Producer¶
TODO
Produce and register samples to try¶
Suggest new parameter sets which optimize the objective.
- class orion.core.worker.producer.Producer(experiment, max_idle_time=None)[source]¶
Produce suggested sets of problem’s parameter space to try out.
It uses an Experiment object to poll for not yet observed trials which have been already evaluated and to register new suggestions (points of the parameter Space) to be evaluated.
- Attributes
is_done
Whether experiment or naive algorithm is done
Methods
backoff
()Wait some time and update algorithm.
produce
(pool_size)Create and register new trials.
register_trial
(new_point)Register a new set of sampled parameters into the DB guaranteeing their uniqueness
register_trials
(new_points)Register new sets of sampled parameters into the DB guaranteeing their uniqueness
suggest
(pool_size)Try suggesting new points with the naive algorithm
update
()Pull all trials to update model with completed ones and naive model with non completed ones.
- property is_done¶
Whether experiment or naive algorithm is done
- register_trial(new_point)[source]¶
Register a new set of sampled parameters into the DB guaranteeing their uniqueness
- Parameters
- new_point: tuple
tuple of values representing the hyperparameters values