In [6]:
import unicodedata, string, re, random, time, math, torch, torch.nn as nn
from torch.autograd import Variable
from torch import optim
import torch.nn.functional as F
import keras, numpy as np
In [7]:
from keras.preprocessing import sequence
The data for this project is a set of many thousands of English to French translation pairs.
This question on Open Data Stack Exchange pointed me to the open translation site http://tatoeba.org/ which has downloads available at http://tatoeba.org/eng/downloads - and better yet, someone did the extra work of splitting language pairs into individual text files here: http://www.manythings.org/anki/
The English to French pairs are too big to include in the repo, so download to data/eng-fra.txt before continuing. The file is a tab separated list of translation pairs:
I am cold. Je suis froid.We'll need a unique index per word to use as the inputs and targets of the networks later. To keep track of all this we will use a helper class called Lang which has word → index (word2index) and index → word (index2word) dictionaries, as well as a count of each word word2count to use to later replace rare words.
In [8]:
SOS_token = 0
EOS_token = 1
class Lang:
def __init__(self, name):
self.name = name
self.word2index = {}
self.word2count = {}
self.index2word = {0: "SOS", 1: "EOS"}
self.n_words = 2 # Count SOS and EOS
def addSentence(self, sentence):
for word in sentence.split(' '):
self.addWord(word)
def addWord(self, word):
if word not in self.word2index:
self.word2index[word] = self.n_words
self.word2count[word] = 1
self.index2word[self.n_words] = word
self.n_words += 1
else:
self.word2count[word] += 1
The files are all in Unicode, to simplify we will turn Unicode characters to ASCII, make everything lowercase, and trim most punctuation.
In [9]:
# Turn a Unicode string to plain ASCII, thanks to http://stackoverflow.com/a/518232/2809427
def unicodeToAscii(s):
return ''.join(
c for c in unicodedata.normalize('NFD', s)
if unicodedata.category(c) != 'Mn'
)
# Lowercase, trim, and remove non-letter characters
def normalizeString(s):
s = unicodeToAscii(s.lower().strip())
s = re.sub(r"([.!?])", r" \1", s)
s = re.sub(r"[^a-zA-Z.!?]+", r" ", s)
return s
To read the data file we will split the file into lines, and then split lines into pairs. The files are all English → Other Language, so if we want to translate from Other Language → English I added the reverse flag to reverse the pairs.
In [10]:
def readLangs(lang1, lang2, reverse=False):
print("Reading lines...")
# Read the file and split into lines
lines = open('data/%s-%s.txt' % (lang1, lang2)).read().strip().split('\n')
# Split every line into pairs and normalize
pairs = [[normalizeString(s) for s in l.split('\t')] for l in lines]
# Reverse pairs, make Lang instances
if reverse:
pairs = [list(reversed(p)) for p in pairs]
input_lang = Lang(lang2)
output_lang = Lang(lang1)
else:
input_lang = Lang(lang1)
output_lang = Lang(lang2)
return input_lang, output_lang, pairs
Since there are a lot of example sentences and we want to train something quickly, we'll trim the data set to only relatively short and simple sentences. Here the maximum length is 10 words (that includes ending punctuation) and we're filtering to sentences that translate to the form "I am" or "He is" etc. (accounting for apostrophes replaced earlier).
In [11]:
MAX_LENGTH = 10
eng_prefixes = (
"i am ", "i m ",
"he is", "he s ",
"she is", "she s",
"you are", "you re ",
"we are", "we re ",
"they are", "they re "
)
def filterPair(p):
return len(p[0].split(' ')) < MAX_LENGTH and \
len(p[1].split(' ')) < MAX_LENGTH and \
p[1].startswith(eng_prefixes)
def filterPairs(pairs):
return [pair for pair in pairs if filterPair(pair)]
The full process for preparing the data is:
In [13]:
def prepareData(lang1, lang2, reverse=False):
input_lang, output_lang, pairs = readLangs(lang1, lang2, reverse)
print("Read %s sentence pairs" % len(pairs))
pairs = filterPairs(pairs)
print("Trimmed to %s sentence pairs" % len(pairs))
print("Counting words...")
