'''

It is recommended to run this script on GPU, as recurrent

networks are quite computationally intensive.

'''

from __future__ import print_function

from keras.models import Sequential

from keras.layers.core import Dense, Activation, Dropout, Merge

from keras.layers.recurrent import GRU, LSTM, SimpleRNN

from keras.datasets.data_utils import get_file

import numpy as np

import random

import sys

import os

path = os.path.abspath("E:/event_logs/contest/training_log_2.txt")#get_file('encoded_log_wil.txt', origin="E:/Git/sequence_modelling/encoded_log_wil.txt")

text = open(path).read().lower()

print('corpus length:', len(text))

# cut the text in semi-redundant sequences of maxlen characters

step = 1

hidden_units = 150

sentences = []

sentences2 = []

next_chars = []

lines = text.splitlines()

lines = map(lambda x: '{'+x+'}',lines)

maxlen = max(map(lambda x: len(x),lines))

chars = map(lambda x : set(x),lines)

chars = set().union(*chars)

print('total chars:', len(chars))

lines = map(lambda x: x.ljust(maxlen),lines)

char_indices = dict((c, i) for i, c in enumerate(chars))

indices_char = dict((i, c) for i, c in enumerate(chars))

for line in lines:

for i in range(0, len(line), step):

sentences.append(line[0: i])

if (i+1) <= len(line):

sentences2.append(line[i+1:len(line)])

next_chars.append(line[i])

print('Vectorization...')

X = np.zeros((len(sentences), maxlen, len(chars)), dtype=np.bool)

y = np.zeros((len(sentences), len(chars)), dtype=np.bool)

for i, sentence in enumerate(sentences):

for t, char in enumerate(sentence):

if char==' ':

continue

X[i, t, char_indices[char]] = 1

if next_chars[i]==' ':

continue

y[i, char_indices[next_chars[i]]] = 1

X2 = np.zeros((len(sentences), maxlen, len(chars)), dtype=np.bool)

for i, sentence in enumerate(sentences2):

for t, char in enumerate(sentence):

if char==' ':

continue

X2[i, t, char_indices[char]] = 1

# build model

def fork (model, n=2):

return forks

left = Sequential()

left.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars))))

left.add(Dropout(0.4))

right = Sequential()

right.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars)), go_backwards=True))

right.add(Dropout(0.4))

model = Sequential()

model.add(Merge([left, right], mode='sum'))

left = Sequential()

left.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars))))

left.add(Dropout(0.4))

right = Sequential()

right.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars)), go_backwards=True))

right.add(Dropout(0.4))

model = Sequential()

model.add(Merge([left, right], mode='sum'))

left = Sequential()

left.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars))))

left.add(Dropout(0.4))

right = Sequential()

right.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars)), go_backwards=True))

right.add(Dropout(0.4))

model = Sequential()

model.add(Merge([left, right], mode='sum'))

left = Sequential()

left.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars))))

left.add(Dropout(0.4))

right = Sequential()

right.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars)), go_backwards=True))

right.add(Dropout(0.4))

model = Sequential()

model.add(Merge([left, right], mode='sum'))

left = Sequential()

left.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars))))

left.add(Dropout(0.4))

right = Sequential()

right.add(LSTM(hidden_units, init='he_normal', return_sequences=False, input_shape=(maxlen, len(chars)), go_backwards=True))

right.add(Dropout(0.4))

model.add(Dense(len(chars)))

model.add(Activation('softmax'))

model.compile(loss='categorical_crossentropy', optimizer='adam')

print("Train...")

model.fit([X,X2], y, batch_size=maxlen, nb_epoch=1)