def get_suggestions(self, params: Dict = None) -> List[Dict]:
if not self.config.num_runs:
raise ValueError("This search strategy requires `num_runs`.")
suggestions = []
params = params or {}
rand_generator = get_random_generator(seed=self.config.seed)
# Validate number of suggestions and total space
all_discrete = True
for v in self.config.params.values():
if v.is_continuous:
all_discrete = False
break
num_runs = self.config.num_runs
if all_discrete:
space = reduce(
mul, [get_length(v) for v in self.config.params.values()])
num_runs = self.config.num_runs if self.config.num_runs <= space else space
while num_runs > 0:
suggestion_params = copy.deepcopy(params)
suggestion_params.update({
k: sample(v, rand_generator=rand_generator)
for k, v in self.config.params.items()
})
suggestion = SuggestionSpec(params=suggestion_params)
if suggestion not in suggestions:
suggestions.append(suggestion)
num_runs -= 1
return [suggestion.params for suggestion in suggestions]
def __init__(self, config, seed=None):
if not isinstance(config, UtilityFunctionConfig):
raise ValueError("Received a non valid configuration.")
self.config = config
self.eps = config.eps
self.kappa = config.kappa
self.acquisition_function = config.acquisition_function
self.random_generator = get_random_generator(seed=seed)
self.gaussian_process = self.get_gaussian_process(
config=config.gaussian_process,
random_generator=self.random_generator)
def get_suggestions(self,
configs: Dict = None,
metrics: Dict = None) -> List[Dict]:
if not self.config.num_runs:
raise ValueError("This search strategy requires `num_runs`.")
suggestions = []
rand_generator = get_random_generator(seed=self.config.seed)
hyperopt_domain = hyperopt.Domain(None,
self._search_space,
pass_expr_memo_ctrl=None)
hyperopt_trials = self._get_previous_observations(
hyperopt_domain=hyperopt_domain, configs=configs, metrics=metrics)
minimize = hyperopt.FMinIter(
self.config.algorithm,
hyperopt_domain,
hyperopt_trials,
max_evals=-1,
rstate=rand_generator,
verbose=0,
)
minimize.catch_eval_exceptions = False
new_ids = minimize.trials.new_trial_ids(self.config.num_runs)
minimize.trials.refresh()
random_state = minimize.rstate.randint(2**31 - 1)
new_trials = self.algorithm(new_ids, minimize.domain, hyperopt_trials,
random_state)
minimize.trials.refresh()
for tid in range(self.config.num_runs):
vals = new_trials[tid]["misc"]["vals"]
suggestion = {}
for param in vals:
observation_value = vals[param][0]
if param in self._param_to_value:
value = self._param_to_value[param][observation_value]
suggestion[param] = value
else:
suggestion[param] = observation_value
suggestions.append(suggestion)
return suggestions