for pair in pairs:
input_lang.addSentence(pair[0])
output_lang.addSentence(pair[1])
print("Counted words:")
print(input_lang.name, input_lang.n_words)
print(output_lang.name, output_lang.n_words)
return input_lang, output_lang, pairs
input_lang, output_lang, pairs = prepareData('eng', 'fra', True)
print(random.choice(pairs))
In [14]:
def indexesFromSentence(lang, sentence):
return [lang.word2index[word] for word in sentence.split(' ')]+[EOS_token]
def variableFromSentence(lang, sentence):
indexes = indexesFromSentence(lang, sentence)
return Variable(torch.LongTensor(indexes).unsqueeze(0))
def variablesFromPair(pair):
input_variable = variableFromSentence(input_lang, pair[0])
target_variable = variableFromSentence(output_lang, pair[1])
return (input_variable, target_variable)
In [15]:
def index_and_pad(lang, dat):
return sequence.pad_sequences([indexesFromSentence(lang, s)
for s in dat], padding='post').astype(np.int64)
In [16]:
fra, eng = list(zip(*pairs))
In [17]:
fra = index_and_pad(input_lang, fra)
eng = index_and_pad(output_lang, eng)
In [18]:
def get_batch(x, y, batch_size=16):
idxs = np.random.permutation(len(x))[:batch_size]
return x[idxs], y[idxs]
In [19]:
class EncoderRNN(nn.Module):
def __init__(self, input_size, hidden_size, n_layers=1):
super(EncoderRNN, self).__init__()
self.hidden_size = hidden_size
self.embedding = nn.Embedding(input_size, hidden_size)
self.gru = nn.GRU(hidden_size, hidden_size, batch_first=True, num_layers=n_layers)
def forward(self, input, hidden):
output, hidden = self.gru(self.embedding(input), hidden)
return output, hidden
# TODO: other inits
def initHidden(self, batch_size):
return Variable(torch.zeros(1, batch_size, self.hidden_size))
In the simplest seq2seq decoder we use only last output of the encoder. This last output is sometimes called the context vector as it encodes context from the entire sequence. This context vector is used as the initial hidden state of the decoder.
At every step of decoding, the decoder is given an input token and hidden state. The initial input token is the start-of-string <SOS> token, and the first hidden state is the context vector (the encoder's last hidden state).
In [20]:
class DecoderRNN(nn.Module):
def __init__(self, hidden_size, output_size, n_layers=1):
super(DecoderRNN, self).__init__()
self.embedding = nn.Embedding(output_size, hidden_size)
self.gru = nn.GRU(hidden_size, hidden_size, batch_first=True, num_layers=n_layers)
# TODO use transpose of embedding
self.out = nn.Linear(hidden_size, output_size)
self.sm = nn.LogSoftmax()
def forward(self, input, hidden):
emb = self.embedding(input).unsqueeze(1)
# NB: Removed relu
res, hidden = self.gru(emb, hidden)
output = self.sm(self.out(res[:,0]))
return output, hidden
If only the context vector is passed betweeen the encoder and decoder, that single vector carries the burden of encoding the entire sentence.
Attention allows the decoder network to "focus" on a different part of the encoder's outputs for every step of the decoder's own outputs. First we calculate a set of attention weights. These will be multiplied by the encoder output vectors to create a weighted combination. The result (called attn_applied in the code) should contain information about that specific part of the input sequence, and thus help the decoder choose the right output words.
Calculating the attention weights is done with another feed-forward layer attn, using the decoder's input and hidden state as inputs. Because there are sentences of all sizes in the training data, to actually create and train this layer we have to choose a maximum sentence length (input length, for encoder outputs) that it can apply to. Sentences of the maximum length will use all the attention weights, while shorter sentences will only use the first few.
In [9]:
class AttnDecoderRNN(nn.Module):
def __init__(self, hidden_size, output_size, n_layers=1, dropout_p=0.1, max_length=MAX_LENGTH):
super(AttnDecoderRNN, self).__init__()
self.hidden_size = hidden_size
self.output_size = output_size
self.n_layers = n_layers
self.dropout_p = dropout_p
self.max_length = max_length
self.embedding = nn.Embedding(self.output_size, self.hidden_size)
self.attn = nn.Linear(self.hidden_size * 2, self.max_length)
self.attn_combine = nn.Linear(self.hidden_size * 2, self.hidden_size)
self.dropout = nn.Dropout(self.dropout_p)
self.gru = nn.GRU(self.hidden_size, self.hidden_size)
self.out = nn.Linear(self.hidden_size, self.output_size)
def forward(self, input, hidden, encoder_output, encoder_outputs):
embedded = self.embedding(input).view(1, 1, -1)
embedded = self.dropout(embedded)
attn_weights = F.softmax(self.attn(torch.cat((embedded[0], hidden[0]), 1)))
attn_applied = torch.bmm(attn_weights.unsqueeze(0), encoder_outputs.unsqueeze(0))
output = torch.cat((embedded[0], attn_applied[0]), 1)
output = self.attn_combine(output).unsqueeze(0)
for i in range(self.n_layers):
output = F.relu(output)
output, hidden = self.gru(output, hidden)
output = F.log_softmax(self.out(output[0]))
return output, hidden, attn_weights
def initHidden(self):
return Variable(torch.zeros(1, 1, self.hidden_size))
Note: There are other forms of attention that work around the length limitation by using a relative position approach. Read about "local attention" in Effective Approaches to Attention-based Neural Machine Translation.
