def solve(self, formula_wrapper):
dimacs, clauses, num_to_name, quantifiers = get_qcnf(formula_wrapper)
is_sat, certificate = pydepqbf.solve(quantifiers, clauses)
if is_sat == QDPLL_RESULT_UNSAT:
return unsat, False
model = MyModel({num_to_name[abs(val)]: BoolVal(True) if val > 0 else BoolVal(False) for val in certificate})
return sat, model
def generalize_cube(z3_formula, model, all_vars):
assigned_vars = set([_v for _v in all_vars if model[_v] is not None])
my_model = MyModel({v: model[v] for v in assigned_vars})
go_over_vars = list(assigned_vars)
for _var in go_over_vars:
model_tag = MyModel(dict(my_model.assignment))
model_tag[_var] = BoolVal(is_false(my_model[_var]))
z3_assigned = simplify(
substitute(z3_formula,
*[(_v, model_tag[_v]) for _v in assigned_vars]))
if is_false(z3_assigned):
continue
s = SatSolverSelector.SatSolverCtor()
res = s.add(z3_assigned)
if res and s.check():
assigned_vars -= {_var}
my_model.unassign(_var)
return my_model
def solve(self, formula):
dimacs, clauses, num_to_name, quantifiers = get_qcnf(formula)
print 'BEFORE CAEQ'
qdimacs = to_qdimacs(dimacs, quantifiers)
path = '__qdimacs.txt'
to_file(path, qdimacs)
out_path = 'qdimacs_out.txt'
cmd_do_caqe = '{caqe} {input} --qdo > {out}'.format(caqe=CAQE_PATH, input=path, out=out_path)
os.system(cmd_do_caqe)
is_sat, certificate = time_me(pydepqbf.solve, [quantifiers, clauses], 'QBF:: ')
print 'CAQE ', is_sat, certificate
del_file(path)
del_file(out_path)
if is_sat == QDPLL_RESULT_UNSAT:
return unsat, False
model = MyModel({num_to_name[abs(val)]: BoolVal(True) if val > 0 else BoolVal(False) for val in certificate})
return sat, model
width_shift_range=0.1,
height_shift_range=0.1,
zoom_range=(0.9, 1.1),
rescale=1.0/255.0
)
validation_generator = validation_datagen.flow_from_directory(
directory=validation_data_dir,
target_size=(Common.IMG_HEIGHT, Common.IMG_WIDTH),
classes=Common.CLASSES,
class_mode='categorical',
batch_size=batch_size,
shuffle=True
)
model = MyModel.get_model()
model.compile(
loss='categorical_crossentropy',
optimizer=optimizers.Adagrad(),
metrics=['accuracy']
)
model.summary()
if train_weights_file is not None:
model.load_weights(train_weights_file)
print('load_weights:', train_weights_file)
checkpoint = ModelCheckpoint(
epoch_weights_file,
save_weights_only=True,
verbose=1
def model(self):
raw_model = self._solver.get_model()
model = MyModel({self._num_to_name[val]: BoolVal(val > 0) for val in raw_model})
return model