In this project, you'll generate your own Simpsons TV scripts using RNNs. You'll be using part of the Simpsons dataset of scripts from 27 seasons. The Neural Network you'll build will generate a new TV script for a scene at Moe's Tavern.
The data is already provided for you. You'll be using a subset of the original dataset. It consists of only the scenes in Moe's Tavern. This doesn't include other versions of the tavern, like "Moe's Cavern", "Flaming Moe's", "Uncle Moe's Family Feed-Bag", etc..
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
data_dir = './data/simpsons/moes_tavern_lines.txt'
text = helper.load_data(data_dir)
# Ignore notice, since we don't use it for analysing the data
text = text[81:]
Play around with view_sentence_range to view different parts of the data.
view_sentence_range = (0, 10)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np
print('Dataset Stats')
print('Roughly the number of unique words: {}'.format(len({word: None for word in text.split()})))
scenes = text.split('\n\n')
print('Number of scenes: {}'.format(len(scenes)))
sentence_count_scene = [scene.count('\n') for scene in scenes]
print('Average number of sentences in each scene: {}'.format(np.average(sentence_count_scene)))
sentences = [sentence for scene in scenes for sentence in scene.split('\n')]
print('Number of lines: {}'.format(len(sentences)))
word_count_sentence = [len(sentence.split()) for sentence in sentences]
print('Average number of words in each line: {}'.format(np.average(word_count_sentence)))
print()
print('The sentences {} to {}:'.format(*view_sentence_range))
print('\n'.join(text.split('\n')[view_sentence_range[0]:view_sentence_range[1]]))
!lspci
Find sentence length average to use it as the RRN sequence length
import nltk
from collections import Counter
from bokeh.charts import Histogram, output_notebook, show
output_notebook()
sentences = set(nltk.sent_tokenize(text[81:]))
sent_c = Counter()
# Get average length
for s in sentences:
sent_c[s] = nltk.word_tokenize(s).__len__()
s_lengths = np.array(list(sent_c.values()))
data = dict(x=s_lengths)
p = Histogram(data, xlabel='sentence lengths', bins=50)
show(p)
print('Mean of sentences length is {}'.format(s_lengths.mean()))
The first thing to do to any dataset is preprocessing. Implement the following preprocessing functions below:
To create a word embedding, you first need to transform the words to ids. In this function, create two dictionaries:
vocab_to_intint_to_vocabReturn these dictionaries in the following tuple (vocab_to_int, int_to_vocab)
import numpy as np
import problem_unittests as tests
def create_lookup_tables(text):
"""
Create lookup tables for vocabulary
:param text: The text of tv scripts split into words
:return: A tuple of dicts (vocab_to_int, int_to_vocab)
"""
vocab = set(text)
vocab_to_int = {word: index for index, word in enumerate(vocab)}
int_to_vocab = {index: word for (word, index) in vocab_to_int.items()}
return vocab_to_int, int_to_vocab
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_create_lookup_tables(create_lookup_tables)
We'll be splitting the script into a word array using spaces as delimiters. However, punctuations like periods and exclamation marks make it hard for the neural network to distinguish between the word "bye" and "bye!".
Implement the function token_lookup to return a dict that will be used to tokenize symbols like "!" into "||Exclamation_Mark||". Create a dictionary for the following symbols where the symbol is the key and value is the token:
This dictionary will be used to token the symbols and add the delimiter (space) around it. This separates the symbols as it's own word, making it easier for the neural network to predict on the next word. Make sure you don't use a token that could be confused as a word. Instead of using the token "dash", try using something like "||dash||".
def token_lookup():
"""
Generate a dict to turn punctuation into a token.
:return: Tokenize dictionary where the key is the punctuation and the value is the token
"""
return {
'.': '||period||',
',': '||comma||',
'"': '||quotation_mark||',
';': '||semicolon||',
'!': '||exclamation_mark||',
'?': '||question_mark||',
'(': '||left_parentheses',
')': '||right_parentheses',
'--': '||dash||',
'\n': '||return||'
}
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_tokenize(token_lookup)
Running the code cell below will preprocess all the data and save it to file.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
# Preprocess Training, Validation, and Testing Data
helper.preprocess_and_save_data(data_dir, token_lookup, create_lookup_tables)
This is your first checkpoint. If you ever decide to come back to this notebook or have to restart the notebook, you can start from here. The preprocessed data has been saved to disk.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
import numpy as np
import problem_unittests as tests
int_text, vocab_to_int, int_to_vocab, token_dict = helper.load_preprocess()
from collections import namedtuple
hyper_params = (('embedding_size', 128),
('lstm_layers', 2),
('keep_prob', 0.7)
)
Hyper = namedtuple('Hyper', map(lambda x: x[0], hyper_params))
HYPER = Hyper(*list(map(lambda x: x[1], hyper_params)))
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer'
print('TensorFlow Version: {}'.format(tf.__version__))
# Check for a GPU
if not tf.test.gpu_device_name():
warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))
Implement the get_inputs() function to create TF Placeholders for the Neural Network. It should create the following placeholders:
name parameter.Return the placeholders in the following the tuple (Input, Targets, LearingRate)
def get_inputs():
"""
Create TF Placeholders for input, targets, and learning rate.
