tf.scatter_nd_update - 'Tensor' object has no attribute '_lazy_read'

[mpekalski]

---

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  yes
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
  Linux Ubuntu 16.04
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
• TensorFlow installed from (source or binary):
  source
• TensorFlow version (use command below):

== cat /etc/issue ===============================================
Linux 3bed2f328777 4.13.0-38-generic #43~16.04.1-Ubuntu SMP Wed Mar 14 17:48:43 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux
VERSION="16.04.5 LTS (Xenial Xerus)"
VERSION_ID="16.04"
VERSION_CODENAME=xenial

== are we in docker =============================================
Yes

== compiler =====================================================
c++ (Ubuntu 5.4.0-6ubuntu1~16.04.10) 5.4.0 20160609
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.


== uname -a =====================================================
Linux 3bed2f328777 4.13.0-38-generic #43~16.04.1-Ubuntu SMP Wed Mar 14 17:48:43 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

== check pips ===================================================
numpy (1.14.5)
protobuf (3.6.1)
tensorflow (1.10.0)

== check for virtualenv =========================================
False

== tensorflow import ============================================
tf.VERSION = 1.10.0
tf.GIT_VERSION = b'unknown'
tf.COMPILER_VERSION = b'unknown'
Sanity check: array([1], dtype=int32)

== env ==========================================================
LD_LIBRARY_PATH /usr/local/cuda/lib64/stubs:/usr/local/cuda/extras/CUPTI/lib64:/usr/local/cuda/lib64/stubs:/usr/local/cuda/lib64/stubs:/usr/local/cuda/extras/CUPTI/lib64:/lib/amd64/server/:/usr/lib/jvm/java-8-openjdk-amd64/jre/lib/amd64/server:/opt/boost/lib:/opt/conda/lib/:/usr/local/cuda/lib64/:/opt/conda/lib/R/lib/:/usr/local/nvidia/lib64/:/usr/local/nvidia/lib:/lib/x86_64-linux-gnu:/usr/local/cuda/extras/CUPTI/lib64:/usr/lib/x86_64-linux-gnu:/opt/opencv/lib
DYLD_LIBRARY_PATH is unset

== nvidia-smi ===================================================
Wed Aug 29 19:57:14 2018       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 396.26                 Driver Version: 396.26                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:01:00.0  On |                  N/A |
|  0%   41C    P0    76W / 250W |    708MiB / 11177MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
+-----------------------------------------------------------------------------+

== cuda libs  ===================================================
/usr/local/cuda-9.2/targets/x86_64-linux/lib/libcudart.so.9.2.148
/usr/local/cuda-9.2/targets/x86_64-linux/lib/libcudart_static.a

1080Ti

• Exact command to reproduce:

import tensorflow as tf

a = tf.random_uniform(minval=0, maxval=9, shape=[], dtype=tf.int32)
img = tf.zeros([10], dtype=tf.int32)
img_z = tf.scatter_nd_update(img, a, tf.zeros_like(a))
                                           

You can collect some of this information using our environment capture script:

https://github.com/tensorflow/tensorflow/tree/master/tools/tf_env_collect.sh

You can obtain the TensorFlow version with

python -c "import tensorflow as tf; print(tf.GIT_VERSION, tf.VERSION)"

Describe the problem

I am not able to update a variable using tf.scatter_nd_update with indices/updates coming from Tensors tf.random_uniform. In another part of my code I use indices and updates coming from tf.nn.top_k and that works fine.

Source code / logs

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-9-ed4c565a2efd> in <module>()
      3 a = tf.random_uniform(minval=0, maxval=10, shape=[], dtype=tf.int32)
      4 img = tf.zeros([9], dtype=tf.int32)
----> 5 img_z = tf.scatter_nd_update(img, a, tf.zeros_like(a))
      6 
      7 with tf.Session() as sess:

/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/state_ops.py in scatter_nd_update(ref, indices, updates, use_locking, name)
    357     return gen_state_ops.scatter_nd_update(
    358         ref, indices, updates, use_locking, name)
--> 359   return ref._lazy_read(gen_state_ops.resource_scatter_nd_update(  # pylint: disable=protected-access
    360       ref.handle, indices, ops.convert_to_tensor(updates, ref.dtype),
    361       name=name))

AttributeError: 'Tensor' object has no attribute '_lazy_read'

[mpekalski]

We can close it. Wrapping img in tf.Variable() solved the issue. Although I am not sure if that is the way to do when in my application img comes from the Dataset and the whole function that ends with scatter_nd_update is a part of the pipeline (called with ds.map()).

I think we can close the ticket.

[IAMAl]

I meet same error when I use Variable on Colabolatory (tf ver:1.14).
Jupyter note shows:
https://github.com/IAMAl/QNet/blob/master/QNet.ipynb

[ruthvik92]

If I wrap img in tf.Variable then I get a new error.

 `ValueError: Shape must be at least rank 1 but is rank 0 for 'ScatterNdUpdate' (op: 'ScatterNdUpdate') with input shapes: [10], [], [].`

I am trying to use tf.scatter_nd_update by setting up a graph and use it in a loop while new values for attributes of tf.scatter_nd_update are supplied using a feed_dict. Any idea how to do this?