To train we run the input sentence through the encoder, and keep track of every output and the latest hidden state. Then the decoder is given the <SOS> token as its first input, and the last hidden state of the decoder as its first hidden state.
"Teacher forcing" is the concept of using the real target outputs as each next input, instead of using the decoder's guess as the next input. Using teacher forcing causes it to converge faster but when the trained network is exploited, it may exhibit instability.
In [21]:
def train(input_variable, target_variable, encoder, decoder,
encoder_optimizer, decoder_optimizer, criterion, max_length=MAX_LENGTH):
batch_size, input_length = input_variable.size()
target_length = target_variable.size()[1]
encoder_hidden = encoder.initHidden(batch_size).cuda()
encoder_optimizer.zero_grad()
decoder_optimizer.zero_grad()
loss = 0
encoder_output, encoder_hidden = encoder(input_variable, encoder_hidden)
decoder_input = Variable(torch.LongTensor([SOS_token]*batch_size)).cuda()
decoder_hidden = encoder_hidden
for di in range(target_length):
decoder_output, decoder_hidden = decoder(decoder_input, decoder_hidden)
#, encoder_output, encoder_outputs)
targ = target_variable[:, di]
# print(decoder_output.size(), targ.size(), target_variable.size())
loss += criterion(decoder_output, targ)
decoder_input = targ
loss.backward()
encoder_optimizer.step()
decoder_optimizer.step()
return loss.data[0] / target_length
In [22]:
def asMinutes(s):
m = math.floor(s / 60)
s -= m * 60
return '%dm %ds' % (m, s)
def timeSince(since, percent):
now = time.time()
s = now - since
es = s / (percent)
rs = es - s
return '%s (- %s)' % (asMinutes(s), asMinutes(rs))
In [23]:
def trainEpochs(encoder, decoder, n_epochs, print_every=1000, plot_every=100,
learning_rate=0.01):
start = time.time()
plot_losses = []
print_loss_total = 0 # Reset every print_every
plot_loss_total = 0 # Reset every plot_every
encoder_optimizer = optim.RMSprop(encoder.parameters(), lr=learning_rate)
decoder_optimizer = optim.RMSprop(decoder.parameters(), lr=learning_rate)
criterion = nn.NLLLoss().cuda()
for epoch in range(1, n_epochs + 1):
training_batch = get_batch(fra, eng)
input_variable = Variable(torch.LongTensor(training_batch[0])).cuda()
target_variable = Variable(torch.LongTensor(training_batch[1])).cuda()
loss = train(input_variable, target_variable, encoder, decoder, encoder_optimizer,
decoder_optimizer, criterion)
print_loss_total += loss
plot_loss_total += loss
if epoch % print_every == 0:
print_loss_avg = print_loss_total / print_every
print_loss_total = 0
print('%s (%d %d%%) %.4f' % (timeSince(start, epoch / n_epochs), epoch,
epoch / n_epochs * 100, print_loss_avg))
if epoch % plot_every == 0:
plot_loss_avg = plot_loss_total / plot_every
plot_losses.append(plot_loss_avg)
plot_loss_total = 0
showPlot(plot_losses)
In [141]:
# TODO: Make this change during training
teacher_forcing_ratio = 0.5
def train(input_variable, target_variable, encoder, decoder, encoder_optimizer,
decoder_optimizer, criterion, max_length=MAX_LENGTH):
encoder_hidden = encoder.initHidden()
encoder_optimizer.zero_grad()
decoder_optimizer.zero_grad()
input_length = input_variable.size()[0]
target_length = target_variable.size()[0]
encoder_outputs = Variable(torch.zeros(max_length, encoder.hidden_size))
loss = 0
for ei in range(input_length):
encoder_output, encoder_hidden = encoder(input_variable[ei], encoder_hidden)
encoder_outputs[ei] = encoder_output[0][0]
decoder_input = Variable(torch.LongTensor([[SOS_token]]))
decoder_hidden = encoder_hidden
use_teacher_forcing = True if random.random() < teacher_forcing_ratio else False
if use_teacher_forcing:
# Teacher forcing: Feed the target as the next input
for di in range(target_length):
decoder_output, decoder_hidden, decoder_attention = decoder(
decoder_input, decoder_hidden, encoder_output, encoder_outputs)
loss += criterion(decoder_output[0], target_variable[di])
decoder_input = target_variable[di] # Teacher forcing
else:
# Without teacher forcing: use its own predictions as the next input
for di in range(target_length):
decoder_output, decoder_hidden, decoder_attention = decoder(
decoder_input, decoder_hidden, encoder_output, encoder_outputs)
topv, topi = decoder_output.data.topk(1)
ni = topi[0][0]
decoder_input = Variable(torch.LongTensor([[ni]]))
loss += criterion(decoder_output[0], target_variable[di])
if ni == EOS_token:
break
loss.backward()
encoder_optimizer.step()
decoder_optimizer.step()
return loss.data[0] / target_length
In [24]:
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
import numpy as np
%matplotlib inline
def showPlot(points):
plt.figure()
fig, ax = plt.subplots()
loc = ticker.MultipleLocator(base=0.2) # this locator puts ticks at regular intervals
ax.yaxis.set_major_locator(loc)
plt.plot(points)
Evaluation is mostly the same as training, but there are no targets so we simply feed the decoder's predictions back to itself for each step. Every time it predicts a word we add it to the output string, and if it predicts the EOS token we stop there. We also store the decoder's attention outputs for display later.
