CODEWITHSUNDEEP
pythonnumpytensorflow
somniumm
somniumm

Reputation: 115

tensorflow placeholder of single dimension

I am having trouble getting this very simple tensorflow code off the ground. I am trying to do a simple line fitting of the form y = theta1 * x + theta2

I created the data for x and y as a numpy float32 array of shape [10], I created their corresponding placeholders as follows:

tf_x = tf.placeholder(tf.float32, [10])
tf_y = tf.placeholder(tf.float32, [10])

I feed them like follows:

sess.run(train, feed_dict={tf_x: x_data, tf_y: y_data})

the full code is bit long so I created a gist: https://gist.github.com/meowmiau/369393f41b679dd95f4ac4e2e16b0782

The issue I am getting is this:

tensorflow.python.framework.errors.InvalidArgumentError: You must feed a value for placeholder tensor 'Placeholder' with dtype float and shape [10]
 [[Node: Placeholder = Placeholder[dtype=DT_FLOAT, shape=[10], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

However, as far as I know there is no mismatch.

Upvotes: 1

Views: 777

Answers (1)

Norbert Bátfai
Norbert Bátfai

Reputation: 46

Try this modification in your code:

for i in range(1000):
  x_data, y_data = gen_data()
  _, e = sess.run([train, err], feed_dict={tf_x: x_data, tf_y: y_data})
  print e

@Meng Sun:

I think to pass the list [train, err] to sess.run() is a possible solution. The following snippet works the same way:

for i in range(1000):
  x_data, y_data = gen_data()
  feed_dict={tf_x: x_data, tf_y: y_data}
  print(sess.run(err, feed_dict=feed_dict))
  sess.run(train, feed_dict=feed_dict)

In your code the two placeholders throw the error "You must feed a value for placeholder tensor" because sess.run(err) was executed without a feed.

Upvotes: 1

Related Questions