def replace_pattern(self, graph: Graph, match: dict):
slice_like = match['slice_like']
anchor_node = slice_like.in_port(0).get_source().node
reshape = slice_like.out_nodes()[0].out_node()
slice_shape = slice_like.out_nodes()[0].shape
anchor_node.value = np.copy(
anchor_node.value[:slice_shape[0], :slice_shape[1], :
slice_shape[2], :slice_shape[3], :slice_shape[4]]
)
anchor_node.shape = slice_shape
val_node = Const(
graph, {
'name':
slice_like.name + '/croped_',
'value':
anchor_node.
value[:slice_shape[0], :slice_shape[1], :slice_shape[2], :
slice_shape[3], :slice_shape[4]],
'shape':
slice_shape
}).create_node_with_data()
slice_like.in_port(0).disconnect()
slice_like.in_port(1).disconnect()
slice_like.out_port(0).disconnect()
reshape.in_port(0).connect(val_node.in_node().out_port(0))
def replace_pattern(self, graph: Graph, match: dict):
#self.pattern()['nodes']
concat_node = match['concat']
if len(concat_node.out_nodes()[0].out_nodes()) == 0:
return
const_values = []
for in_node_index in concat_node.in_nodes():
current_node = concat_node.in_port(in_node_index).get_source().node
for k, v in reversed(self.pattern()['nodes'][:-1]):
if 'op' in v:
assert current_node.op == v['op']
current_node = current_node.in_port(0).get_source().node
if current_node.op == 'Const':
crop_value = current_node.value
crop_value = np.reshape(crop_value, (1, -1))
const_values.append(crop_value)
break
concat_value = np.concatenate(tuple(const_values), axis=1)
concat_value = np.reshape(concat_value, (1, 2, -1))
slice_value = concat_value[0][0]
for i in range(int(concat_value[0][0].size / 4)):
index = i * 4
xmin = slice_value[index] - (slice_value[index + 2] / 2)
ymin = slice_value[index + 1] - (slice_value[index + 3] / 2)
xmax = slice_value[index] + (slice_value[index + 2] / 2)
ymax = slice_value[index + 1] + (slice_value[index + 3] / 2)
slice_value[index] = xmin
slice_value[index + 1] = ymin
slice_value[index + 2] = xmax
slice_value[index + 3] = ymax
val_node = Const(graph, {
'name': concat_node.name + '/const_',
'value': concat_value
}).create_node_with_data()
out_node = concat_node.out_port(0).get_destination().node
concat_node.out_port(0).disconnect()
out_node.in_port(2).connect(val_node.in_node().out_port(0))
def replace_pattern(self, graph: Graph, match: dict):
node = match['reduce']
if node.out_port(0).data.get_value() is not None:
# We leave Reduce* operations located in constant sub-graph as is
# to keep model reshapable with --keep_shape_ops cli key
return
reduce_type = node.type
if reduce_type not in self.pool_method_map:
log.error(
"Reduce type {} is not included in pool_method_map. Please update pool_method_map with new key "
"{}".format(reduce_type, reduce_type))
return
input_data = node.in_node()
output_data = node.out_node()
input_shape = node.in_port(0).data.get_shape()
output_shape = node.out_port(0).data.get_shape()
# normalize node axes to exclude negative indices
axes_data_value = node.in_port(1).data.get_value()
axes = int64_array([
axes_data_value.item()
]) if axes_data_value.size == 1 else axes_data_value
axes = [get_canonical_axis_index(input_shape, a) for a in axes]
axes = sorted(axes)
# Check that values in axes list are consecutive
for idx in range(1, len(axes)):
if axes[idx] != (axes[idx - 1] + 1):
log.error(
"Reduce with not consecutive axes {} is not supported ".
format(axes))
return
# So now we are sure that we can convert Reduce to appropriate operation
# 1. Calculate shape that will be used in reduction
reduction_dim = np.prod([input_shape[idx] for idx in axes])
begin_dims = np.array([input_shape[idx] for idx in range(axes[0])])
end_dim = np.prod([
input_shape[idx] for idx in range(axes[-1] + 1, len(input_shape))
])
# 2. Create reshape with appropriate shape
if len(begin_dims) > 2:
if 0 not in axes:
begin_dims = int64_array(
[begin_dims[0], np.prod(begin_dims[1:])])
else:
begin_dims = int64_array(
[np.prod(begin_dims[0:-1]), begin_dims[-1]])
else:
# Expand begin_dims to 2
begin_dims = int64_array(
np.append(begin_dims, [1] * (2 - len(begin_dims))))
reshape_shape = int64_array([*begin_dims, reduction_dim, end_dim])
pool_window = int64_array([1, 1, reduction_dim, 1])
if end_dim == 1:
new_window = ReduceReplacer.initial_reshape_dim_normalizer(
reduction_dim)
reshape_shape = int64_array([*begin_dims, *new_window])
pool_window = int64_array([1, 1, *new_window])
# 3. Reduce => Reshape->Pooling->Reshape
reshape_op = Reshape(graph, {'name': node.id + '/Reshape'})
reshape_dim_const_data = Const(graph, {
'name': node.id + '/Reshape/Dim',
'value': reshape_shape
}).create_node_with_data()
final_reshape_op = Reshape(graph, {'name': node.id + '/FinalReshape'})
final_reshape_dim_const_data = Const(graph, {
'name': node.id + '/FinalReshape/Dim',
'value': output_shape
}).create_node_with_data()
pooling_op = Pooling(
graph,
dict(name=node.id + '/Pool',
window=pool_window,
output_spatial_shape=None,
batch_dims=int64_array([0]),
channel_dims=int64_array([1]),
exclude_pad='false',
pool_method=self.pool_method_map[reduce_type]))
graph.remove_edge(input_data.id, node.id)
graph.remove_edge(node.id, output_data.id)
if np.array_equal(input_shape, reshape_shape):
input_to_pooling = input_data
else:
input_to_pooling = reshape_op.create_node_with_data(
inputs=[input_data, reshape_dim_const_data])
pooling = pooling_op.create_node_with_data(inputs=[input_to_pooling])
final_reshape_op.create_node_with_data(
inputs=[pooling, final_reshape_dim_const_data],
data_nodes=output_data)
# convert batch dimension to 0 to produce reshape-able IR over the batch dimension
if 0 not in axes:
reshape_dim_const_data.in_node(0).value[0] = 0
final_reshape_dim_const_data.in_node(0).value[0] = 0
# 4. If it is reduction with summation, we need to multiply by size of the reduction slice with Mul op
if reduce_type == 'ReduceSum':
output_data.in_node().insert_node_with_data_after(
output_data, AttributedPower, {
'name': node.name + '/Mul',
'scale': float(reduction_dim)
})
