def main():
if len(sys.argv) < 2:
raise ValueError(f"{sys.argv[0]}: expected file location.")
problem = Problem.from_file(sys.argv[1], delimiter=',')
alns = ALNS(default_rng(problem.instance))
for op in D_OPERATORS:
alns.add_destroy_operator(op)
for op in R_OPERATORS:
alns.add_repair_operator(op)
local_search = LocalSearch()
for op in SOLUTION_OPERATORS:
local_search.add_solution_operator(op)
for op in ROUTE_OPERATORS:
local_search.add_route_operator(op)
alns.on_best(local_search)
init = initial_solution()
result = alns.iterate(init, WEIGHTS, DECAY, CRITERION, ITERATIONS)
# noinspection PyTypeChecker
solution: Solution = result.best_state
solution.to_file(f"solutions/oracs_{problem.instance}.csv")
def main():
args = parse_args()
Problem.from_instance(args.experiment, args.instance)
if args.experiment == "tuning":
generator = rnd.default_rng(args.instance)
else:
# E.g. for exp 72 and inst. 1, this becomes 7201. This way, even for
# inst. 100, there will never be overlap between random number streams
# across experiments.
generator = rnd.default_rng(100 * args.experiment + args.instance)
alns = ALNS(generator) # noqa
for operator in DESTROY_OPERATORS:
if args.exclude == operator.__name__:
continue
alns.add_destroy_operator(operator)
for operator in REPAIR_OPERATORS:
if args.exclude == operator.__name__:
continue
alns.add_repair_operator(operator)
if args.exclude == "reinsert_learner":
alns.on_best(reinsert_learner)
init = initial_solution()
criterion = get_criterion(init.objective())
result = alns.iterate(init, WEIGHTS, DECAY, criterion, ITERATIONS)
location = f"experiments/{args.experiment}/{args.instance}-heuristic.json"
result.best_state.to_file(location) # noqa