>> tf.is_tensor("Hello World")
False
Args:
x: A python object to check.
Returns:
`True` if `x` is a tensor or "tensor-like", `False` if not.
"""
return (isinstance(x, tensor_like._TensorLike) or # pylint: disable=protected-access
ops.is_dense_tensor_like(x) or getattr(x, "is_tensor_like", False))
def _check_is_tensor(x, tensor_name):
"""Returns `x` if it is a `Tensor`, raises TypeError otherwise."""
if not ops.is_dense_tensor_like(x):
raise TypeError('{} must be Tensor, given: {}'.format(tensor_name, x))
return x
def _check_is_tensor_or_operation(x, name):
if not (isinstance(x, tf.Operation) or ops.is_dense_tensor_like(x)):
raise TypeError('{} must be Operation or Tensor, given: {}'.format(
name, x))
def __init__(self, inputs=None, outputs=None, name=None):
"""
Initializing the Model.
Parameters
----------
inputs : Tensor or list of tensors
Input tensor(s), which must come from tl.layers.Input()
outputs : Tensor or list of tensors
Output tensor(s), which must be the output(s) of some TL layers
name : str or None
Name for this network
"""
# Auto naming if the name is not given
global _global_model_name_dict
global _global_model_name_set
if name is None:
prefix = self.__class__.__name__.lower()
if _global_model_name_dict.get(prefix) is not None:
_global_model_name_dict[prefix] += 1
name = prefix + '_' + str(_global_model_name_dict[prefix])
else:
_global_model_name_dict[prefix] = 0
name = prefix
while name in _global_model_name_set:
_global_model_name_dict[prefix] += 1
name = prefix + '_' + str(_global_model_name_dict[prefix])
_global_model_name_set.add(name)
else:
if name in _global_model_name_set:
raise ValueError(
'Model name \'%s\' has already been used by another model. Please change the model name.'
% name)
_global_model_name_set.add(name)
_global_model_name_dict[name] = 0
# Model properties
self.name = name
# Model state: train or test
self.is_train = None
# Model weights
self._weights = None
# Model inputs and outputs
# TODO: note that in dynamic network, inputs and outputs are both None, may cause problem, test needed
self._inputs = inputs
self._outputs = outputs
# Model converted into a Layer
self._model_layer = None
# Layer Node status
self._nodes_fixed = False
# Model layers
self._all_layers = None
if inputs is None and outputs is None:
pass
else:
# check type of inputs and outputs
check_order = ['inputs', 'outputs']
for co, check_argu in enumerate([inputs, outputs]):
if isinstance(check_argu, tf_ops._TensorLike
) or tf_ops.is_dense_tensor_like(check_argu):
pass
elif isinstance(check_argu, list):
if len(check_argu) == 0:
raise ValueError(
"The argument `%s` is detected as an empty list. "
% check_order[co] +
"It should be either Tensor or a list of Tensor.")
for idx in range(len(check_argu)):
if not isinstance(check_argu[idx], tf_ops._TensorLike
) or not tf_ops.is_dense_tensor_like(
check_argu[idx]):
raise TypeError(
"The argument `%s` should be either Tensor or a list of Tensor "
% (check_order[co]) +
"but the %s[%d] is detected as %s" %
(check_order[co], idx, type(check_argu[idx])))
else:
raise TypeError(
"The argument `%s` should be either Tensor or a list of Tensor but received %s"
% (check_order[co], type(check_argu)))
if not _check_tl_layer_tensors(inputs):
raise TypeError(
"The argument `inputs` should be either Tensor or a list of Tensor "
"that come from TensorLayer's Input layer: tl.layers.Input(shape). "
)
if not _check_tl_layer_tensors(outputs):
raise TypeError(
"The argument `outputs` should be either Tensor or a list of Tensor "
"that is/are outputs from some TensorLayer's layers, e.g. tl.layers.Dense, tl.layers.Conv2d."
