def build_cell(self, in_shapes:TensorShapesList, conf_cell:Config,
cell_index:int) ->CellDesc:
stem_shapes, stems = self.build_cell_stems(in_shapes, conf_cell, cell_index)
cell_type = self.get_cell_type(cell_index)
if self.template is None:
node_count = self.get_node_count(cell_index)
in_shape = stem_shapes[0] # input shape to noded is same as cell stem
out_shape = stem_shapes[0] # we ask nodes to keep the output shape same
node_shapes, nodes = self.build_nodes(stem_shapes, conf_cell,
cell_index, cell_type, node_count, in_shape, out_shape)
else:
node_shapes, nodes = self.build_nodes_from_template(stem_shapes, conf_cell, cell_index)
post_op_shape, post_op_desc = self.build_cell_post_op(stem_shapes,
node_shapes, conf_cell, cell_index)
cell_desc = CellDesc(
id=cell_index, cell_type=self.get_cell_type(cell_index),
conf_cell=conf_cell,
stems=stems, stem_shapes=stem_shapes,
nodes=nodes, node_shapes=node_shapes,
post_op=post_op_desc, out_shape=post_op_shape,
trainables_from=self.get_trainables_from(cell_index)
)
# output same shape twice to indicate s0 and s1 inputs for next cell
in_shapes.append([post_op_shape])
return cell_desc
def build_model_stems(self, in_shapes:TensorShapesList,
conf_model_desc:Config)->List[OpDesc]:
# TODO: why do we need stem_multiplier?
# TODO: in original paper stems are always affine
conf_model_stems = self.get_conf_model_stems()
init_node_ch:int = conf_model_stems['init_node_ch']
stem_multiplier:int = conf_model_stems['stem_multiplier']
ops:List[str] = conf_model_stems['ops']
out_channels = init_node_ch*stem_multiplier
conv_params = ConvMacroParams(self.get_ch(in_shapes[-1][0]), # channels of first input tensor
init_node_ch*stem_multiplier)
stems = [OpDesc(name=op_name, params={'conv': conv_params},
in_len=1, trainables=None) \
for op_name in ops]
# get reduction factors done by each stem, typically they should be same but for
# imagenet they can differ
stem_reductions = ModelDescBuilder._stem_reductions(stems)
# Each cell takes input from previous and 2nd previous cells.
# To be consistence we create two outputs for model stems: [[s1, s0], [s0, s1]
# This way when we access first element of each output we get s1, s0.
# Normailly s0==s1 but for networks like imagenet, s0 will have twice the channels
# of s1.
for stem_reduction in stem_reductions:
in_shapes.append([[out_channels, -1, -1.0/stem_reduction, -1.0/stem_reduction]])
return stems
def build_model_pool(self, in_shapes:TensorShapesList, conf_model_desc:Config)\
->OpDesc:
model_post_op = conf_model_desc['model_post_op']
last_shape = in_shapes[-1][0]
in_shapes.append([copy.deepcopy(last_shape)])
return OpDesc(model_post_op,
params={'conv': ConvMacroParams(last_shape[0], last_shape[0])},
in_len=1, trainables=None)