Why does TensorFlow give me an error that I am feeding the incorrect shape and type into a placeholder?

Question

I can't figure this out. I've been going back and forth and I understand that I can copy and paste a working tutorial but I want to understand why this doesn't work.

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from tensorflow.examples.tutorials.mnist import input_data


mnist = input_data.read_data_sets('MNIST_data', one_hot=True)   

#simple constants
learning_rate = .01
batch_size = 100
training_epoch = 10
t = 0
l = t

#gather the data
x_train = mnist.train.images
y_train = mnist.train.labels
batch_count = int(len(x_train)/batch_size)

#Set the variables
Y_ = tf.placeholder(tf.float32, [None,10], name = 'Labels')
X = tf.placeholder(tf.float32,[None,784], name = 'Inputs')
W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))

#Build the graph (Y = WX + b)
Y = tf.nn.softmax(tf.matmul(X,W) + b, name = 'softmax')

cross_entropy = -tf.reduce_mean(Y_ * tf.log(Y)) * 1000.0

correct_prediction = tf.equal(tf.argmax(Y,1), tf.argmax(Y_,1))

accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))

train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy) 

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())


    for epoch in range(training_epoch):
        for i in range(batch_count):
            t += batch_size
            print(y_train[l:t].shape)
            print(x_train[l:t].shape)
            print(y_train[l:t].dtype)
            sess.run(train_step,feed_dict={X: x_train[l:t], Y: y_train[l:t]})
            l = t
        print('Epoch = ', epoch)
    print("Accuracy: ", accuracy.eval(feed_dict={X: x_test, Y_: y_test})) 
    print('Done') 

The error message:

InvalidArgumentError: You must feed a value for placeholder tensor 'Labels_2' with dtype float and shape [?,10]
     [[Node: Labels_2 = Placeholder[dtype=DT_FLOAT, shape=[?,10], _device="/job:localhost/replica:0/task:0/device:GPU:0"]

I also understand that there is more than I need to add to get this to work but I want to struggle through that on my own for now. I'm running this on a jupyter notebook. I am positive that the y_train has a shape (100, 10) and a type float64.

I've been stuck for a few days, so I appreciate the help.

Answer 1

You need to feed the placeholder tensor for Y_ when you call sess.run .

In feed_dict , just change Y: y_train[l:t] to Y_: y_train[l:t] . This will feed the y_train[l:t] into the placeholder.

Answer 2

在此行中将Y更改为Y_：

sess.run(train_step,feed_dict={X: x_train[l:t], Y_: y_train[l:t]})