def _guided_backprop_matching_assert(
self, model: Module, output_layer: Module, test_input: TensorOrTupleOfTensors
):
out = model(test_input)
attrib = GuidedBackprop(model)
neuron_attrib = NeuronGuidedBackprop(model, output_layer)
for i in range(out.shape[1]):
gbp_vals = attrib.attribute(test_input, target=i)
neuron_gbp_vals = neuron_attrib.attribute(test_input, (i,))
assertTensorAlmostEqual(self, gbp_vals, neuron_gbp_vals, delta=0.01)
def _guided_backprop_test_assert(self,
model,
test_input,
expected,
additional_input=None):
guided_backprop = GuidedBackprop(model)
attributions = guided_backprop.attribute(
test_input, target=0, additional_forward_args=additional_input)
for i in range(len(test_input)):
assertTensorAlmostEqual(self,
attributions[i],
expected[i],
delta=0.01)
def _guided_backprop_test_assert(
self,
model: Module,
test_input: TensorOrTupleOfTensorsGeneric,
expected: Tuple[List[List[float]], ...],
additional_input: Any = None,
) -> None:
guided_backprop = GuidedBackprop(model)
attributions = guided_backprop.attribute(
test_input, target=0, additional_forward_args=additional_input
)
for i in range(len(test_input)):
assertTensorAlmostEqual(self, attributions[i], expected[i], delta=0.01)
def __init__(
self, model: Module, layer: Module, device_ids: Union[None, List[int]] = None
) -> None:
r"""
Args:
model (nn.Module): The reference to PyTorch model instance. Model cannot
contain any in-place ReLU submodules; these are not
supported by the register_full_backward_hook PyTorch API
starting from PyTorch v1.9.
layer (Module): Layer for which neuron attributions are computed.
Attributions for a particular neuron in the output of
this layer are computed using the argument neuron_selector
in the attribute method.
Currently, only layers with a single tensor output are
supported.
device_ids (list(int)): Device ID list, necessary only if forward_func
applies a DataParallel model. This allows reconstruction of
intermediate outputs from batched results across devices.
If forward_func is given as the DataParallel model itself,
then it is not necessary to provide this argument.
"""
NeuronAttribution.__init__(self, model, layer, device_ids)
GradientAttribution.__init__(self, model)
self.guided_backprop = GuidedBackprop(model)
def _guided_backprop_matching_assert(
self,
model: Module,
output_layer: Module,
test_input: TensorOrTupleOfTensorsGeneric,
):
out = model(test_input)
model_copy = copy.deepcopy(model)
attrib = GuidedBackprop(model_copy)
self.assertFalse(attrib.multiplies_by_inputs)
neuron_attrib = NeuronGuidedBackprop(model, output_layer)
for i in range(out.shape[1]):
gbp_vals = attrib.attribute(test_input, target=i)
neuron_gbp_vals = neuron_attrib.attribute(test_input, (i, ))
assertTensorAlmostEqual(self,
gbp_vals,
neuron_gbp_vals,
delta=0.01)
def __init__(self,
model: Module,
layer: Module,
device_ids: Union[None, List[int]] = None) -> None:
r"""
Args:
model (nn.Module): The reference to PyTorch model instance.
layer (torch.nn.Module): Layer for which GradCAM attributions are computed.
Currently, only layers with a single tensor output are
supported.
device_ids (list(int)): Device ID list, necessary only if forward_func
applies a DataParallel model. This allows reconstruction of
intermediate outputs from batched results across devices.
If forward_func is given as the DataParallel model itself,
then it is not necessary to provide this argument.
"""
GradientAttribution.__init__(self, model)
self.grad_cam = LayerGradCam(model, layer, device_ids)
self.guided_backprop = GuidedBackprop(model)