-
Notifications
You must be signed in to change notification settings - Fork 1
/
moon-landing.py
233 lines (204 loc) · 8.79 KB
/
moon-landing.py
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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
import argparse
import torch
import gym
from agents import DQNAgent, FixedDQNAgent, DoubleDQNAgent, DuelingDQNAgent
from utils.common import set_seed
from utils.wrapper import EnvWrapper
import os
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--model',
choices=['dqn',
'fixed_dqn',
'double_dqn',
'dueling_dqn'],
default='dqn',
type=str,
help='Model to be used.'
' One between dqn,'
' double_dqn'
' and dueling_dqn (default: %(default)s)')
arg_group = parser.add_mutually_exclusive_group(required=True)
arg_group.add_argument('-t',
'--train',
action='store_true',
help='if present, train the chosen agent;'
' if not, test it')
arg_group.add_argument('-f',
'--file',
type=str,
default=None,
help='Use the weights stored in the given file.'
' Required if in testing mode')
parser.add_argument('-v',
'--verbose',
choices=[0, 1, 2, 3],
default=2,
type=int,
help="Verbose mode."
" One between 0 (no plots, no logs, no video), "
" 1 (yes plots, no logs, no video),"
" 2 (yes plots, yes logs, no video), "
" 3 (yes to all)."
" Considered only in training mode (default: %(default)d)")
parser.add_argument('-r',
'--render',
action='store_true',
help="Render video of the environment."
" Considered only in test mode")
parser.add_argument('-b',
'--batch_size',
type=int,
default=64,
help='size of the batch used to perform training'
' (default: %(default)i)')
parser.add_argument('-m',
'--memory-size',
type=int,
default=int(1e5),
help='maximum size of the replay memory buffer'
' (default: %(default)i)')
parser.add_argument('--gamma',
type=float,
default=0.99,
help='discount rate for the q-values update'
' (default: %(default)0.2f)')
parser.add_argument('--lr',
type=float,
default=.001,
help='learning rate (default: %(default)0.3f)')
parser.add_argument('--episodes',
type=int,
default=2000,
help='Number of episodes to perform training on.'
' Considered only if in training mode'
' (default: %(default)i)')
parser.add_argument('--target-sync-freq',
type=int,
default=500,
help='Number of updates before the network'
' for Q-targets is clones (default: %(default)i)')
parser.add_argument('--learn-freq',
type=int,
default=4,
help='After how many steps'
' the agent should update the weights'
' (default: %(default)i)')
parser.add_argument('--decay',
type=float,
default=0.99,
help='Espilon decay rule for power decay'
' (default: %(default)0.2f)')
args = parser.parse_args()
hyper_params = {
'model': args.model,
'max_timesteps': 1000,
'target_sync_freq': args.target_sync_freq,
'learn_freq': args.learn_freq,
'gamma': args.gamma,
'lr': args.lr,
'num_episodes': args.episodes,
'batch_size': args.batch_size,
'memory_size': args.memory_size,
'verbose': args.verbose,
'max_eps': 1.0,
'min_eps': 0.01,
'decay_rate': args.decay,
# 'decay_rate': 0.998,
'decay_type': 'power_law',
'weights_file': args.file,
'render': args.render
}
print("Chosen parameters:\n")
for key in hyper_params:
print("{}: {}".format(key, hyper_params[key]))
print("\n")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
solved_dir = './solved/'
plots_dir = './plots/'
if not os.path.exists(solved_dir):
os.mkdir(solved_dir)
if not os.path.exists(plots_dir):
os.mkdir(plots_dir)
env = gym.make("LunarLander-v2")
seed = 18
set_seed(seed, env)
env_wrapper = EnvWrapper(env, device)
if args.model == 'dqn':
agent = DQNAgent(env_wrapper.state_dim[0],
env_wrapper.action_dim,
hyper_params['lr'],
hyper_params['gamma'],
hyper_params['memory_size'],
hyper_params['batch_size'],
hyper_params['max_eps'],
hyper_params['min_eps'],
hyper_params['decay_rate'],
device,
decay_type=hyper_params['decay_type'])
elif args.model == "fixed_dqn":
agent = FixedDQNAgent(env_wrapper.state_dim[0],
env_wrapper.action_dim,
hyper_params['lr'],
hyper_params['gamma'],
hyper_params['memory_size'],
hyper_params['batch_size'],
hyper_params['max_eps'],
hyper_params['min_eps'],
hyper_params['decay_rate'],
device,
hyper_params['target_sync_freq'],
decay_type=hyper_params['decay_type'])
elif args.model == "double_dqn":
agent = DoubleDQNAgent(env_wrapper.state_dim[0],
env_wrapper.action_dim,
hyper_params['lr'],
hyper_params['gamma'],
hyper_params['memory_size'],
hyper_params['batch_size'],
hyper_params['max_eps'],
hyper_params['min_eps'],
hyper_params['decay_rate'],
device,
hyper_params['target_sync_freq'],
decay_type=hyper_params['decay_type'])
else: # dueling_dqn
agent = DuelingDQNAgent(env_wrapper.state_dim[0],
env_wrapper.action_dim,
hyper_params['lr'],
hyper_params['gamma'],
hyper_params['memory_size'],
hyper_params['batch_size'],
hyper_params['max_eps'],
hyper_params['min_eps'],
hyper_params['decay_rate'],
device,
hyper_params['target_sync_freq'],
decay_type=hyper_params['decay_type'])
if args.train:
paths = {
'solved_dir': solved_dir,
'plot_dir': plots_dir
}
agent.train(env_wrapper,
paths,
num_episodes=hyper_params['num_episodes'],
max_t=hyper_params['max_timesteps'],
learn_every=hyper_params['learn_freq'],
verbose=hyper_params['verbose'],
avg_period=100,
winning_score=200,
plot_freq=200)
else:
paths = {
'weights': hyper_params['weights_file'],
'plot_dir': plots_dir
}
agent.test(env_wrapper,
paths,
hyper_params['render'],
num_episodes=100,
max_t=1000,
winning_score=200)
if __name__ == "__main__":
main()