#!/usr/bin/env python3
import typing
from typing import Any, cast, List, Tuple, Union
from captum._utils.common import (
_format_tensor_into_tuples,
_reduce_list,
_sort_key_list,
)
from captum._utils.gradient import (
apply_gradient_requirements,
compute_gradients,
undo_gradient_requirements,
)
from captum._utils.typing import (
Literal,
ModuleOrModuleList,
TargetType,
TensorOrTupleOfTensorsGeneric,
)
from captum.attr._core.lrp import LRP
from captum.attr._utils.attribution import LayerAttribution
from torch import Tensor
from torch.nn import Module
[docs]
class LayerLRP(LRP, LayerAttribution):
r"""
Layer-wise relevance propagation is based on a backward propagation
mechanism applied sequentially to all layers of the model. Here, the
model output score represents the initial relevance which is decomposed
into values for each neuron of the underlying layers. The decomposition
is defined by rules that are chosen for each layer, involving its weights
and activations. Details on the model can be found in the original paper
[https://doi.org/10.1371/journal.pone.0130140]. The implementation is
inspired by the tutorial of the same group
[https://doi.org/10.1016/j.dsp.2017.10.011] and the publication by
Ancona et al. [https://openreview.net/forum?id=Sy21R9JAW].
"""
def __init__(self, model: Module, layer: ModuleOrModuleList) -> None:
"""
Args:
model (Module): The forward function of the model or
any modification of it. Custom rules for a given layer need to
be defined as attribute
`module.rule` and need to be of type PropagationRule.
layer (torch.nn.Module or list(torch.nn.Module)): Layer or layers
for which attributions are computed.
The size and dimensionality of the attributions
corresponds to the size and dimensionality of the layer's
input or output depending on whether we attribute to the
inputs or outputs of the layer. If value is None, the
relevance for all layers is returned in attribution.
"""
LayerAttribution.__init__(self, model, layer)
LRP.__init__(self, model)
if hasattr(self.model, "device_ids"):
self.device_ids = cast(List[int], self.model.device_ids)
@typing.overload # type: ignore
def attribute(
self,
inputs: TensorOrTupleOfTensorsGeneric,
target: TargetType = None,
additional_forward_args: Any = None,
return_convergence_delta: Literal[False] = False,
attribute_to_layer_input: bool = False,
verbose: bool = False,
) -> Union[Tensor, Tuple[Tensor, ...], List[Union[Tensor, Tuple[Tensor, ...]]]]: ...
@typing.overload
def attribute(
self,
inputs: TensorOrTupleOfTensorsGeneric,
target: TargetType = None,
additional_forward_args: Any = None,
*,
return_convergence_delta: Literal[True],
attribute_to_layer_input: bool = False,
verbose: bool = False,
) -> Tuple[
Union[Tensor, Tuple[Tensor, ...], List[Union[Tensor, Tuple[Tensor, ...]]]],
Union[Tensor, List[Tensor]],
]: ...
[docs]
def attribute(
self,
inputs: TensorOrTupleOfTensorsGeneric,
target: TargetType = None,
additional_forward_args: Any = None,
return_convergence_delta: bool = False,
attribute_to_layer_input: bool = False,
verbose: bool = False,
) -> Union[
Tensor,
Tuple[Tensor, ...],
List[Union[Tensor, Tuple[Tensor, ...]]],
Tuple[
Union[Tensor, Tuple[Tensor, ...], List[Union[Tensor, Tuple[Tensor, ...]]]],
Union[Tensor, List[Tensor]],
],
]:
r"""
Args:
inputs (Tensor or tuple[Tensor, ...]): Input for which relevance is
propagated.
If model takes a single
tensor as input, a single input tensor should be provided.
If model takes multiple tensors as input, a tuple
of the input tensors should be provided. It is assumed
that for all given input tensors, dimension 0 corresponds
to the number of examples, and if multiple input tensors
are provided, the examples must be aligned appropriately.
target (int, tuple, Tensor, or list, optional): Output indices for
which gradients are computed (for classification cases,
this is usually the target class).
If the network returns a scalar value per example,
no target index is necessary.
For general 2D outputs, targets can be either:
- a single integer or a tensor containing a single
integer, which is applied to all input examples
- a list of integers or a 1D tensor, with length matching
the number of examples in inputs (dim 0). Each integer
is applied as the target for the corresponding example.
For outputs with > 2 dimensions, targets can be either:
- A single tuple, which contains #output_dims - 1
elements. This target index is applied to all examples.
- A list of tuples with length equal to the number of
examples in inputs (dim 0), and each tuple containing
#output_dims - 1 elements. Each tuple is applied as the
target for the corresponding example.
