def mask_propagation(self, model: NNCFNetwork, nx_node, graph: NNCFGraph,
nx_graph: nx.DiGraph):
input_masks = get_input_masks(nx_node, nx_graph)
nx_node['input_masks'] = input_masks
nx_node['output_mask'] = None
nx_node['accept_pruned_input'] = False
def all_inputs_from_convs(self, nx_node, nx_graph, graph):
"""
Return whether all input sources of nx_node is convolutions or not
:param nx_node: node to determine it's sources
:param nx_graph: networkx graph to work with
:param graph: NNCF graph to work with
"""
inputs = [u for u, _ in nx_graph.in_edges(nx_node['key'])]
input_masks = get_input_masks(nx_node, nx_graph)
for i, inp in enumerate(inputs):
# If input has mask -> it went from convolution (source of this node is a convolution)
if input_masks[i] is not None:
continue
nncf_input_node = graph._nx_node_to_nncf_node(nx_graph.nodes[inp])
source_nodes = get_sources_of_node(
nncf_input_node, graph,
Convolution.get_all_op_aliases() +
StopMaskForwardOps.get_all_op_aliases() +
Input.get_all_op_aliases())
sources_types = [
node.op_exec_context.operator_name for node in source_nodes
]
if any([
t in sources_types
for t in StopMaskForwardOps.get_all_op_aliases()
]):
return False
return True
def mask_propagation(self, model: NNCFNetwork, nx_node, graph: NNCFGraph,
nx_graph: nx.DiGraph):
output_mask = None
accept_pruned_input = True
is_depthwise = False
input_masks = get_input_masks(nx_node, nx_graph)
nncf_node = graph._nx_node_to_nncf_node(nx_node)
node_module = model.get_module_by_scope(
nncf_node.op_exec_context.scope_in_model)
if node_module.pre_ops:
output_mask = node_module.pre_ops[
'0'].op.binary_filter_pruning_mask
# In case of group convs we can't prune by output filters
if node_module.groups != 1:
if node_module.weight.size(1) == 1:
# Depthwise case
is_depthwise = True
output_mask = input_masks[0]
else:
accept_pruned_input = False
output_mask = None
nx_node['input_masks'] = input_masks
nx_node['output_mask'] = output_mask
nx_node['accept_pruned_input'] = accept_pruned_input
nx_node['is_depthwise'] = is_depthwise
def mask_propagation(self, model: NNCFNetwork, nx_node, graph: NNCFGraph,
nx_graph: nx.DiGraph):
input_masks = get_input_masks(nx_node, nx_graph)
nx_node['input_masks'] = input_masks
if input_masks[0] is not None:
assert all(
[torch.allclose(input_masks[0], mask) for mask in input_masks])
nx_node['output_mask'] = input_masks[0]
def mask_propagation(self, model: NNCFNetwork, nx_node, graph: NNCFGraph,
nx_graph: nx.DiGraph):
output_mask = None
accept_pruned_input = True
input_masks = get_input_masks(nx_node, nx_graph)
nncf_node = graph._nx_node_to_nncf_node(nx_node)
node_module = model.get_module_by_scope(
nncf_node.op_exec_context.scope_in_model)
if node_module.pre_ops:
output_mask = node_module.pre_ops[
'0'].op.binary_filter_pruning_mask
nx_node['input_masks'] = input_masks
nx_node['output_mask'] = output_mask
nx_node['accept_pruned_input'] = accept_pruned_input