Config

The Config of Problem Setup

Config Name

Description

Possible Values

Default

train_problem

Specify the problem suite for training

[‘bbob-10D’, ‘bbob-30D’, ‘bbob-torch-10D’, ‘bbob-torch-30D’, ‘bbob-noisy-10D’, ‘bbob-noisy-30D’, ‘bbob-noisy-torch-10D’, ‘bbob-noisy-torch-30D’, ‘bbob-surrogate-2D’,’bbob-surrogate-5D’,’bbob-surrogate-10D’, ‘hpo-b’, ‘lsgo’, ‘lsgo-torch’, ‘protein’, ‘protein-torch’, ‘uav’, ‘uav-torch’, ‘mmo’, ‘mmo-torch’, ‘wcci2020’, ‘cec2017mto’, ‘moo-synthetic’]

‘bbob’

test_problem

Specify the problem suite for testing, default to be consistent with training

[‘None’, ‘bbob-10D’, ‘bbob-30D’, ‘bbob-torch-10D’, ‘bbob-torch-30D’, ‘bbob-noisy-10D’, ‘bbob-noisy-30D’, ‘bbob-noisy-torch-10D’, ‘bbob-noisy-torch-30D’, ‘bbob-surrogate-2D’,’bbob-surrogate-5D’,’bbob-surrogate-10D’, ‘hpo-b’, ‘lsgo’, ‘lsgo-torch’, ‘protein’, ‘protein-torch’, ‘uav’, ‘uav-torch’, ‘ne’, ‘mmo’, ‘mmo-torch’, ‘wcci2020’, ‘cec2017mto’, ‘moo-synthetic’]

None

train_difficulty

Difficulty level for training problems

[‘all’, ‘easy’, ‘difficult’]

‘easy’

test_difficulty

Difficulty level for testing problems, default to be consistent with training

[‘all’, ‘easy’, ‘difficult’]

None

user_train_problem_list

User-defined training problem list

List of problems

None

user_test_problem_list

User-defined testing problem list

List of problems

None

device

Device to use

[‘cpu’, ‘cuda’]

‘cpu’

The Config of Training Mode

Config Name

Description

Possible Values

Default

max_learning_step

The maximum learning step for training

int value

1500000

max_epoch

Maximum epoch for training

int value

100

train_batch_size

Batch size of train set

int value

1

agent_save_dir

Save your own trained agent model

string path

‘agent_model/train/’

n_checkpoint

Number of training checkpoints

int value

20

train_parallel_mode

The parallel processing method for batch env step in training

[‘dummy’, ‘subproc’, ‘ray’]

‘subproc’

train_mode

Training mode:’single’ fixed learning step, ‘multi’ fixed number of problems encountered

[‘single’, ‘multi’]

‘single’

end_mode

End mode

[‘step’, ‘epoch’]

‘epoch’

no_tb

Disable tensorboard logging

True or False

False

The Config of Testing Mode

Config Name

Description

Possible Values

Default

test_batch_size

Batch size of test set

int value

1

test_parallel_mode

The parallel processing mode for testing

[‘Full’, ‘Baseline_Problem’, ‘Problem_Testrun’, ‘Batch’,’Seriral’]

‘Batch’

test_run

The run number of test

int value

51

rollout_run

The run number of rollout

int value

10

General Parameters

Config Name

Description

Possible Values

Default

seed

Random seed for training and test to fixed agent

int value

3849

full_meta_data

Store the metadata

True or False

False

log_dir

Logging output

string path

‘output/’

maxFEs

Maximum number of evaluation

int value

20000