Please support Cuda 9.1

[philipstephens]

I tried cloning the alpha zero chess from https://github.com/Zeta36/chess-alpha-zero and almost got everything to work. However when I try "python src/chess_zero/run.py self" I get the message

ImportError: Could not find 'cudart64_90.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 9.0 from this URL: https://developer.nvidia.com/cuda-toolkit

The problem is that I have CUDA 9.1 and cannot seem to get CUDA 9.0 on the website. Would it be hard to make a version of TensorFlow-gpu that works with CUDA 9.1 ?

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  NO

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
  Windows 10

• TensorFlow installed from (source or binary):
  Installed from https://www.tensorflow.org/

• TensorFlow version (use command below):
  1.5

• Python version:
  3.6.3

• Bazel version (if compiling from source):

• GCC/Compiler version (if compiling from source):

• CUDA/cuDNN version:
  9.1

• GPU model and memory:
  NVIDIA GeForce GTX 1050 Ti 8GB Ram

• Exact command to reproduce:
  python src/chess_zero/run.py self

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)"

e:\eDownloads>python -c "import tensorflow as tf; print(tf.GIT_VERSION, tf.VERSION)"
Traceback (most recent call last):
File "C:\Users\User\Anaconda3\lib\site-packages\tensorflow\python\platform\self_check.py", line 75, in preload_check
ctypes.WinDLL(build_info.cudart_dll_name)
File "C:\Users\User\Anaconda3\lib\ctypes_init_.py", line 348, in init
self._handle = _dlopen(self._name, mode)
OSError: [WinError 126] The specified module could not be found

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "", line 1, in
File "C:\Users\User\Anaconda3\lib\site-packages\tensorflow_init_.py", line 24, in
from tensorflow.python import *
File "C:\Users\User\Anaconda3\lib\site-packages\tensorflow\python_init_.py", line 49, in
from tensorflow.python import pywrap_tensorflow
File "C:\Users\User\Anaconda3\lib\site-packages\tensorflow\python\pywrap_tensorflow.py", line 30, in
self_check.preload_check()
File "C:\Users\User\Anaconda3\lib\site-packages\tensorflow\python\platform\self_check.py", line 82, in preload_check
% (build_info.cudart_dll_name, build_info.cuda_version_number))
ImportError: Could not find 'cudart64_90.dll'. TensorFlow requires that this DLL be installed in a directory that is named in your %PATH% environment variable. Download and install CUDA 9.0 from this URL: https://developer.nvidia.com/cuda-toolkit

Describe the problem

Describe the problem clearly here. Be sure to convey here why it's a bug in TensorFlow or a feature request.
I am requesting an update because I have CUDA 9.1 and can't get CUDA 9.0

Source code / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached. Try to provide a reproducible test case that is the bare minimum necessary to generate the problem.

[daisylab]

You can get a legacy release of CUDA, e.g., 9.0, from here;

https://developer.nvidia.com/cuda-toolkit-archive

[gunan]

CC @tfboyd

[philipstephens]

On Mon, Feb 12, 2018 at 4:43 PM Gunhan Gulsoy ***@***.***> wrote: CC @tfboyd <https://github.com/tfboyd> — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#16886 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMt8mNou1lvu8n_BofUBI2StNVjFIWTlks5tUL55gaJpZM4R_d1Q> .

Thank you!

[tfboyd]

The current plan of record is to move to CUDA 9.2 when it comes out. CUDA 9.1 did not have any immediate perf improvements and moving drivers again was a concern. I am not sure if windows is any different but in linux having multiple CUDAs installed is somewhat trivial. We discussed this plan with NVIDIA before making the decision. NVIDIA is working on smoothing out the process and revisiting how these releases will work and land.

I am sorry that there is near term pain.

[ghost]

Following instructions to install Cuda 9.0 on Ubuntu (apt-get) doesn't work, it will still install 9.1.

[szm-R]

Hi everyone, I know this issue is closed but I hope somebody would answer! I have downloaded cuda 9.0 from the same link as @daisylab suggested (it's actually cuda 9.0.176), now that I have installed it, it shows cuda 9.1 instead of 9.0! and obviously tensorflow doesn't work complaining it can't find 9.0 libs. Can someone help me out please?

[imdanboy]

@szm2015
try
sudo apt-get install cuda-9-0
instead of
sudo apt-get install cuda

it works for me