def get_dqn_v0_args(str_):
t = [('lr', '5en4'), ('max_timesteps', '1e5'), ('buffer_size', '5e4'),
('exploration_fraction', '1en1'), ('exploration_final_eps', '0x02'),
('train_freq', '100.'), ('batch_size', '32'),
('learning_starts', '1000'), ('gamma', '1x0'),
('target_network_update_freq', '1e5'), ('prioritized_replay', '0'),
('cnn_model', '0'), ('print_freq', '1000')]
da = utils.DefArgs(t)
args = da.process_string(str_)
vs = vars(args)
for k in [
'batch_size', 'buffer_size', 'learning_starts', 'max_timesteps',
'target_network_update_freq', 'prioritized_replay', 'cnn_model',
'print_freq'
]:
vs[k] = int(utils.str_to_float(vs[k]))
vs['prioritized_replay'] = vs['prioritized_replay'] > 0
for k in [
'lr', 'exploration_fraction', 'exploration_final_eps', 'gamma',
'train_freq'
]:
vs[k] = utils.str_to_float(vs[k])
return vs
def get_greedy_v0_args(str_):
t = [('max_episodes', '5e3')]
da = utils.DefArgs(t)
args = da.process_string(str_)
vs = vars(args)
for k in ['max_episodes']:
vs[k] = int(utils.str_to_float(vs[k]))
return vs
def get_other_args(str_):
t = [('seed', '0'), ('social_network_graph_env', '0'), ('num_valid', '50')]
da = utils.DefArgs(t)
args = da.process_string(str_)
vs = vars(args)
for k in ['seed', 'num_valid']:
vs[k] = int(utils.str_to_float(vs[k]))
vs['social_network_graph_env'] = {
'0': 'SocialNetworkGraphEnv-v1'
}[vs['social_network_graph_env']]
return vs
def get_env_args(str_):
t = [('method_name', 'ours'), ('graph_n', '64'), ('edges_n', '200'),
('n_init_a', '10'), ('threshold_q', '.5'), ('num_seeds', '4'),
('reward_version', '0'), ('seed', '0'),
('graph_obs_version', '0')]
da = utils.DefArgs(t)
args = da.process_string(str_)
vs = vars(args)
for k in [
'graph_n', 'edges_n', 'n_init_a', 'num_seeds',
'reward_version', 'seed', 'graph_obs_version'
]:
vs[k] = int(utils.str_to_float(vs[k]))
for k in ['threshold_q']:
vs[k] = utils.str_to_float(vs[k])
assert (vs['threshold_q'] >= 0
and vs['threshold_q'] <= 1) #valid percentage
return vs