def validate_group_ndims(group_ndims, name=None):
"""
Validate the specified value for `group_ndims` argument.
If the specified `group_ndims` is a dynamic :class:`~tf.Tensor`,
additional assertion will be added to the graph node of `group_ndims`.
Args:
group_ndims: Object to be validated.
name: TensorFlow name scope of the graph nodes. (default
"validate_group_ndims")
Returns:
tf.Tensor or int: The validated `group_ndims`.
Raises:
ValueError: If the specified `group_ndims` cannot pass validation.
"""
@contextlib.contextmanager
def gen_name_scope():
if is_tensor_object(group_ndims):
with tf.name_scope(name, default_name='validate_group_ndims'):
yield
else:
yield
with gen_name_scope():
validator = TensorArgValidator('group_ndims')
group_ndims = validator.require_non_negative(
validator.require_int32(group_ndims)
)
return group_ndims
def validate_n_samples(value, name):
"""
Validate the `n_samples` argument.
Args:
value: An int32 value, a int32 :class:`tf.Tensor`, or :obj:`None`.
name (str): Name of the argument (in error message).
Returns:
int or tf.Tensor: The validated `n_samples` argument value.
Raises:
TypeError or ValueError or None: If the value cannot be validated.
"""
if is_tensor_object(value):
@contextlib.contextmanager
def mkcontext():
with tf.name_scope('validate_n_samples'):
yield
else:
@contextlib.contextmanager
def mkcontext():
yield
if value is not None:
with mkcontext():
validator = TensorArgValidator(name=name)
value = validator.require_positive(validator.require_int32(value))
return value
def iterative_masked_reconstruct(reconstruct, x, mask, iter_count,
back_prop=True, name=None):
"""
用“mask”迭代地重构“x”“iter_count”次。
这个方法将调用:func:`masked_reconstruct``iter_count`次，并将前一次迭代的输出作为下一次迭代的输入`x`。将返回最后一次迭代的输出。
Args:
reconstruct: 重构x的方法
x: 被方法重构的张量
mask: 32位整型，必须对x进行广播，指示每一个x是否要被覆盖掉
iter_count (int or tf.Tensor):迭代次数 必须大于1
back_prop (bool): 是否在所有迭代中支持反向传播?
(default :obj:`True`)
name (str): 此操作在TensorFlow图中的名称。
(default "iterative_masked_reconstruct")
Returns:
tf.Tensor: 迭代重构的x。
"""
with tf.name_scope(name, default_name='iterative_masked_reconstruct'):
# 校验迭代次数
v = TensorArgValidator('iter_count')
iter_count = v.require_positive(v.require_int32(iter_count))
# 覆盖重建
x_r, _ = tf.while_loop(
# 条件
lambda x_i, i: i < iter_count,
# 赋值 标记覆盖处重构
lambda x_i, i: (masked_reconstruct(reconstruct, x_i, mask), i + 1),
[x, tf.constant(0, dtype=tf.int32)],
back_prop=back_prop
)
return x_r
def __init__(self,
distribution,
tensor,
n_samples=None,
group_ndims=0,
is_reparameterized=None,
flow_origin=None,
log_prob=None):
"""
Construct the :class:`StochasticTensor`.
Args:
distribution (tfsnippet.distributions.Distribution): The
distribution of this :class:`StochasticTensor`.
tensor (tf.Tensor or TensorWrapper): The samples or observations
of this :class:`StochasticTensor`.
n_samples (tf.Tensor or int): The number of samples taken in
:class:`Distribution.sample`. If not :obj:`None`, the first
dimension of `tensor` should be the sampling dimension.
group_ndims (int or tf.Tensor): The number of dimensions to be
considered as events group in samples. (default 0)
is_reparameterized (bool): Whether or not the samples are
re-parameterized? If not specified, will inherit from
:attr:`tfsnippet.distributions.Distribution.is_reparameterized`.
log_prob (Tensor or None): Pre-computed log-density of `tensor`,
given `group_ndims`.
flow_origin (StochasticTensor): The original stochastic tensor
from the base distribution of a
:class:`tfsnippet.FlowDistribution`.
"""
from tfsnippet.utils import TensorArgValidator, validate_group_ndims_arg
if is_reparameterized is None:
is_reparameterized = distribution.is_reparameterized
if log_prob is not None and not is_tensor_object(log_prob):
log_prob = tf.convert_to_tensor(log_prob)
n_samples = validate_n_samples_arg(n_samples, 'n_samples')
if n_samples is not None:
with tf.name_scope('validate_n_samples'):
validator = TensorArgValidator('n_samples')
n_samples = validator.require_non_negative(
validator.require_int32(n_samples))
group_ndims = validate_group_ndims_arg(group_ndims)
super(StochasticTensor, self).__init__()
self._self_distribution = distribution
self._self_tensor = tf.convert_to_tensor(tensor)
self._self_n_samples = n_samples
self._self_group_ndims = group_ndims
self._self_is_reparameterized = is_reparameterized
self._self_flow_origin = flow_origin
self._self_log_prob = log_prob
self._self_prob = None
def __init__(self,
distribution,
tensor,
n_samples=None,
group_ndims=0,
is_reparameterized=None):
"""
Construct the :class:`StochasticTensor`.
