def placeholder_scales(self, placeholder: Node):
"""
Helper function to get scales for prior boxes out of input image size:
[1 / im_width, 1 / im_height, 1 / im_width, 1 / im_height]
"""
graph = placeholder.graph
name = placeholder.soft_get('name', placeholder.id)
shape_value = placeholder.soft_get('shape', None)
assert shape_value is not None, \
"[ {} replacer ] Placeholder `{}` should have shape attribute".format(self.replacement_id, name)
assert isinstance(shape_value, np.ndarray), \
"[ {} replacer ] Placeholder `{}` shape attribute should be np.ndarray".format(self.replacement_id, name)
assert shape_value.size == 4, \
"[ {} replacer ] Placeholder `{}` should be 4D. Shape: {}".format(self.replacement_id, name, shape_value)
shape = Shape(graph, {'name': 'input_image_shape'}).create_node()
shape.in_port(0).connect(placeholder.out_port(0))
begin = Const(graph, {'value': int64_array([1])}).create_node()
end = Const(graph, {'value': int64_array([3])}).create_node()
stride = Const(graph, {'value': int64_array([1])}).create_node()
spatial = StridedSlice(graph, {'name': name + '/get_h_w', 'begin_mask': int64_array([1]),
'end_mask': int64_array([1]), 'new_axis_mask': int64_array([0]),
'shrink_axis_mask': int64_array([0]), 'ellipsis_mask': int64_array([0])}).create_node()
spatial.in_port(0).connect(shape.out_port(0))
spatial.in_port(1).connect(begin.out_port(0))
spatial.in_port(2).connect(end.out_port(0))
spatial.in_port(3).connect(stride.out_port(0))
power = Const(graph, {'value': float32_array([-1.])}).create_node()
spatial_scale = Pow(graph, {}).create_node()
spatial_scale.in_port(0).connect(spatial.out_port(0))
spatial_scale.in_port(1).connect(power.out_port(0))
# Power `type_infer` requires inputs to have equal data type
convert_to_fp32 = Cast(graph, {'dst_type': np.float32}).create_node()
spatial_scale.in_port(0).get_connection().insert_node(convert_to_fp32)
order = Const(graph, {'value': int64_array([1, 0])}).create_node()
axis_const = Const(graph, {'value': int64_array(0)}).create_node()
reverse = Gather(graph, {}).create_node()
reverse.in_port(0).connect(spatial_scale.out_port(0))
reverse.in_port(1).connect(order.out_port(0))
axis_const.out_port(0).connect(reverse.in_port(2))
priors_scale_node = Concat(graph, {'axis': 0, 'in_ports_count': 2}).create_node()
priors_scale_node.add_input_port(0, skip_if_exist=True)
priors_scale_node.add_input_port(1, skip_if_exist=True)
priors_scale_node.in_port(0).connect(reverse.out_port(0))
priors_scale_node.in_port(1).connect(reverse.out_port(0))
return priors_scale_node
def replace_op(self, graph: Graph, node: Node):
inp0 = node.in_port(0).get_source().node
inp1 = node.in_port(1).get_source().node
begin_id = Const(graph, {"value": int64_array([1])}).create_node()
end_id = Const(graph, {"value": int64_array([2])}).create_node()
dim1 = StridedSlice(
graph,
dict(
name=inp0.name + "/dim1",
begin_mask=[1],
end_mask=[1],
shrink_axis_mask=[0],
new_axis_mask=[0],
ellipsis_mask=[0],
),
).create_node([inp1, begin_id, end_id])
rows = Div(graph, dict(name=node.name + "/rows")).create_node([inp0, dim1])
inp0 = Cast(
graph, dict(name=inp0.name + "/fp32", dst_type=np.float32)
).create_node([inp0])
dim1 = Cast(
graph, dict(name=dim1.name + "/fp32", dst_type=np.float32)
).create_node([dim1])
cols = FloorMod(graph, dict(name=node.name + "/cols")).create_node([inp0, dim1])
cols = Cast(
graph, dict(name=cols.name + "/i64", dst_type=np.int64)
).create_node([cols])
concat = PackOp(graph, dict(name=node.name + "/merged", axis=0)).create_node(
[rows, cols]
)
return [concat.id]
def replace_sub_graph(self, graph: Graph, match: dict):
node = match['op']
strided_slice_node = StridedSlice(
graph,
dict(name=node.id + '/strided_slice_',
shrink_axis_mask=np.array(
np.zeros(len(node.crop_begin), dtype=np.int64)),
new_axis_mask=np.array(
np.zeros(len(node.crop_begin), dtype=np.int64)),
ellipsis_mask=np.array(
np.zeros(len(node.crop_begin), dtype=np.int64)),
begin_mask=np.array(
np.ones(len(node.crop_begin), dtype=np.int64)),
end_mask=np.array(np.ones(len(node.crop_end),
