Is it possible to get cuda or cuda containers running without cuda-samples?

[pdelbarba]

Currently I have:

    cuda-driver \
    cuda-cudart \
    nvidia-docker \

docker run --rm --runtime nvidia <image>
results in
docker: Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #1: error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: driver error: failed to process request: unknown.

however, including cuda-samples causes it to run fine but appears to add ~500MB to the rootfs size and I'm very disk space sensitive.

Is there a particular document I should be looking at for this? If not, let me know and I'll gladly write something up once I get this working.

UPDATE: I was able to get this working by adding cuda-libraries to my IMAGE_INSTALL in addition to the above. I'm not sure if this is the correct solution and it seems odd that this isn't already a dependency of cuda-driver or cuda-cudart but at least it works. Is this the intended way to include these?

[madisongh]

UPDATE: I was able to get this working by adding cuda-libraries to my IMAGE_INSTALL in addition to the above. I'm not sure if this is the correct solution and it seems odd that this isn't already a dependency of cuda-driver or cuda-cudart but at least it works. Is this the intended way to include these?

Yes. cuda-libraries is an empty package that RDEPENDS on all of the CUDA library packages. Those, in turn depend on the CUDA runtime and driver.