RNNCE.py

import os
import time
import datetime
from tensorflow import flags
import argparse
import tensorflow as tf
import numpy as np
import embed_char as tool

class TextRNN(object):
    """
    <Parameters>
        - sequence_length: 최대 문장 길이
        - num_classes: 클래스 개수
        - vocab_size: 등장 단어 수
        - embedding_size: 각 단어에 해당되는 임베디드 벡터의 차원
        - filter_sizes: convolutional filter들의 사이즈
        - num_filters: 각 filter size 별 filter 수
        - l2_reg_lambda: 각 weights, biases에 대한 l2 regularization 정도
    """

    def __init__(
            self, sequence_length, num_classes, vocab_size,
            embedding_size, filter_sizes, num_filters, batch_size=1, l2_reg_lambda=0.0):
        # Placeholders for input, output and dropout
        self.input_x = tf.placeholder(tf.int32, [None, sequence_length], name="input_x")
        self.input_y = tf.placeholder(tf.float32, [None, num_classes], name="input_y")
        self.batch_size = tf.placeholder(tf.constant(None, dtype=tf.int32))
        # self.dropout_keep_prob = tf.placeholder(tf.float32, name="dropout_keep_prob")

        # Keeping track of l2 regularization loss (optional)
        # l2_loss = tf.constant(0.0)

        # Embedding layer
        """
        <Variable>
            - W: 각 단어의 임베디드 벡터의 성분을 랜덤하게 할당
        """
        with tf.name_scope("embedding"):
        #with tf.device('/cpu:0'), tf.name_scope("embedding"):
            W = tf.Variable(
                tf.random_uniform([vocab_size, embedding_size], -1.0, 1.0, dtype=tf.float32),
                name="W")
            self.embedded_chars = tf.nn.embedding_lookup(W, self.input_x)
            # self.embedded_chars_expanded = tf.expand_dims(self.embedded_chars, -1)

        with tf.name_scope("lstm"):
            cell = tf.nn.rnn_cell.BasicLSTMCell(embedding_size)

            self.cell = cell
            self.initial_state = cell.zero_state(batch_size, tf.float32)

        with tf.name_scope("FullyConnected"):
            self.fc_W = tf.Variable(tf.random_uniform([embedding_size, num_classes], -1.0, 1.0))
            self.fc_b = tf.Variable(tf.constant(0.1, shape=[num_classes]))

        # 출력값
        with tf.name_scope("output"):
            outputs, self.last_state = tf.nn.dynamic_rnn(self.cell, self.embedded_chars,
                                                        initial_state=self.initial_state, dtype=tf.float32)
            # output = outputs[:, -1, :]
            output = tf.reshape(outputs, [-1, embedding_size])

            self.logits = tf.matmul(output, self.fc_W) + self.fc_b
            self.predictions = tf.argmax(self.logits, axis = 1, name="predictions")

        # Loss
        with tf.name_scope("loss"):
            losses = tf.nn.softmax_cross_entropy_with_logits_v2(logits=self.logits, labels=self.input_y)
            self.loss = tf.reduce_mean(losses)

        # Accuracy
        with tf.name_scope("accuracy"):
            correct_predictions = tf.equal(self.predictions, tf.argmax(self.input_y, 1))
            self.accuracy = tf.reduce_mean(tf.cast(correct_predictions, "float"), name="accuracy")

# data loading
data_path = 'data/using/rt-polarity.all'
contents, points = tool.loading_rdata(data_path, eng=True, num=True, punc=False)
contents = tool.cut(contents,cut=4)

# tranform document to vector
max_document_length = 200
x, vocabulary, vocab_size = tool.make_input(contents,max_document_length)
print('사전단어수 : %s' % (vocab_size))
print('사전단어수 : %d' % len(vocabulary))
y = tool.make_output(points, threshold=0.5)

# divide dataset into train/test set
x_train, x_test, y_train, y_test = tool.divide(x,y,train_prop=0.8)



parser = argparse.ArgumentParser()

# Model Hyperparameters
parser.add_argument("--embedding_dim", type=int, default=128, help="Dimensionality of embedded vector (default: 128)")
parser.add_argument("--filter_sizes", type=str, default="3,4,5", help="Comma-separated filter sizes (default: '3,4,5')")
parser.add_argument("--num_filters", type=int, default=128, help="Number of filters per filter size (default: 128)")
parser.add_argument("--dropout_keep_prob", type=float, default=0.5, help="Dropout keep probability (default: 0.5)")
parser.add_argument("--l2_reg_lambda", type=float, default=0.1, help="L2 regularization lambda (default: 0.0)")

# Training parameters
parser.add_argument("--batch_size", type=int, default=64, help="Batch Size (default: 64)")
parser.add_argument("--num_epochs", type=int, default=200, help="Number of training epochs (default: 200)")
parser.add_argument("--evaluate_every", type=int, default=100, help="Evaluate model on dev set after this many steps (default: 100)")
parser.add_argument("--checkpoint_every", type=int, default=100, help="Save model after this many steps (default: 100)")
parser.add_argument("--num_checkpoints", type=int, default=5, help="Number of checkpoints to store (default: 5)")

# Misc Parameters
parser.add_argument("--allow_soft_placement", type=bool, default=True, help="Allow device soft device placement")
parser.add_argument("--log_device_placement", type=bool, default=False, help="Log placement of ops on devices")

FLAGS = parser.parse_args()
#FLAGS._parse_flags()
print("\nParameters:")
#for attr, value in sorted(list(FLAGS)):
#    print("{}={}".format(attr.upper(), value))
print("")

# 3. train the model and test
with tf.Graph().as_default():
    sess = tf.Session()
    with sess.as_default():
        cnn = TextRNN(sequence_length=x_train.shape[1],
                      num_classes=y_train.shape[1],
                      vocab_size=vocab_size,
                      embedding_size=FLAGS.embedding_dim,
                      filter_sizes=list(map(int, FLAGS.filter_sizes.split(","))),
                      num_filters=FLAGS.num_filters,
                      batch_size=FLAGS.batch_size,
                      l2_reg_lambda=FLAGS.l2_reg_lambda)

        # Define Training procedure
        global_step = tf.Variable(0, name="global_step", trainable=False)
        optimizer = tf.train.AdamOptimizer(1e-3)
        # optimizer.minimize(cnn.loss, global_step=global_step)
        grads_and_vars = optimizer.compute_gradients(cnn.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("{}".format(v.name), g)
                sparsity_summary = tf.summary.scalar("{}".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", cnn.loss)
        acc_summary = tf.summary.scalar("accuracy", cnn.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=FLAGS.num_checkpoints)

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

        def train_step(x_batch, y_batch):
            """
            A single training step
            """
            feed_dict = {
                cnn.input_x: x_batch,
                cnn.input_y: y_batch,
                # cnn.dropout_keep_prob: FLAGS.dropout_keep_prob
            }
            _, step, summaries, loss, accuracy = sess.run(
                [train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy],
                feed_dict)
            time_str = datetime.datetime.now().isoformat()
            print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
            train_summary_writer.add_summary(summaries, step)


        def dev_step(x_batch, y_batch, writer=None):
            """
            Evaluates model on a dev set
            """
            feed_dict = {
                cnn.input_x: x_batch,
                cnn.input_y: y_batch,
                # cnn.dropout_keep_prob: 1.0
            }
            step, summaries, loss, accuracy = sess.run(
                [global_step, dev_summary_op, cnn.loss, cnn.accuracy],
                feed_dict)
            time_str = datetime.datetime.now().isoformat()
            print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
            if writer:
                writer.add_summary(summaries, step)


        def batch_iter(data, batch_size, num_epochs, shuffle=True):
            """
            Generates a batch iterator for a dataset.
            """
            data = np.array(data)
            data_size = len(data)
            num_batches_per_epoch = int((len(data) - 1) / batch_size) + 1
            for epoch in range(num_epochs):
                # Shuffle the data at each epoch
                if shuffle:
                    shuffle_indices = np.random.permutation(np.arange(data_size))
                    shuffled_data = data[shuffle_indices]
                else:
                    shuffled_data = data
                for batch_num in range(num_batches_per_epoch):
                    start_index = batch_num * batch_size
                    end_index = min((batch_num + 1) * batch_size, data_size)
                    yield shuffled_data[start_index:end_index]


        # Generate batches
        batches = batch_iter(
            list(zip(x_train, y_train)), FLAGS.batch_size, FLAGS.num_epochs)

        testpoint = 0
        # Training loop. 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 % FLAGS.evaluate_every == 0:
                if testpoint + 100 < len(x_test):
                    testpoint += 100
                else:
                    testpoint = 0
                print("\nEvaluation:")
                dev_step(x_test[testpoint:testpoint+100], y_test[testpoint:testpoint+100], writer=dev_summary_writer)
                print("")
            if current_step % FLAGS.checkpoint_every == 0:
                path = saver.save(sess, checkpoint_prefix, global_step=current_step)
                print("Saved model checkpoint to {}\n".format(path))