tensorflow: AttributeError: 'module' object has no attribute 'global_variables_initializer'

when I was trying a multi gpu training using python cifar10_multi_gpu_train.py --num_gpus=2, I got an error:

usr@linux:~/tensorflow_source/tensorflow/tensorflow/models/image/cifar10$ python cifar10_multi_gpu_train.py --num_gpus=2
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so locally
Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
Traceback (most recent call last):
  File "cifar10_multi_gpu_train.py", line 280, in <module>
    tf.app.run()
  File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 30, in run
    sys.exit(main(sys.argv[:1] + flags_passthrough))
  File "cifar10_multi_gpu_train.py", line 276, in main
    train()
  File "cifar10_multi_gpu_train.py", line 229, in train
    init = tf.global_variables_initializer()
AttributeError: 'module' object has no attribute 'global_variables_initializer'

Seemingly there’s no global_variables_initializer of tf or in other module?

About this issue

Original URL
State: closed
Created 8 years ago
Reactions: 13
Comments: 17

Commits related to this issue

weirdness between 0.11 and 0.12? see tensorflow/tensorflow#5514 — committed to 007/CarND by deleted user 7 years ago

Most upvoted comments

This was introduced lately. Consider upgrading tf or use tf.initialize_all_variables .

(fixed typo)

+64

schmiflo on Jan 3, 2017

@schmiflo you mean tf.initialize_all_variables there is a typo.

+11

liusida on Nov 13, 2016

@liusida I changed global_variables_initializer into initialize_all_variables, and it did work, I haven’t updated my tf to the latest version, maybe it has already had global_variables_initializer

+10

zakizhou on Nov 13, 2016

I solved this by using tf.compat.v1.global_variables_initializer()

mohamedScikitLearn on Jul 9, 2020

@zakizhou yes, i think you are right.

https://github.com/tensorflow/tensorflow/blob/20c3d37ecc9bef0e106002b9d01914efd548e66b/tensorflow/python/ops/variables.py#L1170

@deprecated("2017-03-02", "Use `tf.global_variables_initializer` instead.")
def initialize_all_variables():
  """See `tf.global_variables_initializer`."""
  return global_variables_initializer()

This change was made just 12 days ago

https://github.com/tensorflow/tensorflow/commit/4cbdead95f22de74bcbc72a68c9a38d465202db9#diff-ae1a8f7b66539f000615a4ab7e4b2151

liusida on Nov 16, 2016

@EricBuist if u are using tensorflow version 2.x u can try this: import tensorflow.compat.v1 as tf tf.disable_v2_behavior()

TanihiraDev on Jun 13, 2020

It would be nice if you update your tutorial here https://www.tensorflow.org/tutorials/mnist/beginners/ to use the new method.

solomatov on Jan 10, 2017

@schmiflo Thankyou for your solution. However I got error by just blindly copy and paste the code. There is typo in tf.intialize_all_variables. It should be tf.initialize_all_variables. Could you fix it to prevent confusion?

insikk on Dec 3, 2016

It works! Thank you! I changed global_variables_initializer into initialize_all_variables.

yangliuy on Nov 15, 2016

best answer:)))

Mohammad-DSAI on Oct 4, 2023