import tensorflow as tf
class LSTM(object):
"""docstring for LSTM"""
def __init__(self, n_batch, n_step, n_input, n_output, n_cell, lr):
super(LSTM, self).__init__()
self.batch = n_batch
self.step = n_step
self.n_input = n_input
self.n_output = n_output
self.cellnum = n_cell
self.lr = lr
self.x = tf.placeholder(tf.float32, [None, n_step, n_input])
self.y = tf.placeholder(tf.float32, [None, n_output])
self.add_in_layer()
self.add_lstm_layer()
self.add_out_layer()
self.get_loss()
self.train()
self.evaluate()
def add_in_layer(self):
x_input = tf.reshape(self.x, [-1, self.n_input])
layer_input = tf.layers.dense(x_input, self.cellnum)
self.layer_input = tf.reshape(
layer_input, [-1, self.step, self.cellnum])
def add_lstm_layer(self):
self.cell = tf.contrib.rnn.BasicLSTMCell(self.cellnum)
self.init_state = self.cell.zero_state(self.batch, dtype=tf.float32)
self.output, self.state = tf.nn.dynamic_rnn(
self.cell, self.layer_input, initial_state=self.init_state, time_major=False)
def add_out_layer(self):
self.layer_output = tf.layers.dense(self.state[1], self.n_output)
def get_loss(self):
self.loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
labels=self.y, logits=self.layer_output))
def train(self):
self.train_op = tf.train.AdamOptimizer(self.lr).minimize(self.loss)
def evaluate(self):
self.acc = tf.cast(tf.equal(tf.argmax(self.y, 1), tf.argmax(self.layer_output, 1)), tf.float32)