Tensorflow Error: Attempting to use uninitialized value beta1_power_18

Question

This is my code in tensorflow for simple neural network:

import tensorflow as tf
import numpy as np

class Model:
    def __init__(self,input_neuron=2,hidden_neuron=10,output_neuron=2):
        self.input_neuron = input_neuron
        self.hidden_neuron = hidden_neuron
        self.output_neuron = output_neuron
        self.x = tf.placeholder(tf.float32,[None,self.input_neuron])
        self.y = tf.placeholder(tf.float32,[None,self.output_neuron])
        self.model = self.graph()
        self.sess = tf.InteractiveSession()
        self.sess.run(tf.global_variables_initializer())



    @staticmethod
    def one_hot_encode(y):
        y_ = np.zeros((len(y),2))
        for i in range(len(y)):
            y_[i,y[i][0]]=1

        return y_

    def graph(self):
        w1=tf.Variable(tf.random_normal([self.input_neuron,self.hidden_neuron]))
        l1=tf.nn.relu(tf.matmul(self.x,w1))
        w2=tf.Variable(tf.random_normal([self.hidden_neuron,self.output_neuron]))
        l2=tf.matmul(l1,w2)

        return l2

    def train(self,xTrain,yTrain):
        yTrain = self.one_hot_encode(yTrain)
        loss = tf.reduce_sum(tf.nn.softmax_cross_entropy_with_logits_v2(
                             logits=self.model,labels=self.y))
        train = tf.train.AdamOptimizer(0.1).minimize(loss)
        for epoch in range(100):
            self.sess.run(train,feed_dict={self.x:xTrain,self.y:yTrain})

    def predict(self,xTest):
        prediction = tf.argmax(self.model)
        return self.sess.run(prediction,feed_dict={x:xTest}) 

When I run this using :

model = Model()
xTrain = np.array([[0,0],[0,1],[1,0],[1,1]])
yTrain = np.array([[0],[1],[1],[0]])
model.train(xTrain,yTrain)

I'm getting this error:

FailedPreconditionError (see above for traceback): Attempting to use uninitialized value beta1_power_18

What am I doing wrong?

Answer 1

You do the self.sess.run(tf.global_variables_initializer()) in the __init__ of your Model class, but only in the train() method do you set up the tf.train.AdamOptimizer(). The latter also creates some variables that need to be initialized. Move the

self.sess.run(tf.global_variables_initializer())

line right after

train = tf.train.AdamOptimizer(0.1).minimize(loss)

and it will work.

Full code (tested):

import tensorflow as tf
import numpy as np

class Model:
    def __init__(self,input_neuron=2,hidden_neuron=10,output_neuron=2):
        self.input_neuron = input_neuron
        self.hidden_neuron = hidden_neuron
        self.output_neuron = output_neuron
        self.x = tf.placeholder(tf.float32,[None,self.input_neuron])
        self.y = tf.placeholder(tf.float32,[None,self.output_neuron])
        self.model = self.graph()
        self.sess = tf.InteractiveSession()

    @staticmethod
    def one_hot_encode(y):
        y_ = np.zeros((len(y),2))
        for i in range(len(y)):
            y_[i,y[i][0]]=1

        return y_

    def graph(self):
        w1=tf.Variable(tf.random_normal([self.input_neuron,self.hidden_neuron]))
        l1=tf.nn.relu(tf.matmul(self.x,w1))
        w2=tf.Variable(tf.random_normal([self.hidden_neuron,self.output_neuron]))
        l2=tf.matmul(l1,w2)

        return l2

    def train(self,xTrain,yTrain):
        yTrain = self.one_hot_encode(yTrain)
        loss = tf.reduce_sum(tf.nn.softmax_cross_entropy_with_logits_v2(
                             logits=self.model,labels=self.y))
        train = tf.train.AdamOptimizer(0.1).minimize(loss)
        self.sess.run(tf.global_variables_initializer())

        for epoch in range(100):
            self.sess.run(train,feed_dict={self.x:xTrain,self.y:yTrain})
        print("Training done!")

    def predict(self,xTest):
        prediction = tf.argmax(self.model)
        return self.sess.run(prediction,feed_dict={x:xTest})

model = Model()
xTrain = np.array([[0,0],[0,1],[1,0],[1,1]])
yTrain = np.array([[0],[1],[1],[0]])
model.train(xTrain,yTrain)

Based on your comment, if you don't want to re-initialize the whole network at each call of the train() method, then you need to initialize the network at the __init__() method and use tf.report_uninitialized_variables() to get all the uninitialized ones an initialize only those in train(). I wrote the method initialize_uninitialized() to do that, based on this answer to a question by Salvador Dali.

Full code (tested):

import tensorflow as tf
import numpy as np

class Model:
    def __init__(self,input_neuron=2,hidden_neuron=10,output_neuron=2):
        self.input_neuron = input_neuron
        self.hidden_neuron = hidden_neuron
        self.output_neuron = output_neuron
        self.x = tf.placeholder(tf.float32,[None,self.input_neuron])
        self.y = tf.placeholder(tf.float32,[None,self.output_neuron])
        self.model = self.graph()
        self.sess = tf.InteractiveSession()
        self.sess.run(tf.global_variables_initializer())

    @staticmethod
    def one_hot_encode(y):
        y_ = np.zeros((len(y),2))
        for i in range(len(y)):
            y_[i,y[i][0]]=1

        return y_

    def graph(self):
        w1=tf.Variable(tf.random_normal([self.input_neuron,self.hidden_neuron]))
        l1=tf.nn.relu(tf.matmul(self.x,w1))
        w2=tf.Variable(tf.random_normal([self.hidden_neuron,self.output_neuron]))
        l2=tf.matmul(l1,w2)

        return l2

    def initialize_uninitialized( self ):
        uninitialized_variables = [v for v in tf.global_variables()
            if v.name.split(':')[0] in set(self.sess.run(tf.report_uninitialized_variables())) ]
        self.sess.run( tf.variables_initializer( uninitialized_variables ) )

    def train(self,xTrain,yTrain):
        yTrain = self.one_hot_encode(yTrain)
        loss = tf.reduce_sum(tf.nn.softmax_cross_entropy_with_logits_v2(
                             logits=self.model,labels=self.y))
        train = tf.train.AdamOptimizer(0.1).minimize(loss)
        self.initialize_uninitialized()

        for epoch in range(100):
            self.sess.run(train,feed_dict={self.x:xTrain,self.y:yTrain})
        print("Training done!")

    def predict(self,xTest):
        prediction = tf.argmax(self.model)
        return self.sess.run(prediction,feed_dict={x:xTest}) 

model = Model()
xTrain = np.array([[0,0],[0,1],[1,0],[1,1]])
yTrain = np.array([[0],[1],[1],[0]])
model.train(xTrain,yTrain)