def __init__(self,
batch_shape=torch.Size(),
event_shape=torch.Size(),
validate_args=None):
self._batch_shape = batch_shape
self._event_shape = event_shape
if validate_args is not None:
self._validate_args = validate_args
if self._validate_args:
try:
arg_constraints = self.arg_constraints
except NotImplementedError:
arg_constraints = {}
warnings.warn(
f'{self.__class__} does not define `arg_constraints`. ' +
'Please set `arg_constraints = {}` or initialize the distribution '
+ 'with `validate_args=False` to turn off validation.')
for param, constraint in arg_constraints.items():
if constraints.is_dependent(constraint):
continue # skip constraints that cannot be checked
if param not in self.__dict__ and isinstance(
getattr(type(self), param), lazy_property):
continue # skip checking lazily-constructed args
value = getattr(self, param)
valid = constraint.check(value)
if not valid.all():
raise ValueError(
f"Expected parameter {param} "
f"({type(value).__name__} of shape {tuple(value.shape)}) "
f"of distribution {repr(self)} "
f"to satisfy the constraint {repr(constraint)}, "
f"but found invalid values:\n{value}")
super(Distribution, self).__init__()
def __init__(self,
batch_shape=torch.Size(),
event_shape=torch.Size(),
validate_args=None):
self._batch_shape = batch_shape
self._event_shape = event_shape
if validate_args is not None:
self._validate_args = validate_args
if self._validate_args:
try:
arg_constraints = self.arg_constraints
except NotImplementedError:
arg_constraints = {}
warnings.warn(
f'{self.__class__} does not define `arg_constraints`. ' +
'Please set `arg_constraints = {}` or initialize the distribution '
+ 'with `validate_args=False` to turn off validation.')
for param, constraint in arg_constraints.items():
if constraints.is_dependent(constraint):
continue # skip constraints that cannot be checked
if param not in self.__dict__ and isinstance(
getattr(type(self), param), lazy_property):
continue # skip checking lazily-constructed args
if not constraint.check(getattr(self, param)).all():
raise ValueError(
"The parameter {} has invalid values".format(param))
super(Distribution, self).__init__()
def __init__(self,
batch_shape=torch.Size(),
event_shape=torch.Size(),
validate_args=None):
self._batch_shape = batch_shape
self._event_shape = event_shape
if validate_args is not None:
self._validate_args = validate_args
if self._validate_args:
if not constraints.is_dependent(self.params):
for param, constraint in self.params.items():
if not constraints.is_dependent(constraint):
if not constraint.check(
self.__getattribute__(param)).all():
raise ValueError(
"The parameter {} has invalid values".format(
param))
def __init__(self, batch_shape=torch.Size(), event_shape=torch.Size(), validate_args=None):
self._batch_shape = batch_shape
self._event_shape = event_shape
if validate_args is not None:
self._validate_args = validate_args
if self._validate_args:
for param, constraint in self.arg_constraints.items():
if constraints.is_dependent(constraint):
continue # skip constraints that cannot be checked
if param not in self.__dict__ and isinstance(getattr(type(self), param), lazy_property):
continue # skip checking lazily-constructed args
if not constraint.check(getattr(self, param)).all():
raise ValueError("The parameter {} has invalid values".format(param))
def test_params_contains(self):
for Dist, params in EXAMPLES:
for i, param in enumerate(params):
dist = Dist(**param)
for name, value in param.items():
if isinstance(value, Number):
value = torch.tensor([value])
try:
constraint = dist.arg_constraints[name]
except KeyError:
continue # ignore optional parameters
if is_dependent(constraint):
continue
message = "{} example {}/{} parameter {} = {}".format(
Dist.__name__, i + 1, len(params), name, value)
self.assertTrue(constraint.check(value).all(), msg=message)