-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
245 lines (197 loc) · 7.54 KB
/
train.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
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
import game.wrapped_flappy_bird as game
# from BrainDQN import *
import shutil
import time
import numpy as np
import random
import torch
import torch.nn as nn
import torch.optim as optim
# import PIL.Image as Image
from PIL import Image
from model import QNetwork
from algorithm import DQN
from utils import *
def train_dqn(options):
max_episode = options.max_episode
flappyBird = game.GameState()
print(f'FPS {flappyBird.FPS}')
rpm = ReplayMemory(options.rpm_size, options) # DQN的经验回放池
model = QNetwork()
if options.resume and options.ckpt_path is not None:
print ('load previous model weight: {}'.format(options.ckpt_path))
episode, epsilon = load_checkpoint(options.ckpt_path, model)
else:
epsilon = options.init_e
episode = 0
if options.cuda:
model = model.cuda()
optimizer = optim.Adam(model.parameters(), lr=options.lr)
algorithm = DQN(model, optimizer, epsilon, options)
# 先往经验池里存一些数据,避免最开始训练的时候样本丰富度不够
while len(rpm) < options.rpm_size/4:
run_episode(algorithm, flappyBird, rpm, options)
print(f'observation done {len(rpm)}')
# 开始训练
logname = time.strftime('%Y-%m-%d %M-%I-%S' , time.localtime())
logger = get_logger(f'log/{logname}.log')
best_reward = 0
max_score = 0
begin = time.time()
while episode < max_episode: # 训练max_episode个回合,test部分不计算入episode数量
# train part
reward, loss, score = run_episode(algorithm, flappyBird, rpm, options)
algorithm.epsilon = max(algorithm.final_e, algorithm.epsilon - algorithm.e_decrement)
episode += 1
max_score = max(max_score, score)
if (episode)%10 == 0:
logger.info('episode:[{}/{}]\tscore:{:.3f}\ttrain_reward:{:.5f}\tloss:{:.5f}'.format(
episode, max_episode, score, reward, loss))
# test part
if (episode)%options.evaluate_freq == 0:
eval_reward, score = evaluate(flappyBird, algorithm, options)
mid = time.time()
elapsed = round(mid-begin)
logger.info('episode:[{}/{}]\tscore:{:.3f}\tepsilon:{:.5f}\ttest_reward:{:.5f}\t{}:{}'.format(
episode, max_episode, score, algorithm.epsilon, eval_reward, elapsed//60, elapsed%60))
if eval_reward > best_reward:
save_path = f'ckpt/best_{score}.ckpt'
save_checkpoint({
'episode': episode,
'epsilon': algorithm.epsilon,
'state_dict': model.state_dict(),
}, False, save_path
)
if (episode)%1000 == 0:
save_path = f'ckpt/episode_{episode}.ckpt'
save_checkpoint({
'episode': episode,
'epsilon': algorithm.epsilon,
'state_dict': model.state_dict(),
}, False, save_path
)
# 训练结束,保存模型
save_path = f'ckpt/final_{episode}_{score}.ckpt'
save_checkpoint({
'episode': episode,
'epsilon': algorithm.epsilon,
'state_dict': model.state_dict(),
}, False, save_path)
mid = time.time()
elapsed = round(mid-begin)
logger.info('training completed, {} episiode, {}m {}s'.format(max_episode, elapsed//60, elapsed%60))
print(f'max_score {max_score}')
def run_episode(model, flappyBird, rpm, options):
"""Train DQN
model -- DQN model
flappyBird -- environment
rpm -- replay memory
options -- resume previous model
"""
time_step = 0
total_reward = 0
model.set_train()
rpm.reset()
flappyBird.reset()
action = [1, 0]
o, r, terminal = flappyBird.frame_step(action)
o = preprocess(o)
rpm.store_state(o)
# rpm.append(o, action, r, terminal)
score = 0
loss = 0
while True:
obs = torch.tensor(rpm.current_state).unsqueeze(0)
if options.cuda:
obs = obs.cuda()
action = model.get_action(obs)
# adjust model.epsilon?
score = max(score, flappyBird.score)
o_next, r, terminal = flappyBird.frame_step(action)
total_reward += options.gamma**time_step * r
o_next = preprocess(o_next)
rpm.append(o_next, action, r, terminal)
if time_step % options.learning_freq == 0 and len(rpm) > options.rpm_size/4:
state_batch, action_batch, reward_batch, next_state_batch, done_batch = rpm.sample(options.batch_size)
loss = model.learn(state_batch, action_batch, reward_batch, next_state_batch, done_batch)
model.global_step += 1
if model.global_step % options.update_target_steps == 0:
model.global_step = 0
model.sync_weight()
time_step += 1
if terminal or score > options.max_score:
break
return total_reward, loss, score
def evaluate(flappyBird, model, options):
"""Test the behavor of dqn when training
model -- dqn model
episode -- current training episode
"""
rpm = ReplayMemory(1, options)
model.set_eval()
rewards = []
scores = []
for _ in range(1):
flappyBird.reset()
action = [1, 0]
o, r, terminal = flappyBird.frame_step(action)
o = preprocess(o)
rpm.append(o, action, r, terminal)
time_step, total_reward = 0, 0
score = 0
while True:
prev_o, a, r, o, terminal = rpm.sample(1)
total_reward += options.gamma**time_step * r
action = model.get_optim_action(o)
score = max(score, flappyBird.score)
o, r, terminal = flappyBird.frame_step(action)
if terminal or score > options.max_score:
break
o = preprocess(o)
rpm.append(o, action, r, terminal)
time_step += 1
rewards.append(total_reward.cpu().numpy())
scores.append(score)
return np.mean(rewards), np.mean(scores)
def play_game(options):
"""Play flappy bird with pretrained dqn model
weight -- model file name containing weight of dqn
best -- if the model is best or not
"""
model = QNetwork()
if options.ckpt_path is None:
print ('you should give weight file name.')
return
print ('load previous model weight: {}'.format(options.ckpt_path))
episode, epsilon = load_checkpoint(options.ckpt_path, model)
if options.cuda:
model = model.cuda()
algorithm = DQN(model, optim, epsilon, options)
algorithm.set_eval()
bird_game = game.GameState()
bird_game.FPS = 480
action = [1, 0]
o, r, terminal = bird_game.frame_step(action)
o = preprocess(o)
rpm = ReplayMemory(1, options)
rpm.append(o, action, r, terminal)
start = time.time()
fc = 0
score = 0
while True:
prev_o, a, r, o, terminal = rpm.sample(1)
# q = algorithm(o).cpu().detach().numpy()[0]
score = max(score, bird_game.score)
action = algorithm.get_optim_action(o)
o, r, terminal = bird_game.frame_step(action)
o = preprocess(o)
# img = Image.fromarray((o*255).astype(np.uint8)).convert(mode='L')
# img.save(f'{fc}-{r}-{q.argmax()}.png')
# fc += 1
if terminal or score > options.max_score*2:
break
rpm.append(o, action, r, terminal)
ela = time.time() - start
print(f'Final Score {score}, FPS{bird_game.FPS}, {ela//60}m{ela%60}s')
# if __name__ == "__main__":
# main()