-
Notifications
You must be signed in to change notification settings - Fork 15
/
run_smarts.py
221 lines (202 loc) · 6.55 KB
/
run_smarts.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
# Created by yingwen at 2019-03-16
import json
import os
from multiprocessing import Process
import gym
import argparse
from malib.agents.agent_factory import *
# from malib.environments import DifferentialGame
from malib.logger.utils import set_logger
from malib.samplers.sampler import SingleSampler, MASampler, SingleSampler
from malib.trainers import SATrainer, MATrainer
from malib.utils.random import set_seed
from gym_hiway.env.list_hiway_env import ListHiWayEnv
from malib.environments.wrappers import Wrapper
import numpy as np
from hiway.agent import Agent, AgentType
agent, observation_space, action_space = Agent.from_type(
AgentType.Standard, max_episode_steps=5000
)
def get_agent_by_type(type_name, i, env, hidden_layer_sizes, max_replay_buffer_size):
if type_name == "SAC":
return get_sac_agent(
env,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "PR2":
return get_pr2_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "PR2S":
return get_pr2_soft_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "ROMMEO":
return get_rommeo_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "DDPG":
return get_ddpg_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "BiCNet":
return get_bicnet_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "CommNet":
return get_commnet_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "MADDPG":
return get_maddpg_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
elif type_name == "MFAC":
return get_mfac_agent(
env,
agent_id=i,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
def train_fixed(seed, agent_setting, env_configs, fullly_centralized):
set_seed(seed)
scenario = env_configs["scenario"]
suffix = f"fixed_play/{scenario}/{agent_setting}/{seed}"
set_logger(suffix)
batch_size = 50
training_steps = 25 * 60000
exploration_steps = 100
max_replay_buffer_size = 1e5
hidden_layer_sizes = (100, 100)
max_path_length = 25
agent_num = env_configs["n_agents"]
raw_env = ListHiWayEnv(env_configs)
env = Wrapper(env=raw_env, action_space=raw_env.action_sapce)
if fullly_centralized:
agent = get_agent_by_type(
agent_setting,
0,
env,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
sampler = SingleSampler(
batch_size=batch_size, max_path_length=max_path_length
)
sampler.initialize(env, agent)
extra_experiences = ["target_actions"]
trainer = SATrainer(
env=env,
agent=agent,
sampler=sampler,
steps=training_steps,
exploration_steps=exploration_steps,
training_interval=10,
extra_experiences=extra_experiences,
batch_size=batch_size,
)
else:
agents = []
for i in range(agent_num):
agents.append(
get_agent_by_type(
agent_setting,
i,
env,
hidden_layer_sizes=hidden_layer_sizes,
max_replay_buffer_size=max_replay_buffer_size,
)
)
sampler = MASampler(
agent_num, batch_size=batch_size, max_path_length=max_path_length
)
sampler.initialize(env, agents)
extra_experiences = ["annealing", "recent_experiences", "target_actions"]
trainer = MATrainer(
env=env,
agents=agents,
sampler=sampler,
steps=training_steps,
exploration_steps=exploration_steps,
training_interval=10,
extra_experiences=extra_experiences,
batch_size=batch_size,
)
trainer.run()
def action_fn(action):
action[0:2] = (action[0:2] + 1.0) / 2.0
return action
def main(args):
env_configs = {
"scenario": args.scenario,
"n_agents": args.n_agents,
"headless": args.headless,
"episode_limit": args.episode_limit,
"visdom": False,
"timestep_sec": 0.1,
"action_function": action_fn,
"action_space": gym.spaces.Box(
low=np.array([0, 0, -1]), high=np.array([1, 1, 1]), dtype=np.float32
),
"agent_type": AgentType.Standard,
"algo": args.algo
}
seed = 1 + int(23122134 / (3 + 1))
fullly_centralized = False
if args.algo in ['BiCNet','CommNet']:
fullly_centralized = True
print(env_configs)
train_fixed(seed, args.algo, env_configs, fullly_centralized=fullly_centralized)
if __name__ == "__main__":
parser = argparse.ArgumentParser("hiway-malib-agent-example")
parser.add_argument(
"--scenario",
default="./scenarios/loop",
help="Either the scenario to run (see scenarios/ for some samples you can use) OR a directory of scenarios to sample from",
type=str,
)
parser.add_argument(
"--n_agents",
default=3,
type=int,
)
parser.add_argument(
"--headless",
default=True,
type=bool,
)
parser.add_argument(
"--episode_limit",
default=1000,
type=int,
)
parser.add_argument(
"--algo",
default='MADDPG',
help="MADDPG, DDPG, PR2, ROMMEO, BiCNet, ROMMEO, MFAC, SAC, etc.",
type=str,
)
args = parser.parse_args()
main(args)