The goal of this assignment is to train a Word2Vec skip-gram model over Text8 data.
In [1]:
# These are all the modules we'll be using later. Make sure you can import them
# before proceeding further.
%matplotlib inline
from __future__ import print_function
import collections
import math
import numpy as np
import os
import random
import tensorflow as tf
import zipfile
from matplotlib import pylab
from six.moves import range
from six.moves.urllib.request import urlretrieve
from sklearn.manifold import TSNE
Download the data from the source website if necessary.
In [2]:
url = 'http://mattmahoney.net/dc/'
def maybe_download(filename, expected_bytes):
"""Download a file if not present, and make sure it's the right size."""
if not os.path.exists(filename):
filename, _ = urlretrieve(url + filename, filename)
statinfo = os.stat(filename)
if statinfo.st_size == expected_bytes:
print('Found and verified %s' % filename)
else:
print(statinfo.st_size)
raise Exception(
'Failed to verify ' + filename + '. Can you get to it with a browser?')
return filename
filename = maybe_download('text8.zip', 31344016)
Read the data into a string.
In [3]:
def read_data(filename):
"""Extract the first file enclosed in a zip file as a list of words"""
with zipfile.ZipFile(filename) as f:
data = tf.compat.as_str(f.read(f.namelist()[0])).split()
return data
words = read_data(filename)
print('Data size %d' % len(words))
Build the dictionary and replace rare words with UNK token.
In [4]:
words[0:20]
Out[4]:
In [5]:
vocabulary_size = 50000
def build_dataset(words):
count = [['UNK', -1]]
count.extend(collections.Counter(words).most_common(vocabulary_size - 1))
dictionary = dict()
for word, _ in count:
dictionary[word] = len(dictionary)
data = list()
unk_count = 0
for word in words:
if word in dictionary:
index = dictionary[word]
else:
index = 0 # dictionary['UNK']
unk_count = unk_count + 1
data.append(index)
count[0][1] = unk_count
reverse_dictionary = dict(zip(dictionary.values(), dictionary.keys()))
return data, count, dictionary, reverse_dictionary
data, count, dictionary, reverse_dictionary = build_dataset(words)
print('Most common words (+UNK)', count[:5])
print('Sample data', data[:10])
del words # Hint to reduce memory.
Function to generate a training batch for the skip-gram model.
In [6]:
data_index = 0
def generate_batch(batch_size, num_skips, skip_window):
global data_index
assert batch_size % num_skips == 0
assert num_skips <= 2 * skip_window
batch = np.ndarray(shape=(batch_size), dtype=np.int32)
labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32)
span = 2 * skip_window + 1 # [ skip_window target skip_window ]
# Buffer generated from double-ended queue with fast add/remove
buffer = collections.deque(maxlen=span)
for _ in range(span):
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
for i in range(batch_size // num_skips):
target = skip_window # target label at the center of the buffer
targets_to_avoid = [ skip_window ]
for j in range(num_skips):
while target in targets_to_avoid:
target = random.randint(0, span - 1)
targets_to_avoid.append(target)
batch[i * num_skips + j] = buffer[skip_window]
labels[i * num_skips + j, 0] = buffer[target]
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
return batch, labels
print('data:', [reverse_dictionary[di] for di in data[:8]])
for num_skips, skip_window in [(2, 1), (4, 2)]:
data_index = 0
batch, labels = generate_batch(batch_size=8, num_skips=num_skips, skip_window=skip_window)
print('\nwith num_skips = %d and skip_window = %d:' % (num_skips, skip_window))
print(' batch:', [reverse_dictionary[bi] for bi in batch])
print(' labels:', [reverse_dictionary[li] for li in labels.reshape(8)])
Train a skip-gram model.
In [7]:
# vocabulary is size 50000
batch_size = 128
embedding_size = 128 # Dimension of the embedding vector.
skip_window = 1 # How many words to consider left and right.
num_skips = 2 # How many times to reuse an input to generate a label.
# We pick a random validation set to sample nearest neighbors. here we limit the
# validation samples to the words that have a low numeric ID, which by
# construction are also the most frequent.
valid_size = 16 # Random set of words to evaluate similarity on.
valid_window = 100 # Only pick dev samples in the head of the distribution.
valid_examples = np.array(random.sample(range(valid_window), valid_size))
num_sampled = 64 # Number of negative examples to sample.
