HIDDEN_LAYERS = 64
X = tf.placeholder("float", [None, STRING_LENGTH, len(CHARS)])
y = tf.placeholder("float", [None, len(CHARS)])
X_seq = tf.unstack(X, STRING_LENGTH, 1)
lstm_cell = tf.contrib.rnn.BasicLSTMCell(HIDDEN_LAYERS)
#sequence of 12 chars to output of 7
outputs, states = tf.contrib.rnn.static_rnn(lstm_cell, X_seq, dtype=tf.float32)
final_output = outputs[-1]
weights = tf.get_variable("weights", [HIDDEN_LAYERS, len(CHARS)], initializer=tf.random_normal_initializer())
biases = tf.get_variable("biases", [len(CHARS)], initializer=tf.random_normal_initializer())
prediction = tf.add(tf.matmul(final_output, weights), biases)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))
optimizer = tf.train.AdamOptimizer(1e-2)
train_op = optimizer.minimize(cost)