In [25]:
def evaluate(encoder, decoder, sentence, max_length=MAX_LENGTH):
input_variable = variableFromSentence(input_lang, sentence).cuda()
input_length = input_variable.size()[0]
encoder_hidden = encoder.initHidden(1).cuda()
encoder_output, encoder_hidden = encoder(input_variable, encoder_hidden)
decoder_input = Variable(torch.LongTensor([SOS_token])).cuda()
decoder_hidden = encoder_hidden
decoded_words = []
# decoder_attentions = torch.zeros(max_length, max_length)
for di in range(max_length):
# decoder_output, decoder_hidden, decoder_attention = decoder(
decoder_output, decoder_hidden = decoder(decoder_input, decoder_hidden)
#, encoder_output, encoder_outputs)
# decoder_attentions[di] = decoder_attention.data
topv, topi = decoder_output.data.topk(1)
ni = topi[0][0]
if ni == EOS_token:
decoded_words.append('<EOS>')
break
else:
decoded_words.append(output_lang.index2word[ni])
decoder_input = Variable(torch.LongTensor([ni])).cuda()
return decoded_words,0#, decoder_attentions[:di+1]
In [27]:
def evaluateRandomly(encoder, decoder, n=10):
for i in range(n):
pair = random.choice(pairs)
print('>', pair[0])
print('=', pair[1])
output_words, attentions = evaluate(encoder, decoder, pair[0])
output_sentence = ' '.join(output_words)
print('<', output_sentence)
print('')
In [28]:
#TODO:
# - Test set
# - random teacher forcing
# - attention
# - multi layers
# - bidirectional encoding
In [29]:
hidden_size = 256
encoder1 = EncoderRNN(input_lang.n_words, hidden_size).cuda()
attn_decoder1 = DecoderRNN(hidden_size, output_lang.n_words).cuda()
In [30]:
trainEpochs(encoder1, attn_decoder1, 15000, print_every=500, learning_rate=0.005)
In [107]:
evaluateRandomly(encoder1, attn_decoder1)
A useful property of the attention mechanism is its highly interpretable outputs. Because it is used to weight specific encoder outputs of the input sequence, we can imagine looking where the network is focused most at each time step.
You could simply run plt.matshow(attentions) to see attention output displayed as a matrix, with the columns being input steps and rows being output steps:
In [20]:
output_words, attentions = evaluate(encoder1, attn_decoder1, "je suis trop froid .")
plt.matshow(attentions.numpy())
Out[20]:
For a better viewing experience we will do the extra work of adding axes and labels:
In [21]:
def showAttention(input_sentence, output_words, attentions):
# Set up figure with colorbar
fig = plt.figure()
ax = fig.add_subplot(111)
cax = ax.matshow(attentions.numpy(), cmap='bone')
fig.colorbar(cax)
# Set up axes
ax.set_xticklabels([''] + input_sentence.split(' ') + ['<EOS>'], rotation=90)
ax.set_yticklabels([''] + output_words)
# Show label at every tick
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(1))
plt.show()
def evaluateAndShowAttention(input_sentence):
output_words, attentions = evaluate(encoder1, attn_decoder1, input_sentence)
print('input =', input_sentence)
print('output =', ' '.join(output_words))
showAttention(input_sentence, output_words, attentions)
In [22]:
evaluateAndShowAttention("elle a cinq ans de moins que moi .")
In [23]:
evaluateAndShowAttention("elle est trop petit .")
In [24]:
evaluateAndShowAttention("je ne crains pas de mourir .")
In [25]:
evaluateAndShowAttention("c est un jeune directeur plein de talent .")
In [ ]: