def __init__(
self,
name: str,
window_size: int = 50,
show_local_avg: bool = True,
fmt: str = "{:f}",
show_avg: Optional[bool] = None,
use_running_avg: Optional[bool] = None,
):
if show_avg is not None:
msg = build_deprecation_message(
"The parameter show_avg",
"0.4.0",
info="The average is now always shown.",
)
if show_avg:
warnings.warn(msg)
else:
raise RuntimeError(msg)
if use_running_avg is not None:
msg = build_deprecation_message(
"The parameter use_running_avg",
"0.4.0",
info="It was renamed to show_local_avg.",
)
warnings.warn(msg)
show_local_avg = use_running_avg
super().__init__(name=name, window_size=window_size)
self.show_local_avg = show_local_avg
self.fmt = fmt
def batch_gram_matrix(x: torch.Tensor,
normalize: bool = False) -> torch.Tensor:
msg = build_deprecation_message("The function batch_gram_matrix",
"0.4.0",
info="It was renamed to gram_matrix")
warnings.warn(msg, UserWarning)
return gram_matrix(x, normalize=normalize)
def _pil_resize(
image: Image.Image,
size: Union[int, Tuple[int, int]],
interpolation_mode: str,
**kwargs: Any,
) -> Image.Image:
if is_image_size(size):
size = cast(Tuple[int, int], size)
height, width = size
elif is_edge_size(size):
size = cast(int, size)
height, width = edge_to_image_size(
size, calculate_aspect_ratio((image.height, image.width)))
else:
raise RuntimeError
if kwargs:
msg = build_deprecation_message(
"Passing additional parameters via **resize_kwargs",
"0.4.0",
info="The keyword arguments are ignored.",
)
warnings.warn(msg)
return image.resize((width, height),
resample=_PIL_RESAMPLE_MAP[interpolation_mode])
def __init__(self, *args: Any, **kwargs: Any) -> None:
msg = build_deprecation_message(
"The class GuidedPerceptualLoss",
"1.0",
info="pystiche.loss.PerceptualLoss now also handles guided inputs and targets",
)
warnings.warn(msg)
super().__init__(*args, **kwargs)
def clear_cache(self) -> None:
msg = build_deprecation_message(
"The method clear_cache()",
"0.4.0",
info="It was renamed to empty_storage().",
)
warnings.warn(msg)
self.empty_storage()
def __init__(self, *args: Any, **kwargs: Any) -> None:
msg = build_deprecation_message(
"The class MSEEncodingOperator",
"0.4.0",
info="It was renamed to FeatureReconstructionOperator.",
)
warnings.warn(msg)
super().__init__(*args, **kwargs)
def __init__(self, *args: Any, **kwargs: Any) -> None:
msg = build_deprecation_message(
"The class AlexNetEncoder",
"0.4.0",
info="It was replaced by MultiLayerAlexNetEncoder.",
)
warnings.warn(msg)
super().__init__(*args, **kwargs)
def _vgg_encoder(arch: str, **kwargs: Any) -> VGGMultiLayerEncoder:
msg = build_deprecation_message(
f"The function {arch}_encoder",
"0.4.0",
info=f"It was replaced by {arch}_multi_layer_encoder.",
)
warnings.warn(msg)
return VGGMultiLayerEncoder(arch, **kwargs)
def get_preprocessor(framework: str) -> nn.Module:
msg = build_deprecation_message(
"The function 'get_preprocessor'",
"1.0",
info="It was renamed to 'preprocessing'.",
)
warnings.warn(msg)
return preprocessing(framework)
def alexnet_encoder(*args: Any, **kwargs: Any) -> MultiLayerAlexNetEncoder:
msg = build_deprecation_message(
"The function alexnet_encoder",
"0.4.0",
info="It was replaced by alexnet_multi_layer_encoder.",
)
warnings.warn(msg)
return alexnet_multi_layer_encoder(*args, **kwargs)
def default_epoch_header_fn(*args: Any, **kwargs: Any) -> str:
msg = build_deprecation_message(
"The function default_epoch_header_fn",
"0.4.0",
info="It was renamed to default_epoch_header",
)
warnings.warn(msg)
return default_epoch_header(*args, **kwargs)
def __init__(self, weights: Optional[str] = None, **kwargs: Any) -> None:
if weights is not None:
msg = build_deprecation_message("The parameter weights",
"0.6.0",
info="It was renamed to framework")
warnings.warn(msg, UserWarning)
kwargs["framework"] = weights
super().__init__(**kwargs)
def __getitem__(self, layer: str) -> "SingleLayerEncoder":
msg = build_deprecation_message(
"Extracting a Encoder with bracket indexing",
"0.4.0",
info=("To extract a single layer encoder use "
"MultiLayerEncoder.extract_encoder() instead."),
)
warnings.warn(msg)
