-
Notifications
You must be signed in to change notification settings - Fork 8
/
Copy pathtrain_VesNet_RL.py
201 lines (159 loc) · 5.59 KB
/
train_VesNet_RL.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Oct 14 14:41:25 2021
@author: robotics
"""
import numpy as np
import sys
import torch
import torch.nn.functional as F
import torch.optim as optim
from Env import Env_multi_sim_img
from model import VesNet_RL
from collections import deque
import matplotlib.pyplot as plt
LR = 5e-4
max_step=500
n_episodes=3000
gamma=0.99
gae_lambda=1.0
entropy_coef=0.01
value_loss_coef=0.5
max_grad_norm=50
save_every=50
update_every=20
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
def create_configs_rand(num):
configs=[]
r_min=30
r_max=75
for i in range(num):
offset=np.random.rand()*np.pi/2
size_3d=[750,700,450]
r=np.random.randint(r_min+(r_max-r_min)*i/num,r_min+(r_max-r_min)*(i+1)/num)
c_x=350
c_y=np.random.randint(50+r,225)
c=[c_x,c_y]
config=(c,r,size_3d,offset)
configs.append(config)
return configs
configs=create_configs_rand(10)
env=Env_multi_sim_img(configs=configs, num_channels=4)
model = VesNet_RL(env.num_channels, 5, env.num_actions).to(device)
optimizer = optim.Adam(model.parameters(), lr=LR)
scores_window = deque(maxlen=50)
rewards_window = deque(maxlen=50)
rewards_his=[]
smoothend_rewards=[]
a_file = open("VesNet_RL_ckpt/configs.txt", "w")
for row in configs:
a_file.write(str(row)+'\r')
a_file.close()
plt.clf()
plt.ylabel('Score')
plt.xlabel('Episode #')
done_his=deque(maxlen=100)
reward_max=-sys.float_info.max
best_success_rate=0
i_episode=0
done=True
t=0
state = env.reset(randomVessel=False)
reward_sum=0
finish=False
while i_episode<n_episodes:
if i_episode==1500:
for g in optimizer.param_groups:
g['lr'] = 1e-4
elif i_episode==500:
for g in optimizer.param_groups:
g['lr'] = 3e-4
elif i_episode==0:
for g in optimizer.param_groups:
g['lr'] = 5e-4
if done:
cx = torch.zeros(1, 256).float().to(device)
hx = torch.zeros(1, 256).float().to(device)
else:
cx = cx.detach()
hx = hx.detach()
values = []
log_probs = []
rewards = []
entropies = []
for _ in range(0,update_every):
t+=1
value, logit, (hx, cx) = model((state,(hx, cx)))
prob = F.softmax(logit, dim=-1)
log_prob = F.log_softmax(logit, dim=-1)
entropy = -(log_prob * prob).sum(1, keepdim=True)
entropies.append(entropy)
action = prob.multinomial(num_samples=1).detach()
log_prob = log_prob.gather(1, action)
state, reward, finish_ = env.step(int(action.cpu().detach().numpy()))
finish=finish or finish_
done = t >= max_step
values.append(value)
log_probs.append(log_prob)
rewards.append(reward)
reward_sum+=reward
if done:
break
R = torch.zeros(1, 1).to(device)
if not done:
value, _, _ = model((state, (hx, cx)))
R = value.detach()
values.append(R)
policy_loss = 0
value_loss = 0
gae = torch.zeros(1, 1).to(device)
for i in reversed(range(len(rewards))):
R = gamma * R + rewards[i]
advantage = R - values[i]
value_loss = value_loss + 0.5 * advantage.pow(2)
delta_t = rewards[i] + gamma * values[i + 1] - values[i]
gae = gae * gamma * gae_lambda + delta_t
policy_loss = policy_loss - log_probs[i] * gae.detach() - entropy_coef * entropies[i]
optimizer.zero_grad()
(policy_loss + value_loss_coef * value_loss).backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), max_grad_norm)
optimizer.step()
if done:
done_his.append(finish)
scores_window.append(reward_sum)
rewards_window.append(reward_sum)
rewards_his.append(reward_sum)
smoothend_rewards.append(np.mean(scores_window))
if np.mean(rewards_window)>reward_max:
reward_max=np.mean(rewards_window)
torch.save(model.state_dict(), 'VesNet_RL_ckpt/checkpoint.pth')
a_file = open("VesNet_RL_ckpt/best_ckpt.txt", "w")
a_file.write(str([i_episode,reward_max])+'\r')
a_file.close()
if i_episode%save_every==0:
torch.save(model.state_dict(), 'VesNet_RL_ckpt/checkpoint_latest.pth')
success_rate=len(np.where(np.array(done_his)==1)[0])
if success_rate>best_success_rate:
best_success_rate=success_rate
if i_episode>100:
torch.save(model.state_dict(), 'VesNet_RL_ckpt/checkpoint_best_sr.pth')
a_file = open("VesNet_RL_ckpt/best_ckpt_sr.txt", "w")
a_file.write(str([i_episode,best_success_rate])+'\r')
a_file.close()
print('\r Episode %d Average Score: %.2f Score: %.2f Steps: %d Done: %s Best success rate: %d %% Success rate: %d %%\r' %
(i_episode,np.mean(scores_window),reward_sum,t,finish,best_success_rate,success_rate), end='')
if len(rewards_his)>2:
plt.plot([len(rewards_his)-1,len(rewards_his)], [rewards_his[-2],rewards_his[-1]],'b', alpha=0.3)
plt.plot([len(smoothend_rewards)-1,len(smoothend_rewards)], [smoothend_rewards[-2],smoothend_rewards[-1]],'g')
plt.pause(0.05)
t = 0
if i_episode>50:
state = env.reset(randomVessel=True)
else:
state = env.reset(randomVessel=True)
i_episode+=1
reward_sum=0
finish=False
torch.save(model.state_dict(), 'VesNet_RL_ckpt/checkpoint_latest.pth')
plt.show()