I have to add a row of ones to mnist images dataset, which is batched to 32 samples. Here the code:
(mnist_images, mnist_labels), _ = tf.keras.datasets.mnist.load_data()
dataset = tf.data.Dataset.from_tensor_slices(
(tf.cast(mnist_images[...,tf.newaxis]/255, tf.float32),
tf.cast(mnist_labels,tf.int64)))
dataset = dataset.shuffle(1000).batch(32)
for images,labels in dataset.take(1):
print("Logits: ", mnist_model(images[0:1]).numpy())
b =tf.reshape(images, [784,32], tf.float32)
c = tf.concat(b,tf.ones([1,32], tf.float32),0)
I get the following error, but both are dtype float 32,
ValueError: Tensor conversion requested dtype int32 for Tensor with dtype float32: <tf.Tensor:
shape=(1, 32), dtype=float32, numpy= array([[1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.,1., 1., 1., 1.,
1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1.]],
dtype=float32)>
Is there another way to add a row into images tensor?
It seems like you've just forgotten to use brackets - [ ] .
use:
c = tf.concat([b,tf.ones([1,32], tf.float32)],0)
instead of :
c = tf.concat(b,tf.ones([1,32], tf.float32),0)