dtype=np.int64)))).create_node()
node.in_port(0).get_connection().set_destination(
strided_slice_node.in_port(0))
node.out_port(0).get_connection().set_source(
strided_slice_node.out_port(0))
crop_begin_node = Const(
graph,
dict(value=node.crop_begin,
symbol_dict={'name':
node.id + '/crop_begin_const'})).create_node()
crop_end_node = Const(
graph,
dict(value=node.crop_end,
symbol_dict={'name':
node.id + '/crop_end_const'})).create_node()
strided_slice_node.in_port(1).get_connection().set_source(
crop_begin_node.out_port(0))
strided_slice_node.in_port(2).get_connection().set_source(
crop_end_node.out_port(0))
if len(node.step) > 0:
stride_node = Const(
graph,
dict(value=node.step,
symbol_dict={'name':
node.id + '/steps_const'})).create_node()
strided_slice_node.in_port(3).get_connection().set_source(
stride_node.out_port(0))
def normalize_strided_slice(graph: Graph, node: Node):
input_shape = node.in_port(0).data.get_shape()
input_rank = len(input_shape)
begin = node.in_port(1).data.get_value()
if begin is not None:
slice_rank = len(begin)
else:
slice_rank = input_rank + np.count_nonzero(
node.new_axis_mask) - np.count_nonzero(node.shrink_axis_mask)
StridedSlice.align_mask_with_slice_rank(
node, slice_rank) # if StridedSlice is created after partial_infer
StridedSliceNormalizer.normalize_slices_attr(node)
num_insertions = input_rank - slice_rank + np.count_nonzero(
node.new_axis_mask)
assert num_insertions >= 0, 'slice_rank - num_new_axis must <= input rank. Got instead: ' \
'input_rank = {}, slice_rank = {}, num_new_axis = {}'. \
format(input_rank, slice_rank, np.count_nonzero(node.new_axis_mask))
if np.any(node.ellipsis_mask):
assert np.count_nonzero(
node.ellipsis_mask
) == 1, 'only one ellipsis_mask nonzero value is allowed'
ellipsis_start = np.nonzero(node.ellipsis_mask)[0][0]
# since we don't expect values in begin and end: take the whole range along ellipsis_start
node.begin_mask[ellipsis_start] = 0
node.end_mask[ellipsis_start] = 0
node.ellipsis_mask[ellipsis_start] = 0
insertion_start_idx = ellipsis_start + 1
StridedSliceNormalizer.unroll_ellipsis_for_inputs(
graph, node, ellipsis_start, num_insertions)
elif num_insertions > 0:
insertion_start_idx = slice_rank # insert blank values to mask ends
StridedSliceNormalizer.extend_inputs(node, num_insertions)
if num_insertions > 0:
# insert blank values for ellipsis unrolling and extending
for mask_name in StridedSlice.get_mask_names():
node[mask_name] = np.insert(node[mask_name],
insertion_start_idx,
[0] * num_insertions).astype(int)
def extract(cls, node):
pb = node.pb
bm = int_to_array_bit_mask(pb.attr["begin_mask"].i)
bm = mo_array([1 - b for b in bm], dtype=np.int32)
em = int_to_array_bit_mask(pb.attr["end_mask"].i)
em = mo_array([1 - b for b in em], dtype=np.int32)
attrs = {
'begin_mask':
bm,
'end_mask':
em,
'ellipsis_mask':
int_to_array_bit_mask(pb.attr["ellipsis_mask"].i),
'new_axis_mask':
int_to_array_bit_mask(pb.attr["new_axis_mask"].i),
'shrink_axis_mask':
int_to_array_bit_mask(pb.attr["shrink_axis_mask"].i),
}
StridedSlice.update_node_stat(node, attrs)
return cls.enabled
def extend(op: Node):
for attr in StridedSlice.get_mask_names():
# We can not use op.has_and_set(attr) here as a condition, because it will return False if begin/end is
# 1D tensor and begin_mask/end_mask is equal to 0
if op.has(attr) and op[attr] != '':
Extender.attr_to_list(op, attr)
else:
assert attr not in ['begin_mask', 'end_mask'],\
'{} is not defined for the node {}'.format(attr, op.soft_get('name', op.id))
op[attr] = int64_array([0])
op.begin_mask = int64_array([1 - i for i in op.begin_mask])
op.end_mask = int64_array([1 - i for i in op.end_mask])
def replace_op(self, graph: Graph, node: Node):
if node.module.inverse:
axes = Const(
graph, {
'value': int64_array(range(2, node.module.num_axes - 1))
}).create_node()
dft_node = IDFT(graph, dict(name=node.name,
in_ports_count=2)).create_node(
[node.in_node(0), axes])
# Slice a real part
begin_id = Const(graph, {
'value': int64_array([0, 0])
}).create_node()
end_id = Const(graph, {'value': int64_array([0, 1])}).create_node()
real = StridedSlice(
graph,
dict(name=node.name + '/real',
begin_mask=[0, 0],
end_mask=[0, 1],
shrink_axis_mask=[0, 0],
new_axis_mask=[0],
ellipsis_mask=[1, 0])).create_node(
[dft_node, begin_id, end_id])
squeeze_axis = Const(graph, {'value': -1}).create_node()
res = Squeeze(graph,
dict(name=node.name + '/squeeze')).create_node(
[real, squeeze_axis])
return [res.id]
else:
zero = Const(graph, {'value': 0.0}).create_node()
imag = Mul(graph, dict(name=node.name + '/imag')).create_node(
[node.in_node(0), zero])
cmplx = PackOp(graph,
dict(name=node.name + '/complex',
axis=-1)).create_node([node.in_node(0), imag])
axes = Const(graph, {
'value': int64_array(range(2, node.module.num_axes))
}).create_node()
dft_node = DFT(graph,
dict(name=node.name,
in_ports_count=2)).create_node([cmplx, axes])
return [dft_node.id]
def append_variances(priors_scale_node: Node, variance: list):
graph = priors_scale_node.graph
name = priors_scale_node.name
sp_shape = Shape(graph, {'name': name + '/shape'}).create_node()
priors_scale_node.out_port(0).connect(sp_shape.in_port(0))
begin = Const(graph, {'value': int64_array([-2])}).create_node()
end = Const(graph, {'value': int64_array([-1])}).create_node()
stride = Const(graph, {'value': int64_array([1])}).create_node()
shape_part_for_tiling = StridedSlice(graph, {'name': name + '/get_-2_dim', 'begin_mask': int64_array([1]),
'end_mask': int64_array([1]), 'new_axis_mask': int64_array([0]),
'shrink_axis_mask': int64_array([0]),
'ellipsis_mask': int64_array([0])}).create_node()
sp_shape.out_port(0).connect(shape_part_for_tiling.in_port(0))
begin.out_port(0).connect(shape_part_for_tiling.in_port(1))
end.out_port(0).connect(shape_part_for_tiling.in_port(2))
stride.out_port(0).connect(shape_part_for_tiling.in_port(3))
shape_concat = create_op_node_with_second_input(graph, Concat, int64_array([4]),
{'name': name + '/shape_for_tiling', 'in_ports_count': 2,
'axis': int64_array(0)},
shape_part_for_tiling)
variance = Const(graph, {'name': name + '/variance', 'value': float32_array(variance)}).create_node()
tile = Broadcast(graph, {'name': name + '/variance_tile'}).create_node()
variance.out_port(0).connect(tile.in_port(0))
shape_concat.out_port(0).connect(tile.in_port(1))
reshape_dim = Const(graph, {'value': int64_array([-1, 4])}).create_node()
sp_reshape = Reshape(graph, {'name': name + '/reshape'}).create_node()
sp_reshape.in_port(0).connect(priors_scale_node.out_port(0))
sp_reshape.in_port(1).connect(reshape_dim.out_port(0))
concat = Concat(graph,
{'name': name + '/priors_concat', 'axis': int64_array(0), 'in_ports_count': 2}).create_node()
sp_reshape.out_port(0).connect(concat.in_port(0))
tile.out_port(0).connect(concat.in_port(1))
output_dims = Const(graph, {'value': int64_array([1, 2, -1])}).create_node()
output_node = Reshape(graph, {'name': name + '/3D_priors_wth_variances'}).create_node()
concat.out_port(0).connect(output_node.in_port(0))
output_dims.out_port(0).connect(output_node.in_port(1))
return output_node
def find_and_replace_pattern(self, graph: Graph):
for node in graph.get_op_nodes(op='Slice'):
node_name = node.soft_get('name', node.id)
input_shape = node.in_port(0).data.get_shape()
if node.is_in_port_connected(3):
axes = node.in_port(3).data.get_value().copy()
assert axes is not None, 'The input with axes is not constant for node {}'.format(
node_name)
for i, val in enumerate(axes):
axes[i] = get_canonical_axis_index(input_shape, val)
else:
axes = int64_array(range(len(input_shape)))
ss_begin = create_ss_interval_border(graph,
node.in_port(1).get_source(),
input_shape, axes, node_name)
ss_end = create_ss_interval_border(graph,
node.in_port(2).get_source(),
input_shape, axes, node_name)
node.in_port(1).disconnect()
node.in_port(2).disconnect()
rename_nodes([(ss_begin, node_name + '/Begin'),
(ss_end, node_name + '/End')])
if node.is_in_port_connected(4):
steps = node.in_port(4).data.get_value()
assert steps is not None, 'The input with steps is not constant for node {}'.format(
node_name)
else:
steps = np.ones([axes.size])
ss_begin_mask = np.zeros(len(input_shape), dtype=np.int64)
ss_end_mask = np.zeros(len(input_shape), dtype=np.int64)
ss_step = np.ones(len(input_shape), dtype=np.int64)
for i, axis in enumerate(axes):
ss_begin_mask[axis] = 1
ss_end_mask[axis] = 1
ss_step[axis] = steps[i]
ss_strides = Const(
graph, dict(name=node_name + '/Strides',
value=ss_step)).create_node()
ss = StridedSlice(
graph,
dict(name='ss',
new_axis_mask=np.zeros(len(input_shape), dtype=np.int64),
shrink_axis_mask=np.zeros(len(input_shape),
dtype=np.int64),
ellipsis_mask=np.zeros(len(input_shape), dtype=np.int64),
begin_mask=ss_begin_mask,
end_mask=ss_end_mask)).create_node()
node.in_port(0).get_connection().set_destination(ss.in_port(0))
ss.in_port(1).connect(ss_begin.out_port(0))
ss.in_port(2).connect(ss_end.out_port(0))
ss.in_port(3).connect(ss_strides.out_port(0))
node.out_port(0).get_connection().set_source(ss.out_port(0))
rename_nodes([(node, node_name + '/ShouldBeDeleted'),
(ss, node_name)])
def generate_sub_graph(self, graph: Graph, match: SubgraphMatch):
reshape_classes_node = create_op_node_with_second_input(graph, Reshape, int64_array([0, -1]),
dict(name='do_reshape_classes'),
match.single_input_node(1)[0])
initial_priors_node = match.single_input_node(2)[0]
priors_name = initial_priors_node.soft_get('name', initial_priors_node.id)
# model calculates identical prior boxes for each batch, so we take first slice of them
begin = Const(graph, {'value': mo_array([0, 0, 0], dtype=np.int32)}).create_node()
end = Const(graph, {'value': mo_array([1, 0, 0], dtype=np.int32)}).create_node()
stride = Const(graph, {'value': mo_array([1, 1, 1], dtype=np.int32)}).create_node()
priors_node = StridedSlice(graph, {'name': priors_name + '/0_batch_slice',
'begin_mask': int64_array([1, 1, 1]),
'end_mask': int64_array([1, 0, 0]),
'new_axis_mask': int64_array([0]),
'shrink_axis_mask': int64_array([0]),
'ellipsis_mask': int64_array([0])}).create_node()
initial_priors_node.out_port(0).connect(priors_node.in_port(0))
begin.out_port(0).connect(priors_node.in_port(1))
end.out_port(0).connect(priors_node.in_port(2))
stride.out_port(0).connect(priors_node.in_port(3))
placeholders = graph.get_op_nodes(type='Parameter')
assert len(placeholders) == 1, "{} replacer requires model to have one Placeholder, but current model has " \
"{} placeholders".format(self.replacement_id, len(placeholders))
placeholder = placeholders[0]
# scale prior boxes to the [0, 1] interval
node_with_scales_for_prior_boxes = self.placeholder_scales(placeholder)
priors_scale_node = Mul(graph, {'name': 'scale_priors'}).create_node()
broadcast = Broadcast(graph, {'name': 'scales_broadcast'}).create_node()
shape_of_priors = Shape(graph, {'name': 'priors_shape'}).create_node()
priors_node.out_port(0).connect(shape_of_priors.in_port(0))
broadcast.in_port(1).connect(shape_of_priors.out_port(0))
broadcast.in_port(0).connect(node_with_scales_for_prior_boxes.out_port(0))
priors_scale_node.in_port(0).connect(priors_node.out_port(0))
priors_scale_node.in_port(1).connect(broadcast.out_port(0))
try:
variance = match.custom_replacement_desc.custom_attributes['variance']
except:
raise Error('There is no variance attribute in {} replacement config file `custom_attributes`'
''.format(self.replacement_id))
priors = self.append_variances(priors_scale_node, variance)
# calculate prior boxes widths and heights
split_node = create_op_with_const_inputs(
graph, VariadicSplit, {1: int64_array(2), 2: int64_array([1, 1, 1, 1])}, {'out_ports_count': 4},
priors_scale_node)
priors_width_node = Sub(graph, dict(name=split_node.name + '/sub_2-0_')
).create_node([(split_node, 2), (split_node, 0)])
priors_height_node = Sub(graph, dict(name=split_node.name + '/sub_3-1_')
).create_node([(split_node, 3), (split_node, 1)])
# concat weights and heights into a single tensor and multiple with the box coordinates regression values
# WA with 3 Concats instead of 1 for keeping model reshapable
# concat_width_height_node = Concat(graph, {'name': 'concat_priors_width_height', 'axis': -1,
# 'in_ports_count': 4}).create_node(
# [priors_width_node, priors_height_node, priors_width_node, priors_height_node])
concat_1 = Concat(graph, {'name': 'concat_width_height',
'axis': -1, 'in_ports_count': 2}).create_node([priors_width_node, priors_height_node])
concat_2 = Concat(graph, {'name': 'concat_width_height_width',
'axis': -1, 'in_ports_count': 2}).create_node([concat_1, priors_width_node])
concat_width_height_node = Concat(graph, {'name': 'concat_priors_width_height', 'axis': -1, 'in_ports_count': 2}
).create_node([concat_2, priors_height_node])
applied_width_height_regressions_node = Mul(graph, {'name': 'final_regressions'}).create_node(
[concat_width_height_node, match.single_input_node(0)[0]])
# reshape to 2D tensor as Inference Engine Detection Output layer expects
reshape_regression_node = create_op_node_with_second_input(graph, Reshape, int64_array([0, -1]),
dict(name='reshape_regression'),
applied_width_height_regressions_node)
detection_output_op = DetectionOutput(graph, match.custom_replacement_desc.custom_attributes)
# get nms from the original network
iou_threshold = None
nms_nodes = graph.get_op_nodes(op='NonMaxSuppression')
if len(nms_nodes) > 0:
# it is highly unlikely that for different classes NMS has different
# moreover DetectionOutput accepts only scalar values for iou_threshold (nms_threshold)
iou_threshold = nms_nodes[0].in_node(3).value
if iou_threshold is None:
raise Error('During {} `iou_threshold` was not retrieved from RetinaNet graph'.format(self.replacement_id))
detection_output_node = detection_output_op.create_node(
[reshape_regression_node, reshape_classes_node, priors],
dict(name=detection_output_op.attrs['type'], nms_threshold=iou_threshold, clip_after_nms=1, normalized=1,
variance_encoded_in_target=0, background_label_id=1000))
# As outputs are replaced with a postprocessing node, outgoing tensor names are no longer
# correspond to original tensors and should be removed from output->Result edges
out_nodes = []
for out in range(match.outputs_count()):
out_nodes.append(match.output_node(out)[0])
clear_tensor_names_info(out_nodes)
return {'detection_output_node': detection_output_node}
def replace_sub_graph(self, graph: Graph, match: dict):
node = match['op']
if 1 not in node.in_ports() or node.in_port(1).disconnected():
if node.has_valid('factor') and not node.has_valid('width') and not node.has_valid('height'):
factor = Const(graph, {'value': np.array(node.factor)}).create_node()
shape = Shape(graph, {'name': node.name + '/shape'}).create_node()
begin = Const(graph, {'value': np.array([2])}).create_node()
end = Const(graph, {'value': np.array([4])}).create_node()
stride = Const(graph, {'value': np.array([1])}).create_node()
ss = StridedSlice(graph, {'name': node.name + '/ss_0_port', 'begin_mask': np.array([1]),
'end_mask': np.array([0]), 'new_axis_mask': np.array([0]),
'shrink_axis_mask': np.array([0]),
'ellipsis_mask': np.array([0])}).create_node()
mul = Mul(graph, {'name': node.name + '/factor_mul_'}).create_node()
source = node.in_port(0).get_connection().get_source()
source.connect(shape.in_port(0))
shape.out_port(0).connect(ss.in_port(0))
begin.out_port(0).connect(ss.in_port(1))
end.out_port(0).connect(ss.in_port(2))
stride.out_port(0).connect(ss.in_port(3))
ss.out_port(0).connect(mul.in_port(0))
factor.out_port(0).connect(mul.in_port(1))
node.add_input_port(1, skip_if_exist=True)
assert node.in_port(1).disconnected()
mul.out_port(0).connect(node.in_port(1))
else:
shape = Shape(graph, {'name': node.name + '/shape'}).create_node()
begin = Const(graph, {'value': np.array([2])}).create_node()
end = Const(graph, {'value': np.array([4])}).create_node()
stride = Const(graph, {'value': np.array([1])}).create_node()
ss = StridedSlice(graph, {'name': node.name + '/ss_0_port', 'begin_mask': np.array([1]),
'end_mask': np.array([0]), 'new_axis_mask': np.array([0]),
'shrink_axis_mask': np.array([0]),
'ellipsis_mask': np.array([0])}).create_node()
source = node.in_port(0).get_connection().get_source()
source.connect(shape.in_port(0))
shape.out_port(0).connect(ss.in_port(0))
begin.out_port(0).connect(ss.in_port(1))
end.out_port(0).connect(ss.in_port(2))
stride.out_port(0).connect(ss.in_port(3))
pads_value = node.pads_begin + node.pads_end
pads_const = Const(graph, {'value': np.array(pads_value)}).create_node()
add = Add(graph, {'name': node.name + '/pad_add'}).create_node()
ss.out_port(0).connect(add.in_port(0))
add.in_port(1).connect(pads_const.out_port(0))
if node.soft_get('shrink_factor') != 1 and node.soft_get('zoom_factor') == 1:
shrink_factor = node.shrink_factor
if shrink_factor < 1:
log.error('Shrink factor should be positive in node {}'.format(node.id))
return None
const = Const(graph, {'name': node.name + '/pre_shrink_sub_const',
'value': np.array(-1)}).create_node()
sub = Add(graph, {'name': node.name + '/pre_shrink_sub'}).create_node()
add.out_port(0).connect(sub.in_port(0))
sub.in_port(1).connect(const.out_port(0))
const = Const(graph, {'value': np.array(1 / shrink_factor),
'name': node.name + 'shrink_factor_div_const'}).create_node()
div = Mul(graph, {'name': node.name + 'shrink_factor_div'}).create_node()
sub.out_port(0).connect(div.in_port(0))
div.in_port(1).connect(const.out_port(0))
const = Const(graph, {'name': node.name + '/shrink_factor_add_one_const', 'value': np.array(1)
}).create_node()
add = Add(graph, {'name': node.name + '/shrink_factor_add_one'}).create_node()
div.out_port(0).connect(add.in_port(0))
const.out_port(0).connect(add.in_port(1))
node.add_input_port(1, skip_if_exist=True)
assert node.in_port(1).disconnected()
add.out_port(0).connect(node.in_port(1))
elif node.soft_get('shrink_factor') == 1 and node.soft_get('zoom_factor') != 1:
zoom_factor = node.zoom_factor
if zoom_factor < 1:
log.error('Zoom factor should be positive in node {}'.format(node.id))
return None
node['debug_message'] = 'Interpolate layer replacer may be wrong, please, try to update it in the' \
' file (openvino/tools/mo/front/InterpolateNormalizer.py at the line {}).' \
''.format(inspect.currentframe().f_lineno) + refer_to_faq_msg(100)
# Reshape methods can be different in some cases
# Commented out section represents reshape that used in deeplab-caffe
# Uncomment the following lines, if your model was trained with deeplab-caffe
# or have the same reshape method
# const = Const(graph, {'value': np.array(-1),
# 'name': node.name + 'zoom_factor_deeplab-caffe_sub_const'}).create_node()
# sub = Add(graph, {'name': node.name + 'zoom_factor_deeplab-caffe_sub'}).create_node()
# add.out_port(0).connect(sub.in_port(0))
# const.out_port(0).connect(sub.in_port(1))
#
# const = Const(graph, {'value': np.array(zoom_factor - 1),
# 'name': node.name + 'zoom_factor_deeplab-caffe_mul_const'}).create_node()
# mul = Mul(graph, {'name': node.name + 'zoom_factor_deeplab-caffe_mul'}).create_node()
# sub.out_port(0).connect(mul.in_port(0))
# const.out_port(0).connect(mul.in_port(1))
#
# sum = Add(graph, {'name': node.name + 'zoom_factor_deeplab-caffe_sum'}).create_node()
# add.out_port(0).connect(sum.in_port(0))
# mul.out_port(0).connect(sum.in_port(1))
#
# node.add_input_port(1, skip_if_exist=True)
# assert node.in_port(1).disconnected()
# sum.out_port(0).connect(node.in_port(1))
# Comment out the following lines if you use the reshape method from previous section
const = Const(graph, {'value': np.array(zoom_factor),
'name': node.name + '/zoom_factor_mul_const'}).create_node()
mul = Mul(graph, {'name': node.name + '/zoom_factor_mul'}).create_node()
add.out_port(0).connect(mul.in_port(0))
const.out_port(0).connect(mul.in_port(1))
node.add_input_port(1, skip_if_exist=True)
assert node.in_port(1).disconnected()
mul.out_port(0).connect(node.in_port(1))
elif node.soft_get('width') != 0 and node.soft_get('height') != 0:
const = Const(graph, {'value': np.array([node.height, node.width])}).create_node()
node.add_input_port(1, skip_if_exist=True)
assert node.in_port(1).disconnected()
const.out_port(0).connect(node.in_port(1))
elif node.soft_get('shrink_factor') != 1 and node.soft_get('zoom_factor') != 1:
shrink_factor = node.shrink_factor
zoom_factor = node.zoom_factor
if shrink_factor < 1:
log.error('Shrink factor should be positive in node {}'.format(node.id))
return None
if zoom_factor < 1:
log.error('Zoom factor should be positive in node {}'.format(node.id))
return None
const = Const(graph, {'value': np.array(-1)}).create_node()
sub = Add(graph, {'name': node.name + '/shrink_zoom_factor_sub'}).create_node()
add.out_port(0).connect(sub.in_port(0))
const.out_port(0).connect(sub.in_port(1))
const = Const(graph, {'value': np.array(1 / (shrink_factor + 1))}).create_node()
div = Mul(graph, {'name': node.name + '/shrink_factor_div'}).create_node()
sub.out_port(0).connect(div.in_port(0))
const.out_port(0).connect(div.in_port(1))
const = Const(graph, {'value': np.array(-1),
'name': node.name + 'shrink_zoom_factor_sum_const'}).create_node()
sum = Add(graph, {'name': node.name + '/shrink_zoom_factor_sum'}).create_node()
div.out_port(0).connect(sum.in_port(0))
const.out_port(0).connect(sum.in_port(1))
const = Const(graph, {'value': np.array(zoom_factor - 1)}).create_node()
mul = Mul(graph, {'name': node.name + '/zoom_factor_mul'}).create_node()
sum.out_port(0).connect(mul.in_port(0))
const.out_port(0).connect(mul.in_port(1))
sum = Add(graph, {'name': node.name + '/final_shrink_zoom_factor_sum'}).create_node()
div.out_port(0).connect(sum.in_port(0))
mul.out_port(0).connect(sum.in_port(1))
node.add_input_port(1, skip_if_exist=True)
assert node.in_port(1).disconnected()
sum.out_port(0).connect(node.in_port(1))
else:
if node.soft_get('fw') == 'caffe':
shape = Shape(graph, {'name': node.name + '/shape'}).create_node()
begin = Const(graph, {'value': np.array([2])}).create_node()
end = Const(graph, {'value': np.array([4])}).create_node()
stride = Const(graph, {'value': np.array([1])}).create_node()
ss = StridedSlice(graph, {'name': node.name + '/ss_0_port', 'begin_mask': np.array([1]),
'end_mask': np.array([0]), 'new_axis_mask': np.array([0]),
'shrink_axis_mask': np.array([0]),
'ellipsis_mask': np.array([0])}).create_node()
source = node.in_port(1).get_connection().get_source()
node.in_port(1).disconnect()
source.connect(shape.in_port(0))
shape.out_port(0).connect(ss.in_port(0))
begin.out_port(0).connect(ss.in_port(1))
end.out_port(0).connect(ss.in_port(2))
stride.out_port(0).connect(ss.in_port(3))
ss.out_port(0).connect(node.in_port(1))
def replace_pattern(self, graph: Graph, match: dict):
node = match['pb']
name = node.soft_get('name', node.id)
graph.graph['cmd_params'].static_shape = False
assert len(node.in_ports()) == 2
begin = Const(graph, {
'value': mo_array([2], dtype=np.int32),
'name': name + '/ss_begin'
}).create_node()
end = Const(graph, {
'value': mo_array([4], dtype=np.int32),
'name': name + '/ss_end'
}).create_node()
stride = Const(graph, {
'value': mo_array([1], dtype=np.int32),
'name': name + '/ss_stride'
}).create_node()
shape_0 = Shape(graph, {'name': name + '/0_port'}).create_node()
ss_0 = StridedSlice(
graph, {
'name': name + '/ss_0_port',
'begin_mask': mo_array([1], dtype=np.int32),
'end_mask': mo_array([0], dtype=np.int32),
'new_axis_mask': mo_array([0], dtype=np.int32),
'shrink_axis_mask': mo_array([0], dtype=np.int32),
'ellipsis_mask': mo_array([0], dtype=np.int32)
}).create_node()
shape_0.out_port(0).connect(ss_0.in_port(0))
begin.out_port(0).connect(ss_0.in_port(1))
end.out_port(0).connect(ss_0.in_port(2))
stride.out_port(0).connect(ss_0.in_port(3))
source = node.in_port(0).get_connection().get_source()
node.in_port(0).disconnect()
source.connect(shape_0.in_port(0))
ss_0.out_port(0).connect(node.in_port(0))
shape_1 = Shape(graph, {'name': name + '/1_port'}).create_node()
ss_1 = StridedSlice(
graph, {
'name': name + '/ss_1_port',
'begin_mask': mo_array([1], dtype=np.int32),
'end_mask': mo_array([0], dtype=np.int32),
'new_axis_mask': mo_array([0], dtype=np.int32),
'shrink_axis_mask': mo_array([0], dtype=np.int32),
'ellipsis_mask': mo_array([0], dtype=np.int32)
