tf.unstack did not work with tf 1.8 CudnnGRU tensors

[zheolong]

• 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):

$uname -r
3.10.0-327.el7.x86_64

• Mobile device

Not mobile

• TensorFlow installed from (source or binary):
  anaconda tf 1.8

• TensorFlow version (use command below):

$conda list|grep tensor
tensorboard               1.8.0            py36hf484d3e_0
tensorflow                1.8.0                hb381393_0
tensorflow-base           1.8.0            py36h4df133c_0
tensorflow-gpu            1.8.0                h7b35bdc_0

• Python version:

$python3.6 -V
Python 3.6.2 :: Continuum Analytics, Inc.

• Bazel version (if compiling from source):

$bazel version
Build label: 0.4.5
Build target: bazel-out/local-fastbuild/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar
Build time: Thu Mar 16 12:19:38 2017 (1489666778)
Build timestamp: 1489666778
Build timestamp as int: 1489666778

• CUDA/cuDNN version:

$conda list|grep -i cuda
cudatoolkit               8.0                           3    https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free
cudnn                     7.0.5                 cuda8.0_0

• GPU model and memory:

== cat /etc/issue ===============================================
Linux rvab01298.sqa.ztt 3.10.0-327.el7.x86_64 #1 SMP Thu Nov 19 22:10:57 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux
VERSION="7.2 (Paladin)"
VERSION_ID="7.2"
Qihoo360_BUGZILLA_PRODUCT_VERSION=7.2
Qihoo360_SUPPORT_PRODUCT_VERSION=7.2

== are we in docker =============================================
No

== compiler =====================================================
c++ (GCC) 4.9.2
Copyright (C) 2014 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 rvab01298.sqa.ztt 3.10.0-327.el7.x86_64 #1 SMP Thu Nov 19 22:10:57 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

== check pips ===================================================
numpy (1.13.3)
protobuf (3.5.1)
tensorflow (1.8.0)

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

== tensorflow import ============================================
tf.VERSION = 1.8.0
tf.GIT_VERSION = b'unknown'
tf.COMPILER_VERSION = b'unknown'
Sanity check: array([1], dtype=int32)

== env ==========================================================
LD_LIBRARY_PATH :/usr/local/mpc-0.8.1/lib:/usr/local/gmp-4.3.2/lib:/usr/local/mpfr-2.4.2/lib:/gruntdata/qihoo360/cuda/lib64
DYLD_LIBRARY_PATH is unset

== nvidia-smi ===================================================
Wed Sep 12 13:34:30 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 375.26                 Driver Version: 375.26                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K40m          On   | 0000:02:00.0     Off |                    0 |
| N/A   36C    P0    67W / 235W |   1161MiB / 11439MiB |     39%      Default |
+-------------------------------+----------------------+----------------------+
|   1  Tesla K40m          On   | 0000:03:00.0     Off |                    0 |
| N/A   35C    P0    60W / 235W |     73MiB / 11439MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0     13950    C   bin/arks                                       868MiB |
|    0     27880    C   python3.6                                      288MiB |
|    1     27880    C   python3.6                                       71MiB |
+-----------------------------------------------------------------------------+

== cuda libs  ===================================================
/usr/local/cuda-8.0/doc/man/man7/libcudart.7
/usr/local/cuda-8.0/doc/man/man7/libcudart.so.7
/usr/local/cuda-8.0/lib64/libcudart_static.a
/usr/local/cuda-8.0/lib64/libcudart.so.8.0.61
/usr/local/cuda-7.5/doc/man/man7/libcudart.7
/usr/local/cuda-7.5/doc/man/man7/libcudart.so.7
/usr/local/cuda-7.5/lib64/libcudart.so.7.5.18
/usr/local/cuda-7.5/lib64/libcudart_static.a
/usr/local/cuda-7.5/lib/libcudart.so.7.5.18
/usr/local/cuda-7.5/lib/libcudart_static.a

Describe the problem

tf.unstack did not work as expected. It did not reduce R rank tensor to R-1 rank tensor

Source code / logs

code:

#! /usr/bin/env python
# -*- coding: utf-8 -*-

import sys
import tensorflow as tf
rnn_model = tf.contrib.cudnn_rnn.CudnnGRU(
        num_layers=1,
        num_units=64,
        direction='unidirectional')
rnn_model.build([3, 1, 3])
inputs=[[[1,1,1],[1,1,1],[1,1,1]]]
inputs_tensor= tf.convert_to_tensor(inputs, dtype=tf.float32)
print(tf.shape(inputs_tensor))
rnn_out, rnn_state = rnn_model(inputs_tensor)
print("rnn_state: ", rnn_state)
rnn_layers = tf.unstack(rnn_state)
print("rnn_layers", rnn_layers)

paste the code to file demo.py, then run from linux command line:

$ python3.6 demo.py

output:

Tensor("Shape:0", shape=(3,), dtype=int32)
rnn_state:  (<tf.Tensor 'cudnn_gru/CudnnRNN:1' shape=(1, ?, 64) dtype=float32>,)
rnn_layers [<tf.Tensor 'unstack:0' shape=(1, ?, 64) dtype=float32>]

the rnn_layers should be rnn_layers [<tf.Tensor 'unstack:0' shape=(?, 64) dtype=float32>]

[tensorflowbutler]

Thank you for your post. We noticed you have not filled out the following field in the issue template. Could you update them if they are relevant in your case, or leave them as N/A? Thanks.
Bazel version
Exact command to reproduce

[bignamehyp]

rnn_state is a tuple, not a tensor.

Please debug yourself before posting issue on TensorFlow.

This question is better asked on StackOverflow since it is not a bug or feature request. There is also a larger community that reads questions there.

If you think we've misinterpreted a bug, please comment again with a clear explanation, as well as all of the information requested in the issue template. Thanks!

[zheolong]

@bignamehyp rnn_state is a tuple? but I print it out as 'Tensor', moreover, the doc of tf.contrib.cudnn_rnn.CudnnGRU shows returns as Output tensor(s).

stackoverflow issue added, the link: stackoverflow issue

[zheolong]

Thank you for your post. We noticed you have not filled out the following field in the issue template. Could you update them if they are relevant in your case, or leave them as N/A? Thanks.
Bazel version
Exact command to reproduce

ok, Bazel version and exact command added.

[zheolong]

@bignamehyp rnn_state is a tuple? but I print it out as 'Tensor', moreover, the doc of tf.contrib.cudnn_rnn.CudnnGRU shows returns as Output tensor(s).

stackoverflow issue added, the link: [stackoverflow issue](https://stackoverflow.com/questions/52306358/tf-unstack-did-not-work-with-tf-1-8-cudnngru-tens

rnn_state is a tuple, not a tensor.

Please debug yourself before posting issue on TensorFlow.

This question is better asked on StackOverflow since it is not a bug or feature request. There is also a larger community that reads questions there.

If you think we've misinterpreted a bug, please comment again with a clear explanation, as well as all of the information requested in the issue template. Thanks!

ok , get it done, thx