-
Notifications
You must be signed in to change notification settings - Fork 206
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
cuDNN.version can cause Julia to crash, missing cudnn_ops_infer64_8.dll
#1777
Comments
@staticfloat, re #1755 (comment), you said that loading these submodules would be fine? Is this not the case on Windows? |
Everything I said in that comment should still hold, I'm not entirely sure what is causing the crash here. Do we suspect it is because of mismatched library versions? |
I've been getting the exact same issue as well. Tried deleting the CUDA and cuDNN artifacts and redownloading, but it doesn't resolve itself.
|
I also have the same problem:
Occurred specifically when adding a Flux model into the GPU, then executed Flux.train!. With everything else the same, including loading all other variables onto the GPU, except the Flux model, it doesn't crash. When I add the flux model onto the GPU, julia terminates completely.
|
Should be fixed with JuliaRegistries/General#79340 |
Describe the bug
After upgrading to CUDA ver.
4.0.1
, certain operations will cause Julia to crash.To reproduce
This will show
Could not locate cudnn_ops_infer64_8.dll. Please make sure it is in your library path!
Manifest.toml
Expected behavior
Not crash.
Version info
Details on Julia:
Details on CUDA:
Additional context
The issue was first encountered when using Flux,
ConvTranspose
on newest CUDA (4.0.1
) will cause crash while order version (3.13
) will not. While I don't think this issue exit for everyone, but both my laptop and desktop encounter the same issue.The text was updated successfully, but these errors were encountered: