-
Notifications
You must be signed in to change notification settings - Fork 0
/
hybrid_sac.py
206 lines (140 loc) · 7.58 KB
/
hybrid_sac.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
import gym
import gym_platform
import numpy as np
import matplotlib.pyplot as plt
import random
from agents.agent import Agent
from agents.algorithms.commons.memory import ReplayBuffer
from agents.algorithms.models.sac_model import Policy, SoftQNetwork
from agents.algorithms.commons.utils import state_reduction, action_construction, to_torch_action, gym_to_buffer, to_gym_action
from wrappers.wrapper_gym_platform import ScaledStateWrapper, ScaledParameterisedActionWrapper, PlatformFlattenedActionWrapper
import torch as T
import torch.optim as optim
class HSAC(Agent):
def create_algorithm(self):
"""Create algorithm."""
self.env = gym.make('Platform-v0')
self.env = ScaledStateWrapper(self.env)
self.env = PlatformFlattenedActionWrapper(self.env)
self.env = ScaledParameterisedActionWrapper(self.env)
self.state_size = self.env.observation_space[0].shape[0]
self.action_size = self.env.action_space[0].n
self.rm_f = 5 # nb of features to remove from the state space
self.k_f = self.state_size - self.rm_f #new state size with features removed
self.step_rate_eps = 0.03
self.start_learning = 3000 #in steps
self.alpha = 0.2 #Entropy Regul Coeff
self.gamma = 0.9
self.q_lr = 0.001
self.policy_lr = 0.0001
self.tau = 0.1
self.replay_buffer_size = 10000
self.batch_size = 128
self.memory = ReplayBuffer(self.replay_buffer_size, self.batch_size)
self.target_network_frequency = 1 #TD3 tricks
self.policy_frequency = 1
self.device = T.device("cuda") if T.cuda.is_available() else T.device("cpu")
self.actor = Policy(self.k_f, self.action_size).to(self.device)
self.q1 = SoftQNetwork(self.k_f, self.action_size).to(self.device)
self.q2 = SoftQNetwork(self.k_f, self.action_size).to(self.device)
self.q1_target = SoftQNetwork(self.k_f, self.action_size).to(self.device)
self.q2_target = SoftQNetwork(self.k_f, self.action_size).to(self.device)
self.q1_target.load_state_dict(self.q1.state_dict())
self.q2_target.load_state_dict(self.q2.state_dict())
self.values_optimizer = optim.Adam(list(self.q1.parameters()) + list(self.q2.parameters()), lr=self.q_lr)
self.policy_optimizer = optim.Adam(list(self.actor.parameters()), lr=self.policy_lr)
self._update_target_model()
def train(self):
"""Test algorithm."""
#Initiate variables
episodes = 10000
n_steps = 0
cum_reward_lst = []
mean_cum_rwd_lst = []
for episode in range(episodes):
# Reset environment
state = self.env.reset()
state = state_reduction(state, self.k_f, False)
action = self._act(n_steps, state)
# Initiate variables for each episode
done = False
episode_reward = 0
while not done:
state_, reward, done, _ = self.env.step(action)
state_ = state_reduction(state_, self.k_f, False)
self.memory.remember(state, gym_to_buffer(action), reward, state_, done)
state = state_
if len(self.memory.replay_buffer) > self.batch_size:
self._learn(n_steps)
if self.target_network_frequency % 50 == 0 :
self._update_target_model()
episode_reward += reward
self.target_network_frequency += 1
n_steps += 1
action = self._act(n_steps,state)
cum_reward_lst.append(episode_reward)
if episode % 50 == 0:
mean_r = np.mean(cum_reward_lst[-50:])
print("Episode", episode,"/",episodes, "- Mean Reward:", round(mean_r,2))
mean_cum_rwd_lst.append(mean_r)
# Close the environment
plt.plot(mean_cum_rwd_lst)
plt.legend(["HSAC"])
plt.xlabel("Episode")
plt.ylabel("Mean Cumulated Reward")
plt.show()
self.env.close()
def test(self):
"""Train algorithm."""
print("There is no test yet for {} algorithm, you can only test it.".format(self.name))
def _act(self, n_steps, state):
if n_steps < self.start_learning:
action = self.env.action_space.sample()
action = gym_to_buffer(action)
action = [action[0], action[1:]]
else:
state = T.Tensor(state).to(self.device)
with T.no_grad():
action_c, action_d, _, _, _ = self.actor.sample_normal(state)
action = to_gym_action(action_c, action_d)
return action
def _update_target_model(self):
for param, target_param in zip(self.q1.parameters(), self.q1_target.parameters()):
target_param.data.copy_(self.tau * param.data + (1 - self.tau) * target_param.data)
for param, target_param in zip(self.q2.parameters(), self.q2_target.parameters()):
target_param.data.copy_(self.tau * param.data + (1 - self.tau) * target_param.data)
def _learn(self, n_steps):
states, actions, rewards, next_states, dones = self.memory.sample_hsac()
states = T.Tensor(states).to(self.device)
actions = T.Tensor(actions).to(self.device)
rewards = T.Tensor(rewards).to(self.device)
next_states = T.Tensor(next_states).to(self.device)
dones = T.Tensor(dones).to(self.device)
criterion = T.nn.MSELoss()
with T.no_grad():
next_state_actions_c, _, next_state_log_pi_c, next_state_log_pi_d, next_state_prob_d = self.actor.sample_normal(next_states)
q1_next_target = self.q1_target.forward(next_states, next_state_actions_c)
q2_next_target = self.q2_target.forward(next_states, next_state_actions_c)
min_qf_next_target = next_state_prob_d * (T.min(q1_next_target, q2_next_target) - self.alpha * next_state_prob_d * next_state_log_pi_c - self.alpha * next_state_log_pi_d)
new_q_values = rewards + (self.gamma* (min_qf_next_target.sum(1)).view(-1) * (1 - dones))
s_actions_c, s_actions_d = to_torch_action(actions)
q1_a_values = self.q1.forward(states, s_actions_c).gather(1, s_actions_d.long().view(-1, 1).to(self.device)).squeeze().view(-1)
q2_a_values =self.q2.forward(states, s_actions_c).gather(1, s_actions_d.long().view(-1, 1).to(self.device)).squeeze().view(-1)
q1_loss = criterion(q1_a_values, new_q_values)
q2_loss = criterion(q2_a_values, new_q_values)
q_loss = (q1_loss + q2_loss) / 2
self.values_optimizer.zero_grad()
q_loss.backward()
self.values_optimizer.step()
if n_steps % self.policy_frequency == 0:
for _ in range(self.policy_frequency):
actions_c, actions_d, log_pi_c, log_pi_d, prob_d = self.actor.sample_normal(states)
q1_pi = self.q1.forward(states, actions_c)
q2_pi = self.q2.forward(states, actions_c)
min_q_pi = T.min(q1_pi, q2_pi)
policy_loss_d = (prob_d * (self.alpha * log_pi_d - min_q_pi)).sum(1).mean()
policy_loss_c = (prob_d * (self.alpha * prob_d * log_pi_c - min_q_pi)).sum(1).mean()
policy_loss = policy_loss_d + policy_loss_c
self.policy_optimizer.zero_grad()
policy_loss.backward()
self.policy_optimizer.step()