NotFoundError: Could not find valid device for node when generating a random array of integers in tensorflow 2.0

Question

I am trying to generate a random array of integers in tensorflow 2.0 as below (Please note that I am using Colab):

tf.random.set_seed(123)
var = tf.random.normal(shape=[10, 10], mean=5, stddev=2, dtype=tf.int32, seed=123)
var

This gave me the following error:

NotFoundError: Could not find valid device for node.
Node:{{node RandomStandardNormal}}
All kernels registered for op RandomStandardNormal :
  device='XLA_GPU'; T in [DT_INT32, DT_INT64]; dtype in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF]
  device='XLA_CPU'; T in [DT_INT32, DT_INT64]; dtype in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF]
  device='XLA_CPU_JIT'; T in [DT_INT32, DT_INT64]; dtype in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF]
  device='CPU'; dtype in [DT_DOUBLE]
  device='CPU'; dtype in [DT_FLOAT]
  device='CPU'; dtype in [DT_BFLOAT16]
  device='CPU'; dtype in [DT_HALF]
  device='XLA_GPU_JIT'; T in [DT_INT32, DT_INT64]; dtype in [DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF]
  device='GPU'; T in [DT_INT32]; dtype in [DT_DOUBLE]
  device='GPU'; T in [DT_INT32]; dtype in [DT_FLOAT]
  device='GPU'; T in [DT_INT32]; dtype in [DT_HALF]
 [Op:RandomStandardNormal]

Any help is much appreciated!

Answer 1

you used dtype=tf.int32 in your code which is not supported. you should use one of these types:

types: DT_FLOAT, DT_DOUBLE, DT_BFLOAT16, DT_HALF to use: XLA_GPU

types: [DT_DOUBLE],[DT_FLOAT],[DT_BFLOAT16],[DT_HALF] to use: CPU

types: [DT_DOUBLE],[DT_FLOAT],[DT_HALF] to use: GPU

update:

it seems like I manage to work it out, you can use numpy array with numpy random, then convert it to tensor:

import tensorflow as tf
import numpy as np
numpy.random.seed(123)
s = np.random.normal(5, 2, 100)
s=np.reshape(s, (-1, 10))
data_tf = tf.convert_to_tensor(s, tf.int32)