Default: None
additional_forward_args (tuple, optional): If the forward function
requires additional arguments other than the inputs for
which attributions should not be computed, this argument
can be provided. It must be either a single additional
argument of a Tensor or arbitrary (non-tuple) type or a tuple
containing multiple additional arguments including tensors
or any arbitrary python types. These arguments are provided to
model in order, following the arguments in inputs.
Note that attributions are not computed with respect
to these arguments.
Default: None
return_convergence_delta (bool, optional): Indicates whether to return
convergence delta or not. If `return_convergence_delta`
is set to True convergence delta will be returned in
a tuple following attributions.
Default: False
attribute_to_layer_input (bool, optional): Indicates whether to
compute the attribution with respect to the layer input
or output. If `attribute_to_layer_input` is set to True
then the attributions will be computed with respect to
layer input, otherwise it will be computed with respect
to layer output.
verbose (bool, optional): Indicates whether information on application
of rules is printed during propagation.
Default: False
Returns:
*Tensor* or *tuple[Tensor, ...]* of **attributions** or 2-element tuple of
**attributions**, **delta** or list of **attributions** and **delta**:
- **attributions** (*Tensor* or *tuple[Tensor, ...]*):
The propagated relevance values with respect to each
input feature. Attributions will always
be the same size as the provided inputs, with each value
providing the attribution of the corresponding input index.
If a single tensor is provided as inputs, a single tensor is
returned. If a tuple is provided for inputs, a tuple of
corresponding sized tensors is returned. The sum of attributions
is one and not corresponding to the prediction score as in other
implementations. If attributions for all layers are returned
(layer=None) a list of tensors or tuples of tensors is returned
with entries for each layer.
- **delta** (*Tensor* or list of *Tensor*
returned if return_convergence_delta=True):
Delta is calculated per example, meaning that the number of
elements in returned delta tensor is equal to the number of
examples in input.
If attributions for all layers are returned (layer=None) a list
of tensors is returned with entries for
each layer.
Examples::
>>> # ImageClassifier takes a single input tensor of images Nx3x32x32,
>>> # and returns an Nx10 tensor of class probabilities. It has one
>>> # Conv2D and a ReLU layer.
>>> net = ImageClassifier()
>>> layer_lrp = LayerLRP(net, net.conv1)
>>> input = torch.randn(3, 3, 32, 32)
>>> # Attribution size matches input size: 3x3x32x32
>>> attribution = layer_lrp.attribute(input, target=5)
"""
self.verbose = verbose
self._original_state_dict = self.model.state_dict()
self.layers = []
self._get_layers(self.model)
self._check_and_attach_rules()
self.attribute_to_layer_input = attribute_to_layer_input
self.backward_handles = []
self.forward_handles = []
inputs = _format_tensor_into_tuples(inputs)
gradient_mask = apply_gradient_requirements(inputs)
try:
# 1. Forward pass
output = self._compute_output_and_change_weights(
inputs, target, additional_forward_args
)
self._register_forward_hooks()
# 2. Forward pass + backward pass
_ = compute_gradients(
self._forward_fn_wrapper, inputs, target, additional_forward_args
)
relevances = self._get_output_relevance(output)
finally:
self._restore_model()
undo_gradient_requirements(inputs, gradient_mask)
if return_convergence_delta:
delta: Union[Tensor, List[Tensor]]
if isinstance(self.layer, list):
delta = []
for relevance_layer in relevances:
delta.append(
self.compute_convergence_delta(relevance_layer, output)
)
else:
delta = self.compute_convergence_delta(
cast(Tuple[Tensor, ...], relevances), output
)
return relevances, delta # type: ignore
else:
return relevances # type: ignore
def _get_single_output_relevance(self, layer, output):
if self.attribute_to_layer_input:
normalized_relevances = layer.rule.relevance_input
else:
normalized_relevances = layer.rule.relevance_output
key_list = _sort_key_list(list(normalized_relevances.keys()), self.device_ids)
normalized_relevances = _reduce_list(
[normalized_relevances[device_id] for device_id in key_list]
)
if isinstance(normalized_relevances, tuple):
return tuple(
normalized_relevance
* output.reshape((-1,) + (1,) * (normalized_relevance.dim() - 1))
for normalized_relevance in normalized_relevances
)
else:
return normalized_relevances * output.reshape(
(-1,) + (1,) * (normalized_relevances.dim() - 1)
)
def _get_output_relevance(self, output):
if isinstance(self.layer, list):
relevances = []
for layer in self.layer:
relevances.append(self._get_single_output_relevance(layer, output))
return relevances
else:
return self._get_single_output_relevance(self.layer, output)
@staticmethod
def _convert_list_to_tuple(
relevances: Union[List[Any], Tuple[Any, ...]]
) -> Tuple[Any, ...]:
if isinstance(relevances, list):
return tuple(relevances)
else:
return relevances
