-
Notifications
You must be signed in to change notification settings - Fork 288
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
NVIDIA Driver 555.85 with WSL2 returns error 500 "named symbol not found" when running CUDA apps inside container #520
Comments
This is fixed in nvidia-container-toolkit 1.14.4 and in 1.15.0. https://gitlab.com/nvidia/container-toolkit/libnvidia-container/-/merge_requests/239
|
This change adds the new libnvdxgdmal.so.1 library to the list of files copied from the DriverStore. Signed-off-by: Jakub Bujak <[email protected]>
Add libnvdxgdmal library See merge request nvidia/container-toolkit/container-toolkit!529
Closing this since Docker Desktop 4.31 has been released. |
NVIDIA drivers 555.xx and newer for Windows have added a library called
libnvdxgdmal.so.1
that must be mapped into the container for CUDA to continue working in containers under WSL2.The nvidia-container-toolkit must be updated in order to add the reference to this new library.
If the library is missing from the container because nvidia-container-toolkit is not updated and 555.xx or newer drivers are used, then CUDA initialization will return error 500 "named symbol not found" (CUDA_ERROR_NOT_FOUND).
This is cited in https://www.nvidia.com/en-us/geforce/forums/game-ready-drivers/13/543186/geforce-grd-55585-feedback-thread-released-52124/3453787/ as bug 4668302, "PyTorch-CUDA Docker not compatible with CUDA 12.5/GRD 555.85".
For Docker Desktop users the fix has been released as part of the Docker Desktop 4.31.0 update.
The text was updated successfully, but these errors were encountered: