/
dagger.py
146 lines (116 loc) · 4.58 KB
/
dagger.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
"""
DAgger:
https://arxiv.org/abs/1011.0686
Utilities are from UC Berkeley CS294 homework.
https://github.com/berkeleydeeprlcourse/homework/tree/master/hw1
Usage:
python dagger.py experts/Hopper-v1.pkl Models/behavior_cloning.h5 Hopper-v1 --render --num_rollouts 30
"""
import pickle
import tensorflow as tf
import numpy as np
import tf_util
import gym
import load_policy
import keras
import argparse
import random
from decimal import Decimal
TRAIN_RATIO = 0.8
T = 100
def get_model(observations, actions):
model = keras.Sequential([
keras.layers.Dense(32, activation=tf.nn.relu, input_shape=(1,11)),
keras.layers.Dense(32, activation=tf.nn.relu),
keras.layers.Dense(3,activation='softmax')
])
num_train = int( TRAIN_RATIO*observations.shape[0] )
X_train = observations[:num_train]
Y_train = actions[:num_train]
X_test = observations[num_train:]
Y_test = actions[num_train:]
optimizer = keras.optimizers.SGD(lr=0.01, momentum=0.0, decay=0.0, nesterov=False)
model.compile(optimizer=optimizer, loss='logcosh', metrics=['accuracy'])
model.fit(X_train, Y_train, batch_size=64, nb_epoch=10, verbose=1)
#model.train_on_batch(X_train, Y_train)
return model
def main():
beta = 0.99
parser = argparse.ArgumentParser()
parser.add_argument('expert_policy_file', type=str)
parser.add_argument('agent_policy_file', type=str)
parser.add_argument('envname', type=str)
parser.add_argument('--render', action='store_true')
parser.add_argument('--num_rollouts', type=int, default=20,
help='Number of expert roll outs')
args = parser.parse_args()
print('loading and building expert policy')
policy_fn = load_policy.load_policy(args.expert_policy_file)
print('loaded and built')
with tf.Session() as sess:
tf_util.initialize()
observations = []
expert_actions = []
env = gym.make(args.envname)
#sum_beta = 0
pi = keras.models.load_model(args.agent_policy_file)
print('loaded the agent policy {}'.format(args.agent_policy_file))
returns = []
for i in range(args.num_rollouts):
print('iter', i)
current_observations = []
#current_expert_actions = []
obs = env.reset()
done = False
totalr = 0.
steps = 0
print("beta is {}".format(beta**i))
while not done:
if(random.random() < beta**i):
print("Query expert")
action = policy_fn(obs[None,:])
expert_actions.append(action)
else:
print("Use its own policy")
o = np.array([obs.reshape(1,11)])
action = pi.predict(o)
expert_action = policy_fn(obs[None,:])
expert_actions.append(expert_action)
#print(action.shape)
current_observations.append(obs.reshape(1,11))
#actions.append(action.reshape(1,3))
obs, r, done, _ = env.step(action)
totalr += r
steps += 1
if args.render:
env.render()
if (steps % T)==0:
print("{} steps executed. Update policy".format(steps))
observations = observations + current_observations
pi = get_model(np.array(observations), np.array(expert_actions))
current_observations = []
print("Current rollout has finished. Update policy")
observations = observations + current_observations
pi = get_model(np.array(observations), np.array(expert_actions))
current_observations = []
returns.append(totalr)
print('returns', returns)
print('mean return', np.mean(returns))
print('std of return', np.std(returns))
"""
expert_data = {'observations': np.array(observations),
'actions': np.array(actions)}
obs = expert_data['observations']
act = expert_data['actions']
model = get_model(obs, act)
num_train = int( TRAIN_RATIO*expert_data['observations'].shape[0] )
X_test = obs[num_train:]
Y_test = act[num_train:]
score = model.evaluate(X_test, Y_test, verbose=1)
print("The score of the model is {}".format(score))
print(model.metrics_names)
model.summary()
model.save("Models/behavior_cloning.h5")
"""
if __name__ == '__main__':
main()