graph = tf.Graph()
with graph.as_default(), tf.device('/cpu:0'):
# Input data.
train_dataset = tf.placeholder(tf.int32, shape=[batch_size])
train_labels = tf.placeholder(tf.int32, shape=[batch_size, 1])
valid_dataset = tf.constant(valid_examples, dtype=tf.int32)
# Variables.
embeddings = tf.Variable(
tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0))
softmax_weights = tf.Variable(
tf.truncated_normal([vocabulary_size, embedding_size],
stddev=1.0 / math.sqrt(embedding_size)))
softmax_biases = tf.Variable(tf.zeros([vocabulary_size]))
# Model.
# Look up embeddings for inputs.
embed = tf.nn.embedding_lookup(embeddings, train_dataset)
# Compute the softmax loss, using a sample of the negative labels each time.
loss = tf.reduce_mean(
tf.nn.sampled_softmax_loss(weights=softmax_weights, biases=softmax_biases, inputs=embed,
labels=train_labels, num_sampled=num_sampled, num_classes=vocabulary_size))
# Optimizer.
# Note: The optimizer will optimize the softmax_weights AND the embeddings.
# This is because the embeddings are defined as a variable quantity and the
# optimizer's `minimize` method will by default modify all variable quantities
# that contribute to the tensor it is passed.
# See docs on `tf.train.Optimizer.minimize()` for more details.
optimizer = tf.train.AdagradOptimizer(1.0).minimize(loss)
# Compute the similarity between minibatch examples and all embeddings.
# We use the cosine distance:
norm = tf.sqrt(tf.reduce_sum(tf.square(embeddings), 1, keep_dims=True))
normalized_embeddings = embeddings / norm
valid_embeddings = tf.nn.embedding_lookup(
normalized_embeddings, valid_dataset)
similarity = tf.matmul(valid_embeddings, tf.transpose(normalized_embeddings))
In [8]:
num_steps = 100001
with tf.Session(graph=graph) as session:
tf.global_variables_initializer().run()
print('Initialized')
average_loss = 0
for step in range(num_steps):
batch_data, batch_labels = generate_batch(
batch_size, num_skips, skip_window)
feed_dict = {train_dataset : batch_data, train_labels : batch_labels}
_, l = session.run([optimizer, loss], feed_dict=feed_dict)
average_loss += l
if step % 2000 == 0:
if step > 0:
average_loss = average_loss / 2000
# The average loss is an estimate of the loss over the last 2000 batches.
print('Average loss at step %d: %f' % (step, average_loss))
average_loss = 0
# note that this is expensive (~20% slowdown if computed every 500 steps)
if step % 10000 == 0:
sim = similarity.eval()
for i in range(valid_size):
valid_word = reverse_dictionary[valid_examples[i]]
top_k = 8 # number of nearest neighbors
nearest = (-sim[i, :]).argsort()[1:top_k+1]
log = 'Nearest to %s:' % valid_word
for k in range(top_k):
close_word = reverse_dictionary[nearest[k]]
log = '%s %s,' % (log, close_word)
print(log)
final_embeddings = normalized_embeddings.eval()
In [9]:
num_points = 400
tsne = TSNE(perplexity=30, n_components=2, init='pca', n_iter=5000)
two_d_embeddings = tsne.fit_transform(final_embeddings[1:num_points+1, :])
In [10]:
def plot(embeddings, labels):
assert embeddings.shape[0] >= len(labels), 'More labels than embeddings'
pylab.figure(figsize=(15,15)) # in inches
for i, label in enumerate(labels):
x, y = embeddings[i,:]
pylab.scatter(x, y)
pylab.annotate(label, xy=(x, y), xytext=(5, 2), textcoords='offset points',
ha='right', va='bottom')
pylab.show()
words = [reverse_dictionary[i] for i in range(1, num_points+1)]
plot(two_d_embeddings, words)
An alternative to skip-gram is another Word2Vec model called CBOW (Continuous Bag of Words). In the CBOW model, instead of predicting a context word from a word vector, you predict a word from the sum of all the word vectors in its context. Implement and evaluate a CBOW model trained on the text8 dataset.
So our input vector is a set of words and our output is a predicted word. This is the reverse of the skip-gram?
Swap batch and label?
Is batch still a set of [word, label] pairs? Or is it [[word1, word2, word4, word5], label] with the words from the context window? It is the latter.
Training data for skip-gram model is a set of batch vectors and a set of labels, wherein the batch vectors are 128 in length.
Training data for CBOW is a set of words around a current word.
Function to generate batch for CBOW model.
We need to provide some form of padding? Or just loop around the 'data' structure?
In [22]:
data_index = 0
# Batch needs to output batch of input words-label but excluding word in center of window
def generate_batch(batch_size, context_window):
global data_index
context_size = 2 * context_window
batch = np.ndarray(shape=(batch_size, context_size), dtype=np.int32)
labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32)
span = 2 * context_window + 1 # [ context_window target context_window ]
# Buffer generated from double-ended queue with fast add/remove
buffer = collections.deque(maxlen=span)
# Below builds a buffer and initialises
for _ in range(span):
# This is where the data comes in
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
for i in range(batch_size):
target = context_window # target label at the center of the buffer
buffer_list = list(buffer)
batch[i, :] = buffer_list[:target] + buffer_list[target+1:]
labels[i] = buffer[target]
# When you append to a buffer you add to the front and it pops off the end
buffer.append(data[data_index])
data_index = (data_index + 1) % len(data)
return batch, labels
print('data:', [reverse_dictionary[di] for di in data[:8]])
for context_window in [1, 2, 3, 4]:
data_index = 0
batch, labels = generate_batch(batch_size=8, context_window=context_window)
print('\nwith context_window = %d:' % (context_window))
print('\nbatch shape: {0}'.format(batch.shape))
for be in batch:
print(' batch_entry:', [reverse_dictionary[bindex] for bindex in be])
print(' labels:', [reverse_dictionary[li] for li in labels.reshape(8)])
In [24]:
# vocabulary is size 50000
batch_size = 128
embedding_size = 128 # Dimension of the embedding vector.
context_window = 3 # How many words to consider left and right.
