Frequently Asked Questions

Administrators of on-premise deployments of Deepgram products sometimes encounter the following issues.

Problems installing the NVIDIA drivers

Sometimes loaded kernel modules can interfere with the installation of the NVIDIA drivers. To correct this issue, try unloading nouveau drivers:

bash
# rmmod nouveau

Failed to find any CUDA devices

  1. Ensure that NVIDIA drivers are installed:

    bash
    $ nvidia-smi

    If not, install them before proceeding.

  2. Ensure that nvidia-docker is installed:

    bash
    $ docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi

    If not, install it before proceeding.

  3. If you are using Docker Swarm, ensure that the default Docker runtime is nvidia.

  4. If you have made changes to NVIDIA drivers, restart Docker:

    bash
    # systemctl restart docker
Was this section helpful?
Yes No