CODEWITHSUNDEEP
pythontensorflow
AlexanderSch
AlexanderSch

Reputation: 157

TensorFlow Indices are not valid (out of bounds)

Hi i'am currently trying to run TensorFlow with own image data. But it crashes when i'am trying to run these function: its from mnist.py

def loss_fn(logits, labels):
    batch_size = tf.size(labels)
    labels = tf.expand_dims(labels, 1)
    indices = tf.expand_dims(tf.range(0, batch_size, 1), 1)
    concated = tf.concat(1, [indices, labels])
    onehot_labels = tf.sparse_to_dense(
             concated, tf.pack([batch_size, NUM_CLASSES]), 1.0, 0.0)
    cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits,
                     onehot_labels,name='xentropy')
    loss = tf.reduce_mean(cross_entropy, name='xentropy_mean')
    return loss

with this error:

Compute status: Invalid argument: Indices are not valid (out of bounds).  Shape: dim { size: 100 } dim { size: 447 }

the number 100 is my batch_size and 447 is my number of classes.

i also try to solve that issue like here https://github.com/tensorflow/tensorflow/issues/194 changing the indeces line into this line:

indices = tf.expand_dims(tf.range(0, batch_size, 1), 1)

didn't solve my problem. Does anybody have an idea?

Upvotes: 2

Views: 5863

Answers (1)

jkschin
jkschin

Reputation: 5844

I had this error as well. And I realized my mistake. If you have 10 classes, your label values should range between 0-9, inclusive. The error was reproduced on the TensorFlow CIFAR10 example, used with SVHN dataset. Refer to question and answer below.

TensorFlow CIFAR10 Example

Upvotes: 1

Related Questions