# We pick a random validation set to sample nearest neighbors. here we limit the
# validation samples to the words that have a low numeric ID, which by
# construction are also the most frequent.
valid_size = 16 # Random set of words to evaluate similarity on.
valid_window = 100 # Only pick dev samples in the head of the distribution.
valid_examples = np.array(random.sample(range(valid_window), valid_size))
num_sampled = 64 # Number of negative examples to sample.
graph = tf.Graph()
with graph.as_default(), tf.device('/cpu:0'):
# Input data.
train_dataset = tf.placeholder(tf.int32, shape=[batch_size, 2*context_window])
train_labels = tf.placeholder(tf.int32, shape=[batch_size, 1])
valid_dataset = tf.constant(valid_examples, dtype=tf.int32)
# Variables.
embeddings = tf.Variable(
tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0))
softmax_weights = tf.Variable(
tf.truncated_normal([vocabulary_size, embedding_size],
stddev=1.0 / math.sqrt(embedding_size)))
softmax_biases = tf.Variable(tf.zeros([vocabulary_size]))
# Model - with help from here - http://www.thushv.com/natural_language_processing/word2vec-part-2-nlp-with-deep-learning-with-tensorflow-cbow/
embeds = None
for i in range(2*context_window):
embedding_i = tf.nn.embedding_lookup(embeddings, train_dataset[:,i])
print('embedding %d shape: %s'%(i,embedding_i.get_shape().as_list()))
emb_x,emb_y = embedding_i.get_shape().as_list()
if embeds is None:
embeds = tf.reshape(embedding_i,[emb_x,emb_y,1])
else:
embeds = tf.concat([embeds,tf.reshape(embedding_i,[emb_x,emb_y,1])],2)
assert embeds.get_shape().as_list()[2]==2*context_window
print("Concat embedding size: %s"%embeds.get_shape().as_list())
avg_embed = tf.reduce_mean(embeds,2,keep_dims=False)
print("Avg embedding size: %s"%avg_embed.get_shape().as_list())
# Compute the softmax loss, using a sample of the negative labels each time.
loss = tf.reduce_mean(
tf.nn.sampled_softmax_loss(weights=softmax_weights, biases=softmax_biases, inputs=avg_embed,
labels=train_labels, num_sampled=num_sampled, num_classes=vocabulary_size))
# Optimizer.
# Note: The optimizer will optimize the softmax_weights AND the embeddings.
# This is because the embeddings are defined as a variable quantity and the
# optimizer's `minimize` method will by default modify all variable quantities
# that contribute to the tensor it is passed.
# See docs on `tf.train.Optimizer.minimize()` for more details.
optimizer = tf.train.AdagradOptimizer(1.0).minimize(loss)
# Compute the similarity between minibatch examples and all embeddings.
# We use the cosine distance:
norm = tf.sqrt(tf.reduce_sum(tf.square(embeddings), 1, keep_dims=True))
normalized_embeddings = embeddings / norm
valid_embeddings = tf.nn.embedding_lookup(
normalized_embeddings, valid_dataset)
similarity = tf.matmul(valid_embeddings, tf.transpose(normalized_embeddings))
In [27]:
num_steps = 100001
with tf.Session(graph=graph) as session:
tf.global_variables_initializer().run()
print('Initialized')
average_loss = 0
for step in range(num_steps):
batch_data, batch_labels = generate_batch(
batch_size, context_window)
feed_dict = {train_dataset : batch_data, train_labels : batch_labels}
_, l = session.run([optimizer, loss], feed_dict=feed_dict)
average_loss += l
if step % 2000 == 0:
if step > 0:
average_loss = average_loss / 2000
# The average loss is an estimate of the loss over the last 2000 batches.
print('Average loss at step %d: %f' % (step, average_loss))
average_loss = 0
# note that this is expensive (~20% slowdown if computed every 500 steps)
if step % 10000 == 0:
sim = similarity.eval()
for i in range(valid_size):
valid_word = reverse_dictionary[valid_examples[i]]
top_k = 8 # number of nearest neighbors
nearest = (-sim[i, :]).argsort()[1:top_k+1]
log = 'Nearest to %s:' % valid_word
for k in range(top_k):
close_word = reverse_dictionary[nearest[k]]
log = '%s %s,' % (log, close_word)
print(log)
final_embeddings = normalized_embeddings.eval()
In [ ]: