def process_config_info_pair(self, config: Configuration, info_dict: dict, budget):
self.budget2obvs[budget]["locks"].append(config.get_array().copy())
info_dict = deepcopy(info_dict)
if config.origin is None:
config.origin = "unknown"
info_dict.update({
"origin": config.origin
})
return config.get_dictionary(), info_dict
def merge_configurations(
partial_configs, # type: List[PartialConfig]
cs: ConfigurationSpace) -> Configuration:
complete = {}
for partial_config in partial_configs:
for param, value in partial_config.config.get_dictionary().items():
param = prefixed_name(partial_config.name, param)
complete[param] = value
config = Configuration(cs, complete)
config.origin = Counter([
p.config.origin for p in partial_configs if not p.is_empty()
]).most_common(1)[0][0]
return config
def get_config(self, budget) -> Tuple[dict, dict]:
# get max_budget
# calc by budget2epm
start_time = time()
max_budget = self.get_available_max_budget()
# initial points
if self.initial_points is not None and self.initial_points_index < len(self.initial_points):
while True:
if self.initial_points_index >= len(self.initial_points):
break
initial_point_dict = self.initial_points[self.initial_points_index]
initial_point = Configuration(self.config_space, initial_point_dict)
self.initial_points_index += 1
initial_point.origin = "User Defined"
if not self.is_config_exist(budget, initial_point):
self.logger.debug(f"Using initial points [{self.initial_points_index - 1}]")
return self.process_config_info_pair(initial_point, {}, budget)
config, config_info = self._get_config(budget, max_budget)
self.register_config(config, budget, start_time)
return config, config_info
def get_config(self, budget):
if self.initial_points is not None and self.initial_points_index < len(
self.initial_points):
initial_point_dict = self.initial_points[self.initial_points_index]
initial_point = Configuration(self.config_space,
initial_point_dict)
initial_point.origin = "User Defined"
self.initial_points_index += 1
return self.process_config_info_pair(initial_point, {}, budget)
max_budget = self.get_available_max_budget()
epm = self.budget2epm[max_budget]
if epm is None:
max_sample = 1000
i = 0
info_dict = {"model_based_pick": False}
while i < max_sample:
i += 1
config = self.config_space.sample_configuration()
add_configs_origin(config, "Initial Design")
if self.is_config_exist(max_budget, config):
self.logger.info(
f"The sample already exists and needs to be resampled. "
f"It's the {i}-th time sampling in random sampling. ")
else:
return self.process_config_info_pair(
config, info_dict, budget)
# todo: 收纳这个代码块
seed = self.rng.randint(1, 8888)
self.config_space.seed()
config = self.config_space.sample_configuration()
add_configs_origin(config, "Initial Design")
info_dict.update({
"sampling_different_samples_failed": True,
"seed": seed
})
return self.process_config_info_pair(config, info_dict, budget)
info_dict = {"model_based_pick": True}
# thompson sampling
if self.use_thompson_sampling:
ts_config, ts_info_dict = self.thompson_sampling(
max_budget, info_dict)
if ts_config is not None:
self.logger.info(
"Using thompson sampling near the dominant samples.")
return self.process_config_info_pair(ts_config, ts_info_dict,
budget)
# 让config_evaluator给所有的随机样本打分
configs = self.config_space.sample_configuration(self.n_samples)
losses, configs_sorted = self.evaluate(configs,
max_budget,
return_loss_config=True)
add_configs_origin(configs_sorted, "Random Search (Sorted)")
if self.use_local_search:
start_points = self.get_local_search_initial_points(
max_budget, 10, configs_sorted) # todo 最后把以前跑过的样本删掉
local_losses, local_configs = self.local_search(
start_points,
max_budget) # todo: 判断start_points 与local_configs的关系
add_configs_origin(local_configs, "Local Search")
concat_losses = np.hstack(
[losses.flatten(), local_losses.flatten()])
concat_configs = configs + local_configs
random_var = self.rng.rand(len(concat_losses))
indexes = np.lexsort((random_var.flatten(), concat_losses))
concat_configs_sorted = [concat_configs[i] for i in indexes]
concat_losses = concat_losses[indexes]
else:
concat_losses, concat_configs_sorted = losses, configs_sorted
# 选取获益最大，且没有出现过的一个配置
# todo: 考虑多worker条件下的config加锁
for i, config in enumerate(concat_configs_sorted):
if self.is_config_exist(max_budget, config):
self.logger.info(
f"The sample already exists and needs to be resampled. "
f"It's the {i}-th time sampling in bayesian sampling. ")
# 超过 max_repeated_samples ， 用TS算法采样
if i >= self.max_repeated_samples and self.use_thompson_sampling:
ts_config, ts_info_dict = self.thompson_sampling(
max_budget, info_dict, True)
return self.process_config_info_pair(
ts_config, ts_info_dict, budget)
else:
return self.process_config_info_pair(config, info_dict, budget)
# todo: 收纳这个代码块
seed = self.rng.randint(1, 8888)
self.config_space.seed(seed)
config = self.config_space.sample_configuration()
add_configs_origin(config, "Initial Design")
info_dict.update({
"sampling_different_samples_failed": True,
"seed": seed
})
return self.process_config_info_pair(config, info_dict, budget)
