forked from ShuaibinLi/RL_CARLA
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain02.py
180 lines (150 loc) · 6.37 KB
/
train02.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
import argparse
import datetime
import numpy as np
from parl.utils import logger, tensorboard, ReplayMemory
# from parl.utils import logger, ReplayMemory
from env_utils import ParallelEnv, LocalEnv
from torch_base import TorchModel, TorchSAC, TorchAgent # Choose base wrt which deep-learning framework you are using
# from paddle_base import PaddleModel, PaddleSAC, PaddleAgent
from env_config import EnvConfig
WARMUP_STEPS = 2e3
EVAL_EPISODES = 3
MEMORY_SIZE = int(1e4)
BATCH_SIZE = 256
GAMMA = 0.99
TAU = 0.005
ALPHA = 0.2 # determines the relative importance of entropy term against the reward
ACTOR_LR = 3e-4
CRITIC_LR = 3e-4
# Runs policy for 3 episodes by default and returns average reward
def run_evaluate_episodes(agent, env, eval_episodes):
avg_reward = 0.
for k in range(eval_episodes):
obs = env.reset()
done = False
steps = 0
while not done and steps < env._max_episode_steps:
steps += 1
action = agent.predict(obs)
obs, reward, done, _ = env.step(action)
avg_reward += reward
avg_reward /= eval_episodes
return avg_reward
def main():
logger.info("-----------------Carla_SAC-------------------")
logger.set_dir('./{}_train'.format(args.env))
# Parallel environments for training
train_envs_params = EnvConfig['train_envs_params']
env_num = EnvConfig['env_num']
env_list = ParallelEnv(args.env, args.xparl_addr, train_envs_params)
# env for eval
eval_env_params = EnvConfig['eval_env_params']
eval_env = LocalEnv(args.env, eval_env_params)
obs_dim = eval_env.obs_dim
action_dim = eval_env.action_dim
# Initialize model, algorithm, agent, replay_memory
if args.framework == 'torch':
CarlaModel, SAC, CarlaAgent = TorchModel, TorchSAC, TorchAgent
elif args.framework == 'paddle':
CarlaModel, SAC, CarlaAgent = PaddleModel, PaddleSAC, PaddleAgent
model = CarlaModel(obs_dim, action_dim)
algorithm = SAC(
model,
gamma=GAMMA,
tau=TAU,
alpha=ALPHA,
actor_lr=ACTOR_LR,
critic_lr=CRITIC_LR)
agent = CarlaAgent(algorithm)
rpm = ReplayMemory(
max_size=MEMORY_SIZE, obs_dim=obs_dim, act_dim=action_dim)
total_steps = 0
last_save_steps = 0
test_flag = 0
best_reward = 0
obs_list = env_list.reset()
logger.info("----------------env-reset------------------")
while total_steps < args.train_total_steps:
# Train episode
logger.info("-----------------Train episode-------------------")
if rpm.size() < WARMUP_STEPS:
action_list = [
np.random.uniform(-1, 1, size=action_dim)
for _ in range(env_num)
]
else:
action_list = [agent.sample(obs) for obs in obs_list]
next_obs_list, reward_list, done_list, info_list = env_list.step(
action_list)
# Store data in replay memory
for i in range(env_num):
rpm.append(obs_list[i], action_list[i], reward_list[i],
next_obs_list[i], done_list[i])
obs_list = env_list.get_obs()
total_steps = env_list.total_steps
# Train agent after collecting sufficient data
logger.info("-----------------Train agent after collecting sufficient data-------------------")
if rpm.size() >= WARMUP_STEPS:
batch_obs, batch_action, batch_reward, batch_next_obs, batch_terminal = rpm.sample_batch(
BATCH_SIZE)
agent.learn(batch_obs, batch_action, batch_reward, batch_next_obs,
batch_terminal)
# Save agent
logger.info("-----------------save agent-------------------")
if total_steps > int(1e5) and total_steps > last_save_steps + int(1e4):
agent.save('./{}_model/step_{}_model.ckpt'.format( # 模型存储路径
args.framework, total_steps))
print('model saved')
last_save_steps = total_steps
# print('last_save_steps: ', last_save_steps)
#add current time
# print('current time: ', time.strftime("%Y-%m-%d %H:%M:%S", time.localtime()))
now = datetime.datetime.now()
print('last_save_steps: ', last_save_steps, ' (', now.strftime('%Y-%m-%d %H:%M:%S'), ')')
# Evaluate episode
logger.info("-----------------evaluate-------------------")
if (total_steps + 1) // args.test_every_steps >= test_flag:
while (total_steps + 1) // args.test_every_steps >= test_flag:
test_flag += 1
avg_reward = run_evaluate_episodes(agent, eval_env, EVAL_EPISODES)
tensorboard.add_scalar('eval/episode_reward', avg_reward,
total_steps)
logger.info(
'Total steps {}, Evaluation over {} episodes, Average reward: {}'
.format(total_steps, EVAL_EPISODES, avg_reward))
if avg_reward > best_reward:
best_reward = avg_reward
best_model_path = './{}_model/{}_best.ckpt'.format(args.framework, args.env)
agent.save(best_model_path)
print('best model saved')
logger.info('Saved best model to {}'.format(best_model_path))
# avg_reward = run_evaluate_episodes(agent, eval_env, EVAL_EPISODES)
# if avg_reward > best_reward:
# best_reward = avg_reward
# best_model_path = './model_dir/{}_best'.format(args.env)
# agent.save(best_model_path)
# logger.info('Saved best model to {}'.format(best_model_path))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--xparl_addr",
default='localhost:8080',
help='xparl address for parallel training')
parser.add_argument("--env", default="carla-v2")
parser.add_argument(
'--framework',
# default='paddle',
default='torch',
help='choose deep learning framework: torch or paddle')
parser.add_argument(
"--train_total_steps",
default=5e5,
type=int,
help='max time steps to run environment')
parser.add_argument(
"--test_every_steps",
default=1e3,
type=int,
help='the step interval between two consecutive evaluations')
args = parser.parse_args()
main()