Reinforcement-Learning/Base_agent.cfg at master · vtleonardo/Reinforcement-Learning · GitHub

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
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
######################################################################################################
# Created by Leonardo Viana Teixeira at 20/11/2018                                                   #
######################################################################################################
# Lines starting with # and ; are treated as comments (or with whitespaces+#).
# Lines starting with + are continuation from the previous line.
# It doesn't matter if you use capital letters or not.
# Each line must have only one 'argument = key' pair
# If an argument is not in this file, its default values will be loaded.
# Relative paths to this main folder starts with ..

;#Basic simulation with pong with all commands
;agent_mode = train
;agent_name = DQNPong30
;env = PongNoFrameskip-v4
;include_score = False
;network_model = DQN
;normalize_input = True
;is_recurrent = False
;frame_skip = 4
;num_simul_frames = 1000000
;discount_rate = 0.99
;lr = 1e-4
;epsilon = 1.0
;e_min = 0.02
;decay_mode = linear
;e_lin_decay = 100000
;target_update = 1000
;num_states_stored = 100000
;batch_size = 32
;input_shape = "84,84,1"
;history_size = 4
;num_random_play = 10000
;loss_type = huber
;optimizer = adam
;load_weights = False
;steps_save_weights = 50000
;path_save_weights = ..\Weights
;steps_save_plot = 10000
;path_save_plot = ..\Plot
;to_save_episodes = True
;path_save_episodes = ..\Episodes
;silent_mode = False
;steps_save_episodes = 50
;multi_gpu = False
;gpu_device = 0
;multi_threading = False
;to_render = False
;random_seed = 1

# Simplified version of the basic configuration shown above.
;agent_name = DQNPong30
;num_simul_frames = 1000000
;lr = 1e-4
;e_min = 0.02
;e_lin_decay = 100000
;target_update = 1000
;num_states_stored = 100000
;num_random_play = 10000
;optimizer = adam
;to_save_episodes = True
;random_seed = 1

;# Test a DQN agent in the doom labyrinth_test
;agent_mode = test
;env = Doom
;load_weights = True
;network_model = DQN_regularized
;# Relative paths to this main folder starts with ..
;weights_load_path = ../Weights/Pretrained/Doom/Labyrinth_test/grayh8-fullreg-weights-Doom-labyrinth_test-500000.h5
;config_file_path = ../DoomScenarios/labyrinth_test.cfg
;agent_name = doomh8
;gpu_device = 1
;input_shape = (84,84,1)
;history_size = 8
;to_render = False
;to_save_states = False

;# Test a DRQN agent in the doom labyrinth
;agent_mode = test
;env = Doom
;config_file_path = ../DoomScenarios/labyrinth.cfg
;network_model = DRQN
;is_recurrent = True
;input_shape = "84,84,1"
;history_size = 4
;load_weights = True
;weights_load_path = ../Weights/Pretrained/Doom/Labyrinth/grayh4-LSTM-weights-Doom-labyrinth-5000000.h5
;agent_name = doomh4-lstm-test
;to_render = True
;to_save_states = False

;# Test a DQN agent in the doom labyrinth
;agent_mode = test
;env = Doom
;load_weights = True
;# Relative paths to this main folder starts with ..
;weights_load_path = ../Weights/Pretrained/Doom/Labyrinth/grayh4-weights-Doom-labyrinth-5000000.h5
;agent_name = doomh4
;gpu_device = 1
;network_model = DQN
;input_shape = (84,84,1)
;history_size = 4
;to_render = True
;to_save_states = False

#Training the DRQN agent in the labyrinth map
#agent_mode = train
#env = Doom
#agent_name = grayh4-LSTM
#network_model = DRQN
#is_recurrent = True
#optimizer = adam
#lr = 1e-4
#num_random_play = 50000
#num_states_stored = 250000
#e_lin_decay = 250000
#num_simul_frames = 5000000
#steps_save_weights = 50000
#history_size = 4
#input_shape = (84,84,1)
#to_save_episodes = True
#steps_save_episodes = 100
#multi_threading = True
#gpu_device = 1

;#Training the DRQN agent in the labyrinth_test map
;agent_mode = train
;env = Doom
;agent_name = grayh4-LSTM
;network_model = DRQN
;is_recurrent = True
;load_weights = True
;weights_load_path = ../Weights/Pretrained/Doom/Labyrinth/grayh4-LSTM-weights-Doom-labyrinth-5000000.h5
;config_file_path = ../DoomScenarios/labyrinth_test.cfg
;optimizer = adam
;lr = 1e-4
;num_random_play = 0
;epsilon = 0.05
;num_states_stored = 50000
;num_simul_frames = 500000
;steps_save_weights = 50000
;history_size = 4
;to_save_episodes = True
;steps_save_episodes = 100
;multi_threading = True