:return: Tuple (input, targets, learning rate)
"""
# We use shape [None, None] to feed any batch size and any sequence length
input_placeholder = tf.placeholder(tf.int64, [None, None],name='input')
# Targets are [batch_size, seq_length]
targets_placeholder = tf.placeholder(tf.int64, [None, None], name='targets')
learning_rate_placeholder = tf.placeholder(tf.float32, name='learning_rate')
return input_placeholder, targets_placeholder, learning_rate_placeholder
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_inputs(get_inputs)
Stack one or more BasicLSTMCells in a MultiRNNCell.
rnn_sizezero_state() functiontf.identity()Return the cell and initial state in the following tuple (Cell, InitialState)
def get_init_cell(batch_size, rnn_size):
"""
Create an RNN Cell and initialize it.
:param batch_size: Size of batches
:param rnn_size: Size of RNNs
:return: Tuple (cell, initialize state)
"""
with tf.name_scope('RNN_layers'):
lstm = tf.contrib.rnn.BasicLSTMCell(rnn_size)
# add a dropout wrapper
drop = tf.contrib.rnn.DropoutWrapper(lstm, output_keep_prob=HYPER.keep_prob)
#cell = tf.contrib.rnn.MultiRNNCell([drop] * HYPER.lstm_layers)
cell = tf.contrib.rnn.MultiRNNCell([lstm] * HYPER.lstm_layers)
_initial_state = cell.zero_state(batch_size, tf.float32)
initial_state = tf.identity(_initial_state, name='initial_state')
return cell, initial_state
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_init_cell(get_init_cell)
Apply embedding to input_data using TensorFlow. Return the embedded sequence.
def get_embed(input_data, vocab_size, embed_dim):
"""
Create embedding for <input_data>.
:param input_data: TF placeholder for text input.
:param vocab_size: Number of words in vocabulary.
:param embed_dim: Number of embedding dimensions
:return: Embedded input.
"""
with tf.name_scope('Embedding'):
embeddings = tf.Variable(
tf.random_uniform([vocab_size, embed_dim], -1.0, 1.0)
)
embed = tf.nn.embedding_lookup(embeddings, input_data)
return embed
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_embed(get_embed)
You created a RNN Cell in the get_init_cell() function. Time to use the cell to create a RNN.
tf.nn.dynamic_rnn()tf.identity()Return the outputs and final_state state in the following tuple (Outputs, FinalState)
def build_rnn(cell, inputs):
"""
Create a RNN using a RNN Cell
:param cell: RNN Cell
:param inputs: Input text data
:return: Tuple (Outputs, Final State)
"""
## NOTES
# dynamic rnn automatically takes the seq size in dim=1 [batch_size, max_time, ...] time_major==false (default)
with tf.name_scope('RNN_output'):
outputs, final_state = tf.nn.dynamic_rnn(cell, inputs, dtype=tf.float32)
final_state = tf.identity(final_state, name='final_state')
return outputs, final_state
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_build_rnn(build_rnn)
Apply the functions you implemented above to:
input_data using your get_embed(input_data, vocab_size, embed_dim) function.cell and your build_rnn(cell, inputs) function.vocab_size as the number of outputs.Return the logits and final state in the following tuple (Logits, FinalState)
def build_nn(cell, rnn_size, input_data, vocab_size):
"""
Build part of the neural network
:param cell: RNN cell
:param rnn_size: Size of rnns
:param input_data: Input data
:param vocab_size: Vocabulary size
:return: Tuple (Logits, FinalState)
"""
num_outputs = vocab_size
## Not sure why the unit test was made without taking into
# account we are handling dynamic tensor shape that we need to infer
# at runtime, so I made an if statement just to pass the test case
#
# Some references: https://goo.gl/vD3egn
# https://goo.gl/E8vT2M
if input_data.get_shape().as_list()[1] is not None:
batch_size = input_data.get_shape().as_list()[0]
seq_len = input_data.get_shape().as_list()[1]
# Infer dynamic tensor shape of input
else:
input_dims = tf.shape(input_data)
batch_size = input_dims[0]
seq_len = input_dims[1]
embed = get_embed(input_data, vocab_size, HYPER.embedding_size)
## NOTES
# dynamic rnn automatically takes the seq size in dim=1 [batch_size, max_time, ...] see: time_major==false (default)
## Output shape
## [batch_size, time_step, rnn_size]
raw_rnn_outputs, final_state = build_rnn(cell, embed)
# Put outputs in rows
# make the output into [batch_size*time_step, rnn_size] for easy matmul
with tf.name_scope('sequence_reshape'):
outputs = tf.reshape(raw_rnn_outputs, [-1, rnn_size], name='rnn_output')
