def __init__(self, n_nodes, C_pp, C_p, C, reduction_p, reduction):
"""
Args:
n_nodes: # of intermediate n_nodes
C_pp: C_out[k-2]
C_p : C_out[k-1]
C : C_in[k] (current)
reduction_p: flag for whether the previous cell is reduction cell or not
reduction: flag for whether the current cell is reduction cell or not
"""
super().__init__()
self.reduction = reduction
self.n_nodes = n_nodes
# If previous cell is reduction cell, current input size does not match with
# output size of cell[k-2]. So the output[k-2] should be reduced by preprocessing.
if reduction_p:
self.preproc0 = ops.FactorizedReduce(C_pp, C, affine=False)
else:
self.preproc0 = ops.Conv2dBlock(C_pp, C, 1, 1, 0, affine=False)
self.preproc1 = ops.Conv2dBlock(C_p, C, 1, 1, 0, affine=False)
# generate dag
self.dag = nn.ModuleList()
for i in range(self.n_nodes):
self.dag.append(nn.ModuleList())
for j in range(2 + i): # include 2 input nodes
# reduction should be used only for input node
stride = 2 if reduction and j < 2 else 1
op = ops.MixedOp(C, stride)
self.dag[i].append(op)
def __init__(self, n_big_nodes, cells, start_p, end_p, C):
super().__init__()
self.n_big_nodes = n_big_nodes
self.preproc0 = ops.StdConv(C, C, 1, 1, 0, affine=False)
self.preproc1 = ops.StdConv(C, C, 1, 1, 0, affine=False)
self.DAG = nn.ModuleList()
for i in range(self.n_big_nodes):
self.DAG.append(nn.ModuleList())
if i < 1:
for _ in range(2 + i):
stride = 1
op = ops.MixedOp(C, stride)
self.DAG[i].append(op)
else:
for _ in range(3):
stride = 1
op = ops.MixedOp(C, stride)
self.DAG[i].append(op)
for k in range(start_p, end_p):
self.DAG.append(cells[k])
def __init__(self, in_features, n_nodes):
"""
Args:
n_nodes: # of intermediate n_nodes
"""
super().__init__()
self.n_nodes = n_nodes
self.in_features = in_features
self.out_features = None
# generate dag
self.dag = nn.ModuleList()
for i in range(self.n_nodes):
self.dag.append(nn.ModuleList())
op = ops.MixedOp(in_features)
in_features = op.out_features
self.dag[-1].append(op)
self.out_features = in_features
print(f"new cell(in={self.in_features}, out={self.out_features})")