return self.extract_encoder(layer)
def __delitem__(self, item: Union[str, int]) -> None:
msg = build_deprecation_message(
"Deleting modules via bracket indexing", "0.4.0")
warnings.warn(msg)
if isinstance(item, str):
del self._modules[item]
elif isinstance(item, int):
del self[self._get_child_name_by_idx(item)]
else:
raise TypeError
def encode(self, input_image: torch.Tensor) -> None:
msg = build_deprecation_message(
"The method 'encode'",
"1.0",
info=(
"It is no longer needed to pre-encode the input. "
"See https://github.com/pmeier/pystiche/issues/435 for details"
),
)
warnings.warn(msg)
def __init__(self, criterion: nn.Module) -> None:
msg = build_deprecation_message("The class MLEHandler", "1.0")
warnings.warn(msg)
self.multi_layer_encoders = {
loss.encoder.multi_layer_encoder
for loss in criterion.modules()
if isinstance(loss, Loss)
and not isinstance(loss, LossContainer)
and isinstance(loss.encoder, enc.SingleLayerEncoder)
}
def __init__(
self, named_ops: Sequence[Tuple[str, Loss]], trim: bool = True
) -> None:
msg = build_deprecation_message("The class MultiOperatorLoss", "1.0")
warnings.warn(msg)
super().__init__(named_children=named_ops)
self._mle_handler = MLEHandler(self)
if trim:
self._mle_handler.trim()
def __init__(self,
name: str = "time",
fmt: str = "{:3.1f}",
**kwargs: Any) -> None:
msg = build_deprecation_message(
"The class TimeMeter",
"0.4.0",
info="Please use AverageMeter instead.")
warnings.warn(msg)
super().__init__(name, fmt=fmt, **kwargs)
def patch_matching_loss(
input: torch.Tensor,
target: torch.Tensor,
eps: float = 1e-8,
reduction: str = "mean",
) -> torch.Tensor:
msg = build_deprecation_message(
"The function patch_matching_loss", "0.4.0", info="It was renamed to mrf_loss"
)
warnings.warn(msg, UserWarning)
return mrf_loss(input, target, eps=eps, reduction=reduction)
def __init__(
self,
content_loss: MSEEncodingOperator,
style_loss: GatysEtAl2017StyleLoss,
):
msg = build_deprecation_message(
"The class GatysEtAl2017GuidedPerceptualLoss",
"0.4.0",
info="It can be replaced by pystiche.loss.PerceptualLoss.",
)
warnings.warn(msg)
super().__init__(content_loss, style_loss)
def __init__(self,
*named_ops: Sequence[Tuple[str, Operator]],
trim: bool = True) -> None:
info = (
"Please construct a MultiOperatorLoss with a sequence of named operators."
)
named_children: Optional[Sequence[Tuple[str, Operator]]]
indexed_children: Optional[Sequence[nn.Module]]
if len(named_ops) == 1:
dict_or_seq = named_ops[0]
if isinstance(dict_or_seq, dict):
named_children = tuple(
cast(Dict[str, Operator], dict_or_seq).items())
msg = build_deprecation_message(
"Passing named_ops as dictionary",
"0.4.0",
info=info,
)
warnings.warn(msg)
else:
named_children = named_ops[0]
indexed_children = None
else:
msg = build_deprecation_message(
"Passing a variable number of unnamed operators via *args",
"0.4.0",
info=info,
)
warnings.warn(msg)
named_children = None
indexed_children = cast(Tuple[nn.Module, ...], named_ops)
super().__init__(named_children=named_children,
indexed_children=indexed_children)
self._multi_layer_encoders = self._collect_multi_layer_encoders()
if trim:
for encoder in self._multi_layer_encoders:
encoder.trim()
def __init__(
self,
encoder: Encoder,
patch_size: Union[int, Sequence[int]],
stride: Union[int, Sequence[int]] = 1,
target_transforms: Optional[Iterable[Transform]] = None,
score_weight: float = 1.0,
num_scale_steps: Optional[int] = None,
scale_step_width: Optional[float] = None,
num_rotation_steps: Optional[int] = None,
rotation_step_width: Optional[float] = None,
):
if any(
[
arg is not None
for arg in (
num_scale_steps,
scale_step_width,
num_rotation_steps,
rotation_step_width,
)
]
):
msg = build_deprecation_message(
(
"Parametrizing target transformations with any of "
"num_scale_steps, scale_step_width, num_rotation_steps, or "
"rotation_step_width through the constructor of MRFOperator"
),
"0.4.0",
info=(
"Please provide an iterable of transformations via the parameter "
"target_transforms. You can retain the old functionality with "
"MRFOperator.rotate_and_scale_transforms()."
