-
Notifications
You must be signed in to change notification settings - Fork 0
/
training.py
94 lines (78 loc) · 2.74 KB
/
training.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
'''
Created on Aug 10, 2018
@author: Dragos2811
'''
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import LSTM
from keras.layers import TimeDistributed
from keras.layers import RepeatVector
from keras.models import load_model
from keras.layers.wrappers import Bidirectional
from reader import create
import numpy as np
import random
n_cell1 = 124
n_cell2 = 64
n_epoch = 3
seq_lengh = 64
num_samples = 75000
word_to_ix,ix_to_word,vocab_size,data=create(1,3000)
def no_future_end(n):
for i in range(n,n+seq_lengh):
if data[i] == '/html':
return i+3
return 0
def generate(data,num_samples):
X=np.zeros((num_samples,int(seq_lengh/10),vocab_size))
y=np.zeros((num_samples,int(seq_lengh/10*9), vocab_size)) #out_seq_lenght=1
n = 0
for i in range(num_samples):
if n+seq_lengh >= len(data):
n = random.randint(0,int(len(data)/2))
r = no_future_end(n)
if r != 0:
n = r
for j in range (int(seq_lengh/10)):
X[i][j][word_to_ix[data[n]]] = 1
n += 1
for j in range(int(seq_lengh/10*9)):
y[i][j][word_to_ix[data[n]]] = 1
n += 1
return X,y
try:
model = load_model('./tmp/s%se%sseq%s%sx%s.h5' % (num_samples,n_epoch,seq_lengh,n_cell1,n_cell2))
print("Model loaded")
except ValueError:
print("No model with given hyperparameters found. Creating a new one")
model = Sequential()
model.add(Bidirectional(LSTM(n_cell1, input_shape=(int(seq_lengh/10), vocab_size))))
model.add(RepeatVector(int(seq_lengh/10*9))) #out_seq_lenght
model.add(Bidirectional(LSTM(n_cell2, return_sequences=True)))
model.add(TimeDistributed(Dense(vocab_size, activation= 'softmax' )))
model.compile(loss= 'categorical_crossentropy' , optimizer= 'adam' , metrics=[ 'accuracy' ])
#fit LSTM
X,y=generate(data,num_samples)
model.fit(X, y, epochs=n_epoch, batch_size = 64)
# evaluate LSTM
_,_,_,data=create(3001,4000)
X,y=generate(data,num_samples)
loss, acc = model.evaluate(X, y, verbose=0)
print( 'Loss: %f, Accuracy: %f' % (loss, acc*100))
#saving model
model.save('./tmp/s%se%sseq%s%sx%s.h5' % (num_samples,n_epoch,seq_lengh,n_cell1,n_cell2))
print("model saved")
# evaluate LSTM
_,_,_,data=create(3001,4000)
X,y=generate(data,num_samples)
loss, acc = model.evaluate(X, y, verbose=0)
print( 'Loss: %f, Accuracy: %f' % (loss, acc*100))
#generate html