-
Notifications
You must be signed in to change notification settings - Fork 0
/
train_hacker_model.py
92 lines (76 loc) · 2.5 KB
/
train_hacker_model.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
from keras.models import Sequential, model_from_json
from keras.optimizers import RMSprop
from keras.layers import Dense, Flatten
from keras import backend as K
from keras.utils import to_categorical
import numpy as np
import pandas as pd
hist_size=1
state_size=3
actions = [0,2,3,4,5]
n_actions = 5
batch_size = 32
gamma = 0.99
# Initialize value function
model = Sequential()
model.add(Flatten(input_shape=(state_size, hist_size)))
model.add(Dense(64))
model.add(Dense(64))
model.add(Dense(n_actions))
# Note: pass in_keras=False to use this function with raw numbers of numpy arrays for testing
def huber_loss(a, b, in_keras=True):
error = a - b
quadratic_term = error*error / 2
linear_term = abs(error) - 1/2
use_linear_term = (abs(error) > 1.0)
if in_keras:
# Keras won't let us multiply floats by booleans, so we explicitly cast the booleans to floats
use_linear_term = K.cast(use_linear_term, 'float32')
return use_linear_term * linear_term + (1-use_linear_term) * quadratic_term
opt = RMSprop(lr=0.00025)
model.compile(loss=huber_loss, optimizer=opt)
# Read data
df = pd.read_csv('data.csv')
for i in range(1000):
X = []
ys = []
for j in range(32):
row = df.sample(1)
_, _, s1, s2, s3, a, r, s1_, s2_, s3_ = row.values[0]
s = np.array([s1, s2, s3])
s_ = np.array([s1_, s2_, s3_])
y = r + gamma * np.amax(model.predict(s_.reshape(1, state_size, hist_size))[0])
target_f = model.predict(s.reshape(1, state_size, hist_size))
target_f[0][actions.index(a)] = y
target_f = np.clip(target_f, -10, 10)
ys.append(target_f)
X.append(s.reshape(1, state_size, hist_size))
X = np.array(X).reshape(len(X), state_size, hist_size)
ys = np.array(ys).reshape(len(ys), n_actions)
hist = model.fit(X, ys, batch_size=batch_size, epochs=1, verbose=0)
if i % 10 == 0:
print(i, hist.history)
model_json = model.to_json()
with open("model.json", "w") as json_file:
json_file.write(model_json)
model.save_weights("model.h5")
# for i in batch_idxs:
# s, a, r, s_, d = D_[i]
#
# y = r
# if not d:
# states_ = [x[3] for x in D_[i-(hist_size - 1):i]]
# states_.append(s_)
# stack_ = np.stack(states_, axis=1).reshape(1, state_size, hist_size)
# y = r + gamma * np.amax(model.predict(stack_)[0])
#
# states = [x[0] for x in D_[i-(hist_size-1):i]]
# states.append(s)
# stack = np.stack(states, axis=1).reshape(1, state_size, hist_size)
# X.append(stack)
#
# # Calculate the target vector
# target_f = model.predict(stack)
# target_f[0][a] = y
# target_f = np.clip(target_f, -10, 10)
# ys.append(target_f)