),
)
warnings.warn(msg, UserWarning)
target_transforms = self.scale_and_rotate_transforms(
num_scale_steps=0 if num_scale_steps is None else num_scale_steps,
scale_step_width=5e-2 if scale_step_width is None else scale_step_width,
num_rotate_steps=0
if num_rotation_steps is None
else num_rotation_steps,
rotate_step_width=10.0
if rotation_step_width is None
else rotation_step_width,
)
super().__init__(encoder, score_weight=score_weight)
self.patch_size = to_2d_arg(patch_size)
self.stride = to_2d_arg(stride)
self.target_transforms = target_transforms
def deprecation(fn: Callable) -> Callable:
name = f"{fn.__name__}()"
msg = build_deprecation_message(
f"The function pystiche.{name}",
"0.4.0",
info=f"It was moved to pystiche.meta.{name}.",
)
def wrapper(*args: Any, **kwargs: Any) -> Any:
warnings.warn(msg)
return fn(*args, **kwargs)
return wrapper
def __init__(self, *args: Any, **kwargs: Any) -> None:
msg = build_deprecation_message(
"The class 'Unsupervised'",
"0.7",
info=
"See https://github.com/pmeier/pystiche/issues/458 for details",
)
warnings.warn(msg)
if not isinstance(self, VisionDataset):
raise RuntimeError
# This should be used with double inheritance with a
# torchvision.datasets.VisionDataset
super().__init__(*args, **kwargs) # type: ignore[call-arg]
def add_named_modules(self, modules: Sequence[Tuple[str,
nn.Module]]) -> None:
if isinstance(modules, dict):
msg = build_deprecation_message(
"Adding named_modules from a dictionary",
"0.4",
info=("To achieve the same behavior you can pass "
"tuple(modules.items()) instead."),
)
warnings.warn(msg)
modules = tuple(modules.items())
for name, module in modules:
self.add_module(name, module)
def __init__(
self,
content_loss: MSEEncodingOperator,
style_loss: MultiLayerEncodingOperator,
regularization: TotalVariationOperator,
) -> None:
msg = build_deprecation_message(
"The class JohnsonAlahiLi2016PerceptualLoss",
"0.4.0",
info="It can be replaced by pystiche.loss.PerceptualLoss.",
)
warnings.warn(msg)
super().__init__(content_loss, style_loss, regularization=regularization)
def __getitem__(self, item: Union[str, int]) -> nn.Module:
msg = build_deprecation_message(
"Dynamic access to the modules via bracket indexing",
"0.4.0",
info=
"If you need dynamic access to the operators, use getattr() instead.",
)
warnings.warn(msg)
if isinstance(item, str):
return self._modules[item]
elif isinstance(item, int):
return self[self._get_child_name_by_idx(item)]
else:
raise TypeError
def alexnet_multi_layer_encoder(
weights: str = "torch",
preprocessing: Optional[bool] = None,
internal_preprocessing: bool = True,
allow_inplace: bool = False,
) -> MultiLayerAlexNetEncoder:
if preprocessing is not None:
msg = build_deprecation_message(
"The parameter preprocessing",
"0.4.0",
info="It was replaced by internal_preprocessing.",
)
warnings.warn(msg)
internal_preprocessing = preprocessing
return MultiLayerAlexNetEncoder(weights, internal_preprocessing,
allow_inplace)
def _deprecate(fn: Callable) -> Callable:
name = fn.__name__
msg = build_deprecation_message(
f"The function ops.functional.{name}",
"1.0",
info=(
f"It was moved to loss.functional.{name}. "
f"See https://github.com/pystiche/pystiche/issues/436 for details"
),
)
@functools.wraps(fn)
def wrapper(*args, **kwargs):
warnings.warn(msg)
return fn(*args, **kwargs)
return wrapper
def __op_init__(self, *args: Any, __old_name__: str, __new_name__: str,
**kwargs: Any) -> None:
msg = build_deprecation_message(
f"The class pystiche.ops.{__old_name__}",
"1.0",
info=(
f"It was renamed and moved to pystiche.loss.{__new_name__}. "
f"See https://github.com/pystiche/pystiche/issues/436 for details"
),
)
warnings.warn(msg)
for super_cls in type(self).__mro__:
if _PATTERN.match(str(super_cls)):
break
else:
raise RuntimeError
super_cls.__init__(self, *args, **kwargs)
