Error:55: Could not load dynamic library 'libcudnn.so.7

[Lip651]

System information

• Operating System: 18.04.4 LTS
• TensorFlow installed from: source
• TensorFlow version: 2.1.0
• Python version: 3.6.9
• Installed using: virtualenv, and then pip in order to build the wheel as mentionned in the source installation
• Bazel version: bazel-1.2.1 as mentionned in my configure.py file. No other version was possible
• GCC/Compiler: 7.4.0
• CUDA: 10.2.89
• cuDNN: 7.6.5
• GPU model and memory: GeForce RTX 2070, 8 Gb
• NVIDIA-SMI: 440.33.01
• Driver Version: NVIDIA-SMI 440.33.01

Describe the problem
I have followed carefully the installation guide of cuDNN, and tensorflow from source, didn't get any error during the installation, but when I call tf.test.is_gpu_available() I get one signle error which is:
Could not load dynamic library 'libcudnn.so.7'; dlerror: libcudnn.so.7: cannot open shared object file: No such file or directory
I have looked at the posts already dealing with this question, tried there answers, but nothing changes. I have tried:

• export LD_LIBRARY_PATH=/usr/local/cuda/lib64/ from this post
• sudo sh -c "echo '/usr/local/cuda/lib64\n/usr/local/cuda/lib >> /etc/ld.so.conf.d/nvidia.conf"
sudo ldconfig from here
• export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH from here
  After rebooting each time, I still get the error mentioned above.
• I have also verified where is the libcudnn.so.7 in the root file and it is in:
  /usr/local/cuda-10.2/targets/x86_64-linux/lib
  and important, also in /usr/local/cuda-10.2/targets/x86_64-linux/lib there is a file called:
  libcudnn.so.7.6.5

Please consider that I am not a very experienced Programmer.
Any help would be extremely appreciated. Thank you for your time.

Provide the exact sequence of commands / steps that you executed before running into the problem

Activating my virtual environment, in Python (I have checked that the interpreter is indeed the one in my virtualenv):

from __future__ import absolute_import, division, print_function, unicode_literals
import tensorflow as tf
tf.test.is_gpu_available()

Any other info / logs
The entire error I get calling tf.test.is_gpu_available() is

2020-02-02 20:22:17.488097: I tensorflow/core/platform/profile_utils/cpu_utils.cc:101] CPU Frequency: 3493375000 Hz
2020-02-02 20:22:17.489597: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7fbde003fa20 initialized for platform Host (this does not guarantee that XLA will be used). Devices:
2020-02-02 20:22:17.489647: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
2020-02-02 20:22:17.492686: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcuda.so.1
2020-02-02 20:22:17.598433: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x3d91ff0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:
2020-02-02 20:22:17.598490: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): GeForce RTX 2070, Compute Capability 7.5
2020-02-02 20:22:17.599531: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1558] Found device 0 with properties: 
pciBusID: 0000:09:00.0 name: GeForce RTX 2070 computeCapability: 7.5
coreClock: 1.815GHz coreCount: 36 deviceMemorySize: 7.79GiB deviceMemoryBandwidth: 417.29GiB/s
2020-02-02 20:22:17.599953: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.2
2020-02-02 20:22:17.601575: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10
2020-02-02 20:22:17.603388: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcufft.so.10
2020-02-02 20:22:17.603663: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcurand.so.10
2020-02-02 20:22:17.605504: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusolver.so.10
2020-02-02 20:22:17.606545: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcusparse.so.10
""ERROR IS HERE""
2020-02-02 20:22:17.606679: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libcudnn.so.7'; dlerror: libcudnn.so.7: cannot open shared object file: No such file or directory
""DON'T KNOW IF THIS IS AN ERROR FROM THE PREVIOUS ONE""
2020-02-02 20:22:17.606693: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1595] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices...
2020-02-02 20:22:17.606721: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1099] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-02-02 20:22:17.606729: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1105]      0 
2020-02-02 20:22:17.606736: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1118] 0:   N 

[ymodak]

TF 2.1 supports cuda 10.1. Please roll back to cuda 10.1.
See https://www.tensorflow.org/install/gpu#software_requirements

[s4sarath]

@ymodak - I am using cuda 10.1 but no help

[Lip651]

It indeed worked for me, but it wasn't straight forward

[oanush]

@Lip651 ,
Feel free to close the issue if it is resolved.

[tensorflow-bot]

Are you satisfied with the resolution of your issue?
Yes
No

[NeighborhoodCoding]

how to roll back cuda 10.1.?

[pavlexander]

managed to fix it, my comment is here: #20271 (comment)

[EdisonLeeeee]

tensorflow 2.1 requires CUDA 10.1 from here, it is better to downgrade you cuda toolkit.

[alitokur]

Problem is:
Could not load dynamic library 'libcudnn.so.7'; dlerror: libcudnn.so.7: cannot open shared object file: No such file or directory

Solution :
1-) download related cudnn from here
2-) when you inside the extracted folder:

$ sudo cp include/cudnn.h /usr/local/cuda-10.2/include
$ sudo cp lib64/libcudnn* /usr/local/cuda-10.2/lib64
$ sudo chmod a+r /usr/local/cuda-10.2/lib64/libcudnn*

BE CAREFULL: Check your cuda installation. I mean im using cuda-10.0
So:

$ sudo cp include/cudnn.h /usr/local/cuda-10.0/include
...