/
TrainMaze.py
164 lines (147 loc) · 4.65 KB
/
TrainMaze.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
#coding:UTF-8
import numpy as np
from dqn_dnn.MazeAgent import MazeAgent
from dqn_dnn.MazeModel import MazeModel
from dqn_dnn.DNNPolicyReplayMemory import DNNPolicyReplayMemory,Experience
from parl.algorithms.dqn import DQN
from tqdm import tqdm
from dqn_dnn.MazeEnv import MazeEnv
from parl.utils import logger
import os
def stateToArray(observation):
return np.array([observation//5,observation%5]).astype(np.float32)
def seeMaze(env):
for i in range(0,5):
for j in range(0,5):
pos=i*5+j
if pos==0:
print('S',end=' ')
elif pos==env.row*env.col-1:
print('T',end=' ')
elif pos not in env.wallList:
print('O',end=' ')
else:
print('X',end=' ')
print()
def seeAction(env,agent):
for i in range(0,5):
for j in range(0,5):
pos=i*5+j
if pos not in env.wallList and pos!=env.row*env.col-1:
d=agent.predict(stateToArray(pos))
if d==0:
print('↑',end=' ')
elif d==1:
print('↓',end=' ')
elif d==2:
print('←',end=' ')
else:
print('→',end=' ')
elif pos==env.row*env.col-1:
print('O',end=' ')
else:
print('X',end=' ')
print()
MEMORY_SIZE = int(5e3)
MEMORY_WARMUP_SIZE = MEMORY_SIZE // 5
StateShape=(2,)
UPDATE_FREQ = 2
GAMMA = 0.99
LEARNING_RATE = 1e-3
TOTAL=5e4
batchSize=64
meanReward=0
trainEp=0
def run_train_episode(env, agent, rpm):
global trainEp
global meanReward
total_reward = 0
all_cost = []
state= env.reset2()
step = 0
while True:
step += 1
action = agent.sample(stateToArray(state))
next_state, reward, isOver,_ = env.step(action)
rpm.append(Experience(stateToArray(state), action, reward, isOver,stateToArray(next_state)))
# start training
if rpm.size() > MEMORY_WARMUP_SIZE:
if step % UPDATE_FREQ == 0:
batch_state, batch_action, batch_reward, batch_isOver,batch_next_state = rpm.sample_batch(
batchSize)
cost = agent.learn(batch_state, batch_action, batch_reward,
batch_next_state, batch_isOver)
all_cost.append(float(cost))
total_reward += reward
state = next_state
if isOver:
break
if all_cost:
trainEp+=1
meanReward=meanReward+(total_reward-meanReward)/trainEp
print('trainEpisode:',trainEp)
print('total_reward: {:.3f}, meanReward:{:.3f} mean_cost: {:.3f}'.format(total_reward, meanReward,np.mean(all_cost)))
return total_reward, step
def train():
env = MazeEnv()
rpm = DNNPolicyReplayMemory(max_size=MEMORY_SIZE, state_shape=StateShape)
action_dim = 4
hyperparas = {
'action_dim': action_dim,
'lr': LEARNING_RATE,
'gamma': GAMMA
}
model = MazeModel(act_dim=action_dim)
algorithm = DQN(model, hyperparas)
agent = MazeAgent(algorithm, action_dim)
with tqdm(total=MEMORY_WARMUP_SIZE) as pbar:
while rpm.size() < MEMORY_WARMUP_SIZE:
__, step = run_train_episode(env, agent, rpm)
pbar.update(step)
# train
print('TrainStart!')
total_step = 0
while True:
# start epoch
__, step = run_train_episode(env, agent, rpm)
total_step += step
print('totalStep:{},exploration:{:.3f}'.format(total_step,agent.exploration))
print()
if total_step >= TOTAL:
break
print()
print("训练完毕,每个位置的最佳动作:")
print()
seeMaze(env)
print()
seeAction(env, agent)
save(agent)
def save(agent):
learnDir = os.path.join(logger.get_dir(),'learn_01')
predictDir = os.path.join(logger.get_dir(),'predict_01')
agent.save_params(learnDir,predictDir)
def restore(agent):
learnDir = os.path.join(logger.get_dir(),'learn_01')
predictDir = os.path.join(logger.get_dir(),'predict_01')
logger.info('restore model from {}'.format(learnDir))
agent.load_params(learnDir,predictDir)
def test():
env = MazeEnv()
action_dim = 4
hyperparas = {
'action_dim': action_dim,
'lr': LEARNING_RATE,
'gamma': GAMMA
}
model = MazeModel(act_dim=action_dim)
algorithm = DQN(model, hyperparas)
agent = MazeAgent(algorithm, action_dim)
restore(agent)
print("\n再次加载:")
print()
seeMaze(env)
print()
seeAction(env, agent)
if __name__ == '__main__':
train()
# test()