train.py

#! /usr/bin/env python
import sys

#SELECT WHICH MODEL YOU WISH TO RUN:
from cnn_lstm import CNN_LSTM   #OPTION 0
from lstm_cnn import LSTM_CNN   #OPTION 1
from cnn import CNN             #OPTION 2 (Model by: Danny Britz)
from lstm import LSTM           #OPTION 3
MODEL_TO_RUN = 0


import tensorflow as tf
import numpy as np
import os
import time
import datetime
import batchgen
from tensorflow.contrib import learn

from IPython import embed

# Parameters
# ==================================================

# Data loading params
dev_size = .10

# Model Hyperparameters
embedding_dim  = 32     #128
max_seq_legth = 70 
filter_sizes = [3,4,5]  #3
num_filters = 32
dropout_prob = 0.5 #0.5
l2_reg_lambda = 0.0
use_glove = True #Do we use glove

# Training parameters
batch_size = 128
num_epochs = 10 #200
evaluate_every = 100 #100
checkpoint_every = 100000 #100
num_checkpoints = 1 #Checkpoints to store


# Misc Parameters
allow_soft_placement = True
log_device_placement = False



# Data Preparation
# ==================================================


filename = "tweets.csv"
goodfile = "good_tweets.csv"
badfile = "bad_tweets.csv"


# Load data
print("Loading data...")
x_text, y = batchgen.get_dataset(goodfile, badfile, 5000) #TODO: MAX LENGTH

# Build vocabulary
max_document_length = max([len(x.split(" ")) for x in x_text])
if (not use_glove):
    print "Not using GloVe"
    vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
    x = np.array(list(vocab_processor.fit_transform(x_text)))
else:
    print "Using GloVe"
    embedding_dim = 50
    filename = 'pre-trainer/glove/glove.6B.50d.txt'
    def loadGloVe(filename):
        vocab = []
        embd = []
        file = open(filename,'r')
        for line in file.readlines():
            row = line.strip().split(' ')
            vocab.append(row[0])
            embd.append(row[1:])
        print('Loaded GloVe!')
        file.close()
        return vocab,embd
    vocab,embd = loadGloVe(filename)
    vocab_size = len(vocab)
    embedding_dim = len(embd[0])
    embedding = np.asarray(embd)

    W = tf.Variable(tf.constant(0.0, shape=[vocab_size, embedding_dim]),
                    trainable=False, name="W")
    embedding_placeholder = tf.placeholder(tf.float32, [vocab_size, embedding_dim])
    embedding_init = W.assign(embedding_placeholder)

    session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
    sess = tf.Session(config=session_conf)
    sess.run(embedding_init, feed_dict={embedding_placeholder: embedding})

    from tensorflow.contrib import learn
    #init vocab processor
    vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
    #fit the vocab from glove
    pretrain = vocab_processor.fit(vocab)
    #transform inputs
    x = np.array(list(vocab_processor.transform(x_text)))

    #init vocab processor
    vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
    #fit the vocab from glove
    pretrain = vocab_processor.fit(vocab)
    #transform inputs
    x = np.array(list(vocab_processor.transform(x_text)))


# Randomly shuffle data
np.random.seed(42)
shuffle_indices = np.random.permutation(np.arange(len(y)))
x_shuffled = x[shuffle_indices]
y_shuffled = y[shuffle_indices]

# Split train/test set
# TODO: This is very crude, should use cross-validation
dev_sample_index = -1 * int(dev_size * float(len(y)))
x_train, x_dev = x_shuffled[:dev_sample_index], x_shuffled[dev_sample_index:]
y_train, y_dev = y_shuffled[:dev_sample_index], y_shuffled[dev_sample_index:]
print("Vocabulary Size: {:d}".format(len(vocab_processor.vocabulary_)))
print("Train/Dev split: {:d}/{:d}".format(len(y_train), len(y_dev)))

#embed()


# Training
# ==================================================

with tf.Graph().as_default():
    session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
    sess = tf.Session(config=session_conf)
    with sess.as_default():
        #embed()
        if (MODEL_TO_RUN == 0):
            model = CNN_LSTM(x_train.shape[1],y_train.shape[1],len(vocab_processor.vocabulary_),embedding_dim,filter_sizes,num_filters,l2_reg_lambda)
        elif (MODEL_TO_RUN == 1):
            model = LSTM_CNN(x_train.shape[1],y_train.shape[1],len(vocab_processor.vocabulary_),embedding_dim,filter_sizes,num_filters,l2_reg_lambda)
        elif (MODEL_TO_RUN == 2):
            model = CNN(x_train.shape[1],y_train.shape[1],len(vocab_processor.vocabulary_),embedding_dim,filter_sizes,num_filters,l2_reg_lambda)
        elif (MODEL_TO_RUN == 3):
            model = LSTM(x_train.shape[1],y_train.shape[1],len(vocab_processor.vocabulary_),embedding_dim)
        else:
            print "PLEASE CHOOSE A VALID MODEL!\n0 = CNN_LSTM\n1 = LSTM_CNN\n2 = CNN\n3 = LSTM\n"
            exit();


