Error while importing tensorflow library for deep learning GPU?

Question

I have a GPU of NVIDIA GeForce 940mx GDDR5 2GB on my laptop. I want to use TensorFlow with GPU support.

I tried steps of installing tensorflow from the link https://www.tensorflow.org/install/install_windows

I have installed :

CUDA 9.0 toolkit with all three patches updates available on https://developer.nvidia.com/cuda-90-download-archive?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal
cuDNN 7.1.4 for CUDA toolkit 9.0 from https://developer.nvidia.com/rdp/cudnn-download
pip install tensorflow-gpu

While import tensorflow using:

import tensorflow as tf

I got an error:

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

I have that file in 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin' and my system environment path variable has also been configured to this directory, what else could be the issue?

Please read [Under what circumstances may I add “urgent” or other similar phrases to my question, in order to obtain faster answers?](//meta.stackoverflow.com/q/326569) - the summary is that this is not an ideal way to address volunteers, and is probably counterproductive to obtaining answers. Please refrain from adding this to your questions. — halfer, Aug 05 '18 at 12:16
Sorry @halfer, if I offended you, the way I asked was a bit hurrying/demanding, I agree with that. — tusharRawat, Aug 05 '18 at 12:50
I am not offended. The reason I am keen to discourage begging is that it can be very emotionally manipulative, just like a person on the street asking for money. There are kind people here who spend a lot of their free time here already, and they do not need any pressure to take on more assistance than they have energy for. They are a valuable resource, and the community should look after them. — halfer, Aug 05 '18 at 12:53
`and the community should look after them...` , agree... This also means when somebody answers and get downvoted, like my answer below, the downvoter should gently give a little explanation about the downvote reason. This feature should be implemented [https://meta.stackexchange.com/questions/135/encouraging-people-to-explain-downvotes](https://meta.stackexchange.com/questions/135/encouraging-people-to-explain-downvotes) — BPL, Aug 05 '18 at 13:09

score -1 · Answer 1 · edited Aug 05 '18 at 12:54

I am going to speculate a little bit here so maybe I'm just wrong but I'm going to assume you've got running a command prompt, you install CUDA, you update the env.var PATH adding the CUDA paths and you forgot to restart the command prompt. If it's so, the process where you run Python won't have PATH env.var updated? To make sure about this one, just do python -c 'import os; print(os.environ["PATH"])' and confirm it.

If PATH is ok on the command prompt process, recheck again the CUDA directories, search cudart64_90.dll and make sure the path where that file lives has been added correctly to PATH.

And if the previous step failed, well, I'd say your best chance is opening the Tensorflow file where that dll is being loaded and make some debugging there.

Error while importing tensorflow library for deep learning GPU?

1 Answers1