-
Notifications
You must be signed in to change notification settings - Fork 74k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
DLL load failed for Tensorflow-GPU==1.14, with CUDA 10.0 #42192
Comments
@tueboesen Please, check Your CPU/Python is on 32 bits?Please, refer #36167 and see if it helps you. |
I have the right cuda version and compatibility, and my system is definitely 64 bit, including python. (I have managed to run tensorflow==1.4 just fine, so the problem seems to be with the gpu/cuda part of the setup.
|
@tueboesen Everything seems rights. Can you please share the commands you have used to install TF and GPU inside an environment. Thanks! |
I just went through the process once again, on my desktop computer dedicated for deep learning, just to verify the problem and I ended up with the same issue there. I know this is a rather old version of tensorflow, and I wish I could just use a newer version but unfortunately, I need either version 1.13 or 1.14 in order to run the following code from a research article: My process was the following:
Additional info on this desktop system:
GPU: Nvidia Geforce RTX 2080 Ti A quick pip list on the system gives the following:
Finally it should be noted that I am able to run deep learning codes on both these computers, although they use newer versions of these libraries. I have both cuda 10.1 and 10.0 install right now, but 10.0 is the default in the $PATH environment at the moment due to this. |
There may be various reasons, like your python may be 32 bit, or on windows, we saw that anything downloaded from windows store has issues due to sandboxing, or missing visual studio redistributable. |
@tueboesen |
This issue has been automatically marked as stale because it has no recent activity. It will be closed if no further activity occurs. Thank you. |
Closing as stale. Please reopen if you'd like to work on this further. |
System information
CUDA Version: 10.0
CUDNN Version: 7
OS: Windows 10
Python version: 3.6
GPU: Geforce mx150
I cloned a github repository that requires tensorflow 1.13-1.14, so I just made a new virtual environment and installed tensorflow-gpu==1.14, which installed without any problems.
Then I tried to run a session in order to see that things were working:
However, when I run the above, I get the following error:
I just installed CUDA 10.0, since I realized that this old version of tensorflow would require that version (and restarted afterwards). I have confirmed that CUDA 10.0 is in my path environment, and that it seems to be the version linked in python as shown below:
I'm not really sure what exactly is going wrong here, and the error messages aren't much help either at this point
The text was updated successfully, but these errors were encountered: