def main():
parser = argparse.ArgumentParser()
parser.add_argument("-ckt",
type=str,
help="name of the ircuit, e.g. c17, no extension")
args = parser.parse_args()
circuit = Circuit(args.ckt)
circuit.read_circuit()
circuit.lev()
#observability() need to follow controllability()
# circuit.controllability()
circuit.SCOAP_CC()
# circuit.observability()
circuit.SCOAP_CO()
circuit.STAFAN_CS(100)
circuit.STAFAN_B()
# circuit.STAFAN(1, num_proc=1)
circuit.co_ob_info()
graph = circuit.gen_graph()
# -*- coding: utf-8 -*- from circuit import Circuit from atpg_v0 import ATPG import pdb ckt = "c17" circuit = Circuit(ckt) circuit.read_circuit() circuit.lev() #observability() need to follow controllability() circuit.controllability() circuit.observability() # circuit.STAFAN_CS(7000) # circuit.STAFAN_B() circuit.STAFAN(10000, num_proc=2) circuit.co_ob_info() graph = circuit.gen_graph()
def load_data_with_model():
"""
This function loads the circuit graph, then returns the features, labels, model, split train/test sets
"""
# Load the circuit
circuit = Circuit(options.circuit)
circuit.read_circuit()
circuit.lev()
circuit.STAFAN(10000, num_proc=8)
circuit.co_ob_info()
graph = circuit.gen_graph()
g = DGLGraph(graph)
# Extract the unique gate types present in the circuit for creating the labels
if options.objective == "level":
all_types = []
for n in circuit.nodes_lev:
n_num_trans = circuit.node_ids.index(n.num)
if n.gtype not in all_types:
all_types.append(n.gtype)
features = np.zeros((len(circuit.nodes_lev), len(all_types)))
else:
raise ValueError('The objective ' + options.objective +
' is not available')
labels = np.zeros(len(circuit.nodes_lev), dtype=np.float32)
for n in circuit.nodes_lev:
n_num_trans = circuit.node_ids.index(n.num)
if options.objective == "level":
# features[n_num_trans, all_types.index(n.gtype)] = 1.0
labels[n_num_trans] = circuit.nodes_lev[n_num_trans].lev
if options.objective == "C1":
labels[n_num_trans] = circuit.nodes_lev[n_num_trans].C1
random_bools = np.random.choice(a=[False, True],
size=(len(circuit.nodes_lev)),
p=[0.2, 0.8])
train_mask = th.BoolTensor(random_bools).cuda()
test_mask = th.BoolTensor(np.invert(random_bools)).cuda()
features = th.FloatTensor(features).cuda()
if options.problem == "regression":
labels = th.FloatTensor(labels).cuda()
if options.model == "VanillaGCN":
net = VanillaGCN(feature_dim=features.shape[1],
output_dim=1,
weight_dim=options.weight_dim,
depth=options.depth).cuda()
if options.model == "LSTMGCN":
net = LSTMGCN(feature_dim=features.shape[1],
output_dim=1,
weight_dim=options.weight_dim,
depth=options.depth).cuda()
if options.problem == "classification":
labels = th.LongTensor(labels).cuda()
if options.model == "VanillaGCN":
net = VanillaGCN(feature_dim=features.shape[1],
output_dim=len(np.unique(labels.cpu().numpy())),
weight_dim=options.weight_dim,
depth=options.depth).cuda()
if options.model == "LSTMGCN":
net = LSTMGCN(feature_dim=features.shape[1],
output_dim=len(np.unique(labels.cpu().numpy())),
weight_dim=options.weight_dim,
depth=options.depth).cuda()
print(np.unique(labels.cpu().numpy()))
return g, features, labels, train_mask, test_mask, net
