-
Notifications
You must be signed in to change notification settings - Fork 17
/
Copy pathmain.py
133 lines (118 loc) · 4.66 KB
/
main.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
import gym
import torch as th
import numpy as np
from gumbel_softmax import gumbel_softmax
from model import Critic,Actor
from experience_replay import replay_memory
lr=0.001
tau=0.05
max_t=200
gamma=0.9
memory_size=2000
batchsize=32
warmup=batchsize
env=gym.make('CartPole-v0')
device="cpu"
env=env.unwrapped
n_action=env.action_space.n
n_state=env.observation_space.shape[0]
class DDPG():
def __init__(self):
self.actor=Actor(n_state,n_action).to(device)
self.target_actor=Actor(n_state,n_action).to(device)
self.critic=Critic(n_state,n_action).to(device)
self.target_critic=Critic(n_state,n_action).to(device)
self.memory=replay_memory(memory_size)
self.Aoptimizer=th.optim.Adam(self.actor.parameters(),lr=lr)
self.Coptimizer=th.optim.Adam(self.critic.parameters(),lr=lr)
def choose_action(self,state,eps):
prob=self.actor(th.FloatTensor(state).to(device))
prob=th.nn.functional.softmax(prob,0)
#print(prob)
if np.random.uniform()>eps:
action=th.argmax(prob,dim=0).tolist()
else:
action=np.random.randint(0,n_action)
return action
def actor_learn(self,batch):
b_s=th.FloatTensor(batch[:,0].tolist()).to(device)
b_r=th.FloatTensor(batch[:,1].tolist()).to(device)
b_a=th.FloatTensor(batch[:,2].tolist()).to(device)
differentiable_a=th.nn.functional.gumbel_softmax(th.log(th.nn.functional.softmax(self.actor(b_s),dim=1)),hard=True)
#print(differentiable_a)
#differentiable_a2=th.nn.functional.softmax(th.nn.functional.softmax(self.actor(b_s),dim=1),dim=1)
#index=th.argmax(differentiable_a2,dim=1).unsqueeze(1)
#oh=th.zeros_like(differentiable_a2).scatter_(1,index,1)
#differentiable_a2=(oh-differentiable_a2).detach()+differentiable_a2
loss=-self.critic(b_s,differentiable_a).mean()
self.Aoptimizer.zero_grad()
loss.backward()
self.Aoptimizer.step()
def critic_learn(self,batch):
b_s=th.FloatTensor(batch[:,0].tolist()).to(device)
b_r=th.FloatTensor(batch[:,1].tolist()).to(device)
b_a=th.zeros(batchsize,n_action).scatter_(1,th.LongTensor(batch[:,2].tolist()),1).to(device)
b_s_=th.FloatTensor(batch[:,3].tolist()).to(device)
b_d=th.FloatTensor(batch[:,4].tolist()).to(device)
eval_q=self.critic(b_s,b_a)
next_action=th.nn.functional.softmax(self.target_actor(b_s_),dim=1)
index=th.argmax(next_action,dim=1).unsqueeze(1)
next_action=th.zeros_like(next_action).scatter_(1,index,1).to(device)
print(next_action)
target_q=th.zeros_like(eval_q).to(device)
for i in range(b_d.shape[0]):
target_q[i]=(1-b_d[i,0])*gamma*self.target_critic(b_s_,next_action)[i].detach()+b_r[i]
td_error=eval_q-target_q
loss=(td_error**2).mean()
self.Coptimizer.zero_grad()
loss.backward()
self.Coptimizer.step()
def soft_update(self):
for param,target_param in zip(self.actor.parameters(),self.target_actor.parameters()):
target_param.data.copy_(tau*param.data+(1-tau)*target_param.data)
for param,target_param in zip(self.critic.parameters(),self.target_critic.parameters()):
target_param.data.copy_(tau*param.data+(1-tau)*target_param.data)
for j in range(10):
ddpg=DDPG()
highest=0
for episode in range(300):
s=env.reset()
t=0
total_reward=0
while(t<max_t):
a=ddpg.choose_action(s,0.1)
s_,r,done,_=env.step(a)
total_reward+=r
transition=[s,[r],[a],s_,[done]]
ddpg.memory.store(transition)
if done:
break
s=s_
if(ddpg.memory.size()<warmup):
continue
batch=ddpg.memory.sample(batchsize)
ddpg.critic_learn(batch)
ddpg.actor_learn(batch)
ddpg.soft_update()
t+=1
if episode%10==0:
total_reward=0.0
for i in range(1):
t_s=env.reset()
t_r=0.0
tr=0.0
time=0
while(time<300):
time+=1
t_a=ddpg.choose_action(t_s,0)
ts_,tr,tdone,_=env.step(t_a)
t_r+=tr
if tdone:
break
t_s=ts_
total_reward+=t_r
if total_reward>highest:
highest=total_reward
print("episode:"+format(episode)+",test score:"+format(total_reward))
if(highest>20):
print(format(j+1)+"th round did it")