config.ini

[a3c]
layer_num = 3
hidden_dim = 32
activation_function = torch.relu
last_activation = None
traj_length = 2048
batch_size = 64
trainable_std = True
lr = 3e-4
gamma = 0.98
lambda_ = 0.95
use_gae = True
reward_scaling = 0.1
max_grad_norm = 0.5
asynchronous = True
value_based = False

[dppo]
layer_num = 3
hidden_dim = 32
activation_function = torch.relu
last_activation = None
traj_length = 2048
batch_size = 64
trainable_std = True
lr = 1e-3
gamma = 0.98
lambda_ = 0.95
use_gae = True
reward_scaling = 0.1
asynchronous = False
max_clip = 0.2
critic_coef = 0.5
entropy_coef = 0.01
max_grad_norm = 0.5
train_epoch = 10
value_based = False

[apex]
layer_num = 3
hidden_dim = 64
activation_function = torch.relu
last_activation = None
learner_batch_size = 512
traj_length = 128
trainable_std = False
lr = 3e-4
gamma = 0.99
epsilon = 0.4
alpha = 7
actor_update_cycle = 3
learner_memory_size = 100000
target_update_cycle = 100
value_based = True
n_step = 1

[impala]
layer_num = 3
hidden_dim = 64
activation_function = torch.relu
last_activation = None
trainable_std = True
lr = 1e-4
traj_length = 100
traj_num = 10
train_epoch = 3
learner_memory_size = 2000
entropy_coef = 0.01
gamma = 0.99
rho_bar = 1
c_bar = 1
value_based = False