def main():
data_path = os.path.join("..", "data",
"generated_at_" + str(int(time.time())))
outposts, vehicles, graph = utilities.generate_input(6, data_path)
# data_path = os.path.join("..", "data", "dummy_data")
# data_path = os.path.join("..", "data", "triangle_case")
results_path = os.path.join(data_path, "results")
os.makedirs(results_path, exist_ok=True)
outposts, vehicles, graph = data_importer.import_all_data(data_path)
starting_point = 0
start_time = time.time()
dwave_results = dwave_engine.calculate_routes(outposts, vehicles, graph,
starting_point)
end_time = time.time()
calculation_time = end_time - start_time
print("DWave calculation time:", calculation_time)
if dwave_results is not None:
dwave_results.to_csv(os.path.join(results_path, "dwave_routes.csv"),
sep=";",
index_label="route_id")
visualization.plot_solution(dwave_results,
outposts,
results_path,
file_name="dwave_plot")
starting_point = 0
dummy_results = dummy_engine.calculate_routes(outposts, vehicles, graph,
starting_point)
dummy_results.to_csv(os.path.join(results_path, "dummy_routes.csv"),
sep=";",
index_label="route_id")
visualization.plot_solution(dummy_results,
outposts,
results_path,
file_name="dummy_plot")
start_time = time.time()
or_results = ortools_engine.calculate_routes(
outposts,
vehicles,
graph,
starting_point,
calculation_time=calculation_time)
end_time = time.time()
calculation_time = end_time - start_time
print("ORTools calculation time:", calculation_time)
if or_results is not None:
or_results.to_csv(os.path.join(results_path, "ortools_routes.csv"),
sep=";",
index_label="route_id")
visualization.plot_solution(or_results,
outposts,
results_path,
file_name="ortools_plot")
def benchmark_tsp(problem_source, num_tries, ortools_config, dwave_config):
"""Benchmark D-Wave vs ortools for given problems and configuration.
:param problem_source: an interable with problems to solve
:type problem source: iterable of Problem
:param num_tries: how many times to try to solve given problem
:type num_tries: int
:param ortools_config: configuration for ortools. Currently only the key
'calculation_time` is used, but can be extended should we need to add more
parameters.
:type ortools_config; mapping
:param dwave_config: configuration used with D-Wave solver. Can contains any entries
corresponding to keyword arguments used by DWaveEngine.solve method.
:type dwave_config: mapping
:returns: DataFrame with columns: problem_num, try_num, dwave_time, ortools_time,
dwave_cost, ortools_cost. If any of engines failed to provide a solution fo any of the
problem/try a np.nan is used instead of a corresponding cost.
:rtype: :py:class:`pandas.DataFrame`
"""
dwave_engine = DWaveEngine.default()
results = []
ortools_calculation_time = ortools_config.get('calculation_time', 5)
for i, problem in enumerate(problem_source, 1):
for j in range(1, num_tries + 1):
print('Benchmarking problem {0}, try {1}.'.format(i, j))
start = time()
dwave_solution = dwave_engine.solve(problem, **dwave_config)
dwave_time = time() - start
start = time()
ortools_solution = calculate_routes(
outposts=problem.outposts,
vehicles=problem.vehicles,
graph=problem.graph,
starting_point=problem.starting_point,
calculation_time=ortools_calculation_time)
ortools_time = time() - start
record = Record(problem_num=i,
try_num=j,
dwave_time=dwave_time,
ortools_time=ortools_time,
dwave_cost=np.nan if dwave_solution is None else
dwave_solution.total_cost(),
ortools_cost=np.nan if ortools_solution is None
else ortools_solution['cost'].sum())
results.append(record)
return pd.DataFrame(results)
def benchmark_cvrp(problem_source, num_tries, ortools_config, dwave_config):
"""Benchmark D-Wave vs ortools for given problems and configuration.
:param problem_source: an interable with problems to solve
:type problem source: iterable of Problem
:param num_tries: how many times to try to solve given problem
:type num_tries: int
:param ortools_config: configuration for ortools. Currently only the key
'calculation_time` is used, but can be extended should we need to add more
parameters.
:type ortools_config; mapping
:param dwave_config: configuration used with D-Wave solver. Can contains any entries
corresponding to keyword arguments used by DWaveEngine.solve method.
:type dwave_config: mapping
:returns: DataFrame with columns: problem_num, try_num, dwave_time, ortools_time,
dwave_cost, ortools_cost. If any of engines failed to provide a solution fo any of the
problem/try a np.nan is used instead of a corresponding cost.
:rtype: :py:class:`pandas.DataFrame`
"""
dwave_engine = DWaveEngine.default()
qbsolv_engine = QBSolvEngine.default()
results = []
ortools_calculation_time = ortools_config.get('calculation_time', 5)
for i, problem in enumerate(problem_source, 1):
optimal_cost = utilities.compute_all_cvrp_solutions(problem.outposts, problem.vehicles, problem.graph)[0][1]
for j in range(1, num_tries+1):
print('Benchmarking problem {0}, try {1}.'.format(i, j))
number_of_partitions = len(generate_partitions(problem.outposts, problem.vehicles))
start = time()
dwave_solution, number_of_samples, info = dwave_engine.solve(problem, **dwave_config)
dwave_time = time() - start
start = time()
qbsolv_solution = qbsolv_engine.solve(problem, **dwave_config)
qbsolv_time = time() - start
if number_of_samples is not None:
sample_percentage = number_of_samples / dwave_config['num_reads'] * 100
else:
sample_percentage = np.nan
start = time()
ortools_solution = calculate_routes(outposts=problem.outposts,
vehicles=problem.vehicles,
graph=problem.graph,
starting_point=problem.starting_point,
calculation_time=ortools_calculation_time)
ortools_time = time() - start
if ortools_solution['cost'].sum() < optimal_cost:
ortools_solution = None
record = Record(
problem_num=i,
try_num=j,
number_of_partitions=number_of_partitions,
dwave_time=dwave_time,
dwave_qpu_time=info['timing']['total_qpu_time'],
qbsolv_time=qbsolv_time,
ortools_time=ortools_time,
sample_percentage=sample_percentage,
dwave_cost=np.nan if dwave_solution is None else dwave_solution.total_cost(),
qbsolv_cost=np.nan if qbsolv_solution is None else qbsolv_solution.total_cost(),
ortools_cost=np.nan if ortools_solution is None else ortools_solution['cost'].sum(),
optimal_cost=optimal_cost)
results.append(record)
return pd.DataFrame(results)