Args:
distribution (tfsnippet.distributions.Distribution): The
distribution of this :class:`StochasticTensor`.
tensor (tf.Tensor or TensorWrapper): The samples or observations
of this :class:`StochasticTensor`.
n_samples (tf.Tensor or int): The number of samples taken in
:class:`Distribution.sample`. If not :obj:`None`, the first
dimension of `tensor` should be the sampling dimension.
group_ndims (int or tf.Tensor): The number of dimensions to be
considered as events group in samples. (default 0)
is_reparameterized (bool): Whether or not the samples are
re-parameterized? If not specified, will inherit from
:attr:`tfsnippet.distributions.Distribution.is_reparameterized`.
"""
from tfsnippet.distributions import validate_group_ndims
if is_reparameterized is None:
is_reparameterized = distribution.is_reparameterized
n_samples = validate_n_samples(n_samples, 'n_samples')
if n_samples is not None:
with tf.name_scope('validate_n_samples'):
validator = TensorArgValidator('n_samples')
n_samples = validator.require_non_negative(
validator.require_int32(n_samples))
group_ndims = validate_group_ndims(group_ndims)
super(StochasticTensor, self).__init__()
self._self_distribution = distribution
self._self_tensor = tf.convert_to_tensor(tensor)
self._self_n_samples = n_samples
self._self_group_ndims = group_ndims
self._self_is_reparameterized = is_reparameterized
self._self_log_prob = None
self._self_prob = None
def iterative_masked_reconstruct(reconstruct,
x,
mask,
iter_count,
back_prop=True,
name=None):
"""
Iteratively reconstruct `x` with `mask` for `iter_count` times.
This method will call :func:`masked_reconstruct` for `iter_count` times,
with the output from previous iteration as the input `x` for the next
iteration. The output of the final iteration would be returned.
Args:
reconstruct: Function for reconstructing `x`.
x: The tensor to be reconstructed by `func`.
mask: int32 mask, must be broadcastable against `x`.
Indicating whether or not to mask each element of `x`.
iter_count (int or tf.Tensor):
Number of iterations(must be greater than 1).
back_prop (bool): Whether or not to support back-propagation through
all the iterations? (default :obj:`True`)
name (str): Name of this operation in TensorFlow graph.
(default "iterative_masked_reconstruct")
Returns:
tf.Tensor: The iteratively reconstructed `x`.
"""
with tf.name_scope(name, default_name='iterative_masked_reconstruct'):
# validate the iteration count
v = TensorArgValidator('iter_count')
iter_count = v.require_positive(v.require_int32(iter_count))
# do the masked reconstructions
x_r, _ = tf.while_loop(
lambda x_i, i: i < iter_count,
lambda x_i, i: (masked_reconstruct(reconstruct, x_i, mask), i + 1),
[x, tf.constant(0, dtype=tf.int32)],
back_prop=back_prop)
return x_r
def test_require_int32(self):
v = TensorArgValidator('xyz')
# test static values
for o in [0, 1, -1]:
self.assertEqual(v.require_int32(o), o)
for o in [object(), None, (), [], 1.2, LONG_MAX]:
with pytest.raises(TypeError,
match='xyz cannot be converted to int32'):
_ = v.require_int32(o)
# test dynamic values
with self.test_session():
for o in [0, 1, -1]:
self.assertEqual(
v.require_int32(tf.constant(o, dtype=tf.int32)).eval(), o)
for o in [
tf.constant(1.2, dtype=tf.float32),
tf.constant(LONG_MAX, dtype=tf.int64)
]:
with pytest.raises(TypeError,
match='xyz cannot be converted to int32'):
_ = v.require_int32(o)
def test_require_non_negative(self):
v = TensorArgValidator('xyz')
# test static values
for o in [0, 0., 1e-7, 1, 1.]:
self.assertEqual(v.require_non_negative(o), o)
for o in [-1., -1, -1e-7]:
with pytest.raises(ValueError, match='xyz must be non-negative'):
_ = v.require_non_negative(o)
# test dynamic values
with self.test_session():
for o, dtype in zip(
[0, 0., 1e-7, 1, 1.],
[tf.int32, tf.float32, tf.float32, tf.int32, tf.float32]):
self.assertAllClose(
v.require_non_negative(tf.constant(o, dtype=dtype)).eval(),
o)
for o, dtype in zip([-1., -1, -1e-7],
[tf.float32, tf.int32, tf.float32]):
with pytest.raises(Exception,
match='xyz must be non-negative'):
_ = v.require_non_negative(tf.constant(
o, dtype=dtype)).eval()