-
Notifications
You must be signed in to change notification settings - Fork 10
/
config.ini
105 lines (97 loc) · 1.82 KB
/
config.ini
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
[demonstrations_location]
expert_state_location = ./expert_data/hopper_expert_states.npy
expert_action_location = ./expert_data/hopper_expert_actions.npy
expert_next_state_location = ./expert_data/hopper_expert_next_states.npy
expert_done_location = ./expert_data/hopper_expert_done.npy
[airl]
state_only = True
lr = 0.0003
is_airl = True
layer_num = 3
activation_function = torch.tanh
last_activation = None
batch_size = 512
hidden_dim = 64
gamma = 0.99
[vairl]
state_only = True
lr = 0.0003
is_airl = True
epoch = 3
beta = 0
dual_stepsize = 1e-5
mutual_info_constraint = 0.5
batch_size = 512
hidden_dim = 64
z_dim = 4
gamma = 0.99
[gail]
is_airl = False
lr = 0.0003
layer_num = 3
activation_function = torch.tanh
last_activation = torch.sigmoid
batch_size = 512
hidden_dim = 64
[vail]
is_airl = False
lr = 0.0003
epoch = 3
dual_stepsize = 1e-5
mutual_info_constraint = 0.5
batch_size = 512
hidden_dim = 64
z_dim = 4
beta = 0
[eairl]
hidden_dim = 64
lr = 0.0003
beta = 1
gamma = 0.99
state_only = False
layer_num = 3
activation_function = torch.tanh
last_activation = None
is_airl = True
batch_size = 512
update_cycle = 5
i_lambda = 0.001
trainable_std = False
[sqil]
lambda_ = 1
is_airl = False
batch_size = 64
[ppo]
entropy_coef = 1e-2
critic_coef = 0.5
max_grad_norm = 0.5
actor_lr = 0.0003
critic_lr = 0.0003
gamma = 0.99
lambda_ = 0.95
max_clip = 0.2
train_epoch = 10
hidden_dim = 64
batch_size = 64
layer_num = 3
traj_length = 2048
activation_function = torch.tanh
last_activation = None
trainable_std = False
on_policy = True
[sac]
alpha_init = 0.2
gamma = 0.99
q_lr = 3e-4
actor_lr = 3e-4
alpha_lr = 3e-4
soft_update_rate = 0.005
hidden_dim = 256
learn_start_size = 1000
memory_size = 1e+6
batch_size = 64
layer_num = 3
activation_function = torch.relu
last_activation = None
trainable_std = True
on_policy = False