def get_pytorch_optimizers() -> Dict[str, Type[Optimizer]]:
optimizers = inheritors(Optimizer)
optimizers_dict = {
opt.__name__: opt
for opt in optimizers if _is_pytorch_optimizer(opt)
}
return optimizers_dict
def get_pytorch_losses() -> Dict[str, Type[_Loss]]:
losses = inheritors(_Loss)
losses_dict = {
loss.__name__: loss
for loss in losses if _is_pytorch_loss(loss)
}
return losses_dict
def get_pytorch_losses():
losses = inheritors(_Loss)
losses_dict = {
l.__name__: l
for l in losses if not l.__name__.startswith('_')
} # filter _WeightedLoss
return losses_dict
def test_inheritors():
class ParentClass:
pass
class ChildClass1(ParentClass):
pass
classes = inheritors(ParentClass)
assert len(classes) == 1
assert list(classes)[0] is ChildClass1
class ChildClass2(ParentClass):
pass
class ChildClass3(ParentClass):
pass
classes = inheritors(ParentClass)
assert len(classes) == 3
for cls in [ChildClass1, ChildClass2, ChildClass3]:
assert cls in classes
def attach(self, engine, handler_kwargs_dict=None):
if handler_kwargs_dict is None:
handler_kwargs_dict = dict()
for event_enum in inheritors(EventEnum):
for key, event in event_enum.__members__.items():
if hasattr(self, event.value):
handler = getattr(self, event.value)
if isinstance(handler, Callable):
handler_kwargs = handler_kwargs_dict.get(event, dict())
engine.add_event_handler(event, handler,
**handler_kwargs)
else:
raise TypeError(
f"Attribute {event.value} is not callable.")
def attach(self, engine: Engine):
"""Attach callback to the :class:`argus.engine.Engine`.
Args:
engine (Engine): The engine to which the callback will be attached.
"""
for event_enum in inheritors(EventEnum):
for key, event in event_enum.__members__.items():
if hasattr(self, event.value):
handler = getattr(self, event.value)
if isinstance(handler, (FunctionType, MethodType)):
engine.add_event_handler(event, handler)
else:
raise TypeError(
f"Attribute {event.value} is not callable.")
def get_pytorch_optimizers():
optimizers = inheritors(Optimizer)
optimizers_dict = {opt.__name__: opt for opt in optimizers}
return optimizers_dict