ValueError: Tensor conversion requested dtype float32 for Tensor with dtype int32

neural-network tensorflow conv-neural-network
anamar picture anamar · Feb 13, 2016 · Viewed 25.7k times · Source

I get following error 

ValueError: Tensor conversion requested dtype float32 for Tensor with dtype int32: 'Tensor("Placeholder_1:0", shape=TensorShape([Dimension(128), Dimension(2)]), dtype=int32)'

when I try to calculate cross entropy loss

losses = tf.nn.softmax_cross_entropy_with_logits(scores, input_y)

I use Python 3.4.3.

Any ideas why?

Answer

mrry picture mrry · Feb 13, 2016

It sounds like you have defined input_y—which I am assuming is a tf.placeholder()—as having type tf.int32. Either change this to tf.float32 or add a cast: tf.cast(input_y, tf.float32) or tf.to_float(input_y).

Related questions

Ordering of batch normalization and dropout?

The original question was in regard to TensorFlow implementations specifically. However, the answers are for implementations in general. This general answer is also the correct answer for TensorFlow. When using batch normalization and dropout in TensorFlow (specifically using the contrib.…

Read More
Tensorflow Strides Argument

I am trying to understand the strides argument in tf.nn.avg_pool, tf.nn.max_pool, tf.nn.conv2d. The documentation repeatedly says strides: A list of ints that has length >= 4. The stride of the sliding window …

Read More
How-to run TensorFlow on multiple core and threads

I should start saying that I am completely new to any kind of parallelism/multithreading/multiprocessing programming. Now, I have the chance to run my TensorFlow CNN on 32 cores (each with 2 hyperthreads). I've spent a lot of time trying to …

Read More