# Question, why are we using linear activation and not softmax ?
# My Guess: because seq2seq.sequence_loss has an efficient way to calculate the loss directly from logits
with tf.name_scope('logits'):
linear_w = tf.Variable(tf.truncated_normal((rnn_size, num_outputs), stddev=0.05), name='linear_w')
linear_b = tf.Variable(tf.zeros(num_outputs), name='linear_b')
logits = tf.matmul(outputs, linear_w) + linear_b
# Reshape the logits back into the original input shape -> [batch_size, seq_len, num_classes]
# We do this beceause the loss function seq2seq.sequence_loss takes as logits a shape of [batch_size,seq_len,num_decoded_symbols]
with tf.name_scope('logits_reshape_to_loss'):
logits = tf.reshape(logits, [batch_size, seq_len, num_outputs], name='logits')
print('logits after reshape: ', logits)
return logits, final_state
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_build_nn(build_nn)
Implement get_batches to create batches of input and targets using int_text. The batches should be a Numpy array with the shape (number of batches, 2, batch size, sequence length). Each batch contains two elements:
[batch size, sequence length][batch size, sequence length]If you can't fill the last batch with enough data, drop the last batch.
For exmple, get_batches([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15], 2, 3) would return a Numpy array of the following:
[
# First Batch
[
# Batch of Input
[[ 1 2 3], [ 7 8 9]],
# Batch of targets
[[ 2 3 4], [ 8 9 10]]
],
# Second Batch
[
# Batch of Input
[[ 4 5 6], [10 11 12]],
# Batch of targets
[[ 5 6 7], [11 12 13]]
]
]t = np.array([1,2,3,4])
np.random.shuffle(t)
t
def get_batches(int_text, batch_size, seq_length):
"""
Return batches of input and target
:param int_text: Text with the words replaced by their ids
:param batch_size: The size of batch
:param seq_length: The length of sequence
:return: Batches as a Numpy array
"""
slice_size = int(batch_size * seq_length)
n_batches = int(len(int_text)/slice_size)
assert n_batches > 0, 'Maybe your batch size is too big ?'
print('n_batches {}\n'.format(n_batches))
# input part
_inputs = np.array(int_text[:n_batches*slice_size])
# target part
_targets = np.array(int_text[1:n_batches*slice_size + 1])
# Group inputs, targets into n_batches x seq_len items
inputs, targets = np.reshape(_inputs, (-1, n_batches, seq_length)), np.reshape(_targets, (-1, n_batches, seq_length))
# Look through the n_batches axes
#print(inputs[:,0,:])
# Swatch the 0 axis with n_batches axes, now 0 axis is n_batches axis
inputs, targets = np.swapaxes(inputs, 0,1), np.swapaxes(targets, 0,1)
# stack inputs and targets on columns
batches = np.column_stack((inputs, targets))
# Reshape into final batches output
batches = batches.reshape((-1, 2, batch_size, seq_length))
return batches
#Check first batch result
#print(get_batches(np.arange(1,200,1), 20, 3)[0])
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_batches(get_batches)
Tune the following parameters:
num_epochs to the number of epochs.batch_size to the batch size.rnn_size to the size of the RNNs.seq_length to the length of sequence.learning_rate to the learning rate.show_every_n_batches to the number of batches the neural network should print progress.# Number of Epochs
num_epochs = 100
# Batch Size
batch_size = 256
# RNN Size
rnn_size = 128
# Sequence Length
# Use the mean of sentences length as sequence length
seq_length = int(s_lengths.mean())
# Learning Rate
learning_rate = 1e-2
# Show stats for every n number of batches
show_every_n_batches = 10
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
save_dir = './save'
Build the graph using the neural network you implemented.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from tensorflow.contrib import seq2seq
train_graph = tf.Graph()
with train_graph.as_default():
vocab_size = len(int_to_vocab)
input_text, targets, lr = get_inputs()
input_data_shape = tf.shape(input_text)
cell, initial_state = get_init_cell(input_data_shape[0], rnn_size)
logits, final_state = build_nn(cell, rnn_size, input_text, vocab_size)
# Probabilities for generating words
probs = tf.nn.softmax(logits, name='probs')
# Loss function
cost = seq2seq.sequence_loss(
logits,
targets,
tf.ones([input_data_shape[0], input_data_shape[1]]))
# Optimizer
optimizer = tf.train.AdamOptimizer(lr)
# Gradient Clipping
gradients = optimizer.compute_gradients(cost)
capped_gradients = [(tf.clip_by_value(grad, -1., 1.), var) for grad, var in gradients]
train_op = optimizer.apply_gradients(capped_gradients)
# write out the graph for tensorboard
with tf.Session(graph=train_graph) as sess:
file_writer = tf.summary.FileWriter('./logs/1', sess.graph)
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
batches = get_batches(int_text, batch_size, seq_length)
with tf.Session(graph=train_graph) as sess:
sess.run(tf.global_variables_initializer())
for epoch_i in range(num_epochs):
state = sess.run(initial_state, {input_text: batches[0][0]})
for batch_i, (x, y) in enumerate(batches):
feed = {
input_text: x,
targets: y,
initial_state: state,
lr: learning_rate}
train_loss, state, _ = sess.run([cost, final_state, train_op], feed)
# Show every <show_every_n_batches> batches
if (epoch_i * len(batches) + batch_i) % show_every_n_batches == 0:
print('Epoch {:>3} Batch {:>4}/{} train_loss = {:.3f}'.format(
epoch_i,
batch_i,
len(batches),
train_loss))
# Save Model
saver = tf.train.Saver()
saver.save(sess, save_dir)
print('Model Trained and Saved')
Save seq_length and save_dir for generating a new TV script.