}).create_node()
shape_1.out_port(0).connect(ss_1.in_port(0))
begin.out_port(0).connect(ss_1.in_port(1))
end.out_port(0).connect(ss_1.in_port(2))
stride.out_port(0).connect(ss_1.in_port(3))
source = node.in_port(1).get_connection().get_source()
node.in_port(1).disconnect()
source.connect(shape_1.in_port(0))
ss_1.out_port(0).connect(node.in_port(1))
ss_0['force_precision_in_ports'] = {1: 'int64', 2: 'int64', 3: 'int64'}
ss_1['force_precision_in_ports'] = {1: 'int64', 2: 'int64', 3: 'int64'}
node['need_shape_inference'] = True
node['override_output_shape'] = True
node['V10_infer'] = True
unsqueeze = create_op_node_with_second_input(
graph, Unsqueeze, int64_array([0]), {'name': name + '/unsqueeze'})
naked_priorbox_name = name + '/naked_not_unsqueezed'
rename_nodes([(node, naked_priorbox_name), (unsqueeze, name)])
node.out_port(0).get_connection().set_source(unsqueeze.out_port(0))
node.out_port(0).connect(unsqueeze.in_port(0))
def replace_pattern(self, graph: Graph, match: [str, Node]):
node = match['crop']
assert node.has_valid('axis')
node_axis = self.list_to_ndarray(node.axis)
in_shape = node.in_port(0).data.get_shape()
shape_rank = in_shape.size
axis_mask = int64_array(
[1 if i in node_axis else 0 for i in range(shape_rank)])
begin_mask = axis_mask.copy()
end_mask = axis_mask.copy()
ss = StridedSlice(
graph, {
'name': node.soft_get('name', node.id) + '/strided_slice',
'begin_mask': begin_mask,
'end_mask': end_mask,
'new_axis_mask': np.zeros(len(end_mask)),
'shrink_axis_mask': np.zeros(len(end_mask)),
'ellipsis_mask': np.zeros(len(end_mask))
}).create_node()
if len(node.in_nodes()) == 2 and node.has_valid('offset'):
# Crop Type 1
begin = Const(
graph, {
'value':
self.mask_normalizer(shape_rank, node_axis, node.offset),
'name':
ss.name + '/begin'
}).create_node()
shape = Shape(graph, {
'name': ss.name + '/shape_of_crop'
}).create_node()
end = Add(graph, {'name': ss.name + '/end'}).create_node()
node.in_port(1).get_connection().get_source().connect(
shape.in_port(0))
node.in_port(1).disconnect()
shape.out_port(0).connect(end.in_port(0))
begin.out_port(0).connect(end.in_port(1))
elif node.has_valid('dim') and node.has_valid('offset'):
# Crop Type 2
node_dim = self.list_to_ndarray(node.dim)
node_offset = self.list_to_ndarray(node.offset)
assert node_dim.size == node_offset.size == node_axis.size
begin = Const(
graph, {
'value':
self.mask_normalizer(shape_rank, node_axis, node_offset),
'name':
ss.name + '/begin'
}).create_node()
end_values = mo_array(
[node_offset[i] + node_dim[i] for i in range(len(node_dim))])
end = Const(
graph, {
'value':
self.mask_normalizer(shape_rank, node_axis, end_values),
'name':
ss.name + '/end'
}).create_node()
elif node.has_valid('crop_begin') and node.has_valid('crop_end'):
# Crop Type 3
node_crop_begin = self.list_to_ndarray(node.crop_begin)
node_crop_end = self.list_to_ndarray(node.crop_end)
assert len(node_crop_begin) == len(node_crop_end) == len(node_axis)
begin = Const(
graph, {
'value':
self.mask_normalizer(shape_rank, node_axis,
node_crop_begin),
'name':
ss.name + '/begin'
}).create_node()
shape = Shape(graph, {'name': ss.name + '/shape'}).create_node()
end = Add(graph, {'name': ss.name + '/end'}).create_node()
const = Const(
graph, {
'value':
-1 *
self.mask_normalizer(shape_rank, node_axis, node_crop_end),
'name':
ss.name + '/const'
}).create_node()
node.in_port(0).get_connection().get_source().connect(
shape.in_port(0))
shape.out_port(0).connect(end.in_port(0))
const.out_port(0).connect(end.in_port(1))
else:
raise Exception("Unknown type of Crop")
source = node.in_port(0).get_connection().get_source()
stride = Const(
graph, {
'value': np.ones(shape_rank, dtype=np.int64),
'name': ss.name + '/stride'
}).create_node()
source.connect(ss.in_port(0))
begin.out_port(0).connect(ss.in_port(1))
end.out_port(0).connect(ss.in_port(2))
stride.out_port(0).connect(ss.in_port(3))
node.in_port(0).disconnect()
node.out_port(0).get_connection().set_source(ss.out_port(0))
ss['force_precision_in_ports'] = {1: 'int64', 2: 'int64', 3: 'int64'}
