84 lines
2.6 KiB
Python
84 lines
2.6 KiB
Python
from keras.models import Sequential
|
|
from keras.layers import Dense
|
|
import numpy as np
|
|
import sys, os, string
|
|
|
|
characters = string.printable
|
|
char_indices = dict((c, i) for i, c in enumerate(characters))
|
|
indices_char = dict((i, c) for i, c in enumerate(characters))
|
|
|
|
INPUT_VOCAB_SIZE = len(characters)
|
|
WINDOW_SIZE = 3
|
|
|
|
def encode_one_hot(line):
|
|
line = " " + line + " "
|
|
x = np.zeros((len(line), INPUT_VOCAB_SIZE))
|
|
for i, c in enumerate(line):
|
|
index = char_indices[c] if c in characters else char_indices[' ']
|
|
x[i][index] = 1
|
|
return x
|
|
|
|
def decode_one_hot(x):
|
|
s = []
|
|
for onehot in x:
|
|
one_index = np.argmax(onehot)
|
|
s.append(indices_char[one_index])
|
|
return ''.join(s)
|
|
|
|
def prepare_for_window(x):
|
|
# All slices of size WINDOW_SIZE, sliding through x
|
|
ind = [np.array(np.arange(i, i+WINDOW_SIZE)) for i in range(x.shape[0] - WINDOW_SIZE + 1)]
|
|
ind = np.array(ind, dtype=np.int32)
|
|
x_window = x[ind]
|
|
# Reshape it back to a 2-d tensor
|
|
return x_window.reshape(x_window.shape[0], x_window.shape[1]*x_window.shape[2])
|
|
|
|
def normalization_layer_set_weights(n_layer):
|
|
wb = []
|
|
w = np.zeros((WINDOW_SIZE*INPUT_VOCAB_SIZE, INPUT_VOCAB_SIZE))
|
|
b = np.zeros((INPUT_VOCAB_SIZE))
|
|
# Let lower case letters go through
|
|
for c in string.ascii_lowercase:
|
|
i = char_indices[c]
|
|
w[INPUT_VOCAB_SIZE+i, i] = 1
|
|
# Map capitals to lower case
|
|
for c in string.ascii_uppercase:
|
|
i = char_indices[c]
|
|
il = char_indices[c.lower()]
|
|
w[INPUT_VOCAB_SIZE+i, il] = 1
|
|
# Map all non-letters to space
|
|
sp_idx = char_indices[' ']
|
|
non_letters = [c for c in list(characters) if c not in list(string.ascii_letters)]
|
|
for c in non_letters:
|
|
i = char_indices[c]
|
|
w[INPUT_VOCAB_SIZE+i, sp_idx] = 1
|
|
# Map single letters to space
|
|
for c in non_letters:
|
|
i = char_indices[c]
|
|
w[i, sp_idx] = 0.75
|
|
w[INPUT_VOCAB_SIZE*2+i, sp_idx] = 0.75
|
|
|
|
wb.append(w)
|
|
wb.append(b)
|
|
n_layer.set_weights(wb)
|
|
return n_layer
|
|
|
|
def build_model():
|
|
# Normalize characters using a dense layer
|
|
model = Sequential()
|
|
model.add(Dense(INPUT_VOCAB_SIZE,
|
|
input_shape=(WINDOW_SIZE*INPUT_VOCAB_SIZE,),
|
|
activation='softmax'))
|
|
return model
|
|
|
|
model = build_model()
|
|
model.summary()
|
|
normalization_layer_set_weights(model.layers[0])
|
|
|
|
with open(sys.argv[1]) as f:
|
|
for line in f:
|
|
if line.isspace(): continue
|
|
batch = prepare_for_window(encode_one_hot(line))
|
|
preds = model.predict(batch)
|
|
normal = decode_one_hot(preds)
|
|
print(normal) |