utils.py

import numpy as np
import ipdb
from theano import tensor as T, config
import os, pickle, RNN

def oneHot( word, nb_class):
 onehot = np.zeros(nb_class, dtype=np.float32)
 onehot[word] = 1
 return onehot

def oneHots(xs, nb_class):
 return np.array([oneHot(x, nb_class) for x in xs]).astype(np.float32)


def t_crossEntropy(p, q):
 return - T.sum(p * T.log(q))

def crossEntropy(ps, qs):
 #ipdb.set_trace()
 return -1* sum([np.dot(p, [np.math.log(i, 2) for i in q]) for p, q in zip(ps, qs)])


def save_everything(saving_path, rnn, metadata):

 if not os.path.exists(saving_path):
 os.mkdir(saving_path)

 print "saving to... {}".format(saving_path)

 rnn_file_name = os.path.join(saving_path, "rnn.pkl")
 metadata_name = os.path.join(saving_path, "metadata")

 rnn.save(rnn_file_name)
 pickle.dump(metadata, open(metadata_name, 'w'))

def load_everything(loading_path):


 rnn_file_name = os.path.join(loading_path, "rnn.pkl")
 metadata_name = os.path.join(loading_path, "metadata")

 # ATTENTION, LSTM ne marche pas, je sais.
 #ipdb.set_trace()
 rnn = None

 #if is_LSTM:
 # rnn = RNN.LSTM()
 #else:
 # rnn = RNN.RNN()
 rnn = RNN.MLP()

 rnn.load(rnn_file_name)

 metadata = pickle.load(open(metadata_name))

 return rnn, metadata


def hotify_minibatch(minibatch, v_size, pad_before=1):
 """
 Makes sure all the sentences in the minibatch are the same length. Also add an empty word at the beginning.
 Plus make the sentences 1-hot.
 :param minibatch: a list of sentences
 :return: a padded list of sentences.
 """
 max_len = max([len(x) for x in minibatch])
 sentences = []

 # ipdb.set_trace()
 for sentence in minibatch:
 sentence = oneHots(sentence, v_size) # one hot representation
 sentence = np.pad(sentence, ((pad_before, max_len - len(sentence)), (0, 0)),
 'constant', constant_values=(0)) # padding to the max length
 sentences.append(sentence)

 sentences = np.array(sentences).astype(config.floatX)
 sentences = sentences.transpose((1, 0, 2))
 return sentences