        # Define Training procedure
        global_step = tf.Variable(0, name="global_step", trainable=False)
        optimizer = tf.train.AdamOptimizer(1e-3)
        grads_and_vars = optimizer.compute_gradients(model.loss)
        train_op = optimizer.apply_gradients(grads_and_vars, global_step=global_step)

        # Keep track of gradient values and sparsity (optional)
        grad_summaries = []
        for g, v in grads_and_vars:
            if g is not None:
                grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name), g)
                sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name), tf.nn.zero_fraction(g))
                grad_summaries.append(grad_hist_summary)
                grad_summaries.append(sparsity_summary)
        grad_summaries_merged = tf.summary.merge(grad_summaries)

        # Output directory for models and summaries
        timestamp = str(int(time.time()))
        out_dir = os.path.abspath(os.path.join(os.path.curdir, "runs", timestamp))
        print("Writing to {}\n".format(out_dir))

        # Summaries for loss and accuracy
        loss_summary = tf.summary.scalar("loss", model.loss)
        acc_summary = tf.summary.scalar("accuracy", model.accuracy)

        # Train Summaries
        train_summary_op = tf.summary.merge([loss_summary, acc_summary, grad_summaries_merged])
        train_summary_dir = os.path.join(out_dir, "summaries", "train")
        train_summary_writer = tf.summary.FileWriter(train_summary_dir, sess.graph)

        # Dev summaries
        dev_summary_op = tf.summary.merge([loss_summary, acc_summary])
        dev_summary_dir = os.path.join(out_dir, "summaries", "dev")
        dev_summary_writer = tf.summary.FileWriter(dev_summary_dir, sess.graph)

        # Checkpoint directory. Tensorflow assumes this directory already exists so we need to create it
        checkpoint_dir = os.path.abspath(os.path.join(out_dir, "checkpoints"))
        checkpoint_prefix = os.path.join(checkpoint_dir, "model")
        if not os.path.exists(checkpoint_dir):
            os.makedirs(checkpoint_dir)
        saver = tf.train.Saver(tf.global_variables(), max_to_keep=num_checkpoints)

        # Write vocabulary
        vocab_processor.save(os.path.join(out_dir, "vocab"))

        # Initialize all variables
        sess.run(tf.global_variables_initializer())

        #TRAINING STEP
        def train_step(x_batch, y_batch,save=False):
            feed_dict = {
              model.input_x: x_batch,
              model.input_y: y_batch,
              model.dropout_keep_prob: dropout_prob
            }
            _, step, summaries, loss, accuracy = sess.run(
                [train_op, global_step, train_summary_op, model.loss, model.accuracy],
                feed_dict)
            time_str = datetime.datetime.now().isoformat()
            print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
            if save:
                train_summary_writer.add_summary(summaries, step)

        #EVALUATE MODEL
        def dev_step(x_batch, y_batch, writer=None,save=False):
            feed_dict = {
              model.input_x: x_batch,
              model.input_y: y_batch,
              model.dropout_keep_prob: 0.5
            }
            step, summaries, loss, accuracy = sess.run(
                [global_step, dev_summary_op, model.loss, model.accuracy],
                feed_dict)
            time_str = datetime.datetime.now().isoformat()
            print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
            if save:
                if writer:
                    writer.add_summary(summaries, step)

        #CREATE THE BATCHES GENERATOR
        batches = batchgen.gen_batch(list(zip(x_train, y_train)), batch_size, num_epochs)
        
        #TRAIN FOR EACH BATCH
        for batch in batches:
            x_batch, y_batch = zip(*batch)
            train_step(x_batch, y_batch)
            current_step = tf.train.global_step(sess, global_step)
            if current_step % evaluate_every == 0:
                print("\nEvaluation:")
                dev_step(x_dev, y_dev, writer=dev_summary_writer)
                print("")
            if current_step % checkpoint_every == 0:
                path = saver.save(sess, checkpoint_prefix, global_step=current_step)
                print("Saved model checkpoint to {}\n".format(path))
        dev_step(x_dev, y_dev, writer=dev_summary_writer)