)
# build network graph
self._node_by_depth, self._all_layers = self._construct_graph()
self._fix_nodes_for_layers()
def _check_is_tensor(x, tensor_name):
"""Returns `x` if it is a `Tensor`, raises TypeError otherwise."""
if not ops.is_dense_tensor_like(x):
raise TypeError('{} must be Tensor, given: {}'.format(tensor_name, x))
return x
def _check_is_tensor_or_operation(x, name):
if not (isinstance(x, ops.Operation) or ops.is_dense_tensor_like(x)):
raise TypeError('{} must be Operation or Tensor, given: {}'.format(name, x))
def is_tensor(x):
return isinstance(
x, tf_ops._TensorLike) or tf_ops.is_dense_tensor_like(x)
def is_tensor(node):
try:
# TODO(b/154650521): Use tf.Tensor instead of core.Tensor.
return isinstance(node, core.Tensor)
except NameError:
return tf_ops.is_dense_tensor_like(node)
def __init__(self, inputs=None, outputs=None, name=None):
"""
Initializing the Model.
Parameters
----------
inputs : Tensor or list of tensors
Input tensor(s), which must come from tl.layers.Input()
outputs : Tensor or list of tensors
Output tensor(s), which must be the output(s) of some TL layers
name : str or None
Name for this network
"""
# Auto naming if the name is not given
global _global_model_name_dict
global _global_model_name_set
if name is None:
prefix = self.__class__.__name__.lower()
if _global_model_name_dict.get(prefix) is not None:
_global_model_name_dict[prefix] += 1
name = prefix + '_' + str(_global_model_name_dict[prefix])
else:
_global_model_name_dict[prefix] = 0
name = prefix
while name in _global_model_name_set:
_global_model_name_dict[prefix] += 1
name = prefix + '_' + str(_global_model_name_dict[prefix])
_global_model_name_set.add(name)
else:
if name in _global_model_name_set:
raise ValueError(
'Model name \'%s\' has already been used by another model. Please change the model name.' % name
)
_global_model_name_set.add(name)
_global_model_name_dict[name] = 0
# Model properties
self.name = name
# Model state: train or test
self.is_train = None
# Model weights
self._weights = None
# Model inputs and outputs
# TODO: note that in dynamic network, inputs and outputs are both None, may cause problem, test needed
self._inputs = inputs
self._outputs = outputs
# Model converted into a Layer
self._model_layer = None
# Layer Node status
self._nodes_fixed = False
# Model layers
self._all_layers = None
if inputs is None and outputs is None:
pass
else:
# check type of inputs and outputs
check_order = ['inputs', 'outputs']
for co, check_argu in enumerate([inputs, outputs]):
if isinstance(check_argu, tf_ops._TensorLike) or tf_ops.is_dense_tensor_like(check_argu):
pass
elif isinstance(check_argu, list):
if len(check_argu) == 0:
raise ValueError(
"The argument `%s` is detected as an empty list. " % check_order[co] +
"It should be either Tensor or a list of Tensor."
)
for idx in range(len(check_argu)):
if not isinstance(check_argu[idx], tf_ops._TensorLike) or not tf_ops.is_dense_tensor_like(
check_argu[idx]):
raise TypeError(
"The argument `%s` should be either Tensor or a list of Tensor " % (check_order[co]) +
"but the %s[%d] is detected as %s" % (check_order[co], idx, type(check_argu[idx]))
)
else:
raise TypeError(
"The argument `%s` should be either Tensor or a list of Tensor but received %s" %
(check_order[co], type(check_argu))
)
if not _check_tl_layer_tensors(inputs):
raise TypeError(
"The argument `inputs` should be either Tensor or a list of Tensor "
"that come from TensorLayer's Input layer: tl.layers.Input(shape). "
)
if not _check_tl_layer_tensors(outputs):
raise TypeError(
"The argument `outputs` should be either Tensor or a list of Tensor "
"that is/are outputs from some TensorLayer's layers, e.g. tl.layers.Dense, tl.layers.Conv2d."
)
# build network graph
self._node_by_depth, self._all_layers = self._construct_graph()
self._fix_nodes_for_layers()
def is_tensor(x):
return isinstance(x, tf_ops._TensorLike) or \
tf_ops.is_dense_tensor_like(x) or \
isinstance(x, tf.SparseTensor)
def is_tensor(node):
# return (isinstance(node, (tf.Tensor, tf.Variable))
# or resource_variable_ops.is_resource_variable(node))
return tf_ops.is_dense_tensor_like(node)
