dense, shallow, monolithic

This commit is contained in:
Crista Lopes
2019-12-30 10:36:16 -08:00
parent 40c292a4f7
commit 78df7731b3

View File

@@ -0,0 +1,90 @@
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)
LINE_SIZE = 80
def encode_one_hot(line):
x = np.zeros((1, LINE_SIZE, INPUT_VOCAB_SIZE))
sp_idx = char_indices[' ']
for i, c in enumerate(line):
index = char_indices[c] if c in characters else sp_idx
x[0][i][index] = 1
# Pad with spaces
for i in range(len(line), LINE_SIZE):
x[0][i][sp_idx] = 1
return x.reshape([1, LINE_SIZE*INPUT_VOCAB_SIZE])
def decode_one_hot(y):
s = []
x = y.reshape([1, LINE_SIZE, INPUT_VOCAB_SIZE])
for onehot in x[0]:
one_index = np.argmax(onehot)
s.append(indices_char[one_index])
return ''.join(s)
def normalization_layer_set_weights(n_layer):
print(n_layer.get_weights()[0].shape)
print(n_layer.get_weights()[1].shape)
wb = []
w = np.zeros((LINE_SIZE*INPUT_VOCAB_SIZE, LINE_SIZE*INPUT_VOCAB_SIZE), dtype=np.float32)
b = np.zeros((LINE_SIZE*INPUT_VOCAB_SIZE), dtype=np.float32)
# Let lower case letters go through
for r in range(0, LINE_SIZE*INPUT_VOCAB_SIZE, INPUT_VOCAB_SIZE):
for c in string.ascii_lowercase:
i = char_indices[c]
w[r+i, r+i] = 1
# Map capitals to lower case
for c in string.ascii_uppercase:
i = char_indices[c]
il = char_indices[c.lower()]
w[r+i, r+il] = 1
# Map all non-letters to space
sp_idx = char_indices[' ']
for c in [c for c in list(string.printable) if c not in list(string.ascii_letters)]:
i = char_indices[c]
w[r+i, r+sp_idx] = 1
# Map single letters to space
previous_c = r-INPUT_VOCAB_SIZE
next_c = r+INPUT_VOCAB_SIZE
for c in [c for c in list(string.printable) if c not in list(string.ascii_letters)]:
i = char_indices[c]
if r > 0 and r < (LINE_SIZE-1)*INPUT_VOCAB_SIZE:
w[previous_c+i, r+sp_idx] = 0.75
w[next_c+i, r+sp_idx] = 0.75
if r == 0:
w[next_c+i, r+sp_idx] = 1.5
if r == (LINE_SIZE-1)*INPUT_VOCAB_SIZE:
w[previous_c+i, r+sp_idx] = 1.5
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()
dense_layer = Dense(LINE_SIZE*INPUT_VOCAB_SIZE,
input_shape=(LINE_SIZE*INPUT_VOCAB_SIZE,),
activation='softmax')
model.add(dense_layer)
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 = encode_one_hot(line)
preds = model.predict(batch)
normal = decode_one_hot(preds)
print(normal)