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
# Save parameters for checkpoint
helper.save_params((seq_length, save_dir))
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import tensorflow as tf
import numpy as np
import helper
import problem_unittests as tests
_, vocab_to_int, int_to_vocab, token_dict = helper.load_preprocess()
seq_length, load_dir = helper.load_params()
Get tensors from loaded_graph using the function get_tensor_by_name(). Get the tensors using the following names:
Return the tensors in the following tuple (InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor)
def get_tensors(loaded_graph):
"""
Get input, initial state, final state, and probabilities tensor from <loaded_graph>
:param loaded_graph: TensorFlow graph loaded from file
:return: Tuple (InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor)
"""
t_input = loaded_graph.get_tensor_by_name('input:0')
t_initial_state = loaded_graph.get_tensor_by_name('initial_state:0')
t_final_state = loaded_graph.get_tensor_by_name('final_state:0')
t_probs = loaded_graph.get_tensor_by_name('probs:0')
return t_input, t_initial_state, t_final_state, t_probs
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_tensors(get_tensors)
Implement the pick_word() function to select the next word using probabilities.
def pick_word(probabilities, int_to_vocab):
"""
Pick the next word in the generated text
:param probabilities: Probabilites of the next word
:param int_to_vocab: Dictionary of word ids as the keys and words as the values
:return: String of the predicted word
"""
word_p = np.random.choice(probabilities,p=probabilities)
word = probabilities.tolist().index(word_p)
return int_to_vocab[word]
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_pick_word(pick_word)
This will generate the TV script for you. Set gen_length to the length of TV script you want to generate.
gen_length = 200
# homer_simpson, moe_szyslak, or Barney_Gumble
prime_word = 'moe_szyslak'
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
loaded_graph = tf.Graph()
with tf.Session(graph=loaded_graph) as sess:
# Load saved model
loader = tf.train.import_meta_graph(load_dir + '.meta')
loader.restore(sess, load_dir)
# Get Tensors from loaded model
input_text, initial_state, final_state, probs = get_tensors(loaded_graph)
# Sentences generation setup
gen_sentences = [prime_word + ':']
prev_state = sess.run(initial_state, {input_text: np.array([[1]])})
# Generate sentences
for n in range(gen_length):
# Dynamic Input
dyn_input = [[vocab_to_int[word] for word in gen_sentences[-seq_length:]]]
dyn_seq_length = len(dyn_input[0])
# Get Prediction
probabilities, prev_state = sess.run(
[probs, final_state],
{input_text: dyn_input, initial_state: prev_state})
pred_word = pick_word(probabilities[dyn_seq_length-1], int_to_vocab)
gen_sentences.append(pred_word)
# Remove tokens
tv_script = ' '.join(gen_sentences)
for key, token in token_dict.items():
ending = ' ' if key in ['\n', '(', '"'] else ''
tv_script = tv_script.replace(' ' + token.lower(), key)
tv_script = tv_script.replace('\n ', '\n')
tv_script = tv_script.replace('( ', '(')
print(tv_script)
It's ok if the TV script doesn't make any sense. We trained on less than a megabyte of text. In order to get good results, you'll have to use a smaller vocabulary or get more data. Luckly there's more data! As we mentioned in the begging of this project, this is a subset of another dataset. We didn't have you train on all the data, because that would take too long. However, you are free to train your neural network on all the data. After you complete the project, of course.
When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_tv_script_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.